id sid tid token lemma pos fcis-7983 1 1 frontiers frontier NOUN fcis-7983 1 2 in in ADP fcis-7983 1 3 computing computing NOUN fcis-7983 1 4 and and CCONJ fcis-7983 1 5 intelligent intelligent ADJ fcis-7983 1 6 systems system NOUN fcis-7983 1 7 issn issn VERB fcis-7983 1 8 : : PUNCT fcis-7983 1 9 2832 2832 NUM fcis-7983 1 10 - - SYM fcis-7983 1 11 6024 6024 NUM fcis-7983 1 12 | | NOUN fcis-7983 1 13 vol vol NOUN fcis-7983 1 14 . . PROPN fcis-7983 2 1 3 3 NUM fcis-7983 2 2 , , PUNCT fcis-7983 2 3 no no INTJ fcis-7983 2 4 . . NOUN fcis-7983 2 5 3 3 NUM fcis-7983 2 6 , , PUNCT fcis-7983 2 7 2023 2023 NUM fcis-7983 2 8 6 6 NUM fcis-7983 2 9 named name VERB fcis-7983 2 10 entity entity NOUN fcis-7983 2 11 recognition recognition NOUN fcis-7983 2 12 of of ADP fcis-7983 2 13 electronic electronic ADJ fcis-7983 2 14 medical medical ADJ fcis-7983 2 15 records record NOUN fcis-7983 2 16 based base VERB fcis-7983 2 17 on on ADP fcis-7983 2 18 multi multi ADJ fcis-7983 2 19 - - ADJ fcis-7983 2 20 feature feature ADJ fcis-7983 2 21 fusion fusion NOUN fcis-7983 2 22 xiaoqin xiaoqin ADJ fcis-7983 2 23 tan tan PROPN fcis-7983 2 24 * * PROPN fcis-7983 2 25 school school NOUN fcis-7983 2 26 of of ADP fcis-7983 2 27 electrical electrical ADJ fcis-7983 2 28 engineering engineering NOUN fcis-7983 2 29 , , PUNCT fcis-7983 2 30 southwest southwest PROPN fcis-7983 2 31 minzu minzu PROPN fcis-7983 2 32 university university PROPN fcis-7983 2 33 , , PUNCT fcis-7983 2 34 chengdu chengdu PROPN fcis-7983 2 35 , , PUNCT fcis-7983 2 36 china china PROPN fcis-7983 2 37 abstract abstract NOUN fcis-7983 2 38 : : PUNCT fcis-7983 2 39 named name VERB fcis-7983 2 40 entity entity NOUN fcis-7983 2 41 recognition recognition NOUN fcis-7983 2 42 ( ( PUNCT fcis-7983 2 43 ner ner NOUN fcis-7983 2 44 ) ) PUNCT fcis-7983 2 45 is be AUX fcis-7983 2 46 a a DET fcis-7983 2 47 very very ADV fcis-7983 2 48 basic basic ADJ fcis-7983 2 49 task task NOUN fcis-7983 2 50 in in ADP fcis-7983 2 51 natural natural ADJ fcis-7983 2 52 language language NOUN fcis-7983 2 53 processing processing NOUN fcis-7983 2 54 ( ( PUNCT fcis-7983 2 55 nlp nlp NOUN fcis-7983 2 56 ) ) PUNCT fcis-7983 2 57 . . PUNCT fcis-7983 3 1 the the DET fcis-7983 3 2 paper paper NOUN fcis-7983 3 3 studies study VERB fcis-7983 3 4 the the DET fcis-7983 3 5 problem problem NOUN fcis-7983 3 6 of of ADP fcis-7983 3 7 named name VERB fcis-7983 3 8 entity entity NOUN fcis-7983 3 9 recognition recognition NOUN fcis-7983 3 10 in in ADP fcis-7983 3 11 chinese chinese ADJ fcis-7983 3 12 electronic electronic ADJ fcis-7983 3 13 medical medical ADJ fcis-7983 3 14 records record NOUN fcis-7983 3 15 , , PUNCT fcis-7983 3 16 and and CCONJ fcis-7983 3 17 proposes propose VERB fcis-7983 3 18 a a DET fcis-7983 3 19 method method NOUN fcis-7983 3 20 based base VERB fcis-7983 3 21 on on ADP fcis-7983 3 22 the the DET fcis-7983 3 23 bert bert PROPN fcis-7983 3 24 - - PUNCT fcis-7983 3 25 bi bi ADJ fcis-7983 3 26 - - ADJ fcis-7983 3 27 lstmcrf lstmcrf ADJ fcis-7983 3 28 model model NOUN fcis-7983 3 29 . . PUNCT fcis-7983 4 1 in in ADP fcis-7983 4 2 addition addition NOUN fcis-7983 4 3 , , PUNCT fcis-7983 4 4 the the DET fcis-7983 4 5 model model NOUN fcis-7983 4 6 incorporates incorporate VERB fcis-7983 4 7 the the DET fcis-7983 4 8 functionality functionality NOUN fcis-7983 4 9 of of ADP fcis-7983 4 10 radical radical ADJ fcis-7983 4 11 components component NOUN fcis-7983 4 12 and and CCONJ fcis-7983 4 13 dictionaries dictionary NOUN fcis-7983 4 14 to to PART fcis-7983 4 15 improve improve VERB fcis-7983 4 16 the the DET fcis-7983 4 17 recognition recognition NOUN fcis-7983 4 18 accuracy accuracy NOUN fcis-7983 4 19 . . PUNCT fcis-7983 5 1 the the DET fcis-7983 5 2 complexity complexity NOUN fcis-7983 5 3 of of ADP fcis-7983 5 4 chinese chinese ADJ fcis-7983 5 5 medical medical ADJ fcis-7983 5 6 record record NOUN fcis-7983 5 7 entities entity NOUN fcis-7983 5 8 , , PUNCT fcis-7983 5 9 the the DET fcis-7983 5 10 ambiguity ambiguity NOUN fcis-7983 5 11 in in ADP fcis-7983 5 12 language language NOUN fcis-7983 5 13 expression expression NOUN fcis-7983 5 14 , , PUNCT fcis-7983 5 15 and and CCONJ fcis-7983 5 16 the the DET fcis-7983 5 17 lack lack NOUN fcis-7983 5 18 of of ADP fcis-7983 5 19 adequate adequate ADJ fcis-7983 5 20 labeled label VERB fcis-7983 5 21 data datum NOUN fcis-7983 5 22 make make VERB fcis-7983 5 23 traditional traditional ADJ fcis-7983 5 24 rule rule NOUN fcis-7983 5 25 - - PUNCT fcis-7983 5 26 based base VERB fcis-7983 5 27 or or CCONJ fcis-7983 5 28 machine machine NOUN fcis-7983 5 29 learning learning NOUN fcis-7983 5 30 methods method NOUN fcis-7983 5 31 less less ADV fcis-7983 5 32 effective effective ADJ fcis-7983 5 33 . . PUNCT fcis-7983 6 1 to to PART fcis-7983 6 2 address address VERB fcis-7983 6 3 this this DET fcis-7983 6 4 problem problem NOUN fcis-7983 6 5 , , PUNCT fcis-7983 6 6 we we PRON fcis-7983 6 7 adopt adopt VERB fcis-7983 6 8 the the DET fcis-7983 6 9 bert bert NOUN fcis-7983 6 10 - - PUNCT fcis-7983 6 11 bi bi ADJ fcis-7983 6 12 - - ADJ fcis-7983 6 13 lstm lstm ADJ fcis-7983 6 14 - - PUNCT fcis-7983 6 15 crf crf NOUN fcis-7983 6 16 model model NOUN fcis-7983 6 17 , , PUNCT fcis-7983 6 18 which which PRON fcis-7983 6 19 effectively effectively ADV fcis-7983 6 20 captures capture VERB fcis-7983 6 21 contextual contextual ADJ fcis-7983 6 22 information information NOUN fcis-7983 6 23 and and CCONJ fcis-7983 6 24 semantic semantic ADJ fcis-7983 6 25 relationships relationship NOUN fcis-7983 6 26 to to PART fcis-7983 6 27 improve improve VERB fcis-7983 6 28 entity entity NOUN fcis-7983 6 29 recognition recognition NOUN fcis-7983 6 30 accuracy accuracy NOUN fcis-7983 6 31 . . PUNCT fcis-7983 7 1 furthermore furthermore ADV fcis-7983 7 2 , , PUNCT fcis-7983 7 3 to to PART fcis-7983 7 4 further far ADV fcis-7983 7 5 enhance enhance VERB fcis-7983 7 6 the the DET fcis-7983 7 7 model model NOUN fcis-7983 7 8 's 's PART fcis-7983 7 9 performance performance NOUN fcis-7983 7 10 , , PUNCT fcis-7983 7 11 we we PRON fcis-7983 7 12 introduce introduce VERB fcis-7983 7 13 the the DET fcis-7983 7 14 functionality functionality NOUN fcis-7983 7 15 of of ADP fcis-7983 7 16 radical radical ADJ fcis-7983 7 17 components component NOUN fcis-7983 7 18 and and CCONJ fcis-7983 7 19 dictionaries dictionary NOUN fcis-7983 7 20 . . PUNCT fcis-7983 8 1 radical radical ADJ fcis-7983 8 2 components component NOUN fcis-7983 8 3 are be AUX fcis-7983 8 4 an an DET fcis-7983 8 5 important important ADJ fcis-7983 8 6 component component NOUN fcis-7983 8 7 of of ADP fcis-7983 8 8 chinese chinese ADJ fcis-7983 8 9 characters character NOUN fcis-7983 8 10 and and CCONJ fcis-7983 8 11 can can AUX fcis-7983 8 12 be be AUX fcis-7983 8 13 used use VERB fcis-7983 8 14 to to PART fcis-7983 8 15 assist assist VERB fcis-7983 8 16 in in ADP fcis-7983 8 17 identifying identify VERB fcis-7983 8 18 entities entity NOUN fcis-7983 8 19 and and CCONJ fcis-7983 8 20 improve improve VERB fcis-7983 8 21 the the DET fcis-7983 8 22 model model NOUN fcis-7983 8 23 's 's PART fcis-7983 8 24 generalization generalization NOUN fcis-7983 8 25 ability ability NOUN fcis-7983 8 26 . . PUNCT fcis-7983 9 1 we we PRON fcis-7983 9 2 also also ADV fcis-7983 9 3 utilize utilize VERB fcis-7983 9 4 medical medical ADJ fcis-7983 9 5 dictionaries dictionary NOUN fcis-7983 9 6 to to PART fcis-7983 9 7 assist assist VERB fcis-7983 9 8 in in ADP fcis-7983 9 9 entity entity NOUN fcis-7983 9 10 recognition recognition NOUN fcis-7983 9 11 . . PUNCT fcis-7983 10 1 these these DET fcis-7983 10 2 dictionaries dictionary NOUN fcis-7983 10 3 contain contain VERB fcis-7983 10 4 rich rich ADJ fcis-7983 10 5 medical medical ADJ fcis-7983 10 6 terms term NOUN fcis-7983 10 7 and and CCONJ fcis-7983 10 8 vocabulary vocabulary NOUN fcis-7983 10 9 , , PUNCT fcis-7983 10 10 which which PRON fcis-7983 10 11 can can AUX fcis-7983 10 12 effectively effectively ADV fcis-7983 10 13 help help VERB fcis-7983 10 14 the the DET fcis-7983 10 15 model model NOUN fcis-7983 10 16 identify identify VERB fcis-7983 10 17 entities entity NOUN fcis-7983 10 18 . . PUNCT fcis-7983 11 1 the the DET fcis-7983 11 2 proposed propose VERB fcis-7983 11 3 method method NOUN fcis-7983 11 4 is be AUX fcis-7983 11 5 evaluated evaluate VERB fcis-7983 11 6 on on ADP fcis-7983 11 7 the the DET fcis-7983 11 8 public public ADJ fcis-7983 11 9 dataset dataset NOUN fcis-7983 11 10 ccks2019 ccks2019 PROPN fcis-7983 11 11 , , PUNCT fcis-7983 11 12 and and CCONJ fcis-7983 11 13 the the DET fcis-7983 11 14 experimental experimental ADJ fcis-7983 11 15 results result NOUN fcis-7983 11 16 demonstrate demonstrate VERB fcis-7983 11 17 that that SCONJ fcis-7983 11 18 it it PRON fcis-7983 11 19 outperforms outperform VERB fcis-7983 11 20 traditional traditional ADJ fcis-7983 11 21 methods method NOUN fcis-7983 11 22 , , PUNCT fcis-7983 11 23 achieving achieve VERB fcis-7983 11 24 an an DET fcis-7983 11 25 f1 f1 ADJ fcis-7983 11 26 score score NOUN fcis-7983 11 27 improvement improvement NOUN fcis-7983 11 28 of of ADP fcis-7983 11 29 nearly nearly ADV fcis-7983 11 30 7 7 NUM fcis-7983 11 31 percentage percentage NOUN fcis-7983 11 32 points point NOUN fcis-7983 11 33 , , PUNCT fcis-7983 11 34 and and CCONJ fcis-7983 11 35 achieves achieve VERB fcis-7983 11 36 good good ADJ fcis-7983 11 37 experimental experimental ADJ fcis-7983 11 38 results result NOUN fcis-7983 11 39 . . PUNCT fcis-7983 12 1 keywords keyword NOUN fcis-7983 12 2 : : PUNCT fcis-7983 12 3 chinese chinese ADJ fcis-7983 12 4 electronic electronic ADJ fcis-7983 12 5 medical medical ADJ fcis-7983 12 6 records record NOUN fcis-7983 12 7 ; ; PUNCT fcis-7983 12 8 named name VERB fcis-7983 12 9 entity entity NOUN fcis-7983 12 10 recognition recognition NOUN fcis-7983 12 11 ; ; PUNCT fcis-7983 12 12 bert bert NOUN fcis-7983 12 13 - - PUNCT fcis-7983 12 14 bi bi ADJ fcis-7983 12 15 - - ADJ fcis-7983 12 16 lstm lstm ADJ fcis-7983 12 17 - - PUNCT fcis-7983 12 18 crf crf NOUN fcis-7983 12 19 model model NOUN fcis-7983 12 20 ; ; PUNCT fcis-7983 12 21 deep deep ADJ fcis-7983 12 22 learning learning NOUN fcis-7983 12 23 . . PUNCT fcis-7983 13 1 1 1 X fcis-7983 13 2 . . X fcis-7983 13 3 introduction introduction NOUN fcis-7983 13 4 named name VERB fcis-7983 13 5 entity entity NOUN fcis-7983 13 6 recognition recognition NOUN fcis-7983 13 7 ( ( PUNCT fcis-7983 13 8 ner ner NOUN fcis-7983 13 9 ) ) PUNCT fcis-7983 13 10 is be AUX fcis-7983 13 11 the the DET fcis-7983 13 12 extraction extraction NOUN fcis-7983 13 13 of of ADP fcis-7983 13 14 required require VERB fcis-7983 13 15 entities entity NOUN fcis-7983 13 16 from from ADP fcis-7983 13 17 structured structured ADJ fcis-7983 13 18 or or CCONJ fcis-7983 13 19 unstructured unstructured ADJ fcis-7983 13 20 text text NOUN fcis-7983 13 21 . . PUNCT fcis-7983 14 1 in in ADP fcis-7983 14 2 1996 1996 NUM fcis-7983 14 3 , , PUNCT fcis-7983 14 4 the the DET fcis-7983 14 5 muc-6 muc-6 NUM fcis-7983 14 6 conference conference NOUN fcis-7983 14 7 first first ADV fcis-7983 14 8 proposed propose VERB fcis-7983 14 9 the the DET fcis-7983 14 10 task task NOUN fcis-7983 14 11 of of ADP fcis-7983 14 12 named name VERB fcis-7983 14 13 entity entity NOUN fcis-7983 14 14 recognition recognition NOUN fcis-7983 14 15 and and CCONJ fcis-7983 14 16 specified specify VERB fcis-7983 14 17 the the DET fcis-7983 14 18 main main ADJ fcis-7983 14 19 task task NOUN fcis-7983 14 20 as as ADP fcis-7983 14 21 identifying identify VERB fcis-7983 14 22 entities entity NOUN fcis-7983 14 23 in in ADP fcis-7983 14 24 the the DET fcis-7983 14 25 text text NOUN fcis-7983 14 26 to to PART fcis-7983 14 27 be be AUX fcis-7983 14 28 processed process VERB fcis-7983 14 29 . . PUNCT fcis-7983 15 1 with with ADP fcis-7983 15 2 the the DET fcis-7983 15 3 widespread widespread ADJ fcis-7983 15 4 application application NOUN fcis-7983 15 5 of of ADP fcis-7983 15 6 chinese chinese ADJ fcis-7983 15 7 electronic electronic ADJ fcis-7983 15 8 medical medical ADJ fcis-7983 15 9 records record NOUN fcis-7983 15 10 , , PUNCT fcis-7983 15 11 it it PRON fcis-7983 15 12 has have AUX fcis-7983 15 13 become become VERB fcis-7983 15 14 increasingly increasingly ADV fcis-7983 15 15 important important ADJ fcis-7983 15 16 to to PART fcis-7983 15 17 automate automate VERB fcis-7983 15 18 the the DET fcis-7983 15 19 processing processing NOUN fcis-7983 15 20 and and CCONJ fcis-7983 15 21 mining mining NOUN fcis-7983 15 22 of of ADP fcis-7983 15 23 the the DET fcis-7983 15 24 rich rich ADJ fcis-7983 15 25 medical medical ADJ fcis-7983 15 26 information information NOUN fcis-7983 15 27 contained contain VERB fcis-7983 15 28 within within ADP fcis-7983 15 29 them they PRON fcis-7983 15 30 . . PUNCT fcis-7983 16 1 among among ADP fcis-7983 16 2 them they PRON fcis-7983 16 3 , , PUNCT fcis-7983 16 4 named name VERB fcis-7983 16 5 entity entity NOUN fcis-7983 16 6 recognition recognition NOUN fcis-7983 16 7 is be AUX fcis-7983 16 8 an an DET fcis-7983 16 9 important important ADJ fcis-7983 16 10 part part NOUN fcis-7983 16 11 of of ADP fcis-7983 16 12 automated automate VERB fcis-7983 16 13 processing processing NOUN fcis-7983 16 14 of of ADP fcis-7983 16 15 electronic electronic ADJ fcis-7983 16 16 medical medical ADJ fcis-7983 16 17 records record NOUN fcis-7983 16 18 , , PUNCT fcis-7983 16 19 whose whose DET fcis-7983 16 20 purpose purpose NOUN fcis-7983 16 21 is be AUX fcis-7983 16 22 to to PART fcis-7983 16 23 automatically automatically ADV fcis-7983 16 24 identify identify VERB fcis-7983 16 25 entities entity NOUN fcis-7983 16 26 in in ADP fcis-7983 16 27 text text NOUN fcis-7983 16 28 and and CCONJ fcis-7983 16 29 label label VERB fcis-7983 16 30 them they PRON fcis-7983 16 31 as as ADP fcis-7983 16 32 different different ADJ fcis-7983 16 33 categories category NOUN fcis-7983 16 34 , , PUNCT fcis-7983 16 35 such such ADJ fcis-7983 16 36 as as ADP fcis-7983 16 37 diseases disease NOUN fcis-7983 16 38 , , PUNCT fcis-7983 16 39 drugs drug NOUN fcis-7983 16 40 , , PUNCT fcis-7983 16 41 surgeries surgery NOUN fcis-7983 16 42 , , PUNCT fcis-7983 16 43 examinations examination NOUN fcis-7983 16 44 , , PUNCT fcis-7983 16 45 and and CCONJ fcis-7983 16 46 tests test NOUN fcis-7983 16 47 . . PUNCT fcis-7983 17 1 however however ADV fcis-7983 17 2 , , PUNCT fcis-7983 17 3 named name VERB fcis-7983 17 4 entity entity NOUN fcis-7983 17 5 recognition recognition NOUN fcis-7983 17 6 in in ADP fcis-7983 17 7 chinese chinese ADJ fcis-7983 17 8 electronic electronic ADJ fcis-7983 17 9 medical medical ADJ fcis-7983 17 10 records record NOUN fcis-7983 17 11 faces face VERB fcis-7983 17 12 challenges challenge NOUN fcis-7983 17 13 such such ADJ fcis-7983 17 14 as as ADP fcis-7983 17 15 numerous numerous ADJ fcis-7983 17 16 entity entity NOUN fcis-7983 17 17 types type NOUN fcis-7983 17 18 , , PUNCT fcis-7983 17 19 complex complex ADJ fcis-7983 17 20 expressions expression NOUN fcis-7983 17 21 , , PUNCT fcis-7983 17 22 and and CCONJ fcis-7983 17 23 ambiguity ambiguity NOUN fcis-7983 17 24 , , PUNCT fcis-7983 17 25 and and CCONJ fcis-7983 17 26 traditional traditional ADJ fcis-7983 17 27 methods method NOUN fcis-7983 17 28 have have VERB fcis-7983 17 29 poor poor ADJ fcis-7983 17 30 performance performance NOUN fcis-7983 17 31 . . PUNCT fcis-7983 18 1 therefore therefore ADV fcis-7983 18 2 , , PUNCT fcis-7983 18 3 a a DET fcis-7983 18 4 named name VERB fcis-7983 18 5 entity entity NOUN fcis-7983 18 6 recognition recognition NOUN fcis-7983 18 7 method method NOUN fcis-7983 18 8 based base VERB fcis-7983 18 9 on on ADP fcis-7983 18 10 the the DET fcis-7983 18 11 bert bert PROPN fcis-7983 18 12 - - PUNCT fcis-7983 18 13 bi bi ADJ fcis-7983 18 14 - - ADJ fcis-7983 18 15 lstm lstm ADJ fcis-7983 18 16 - - PUNCT fcis-7983 18 17 crf crf NOUN fcis-7983 18 18 model model NOUN fcis-7983 18 19 is be AUX fcis-7983 18 20 proposed propose VERB fcis-7983 18 21 , , PUNCT fcis-7983 18 22 which which PRON fcis-7983 18 23 combines combine VERB fcis-7983 18 24 the the DET fcis-7983 18 25 functions function NOUN fcis-7983 18 26 of of ADP fcis-7983 18 27 radical radical ADJ fcis-7983 18 28 components component NOUN fcis-7983 18 29 and and CCONJ fcis-7983 18 30 dictionaries dictionary NOUN fcis-7983 18 31 to to PART fcis-7983 18 32 improve improve VERB fcis-7983 18 33 the the DET fcis-7983 18 34 accuracy accuracy NOUN fcis-7983 18 35 and and CCONJ fcis-7983 18 36 generalization generalization NOUN fcis-7983 18 37 ability ability NOUN fcis-7983 18 38 of of ADP fcis-7983 18 39 chinese chinese ADJ fcis-7983 18 40 electronic electronic ADJ fcis-7983 18 41 medical medical ADJ fcis-7983 18 42 record record NOUN fcis-7983 18 43 named name VERB fcis-7983 18 44 entity entity NOUN fcis-7983 18 45 recognition recognition NOUN fcis-7983 18 46 . . PUNCT fcis-7983 19 1 specifically specifically ADV fcis-7983 19 2 , , PUNCT fcis-7983 19 3 this this DET fcis-7983 19 4 method method NOUN fcis-7983 19 5 first first ADV fcis-7983 19 6 uses use VERB fcis-7983 19 7 a a DET fcis-7983 19 8 pre pre ADJ fcis-7983 19 9 - - ADJ fcis-7983 19 10 trained train VERB fcis-7983 19 11 bert bert NOUN fcis-7983 19 12 model model NOUN fcis-7983 19 13 to to PART fcis-7983 19 14 obtain obtain VERB fcis-7983 19 15 contextual contextual ADJ fcis-7983 19 16 information information NOUN fcis-7983 19 17 of of ADP fcis-7983 19 18 the the DET fcis-7983 19 19 text text NOUN fcis-7983 19 20 , , PUNCT fcis-7983 19 21 then then ADV fcis-7983 19 22 uses use VERB fcis-7983 19 23 a a DET fcis-7983 19 24 bilstm bilstm NOUN fcis-7983 19 25 network network NOUN fcis-7983 19 26 for for ADP fcis-7983 19 27 feature feature NOUN fcis-7983 19 28 extraction extraction NOUN fcis-7983 19 29 and and CCONJ fcis-7983 19 30 sequence sequence NOUN fcis-7983 19 31 labeling labeling NOUN fcis-7983 19 32 , , PUNCT fcis-7983 19 33 and and CCONJ fcis-7983 19 34 finally finally ADV fcis-7983 19 35 uses use VERB fcis-7983 19 36 a a DET fcis-7983 19 37 crf crf NOUN fcis-7983 19 38 model model NOUN fcis-7983 19 39 to to PART fcis-7983 19 40 constrain constrain VERB fcis-7983 19 41 the the DET fcis-7983 19 42 labeling labeling NOUN fcis-7983 19 43 results result NOUN fcis-7983 19 44 to to PART fcis-7983 19 45 obtain obtain VERB fcis-7983 19 46 the the DET fcis-7983 19 47 final final ADJ fcis-7983 19 48 entity entity NOUN fcis-7983 19 49 recognition recognition NOUN fcis-7983 19 50 results result VERB fcis-7983 19 51 . . PUNCT fcis-7983 20 1 the the DET fcis-7983 20 2 ccks2019 ccks2019 PROPN fcis-7983 20 3 chinese chinese ADJ fcis-7983 20 4 electronic electronic ADJ fcis-7983 20 5 medical medical ADJ fcis-7983 20 6 record record NOUN fcis-7983 20 7 dataset dataset NOUN fcis-7983 20 8 was be AUX fcis-7983 20 9 used use VERB fcis-7983 20 10 for for ADP fcis-7983 20 11 experimental experimental ADJ fcis-7983 20 12 evaluation evaluation NOUN fcis-7983 20 13 , , PUNCT fcis-7983 20 14 and and CCONJ fcis-7983 20 15 comparison comparison NOUN fcis-7983 20 16 and and CCONJ fcis-7983 20 17 analysis analysis NOUN fcis-7983 20 18 were be AUX fcis-7983 20 19 performed perform VERB fcis-7983 20 20 with with ADP fcis-7983 20 21 traditional traditional ADJ fcis-7983 20 22 methods method NOUN fcis-7983 20 23 . . PUNCT fcis-7983 21 1 the the DET fcis-7983 21 2 experimental experimental ADJ fcis-7983 21 3 results result NOUN fcis-7983 21 4 show show VERB fcis-7983 21 5 that that SCONJ fcis-7983 21 6 the the DET fcis-7983 21 7 proposed propose VERB fcis-7983 21 8 named name VERB fcis-7983 21 9 entity entity NOUN fcis-7983 21 10 recognition recognition NOUN fcis-7983 21 11 method method NOUN fcis-7983 21 12 achieves achieve VERB fcis-7983 21 13 good good ADJ fcis-7983 21 14 performance performance NOUN fcis-7983 21 15 , , PUNCT fcis-7983 21 16 with with ADP fcis-7983 21 17 significant significant ADJ fcis-7983 21 18 improvements improvement NOUN fcis-7983 21 19 in in ADP fcis-7983 21 20 precision precision NOUN fcis-7983 21 21 , , PUNCT fcis-7983 21 22 recall recall NOUN fcis-7983 21 23 , , PUNCT fcis-7983 21 24 and and CCONJ fcis-7983 21 25 f1 f1 NOUN fcis-7983 21 26 - - PUNCT fcis-7983 21 27 score score NOUN fcis-7983 21 28 compared compare VERB fcis-7983 21 29 to to ADP fcis-7983 21 30 other other ADJ fcis-7983 21 31 methods method NOUN fcis-7983 21 32 . . PUNCT fcis-7983 22 1 the the DET fcis-7983 22 2 proposed propose VERB fcis-7983 22 3 method method NOUN fcis-7983 22 4 can can AUX fcis-7983 22 5 not not PART fcis-7983 22 6 only only ADV fcis-7983 22 7 effectively effectively ADV fcis-7983 22 8 solve solve VERB fcis-7983 22 9 the the DET fcis-7983 22 10 problem problem NOUN fcis-7983 22 11 of of ADP fcis-7983 22 12 named name VERB fcis-7983 22 13 entity entity NOUN fcis-7983 22 14 recognition recognition NOUN fcis-7983 22 15 in in ADP fcis-7983 22 16 chinese chinese ADJ fcis-7983 22 17 electronic electronic ADJ fcis-7983 22 18 medical medical ADJ fcis-7983 22 19 records record NOUN fcis-7983 22 20 but but CCONJ fcis-7983 22 21 also also ADV fcis-7983 22 22 achieve achieve VERB fcis-7983 22 23 better well ADJ fcis-7983 22 24 results result NOUN fcis-7983 22 25 in in ADP fcis-7983 22 26 practical practical ADJ fcis-7983 22 27 applications application NOUN fcis-7983 22 28 , , PUNCT fcis-7983 22 29 and and CCONJ fcis-7983 22 30 has have VERB fcis-7983 22 31 certain certain ADJ fcis-7983 22 32 promotional promotional ADJ fcis-7983 22 33 and and CCONJ fcis-7983 22 34 application application NOUN fcis-7983 22 35 value value NOUN fcis-7983 22 36 . . PUNCT fcis-7983 23 1 2 2 X fcis-7983 23 2 . . X fcis-7983 23 3 related relate VERB fcis-7983 23 4 work work NOUN fcis-7983 23 5 named name VERB fcis-7983 23 6 entity entity NOUN fcis-7983 23 7 recognition recognition NOUN fcis-7983 23 8 methods method NOUN fcis-7983 23 9 can can AUX fcis-7983 23 10 be be AUX fcis-7983 23 11 classified classify VERB fcis-7983 23 12 into into ADP fcis-7983 23 13 three three NUM fcis-7983 23 14 categories category NOUN fcis-7983 23 15 : : PUNCT fcis-7983 23 16 rule rule NOUN fcis-7983 23 17 - - PUNCT fcis-7983 23 18 based base VERB fcis-7983 23 19 methods method NOUN fcis-7983 23 20 , , PUNCT fcis-7983 23 21 traditional traditional ADJ fcis-7983 23 22 machine machine NOUN fcis-7983 23 23 learning learning NOUN fcis-7983 23 24 methods method NOUN fcis-7983 23 25 , , PUNCT fcis-7983 23 26 and and CCONJ fcis-7983 23 27 deep deep ADJ fcis-7983 23 28 learning learning NOUN fcis-7983 23 29 - - PUNCT fcis-7983 23 30 based base VERB fcis-7983 23 31 methods method NOUN fcis-7983 23 32 . . PUNCT fcis-7983 24 1 among among ADP fcis-7983 24 2 them they PRON fcis-7983 24 3 , , PUNCT fcis-7983 24 4 deep deep ADJ fcis-7983 24 5 learning learning NOUN fcis-7983 24 6 - - PUNCT fcis-7983 24 7 based base VERB fcis-7983 24 8 methods method NOUN fcis-7983 24 9 have have AUX fcis-7983 24 10 become become VERB fcis-7983 24 11 a a DET fcis-7983 24 12 research research NOUN fcis-7983 24 13 hotspot hotspot NOUN fcis-7983 24 14 in in ADP fcis-7983 24 15 recent recent ADJ fcis-7983 24 16 years year NOUN fcis-7983 24 17 due due ADP fcis-7983 24 18 to to ADP fcis-7983 24 19 their their PRON fcis-7983 24 20 ability ability NOUN fcis-7983 24 21 to to PART fcis-7983 24 22 automatically automatically ADV fcis-7983 24 23 capture capture VERB fcis-7983 24 24 input input NOUN fcis-7983 24 25 sentence sentence NOUN fcis-7983 24 26 features feature NOUN fcis-7983 24 27 and and CCONJ fcis-7983 24 28 achieve achieve VERB fcis-7983 24 29 end end NOUN fcis-7983 24 30 - - PUNCT fcis-7983 24 31 to to ADP fcis-7983 24 32 - - PUNCT fcis-7983 24 33 end end NOUN fcis-7983 24 34 named name VERB fcis-7983 24 35 entity entity NOUN fcis-7983 24 36 recognition recognition NOUN fcis-7983 24 37 . . PUNCT fcis-7983 25 1 some some DET fcis-7983 25 2 recent recent ADJ fcis-7983 25 3 research research NOUN fcis-7983 25 4 work work NOUN fcis-7983 25 5 has have AUX fcis-7983 25 6 applied apply VERB fcis-7983 25 7 deep deep ADJ fcis-7983 25 8 learning learning NOUN fcis-7983 25 9 - - PUNCT fcis-7983 25 10 based base VERB fcis-7983 25 11 methods method NOUN fcis-7983 25 12 to to ADP fcis-7983 25 13 the the DET fcis-7983 25 14 named name VERB fcis-7983 25 15 entity entity NOUN fcis-7983 25 16 recognition recognition NOUN fcis-7983 25 17 task task NOUN fcis-7983 25 18 . . PUNCT fcis-7983 26 1 huang huang PROPN fcis-7983 26 2 et et PROPN fcis-7983 26 3 al al PROPN fcis-7983 26 4 applied apply VERB fcis-7983 26 5 the the DET fcis-7983 26 6 bidirectional bidirectional ADJ fcis-7983 26 7 long long ADJ fcis-7983 26 8 short short ADJ fcis-7983 26 9 - - PUNCT fcis-7983 26 10 term term NOUN fcis-7983 26 11 memory memory NOUN fcis-7983 26 12 network network NOUN fcis-7983 26 13 ( ( PUNCT fcis-7983 26 14 bilstm bilstm NOUN fcis-7983 26 15 ) ) PUNCT fcis-7983 26 16 and and CCONJ fcis-7983 26 17 conditional conditional ADJ fcis-7983 26 18 random random ADJ fcis-7983 26 19 field field NOUN fcis-7983 26 20 network network NOUN fcis-7983 26 21 to to ADP fcis-7983 26 22 named name VERB fcis-7983 26 23 entity entity NOUN fcis-7983 26 24 labeling labeling NOUN fcis-7983 26 25 . . PUNCT fcis-7983 27 1 however however ADV fcis-7983 27 2 , , PUNCT fcis-7983 27 3 bilstm bilstm NOUN fcis-7983 27 4 has have AUX fcis-7983 27 5 limited limit VERB fcis-7983 27 6 encoding encoding NOUN fcis-7983 27 7 ability ability NOUN fcis-7983 27 8 for for ADP fcis-7983 27 9 long long ADJ fcis-7983 27 10 sequences sequence NOUN fcis-7983 27 11 and and CCONJ fcis-7983 27 12 performs perform VERB fcis-7983 27 13 poorly poorly ADV fcis-7983 27 14 in in ADP fcis-7983 27 15 terms term NOUN fcis-7983 27 16 of of ADP fcis-7983 27 17 computational computational ADJ fcis-7983 27 18 speed speed NOUN fcis-7983 27 19 . . PUNCT fcis-7983 28 1 strubell strubell VERB fcis-7983 28 2 et et PROPN fcis-7983 28 3 al al PROPN fcis-7983 28 4 . . PROPN fcis-7983 29 1 used use VERB fcis-7983 29 2 convolutional convolutional ADJ fcis-7983 29 3 neural neural ADJ fcis-7983 29 4 networks network NOUN fcis-7983 29 5 ( ( PUNCT fcis-7983 29 6 cnn cnn PROPN fcis-7983 29 7 ) ) PUNCT fcis-7983 29 8 for for ADP fcis-7983 29 9 named name VERB fcis-7983 29 10 entity entity NOUN fcis-7983 29 11 recognition recognition NOUN fcis-7983 29 12 . . PUNCT fcis-7983 30 1 compared compare VERB fcis-7983 30 2 with with ADP fcis-7983 30 3 recurrent recurrent ADJ fcis-7983 30 4 neural neural ADJ fcis-7983 30 5 networks network NOUN fcis-7983 30 6 like like ADP fcis-7983 30 7 bilstm bilstm NOUN fcis-7983 30 8 , , PUNCT fcis-7983 30 9 cnn cnn PROPN fcis-7983 30 10 has have VERB fcis-7983 30 11 faster fast ADJ fcis-7983 30 12 computational computational ADJ fcis-7983 30 13 speed speed NOUN fcis-7983 30 14 . . PUNCT fcis-7983 31 1 however however ADV fcis-7983 31 2 , , PUNCT fcis-7983 31 3 cnn cnn PROPN fcis-7983 31 4 may may AUX fcis-7983 31 5 lose lose VERB fcis-7983 31 6 a a DET fcis-7983 31 7 large large ADJ fcis-7983 31 8 amount amount NOUN fcis-7983 31 9 of of ADP fcis-7983 31 10 global global ADJ fcis-7983 31 11 information information NOUN fcis-7983 31 12 despite despite SCONJ fcis-7983 31 13 its its PRON fcis-7983 31 14 good good ADJ fcis-7983 31 15 local local ADJ fcis-7983 31 16 capturing capturing NOUN fcis-7983 31 17 ability ability NOUN fcis-7983 31 18 . . PUNCT fcis-7983 32 1 previous previous ADJ fcis-7983 32 2 research research NOUN fcis-7983 32 3 has have AUX fcis-7983 32 4 improved improve VERB fcis-7983 32 5 the the DET fcis-7983 32 6 transformer transformer NOUN fcis-7983 32 7 encoder encoder NOUN fcis-7983 32 8 by by ADP fcis-7983 32 9 supplementing supplement VERB fcis-7983 32 10 directional directional ADJ fcis-7983 32 11 information information NOUN fcis-7983 32 12 , , PUNCT fcis-7983 32 13 enhancing enhance VERB fcis-7983 32 14 its its PRON fcis-7983 32 15 encoding encoding NOUN fcis-7983 32 16 ability ability NOUN fcis-7983 32 17 . . PUNCT fcis-7983 33 1 however however ADV fcis-7983 33 2 , , PUNCT fcis-7983 33 3 shallow shallow ADJ fcis-7983 33 4 transformer transformer NOUN fcis-7983 33 5 encoders encoder NOUN fcis-7983 33 6 have have VERB fcis-7983 33 7 the the DET fcis-7983 33 8 disadvantage disadvantage NOUN fcis-7983 33 9 of of ADP fcis-7983 33 10 insufficient insufficient ADJ fcis-7983 33 11 structural structural ADJ fcis-7983 33 12 ability ability NOUN fcis-7983 33 13 at at ADP fcis-7983 33 14 the the DET fcis-7983 33 15 encoding encoding NOUN fcis-7983 33 16 layer layer NOUN fcis-7983 33 17 level level NOUN fcis-7983 33 18 . . PUNCT fcis-7983 34 1 due due ADP fcis-7983 34 2 to to ADP fcis-7983 34 3 the the DET fcis-7983 34 4 unique unique ADJ fcis-7983 34 5 characteristics characteristic NOUN fcis-7983 34 6 of of ADP fcis-7983 34 7 the the DET fcis-7983 34 8 chinese chinese ADJ fcis-7983 34 9 language language NOUN fcis-7983 34 10 , , PUNCT fcis-7983 34 11 chinese chinese PROPN fcis-7983 34 12 named name VERB fcis-7983 34 13 entity entity NOUN fcis-7983 34 14 recognition recognition NOUN fcis-7983 34 15 is be AUX fcis-7983 34 16 more more ADV fcis-7983 34 17 challenging challenging ADJ fcis-7983 34 18 than than ADP fcis-7983 34 19 english english PROPN fcis-7983 34 20 named name VERB fcis-7983 34 21 entity entity NOUN fcis-7983 34 22 recognition recognition NOUN fcis-7983 34 23 , , PUNCT fcis-7983 34 24 even even ADV fcis-7983 34 25 though though SCONJ fcis-7983 34 26 it it PRON fcis-7983 34 27 has have AUX fcis-7983 34 28 been be AUX fcis-7983 34 29 developed develop VERB fcis-7983 34 30 earlier early ADV fcis-7983 34 31 . . PUNCT fcis-7983 35 1 yang yang PROPN fcis-7983 35 2 et et PROPN fcis-7983 35 3 al al PROPN fcis-7983 35 4 . . PROPN fcis-7983 35 5 used use VERB fcis-7983 35 6 a a DET fcis-7983 35 7 segmentation segmentation NOUN fcis-7983 35 8 tool tool NOUN fcis-7983 35 9 to to PART fcis-7983 35 10 segment segment VERB fcis-7983 35 11 chinese chinese ADJ fcis-7983 35 12 sentences sentence NOUN fcis-7983 35 13 before before ADP fcis-7983 35 14 labeling label VERB fcis-7983 35 15 word word NOUN fcis-7983 35 16 sequences sequence NOUN fcis-7983 35 17 . . PUNCT fcis-7983 36 1 however however ADV fcis-7983 36 2 , , PUNCT fcis-7983 36 3 word word NOUN fcis-7983 36 4 segmentation segmentation NOUN fcis-7983 36 5 tools tool NOUN fcis-7983 36 6 inevitably inevitably ADV fcis-7983 36 7 make make VERB fcis-7983 36 8 errors error NOUN fcis-7983 36 9 in in ADP fcis-7983 36 10 word word NOUN fcis-7983 36 11 segmentation segmentation NOUN fcis-7983 36 12 , , PUNCT fcis-7983 36 13 leading lead VERB fcis-7983 36 14 to to ADP fcis-7983 36 15 errors error NOUN fcis-7983 36 16 in in ADP fcis-7983 36 17 identifying identify VERB fcis-7983 36 18 entity entity NOUN fcis-7983 36 19 boundaries boundary NOUN fcis-7983 36 20 . . PUNCT fcis-7983 37 1 this this PRON fcis-7983 37 2 is be AUX fcis-7983 37 3 because because SCONJ fcis-7983 37 4 , , PUNCT fcis-7983 37 5 unlike unlike ADP fcis-7983 37 6 in in ADP fcis-7983 37 7 english english PROPN fcis-7983 37 8 where where SCONJ fcis-7983 37 9 delimiters delimiter NOUN fcis-7983 37 10 are be AUX fcis-7983 37 11 used use VERB fcis-7983 37 12 to to PART fcis-7983 37 13 identify identify VERB fcis-7983 37 14 word word NOUN fcis-7983 37 15 boundaries boundary NOUN fcis-7983 37 16 , , PUNCT fcis-7983 37 17 chinese chinese ADJ fcis-7983 37 18 words word NOUN fcis-7983 37 19 do do AUX fcis-7983 37 20 not not PART fcis-7983 37 21 have have VERB fcis-7983 37 22 natural natural ADJ fcis-7983 37 23 boundaries boundary NOUN fcis-7983 37 24 , , PUNCT fcis-7983 37 25 making make VERB fcis-7983 37 26 segmentation segmentation NOUN fcis-7983 37 27 7 7 NUM fcis-7983 37 28 more more ADV fcis-7983 37 29 difficult difficult ADJ fcis-7983 37 30 . . PUNCT fcis-7983 38 1 word word NOUN fcis-7983 38 2 - - PUNCT fcis-7983 38 3 enhancement enhancement NOUN fcis-7983 38 4 methods method NOUN fcis-7983 38 5 can can AUX fcis-7983 38 6 reduce reduce VERB fcis-7983 38 7 segmentation segmentation NOUN fcis-7983 38 8 errors error NOUN fcis-7983 38 9 and and CCONJ fcis-7983 38 10 increase increase VERB fcis-7983 38 11 chinese chinese ADJ fcis-7983 38 12 semantic semantic ADJ fcis-7983 38 13 and and CCONJ fcis-7983 38 14 boundary boundary ADJ fcis-7983 38 15 information information NOUN fcis-7983 38 16 , , PUNCT fcis-7983 38 17 but but CCONJ fcis-7983 38 18 wu wu PROPN fcis-7983 38 19 et et PROPN fcis-7983 38 20 al al PROPN fcis-7983 38 21 . . PROPN fcis-7983 38 22 argued argue VERB fcis-7983 38 23 that that SCONJ fcis-7983 38 24 this this DET fcis-7983 38 25 method method NOUN fcis-7983 38 26 ignores ignore VERB fcis-7983 38 27 the the DET fcis-7983 38 28 information information NOUN fcis-7983 38 29 in in ADP fcis-7983 38 30 chinese chinese ADJ fcis-7983 38 31 character character NOUN fcis-7983 38 32 structures structure NOUN fcis-7983 38 33 and and CCONJ fcis-7983 38 34 proposed propose VERB fcis-7983 38 35 a a DET fcis-7983 38 36 multi multi ADJ fcis-7983 38 37 - - ADJ fcis-7983 38 38 dimensional dimensional ADJ fcis-7983 38 39 data datum NOUN fcis-7983 38 40 embedding embed VERB fcis-7983 38 41 method method NOUN fcis-7983 38 42 that that PRON fcis-7983 38 43 combines combine VERB fcis-7983 38 44 chinese chinese ADJ fcis-7983 38 45 character character NOUN fcis-7983 38 46 features feature NOUN fcis-7983 38 47 and and CCONJ fcis-7983 38 48 radical radical ADJ fcis-7983 38 49 information information NOUN fcis-7983 38 50 for for ADP fcis-7983 38 51 improvement improvement NOUN fcis-7983 38 52 . . PUNCT fcis-7983 39 1 studies study NOUN fcis-7983 39 2 have have AUX fcis-7983 39 3 shown show VERB fcis-7983 39 4 that that SCONJ fcis-7983 39 5 character character NOUN fcis-7983 39 6 - - PUNCT fcis-7983 39 7 level level NOUN fcis-7983 39 8 named name VERB fcis-7983 39 9 entity entity NOUN fcis-7983 39 10 recognition recognition NOUN fcis-7983 39 11 is be AUX fcis-7983 39 12 more more ADV fcis-7983 39 13 effective effective ADJ fcis-7983 39 14 than than ADP fcis-7983 39 15 word word NOUN fcis-7983 39 16 - - PUNCT fcis-7983 39 17 level level NOUN fcis-7983 39 18 [ [ X fcis-7983 39 19 9,10 9,10 NUM fcis-7983 39 20 ] ] X fcis-7983 39 21 . . PUNCT fcis-7983 40 1 however however ADV fcis-7983 40 2 , , PUNCT fcis-7983 40 3 a a DET fcis-7983 40 4 clear clear ADJ fcis-7983 40 5 disadvantage disadvantage NOUN fcis-7983 40 6 of of ADP fcis-7983 40 7 character character NOUN fcis-7983 40 8 - - PUNCT fcis-7983 40 9 based base VERB fcis-7983 40 10 named name VERB fcis-7983 40 11 entity entity NOUN fcis-7983 40 12 recognition recognition NOUN fcis-7983 40 13 is be AUX fcis-7983 40 14 the the DET fcis-7983 40 15 loss loss NOUN fcis-7983 40 16 of of ADP fcis-7983 40 17 rich rich ADJ fcis-7983 40 18 information information NOUN fcis-7983 40 19 in in ADP fcis-7983 40 20 words word NOUN fcis-7983 40 21 . . PUNCT fcis-7983 41 1 therefore therefore ADV fcis-7983 41 2 , , PUNCT fcis-7983 41 3 fully fully ADV fcis-7983 41 4 integrating integrate VERB fcis-7983 41 5 dictionary dictionary ADJ fcis-7983 41 6 information information NOUN fcis-7983 41 7 into into ADP fcis-7983 41 8 character character NOUN fcis-7983 41 9 models model NOUN fcis-7983 41 10 is be AUX fcis-7983 41 11 a a DET fcis-7983 41 12 major major ADJ fcis-7983 41 13 research research NOUN fcis-7983 41 14 focus focus NOUN fcis-7983 41 15 in in ADP fcis-7983 41 16 chinese chinese PROPN fcis-7983 41 17 named name VERB fcis-7983 41 18 entity entity NOUN fcis-7983 41 19 recognition recognition NOUN fcis-7983 41 20 . . PUNCT fcis-7983 42 1 the the DET fcis-7983 42 2 lattice lattice NOUN fcis-7983 42 3 - - PUNCT fcis-7983 42 4 lstm lstm ADJ fcis-7983 42 5 model model NOUN fcis-7983 42 6 proposed propose VERB fcis-7983 42 7 in in ADP fcis-7983 42 8 literature literature NOUN fcis-7983 42 9 [ [ X fcis-7983 42 10 11 11 NUM fcis-7983 42 11 ] ] PUNCT fcis-7983 42 12 improves improve VERB fcis-7983 42 13 model model NOUN fcis-7983 42 14 recognition recognition NOUN fcis-7983 42 15 capabilities capability NOUN fcis-7983 42 16 by by ADP fcis-7983 42 17 combining combine VERB fcis-7983 42 18 dictionary dictionary ADJ fcis-7983 42 19 information information NOUN fcis-7983 42 20 with with ADP fcis-7983 42 21 the the DET fcis-7983 42 22 model model NOUN fcis-7983 42 23 . . PUNCT fcis-7983 43 1 specifically specifically ADV fcis-7983 43 2 , , PUNCT fcis-7983 43 3 the the DET fcis-7983 43 4 model model NOUN fcis-7983 43 5 uses use VERB fcis-7983 43 6 the the DET fcis-7983 43 7 gate gate NOUN fcis-7983 43 8 mechanism mechanism NOUN fcis-7983 43 9 of of ADP fcis-7983 43 10 long long ADJ fcis-7983 43 11 short short ADJ fcis-7983 43 12 - - PUNCT fcis-7983 43 13 term term NOUN fcis-7983 43 14 neural neural ADJ fcis-7983 43 15 networks network NOUN fcis-7983 43 16 to to PART fcis-7983 43 17 automatically automatically ADV fcis-7983 43 18 match match VERB fcis-7983 43 19 each each DET fcis-7983 43 20 character character NOUN fcis-7983 43 21 in in ADP fcis-7983 43 22 the the DET fcis-7983 43 23 sentence sentence NOUN fcis-7983 43 24 with with ADP fcis-7983 43 25 the the DET fcis-7983 43 26 corresponding corresponding ADJ fcis-7983 43 27 word word NOUN fcis-7983 43 28 and and CCONJ fcis-7983 43 29 incorporates incorporate VERB fcis-7983 43 30 the the DET fcis-7983 43 31 word word NOUN fcis-7983 43 32 information information NOUN fcis-7983 43 33 that that PRON fcis-7983 43 34 is be AUX fcis-7983 43 35 most most ADV fcis-7983 43 36 compatible compatible ADJ fcis-7983 43 37 with with ADP fcis-7983 43 38 the the DET fcis-7983 43 39 sentence sentence NOUN fcis-7983 43 40 semantics semantic NOUN fcis-7983 43 41 into into ADP fcis-7983 43 42 the the DET fcis-7983 43 43 sentence sentence NOUN fcis-7983 43 44 representation representation NOUN fcis-7983 43 45 . . PUNCT fcis-7983 44 1 li li PROPN fcis-7983 44 2 et et PROPN fcis-7983 44 3 al al PROPN fcis-7983 45 1 [ [ X fcis-7983 45 2 12 12 NUM fcis-7983 45 3 ] ] PUNCT fcis-7983 45 4 . . PUNCT fcis-7983 46 1 proved prove VERB fcis-7983 46 2 chinese chinese ADJ fcis-7983 46 3 glyph glyph NOUN fcis-7983 46 4 embedding embed VERB fcis-7983 46 5 by by ADP fcis-7983 46 6 using use VERB fcis-7983 46 7 the the DET fcis-7983 46 8 " " PUNCT fcis-7983 46 9 wubi wubi NOUN fcis-7983 46 10 " " PUNCT fcis-7983 46 11 stroke stroke NOUN fcis-7983 46 12 coding code VERB fcis-7983 46 13 method method NOUN fcis-7983 46 14 to to PART fcis-7983 46 15 represent represent VERB fcis-7983 46 16 chinese chinese ADJ fcis-7983 46 17 character character NOUN fcis-7983 46 18 structural structural ADJ fcis-7983 46 19 patterns pattern NOUN fcis-7983 46 20 and and CCONJ fcis-7983 46 21 improve improve VERB fcis-7983 46 22 overall overall ADJ fcis-7983 46 23 performance performance NOUN fcis-7983 46 24 in in ADP fcis-7983 46 25 chinese chinese PROPN fcis-7983 46 26 named name VERB fcis-7983 46 27 entity entity NOUN fcis-7983 46 28 recognition recognition NOUN fcis-7983 46 29 . . PUNCT fcis-7983 47 1 xu xu PROPN fcis-7983 47 2 et et PROPN fcis-7983 48 1 al al PROPN fcis-7983 49 1 [ [ X fcis-7983 49 2 13 13 NUM fcis-7983 49 3 ] ] PUNCT fcis-7983 49 4 . . PUNCT fcis-7983 50 1 rgued rgue VERB fcis-7983 50 2 that that SCONJ fcis-7983 50 3 the the DET fcis-7983 50 4 feature feature NOUN fcis-7983 50 5 information information NOUN fcis-7983 50 6 contained contain VERB fcis-7983 50 7 in in ADP fcis-7983 50 8 chinese chinese ADJ fcis-7983 50 9 character character NOUN fcis-7983 50 10 radicals radical NOUN fcis-7983 50 11 can can AUX fcis-7983 50 12 also also ADV fcis-7983 50 13 help help VERB fcis-7983 50 14 improve improve VERB fcis-7983 50 15 the the DET fcis-7983 50 16 ability ability NOUN fcis-7983 50 17 to to PART fcis-7983 50 18 recognize recognize VERB fcis-7983 50 19 named name VERB fcis-7983 50 20 entities entity NOUN fcis-7983 50 21 . . PUNCT fcis-7983 51 1 this this DET fcis-7983 51 2 work work NOUN fcis-7983 51 3 proposes propose VERB fcis-7983 51 4 using use VERB fcis-7983 51 5 three three NUM fcis-7983 51 6 different different ADJ fcis-7983 51 7 levels level NOUN fcis-7983 51 8 of of ADP fcis-7983 51 9 embedding embed VERB fcis-7983 51 10 , , PUNCT fcis-7983 51 11 i.e. i.e. X fcis-7983 51 12 , , PUNCT fcis-7983 51 13 character character NOUN fcis-7983 51 14 , , PUNCT fcis-7983 51 15 word word NOUN fcis-7983 51 16 , , PUNCT fcis-7983 51 17 and and CCONJ fcis-7983 51 18 radical radical ADJ fcis-7983 51 19 , , PUNCT fcis-7983 51 20 in in ADP fcis-7983 51 21 the the DET fcis-7983 51 22 model model NOUN fcis-7983 51 23 to to PART fcis-7983 51 24 enrich enrich VERB fcis-7983 51 25 the the DET fcis-7983 51 26 character character NOUN fcis-7983 51 27 representation representation NOUN fcis-7983 51 28 in in ADP fcis-7983 51 29 the the DET fcis-7983 51 30 sentence sentence NOUN fcis-7983 51 31 and and CCONJ fcis-7983 51 32 validates validate VERB fcis-7983 51 33 the the DET fcis-7983 51 34 effectiveness effectiveness NOUN fcis-7983 51 35 of of ADP fcis-7983 51 36 radical radical ADJ fcis-7983 51 37 information information NOUN fcis-7983 51 38 . . PUNCT fcis-7983 52 1 the the DET fcis-7983 52 2 success success NOUN fcis-7983 52 3 of of ADP fcis-7983 52 4 these these DET fcis-7983 52 5 works work NOUN fcis-7983 52 6 also also ADV fcis-7983 52 7 confirms confirm VERB fcis-7983 52 8 the the DET fcis-7983 52 9 effectiveness effectiveness NOUN fcis-7983 52 10 of of ADP fcis-7983 52 11 multi multi ADJ fcis-7983 52 12 - - ADJ fcis-7983 52 13 level level ADJ fcis-7983 52 14 feature feature NOUN fcis-7983 52 15 information information NOUN fcis-7983 52 16 in in ADP fcis-7983 52 17 chinese chinese PROPN fcis-7983 52 18 . . PUNCT fcis-7983 53 1 to to PART fcis-7983 53 2 further far ADV fcis-7983 53 3 improve improve VERB fcis-7983 53 4 the the DET fcis-7983 53 5 performance performance NOUN fcis-7983 53 6 of of ADP fcis-7983 53 7 the the DET fcis-7983 53 8 model model NOUN fcis-7983 53 9 , , PUNCT fcis-7983 53 10 this this DET fcis-7983 53 11 paper paper NOUN fcis-7983 53 12 introduces introduce VERB fcis-7983 53 13 external external ADJ fcis-7983 53 14 information information NOUN fcis-7983 53 15 , , PUNCT fcis-7983 53 16 i.e. i.e. X fcis-7983 53 17 , , PUNCT fcis-7983 53 18 dictionary dictionary ADJ fcis-7983 53 19 and and CCONJ fcis-7983 53 20 radical radical ADJ fcis-7983 53 21 information information NOUN fcis-7983 53 22 , , PUNCT fcis-7983 53 23 into into ADP fcis-7983 53 24 the the DET fcis-7983 53 25 model model NOUN fcis-7983 53 26 . . PUNCT fcis-7983 54 1 3 3 X fcis-7983 54 2 . . X fcis-7983 54 3 bert bert NOUN fcis-7983 54 4 - - PUNCT fcis-7983 54 5 bi bi ADJ fcis-7983 54 6 - - ADJ fcis-7983 54 7 lstm lstm ADJ fcis-7983 54 8 - - PUNCT fcis-7983 54 9 crf crf NOUN fcis-7983 54 10 model model NOUN fcis-7983 54 11 this this DET fcis-7983 54 12 paper paper NOUN fcis-7983 54 13 proposes propose VERB fcis-7983 54 14 a a DET fcis-7983 54 15 multi multi ADJ fcis-7983 54 16 - - ADJ fcis-7983 54 17 feature feature ADJ fcis-7983 54 18 fusion fusion NOUN fcis-7983 54 19 model model NOUN fcis-7983 54 20 for for ADP fcis-7983 54 21 chinese chinese ADJ fcis-7983 54 22 electronic electronic ADJ fcis-7983 54 23 medical medical ADJ fcis-7983 54 24 record record NOUN fcis-7983 54 25 named name VERB fcis-7983 54 26 entity entity NOUN fcis-7983 54 27 recognition recognition NOUN fcis-7983 54 28 , , PUNCT fcis-7983 54 29 which which PRON fcis-7983 54 30 consists consist VERB fcis-7983 54 31 of of ADP fcis-7983 54 32 two two NUM fcis-7983 54 33 parts part NOUN fcis-7983 54 34 . . PUNCT fcis-7983 55 1 the the DET fcis-7983 55 2 first first ADJ fcis-7983 55 3 part part NOUN fcis-7983 55 4 is be AUX fcis-7983 55 5 a a DET fcis-7983 55 6 bert bert NOUN fcis-7983 55 7 - - PUNCT fcis-7983 55 8 bilstm bilstm NOUN fcis-7983 55 9 - - PUNCT fcis-7983 55 10 crf crf NOUN fcis-7983 55 11 model model NOUN fcis-7983 55 12 , , PUNCT fcis-7983 55 13 and and CCONJ fcis-7983 55 14 the the DET fcis-7983 55 15 second second ADJ fcis-7983 55 16 part part NOUN fcis-7983 55 17 incorporates incorporate VERB fcis-7983 55 18 radical radical ADJ fcis-7983 55 19 components component NOUN fcis-7983 55 20 and and CCONJ fcis-7983 55 21 domain domain NOUN fcis-7983 55 22 dictionaries dictionary NOUN fcis-7983 55 23 . . PUNCT fcis-7983 56 1 the the DET fcis-7983 56 2 bert bert PROPN fcis-7983 56 3 - - PUNCT fcis-7983 56 4 bi bi ADJ fcis-7983 56 5 - - ADJ fcis-7983 56 6 lstmcrf lstmcrf ADJ fcis-7983 56 7 model model NOUN fcis-7983 56 8 is be AUX fcis-7983 56 9 a a DET fcis-7983 56 10 deep deep ADJ fcis-7983 56 11 learning learning NOUN fcis-7983 56 12 - - PUNCT fcis-7983 56 13 based base VERB fcis-7983 56 14 text text NOUN fcis-7983 56 15 sequence sequence NOUN fcis-7983 56 16 labeling labeling NOUN fcis-7983 56 17 model model NOUN fcis-7983 56 18 that that PRON fcis-7983 56 19 integrates integrate VERB fcis-7983 56 20 bert bert PROPN fcis-7983 57 1 [ [ X fcis-7983 57 2 14,15 14,15 NUM fcis-7983 57 3 ] ] PUNCT fcis-7983 57 4 . . PUNCT fcis-7983 58 1 brectional brectional ADJ fcis-7983 58 2 long long ADJ fcis-7983 58 3 shortterm shortterm PROPN fcis-7983 58 4 memory memory NOUN fcis-7983 58 5 networks network NOUN fcis-7983 58 6 ( ( PUNCT fcis-7983 58 7 bi bi NOUN fcis-7983 58 8 - - ADJ fcis-7983 58 9 lstm lstm ADJ fcis-7983 58 10 ) ) PUNCT fcis-7983 58 11 , , PUNCT fcis-7983 58 12 and and CCONJ fcis-7983 58 13 conditional conditional ADJ fcis-7983 58 14 random random ADJ fcis-7983 58 15 fields field NOUN fcis-7983 58 16 ( ( PUNCT fcis-7983 58 17 crf crf NOUN fcis-7983 58 18 ) ) PUNCT fcis-7983 58 19 , , PUNCT fcis-7983 58 20 mainly mainly ADV fcis-7983 58 21 used use VERB fcis-7983 58 22 for for ADP fcis-7983 58 23 named name VERB fcis-7983 58 24 entity entity NOUN fcis-7983 58 25 recognition recognition NOUN fcis-7983 58 26 ( ( PUNCT fcis-7983 58 27 ner ner NOUN fcis-7983 58 28 ) ) PUNCT fcis-7983 58 29 tasks task NOUN fcis-7983 58 30 . . PUNCT fcis-7983 59 1 the the DET fcis-7983 59 2 model model NOUN fcis-7983 59 3 structure structure NOUN fcis-7983 59 4 is be AUX fcis-7983 59 5 shown show VERB fcis-7983 59 6 in in ADP fcis-7983 59 7 figure figure NOUN fcis-7983 59 8 1 1 NUM fcis-7983 59 9 . . PUNCT fcis-7983 60 1 fig.1 fig.1 PROPN fcis-7983 60 2 named name VERB fcis-7983 60 3 entity entity NOUN fcis-7983 60 4 recognition recognition NOUN fcis-7983 60 5 framework framework NOUN fcis-7983 60 6 based base VERB fcis-7983 60 7 on on ADP fcis-7983 60 8 bert bert NOUN fcis-7983 60 9 - - PUNCT fcis-7983 60 10 bilstm bilstm NOUN fcis-7983 60 11 - - PUNCT fcis-7983 60 12 crf crf NOUN fcis-7983 60 13 in in ADP fcis-7983 60 14 the the DET fcis-7983 60 15 bert bert PROPN fcis-7983 60 16 - - PUNCT fcis-7983 60 17 bi bi ADJ fcis-7983 60 18 - - ADJ fcis-7983 60 19 lstm lstm ADJ fcis-7983 60 20 - - PUNCT fcis-7983 60 21 crf crf NOUN fcis-7983 60 22 model model NOUN fcis-7983 60 23 , , PUNCT fcis-7983 60 24 the the DET fcis-7983 60 25 input input NOUN fcis-7983 60 26 text text NOUN fcis-7983 60 27 sequence sequence NOUN fcis-7983 60 28 is be AUX fcis-7983 60 29 first first ADV fcis-7983 60 30 processed process VERB fcis-7983 60 31 by by ADP fcis-7983 60 32 the the DET fcis-7983 60 33 bert bert PROPN fcis-7983 60 34 model model NOUN fcis-7983 60 35 to to PART fcis-7983 60 36 obtain obtain VERB fcis-7983 60 37 contextualized contextualized ADJ fcis-7983 60 38 word word NOUN fcis-7983 60 39 embeddings embedding NOUN fcis-7983 60 40 for for ADP fcis-7983 60 41 each each DET fcis-7983 60 42 word word NOUN fcis-7983 60 43 . . PUNCT fcis-7983 61 1 these these DET fcis-7983 61 2 embeddings embedding NOUN fcis-7983 61 3 are be AUX fcis-7983 61 4 then then ADV fcis-7983 61 5 fed feed VERB fcis-7983 61 6 into into ADP fcis-7983 61 7 the the DET fcis-7983 61 8 bi bi ADJ fcis-7983 61 9 - - ADJ fcis-7983 61 10 lstm lstm ADJ fcis-7983 61 11 layer layer NOUN fcis-7983 61 12 to to PART fcis-7983 61 13 learn learn VERB fcis-7983 61 14 feature feature NOUN fcis-7983 61 15 representations representation NOUN fcis-7983 61 16 that that PRON fcis-7983 61 17 capture capture VERB fcis-7983 61 18 contextual contextual ADJ fcis-7983 61 19 information information NOUN fcis-7983 61 20 . . PUNCT fcis-7983 62 1 finally finally ADV fcis-7983 62 2 , , PUNCT fcis-7983 62 3 the the DET fcis-7983 62 4 crf crf NOUN fcis-7983 62 5 layer layer NOUN fcis-7983 62 6 labels label VERB fcis-7983 62 7 the the DET fcis-7983 62 8 output output NOUN fcis-7983 62 9 of of ADP fcis-7983 62 10 the the DET fcis-7983 62 11 bi bi ADJ fcis-7983 62 12 - - ADJ fcis-7983 62 13 lstm lstm ADJ fcis-7983 62 14 layer layer NOUN fcis-7983 62 15 to to PART fcis-7983 62 16 obtain obtain VERB fcis-7983 62 17 the the DET fcis-7983 62 18 final final ADJ fcis-7983 62 19 output output NOUN fcis-7983 62 20 sequence sequence NOUN fcis-7983 62 21 . . PUNCT fcis-7983 63 1 by by ADP fcis-7983 63 2 using use VERB fcis-7983 63 3 bert bert PROPN fcis-7983 63 4 as as ADP fcis-7983 63 5 the the DET fcis-7983 63 6 input input NOUN fcis-7983 63 7 layer layer NOUN fcis-7983 63 8 , , PUNCT fcis-7983 63 9 the the DET fcis-7983 63 10 model model NOUN fcis-7983 63 11 can can AUX fcis-7983 63 12 learn learn VERB fcis-7983 63 13 contextual contextual ADJ fcis-7983 63 14 representations representation NOUN fcis-7983 63 15 and and CCONJ fcis-7983 63 16 apply apply VERB fcis-7983 63 17 them they PRON fcis-7983 63 18 to to ADP fcis-7983 63 19 tasks task NOUN fcis-7983 63 20 such such ADJ fcis-7983 63 21 as as ADP fcis-7983 63 22 text text NOUN fcis-7983 63 23 classification classification NOUN fcis-7983 63 24 and and CCONJ fcis-7983 63 25 named name VERB fcis-7983 63 26 entity entity NOUN fcis-7983 63 27 recognition recognition NOUN fcis-7983 63 28 . . PUNCT fcis-7983 64 1 the the DET fcis-7983 64 2 bi bi ADJ fcis-7983 64 3 - - ADJ fcis-7983 64 4 lstm lstm ADJ fcis-7983 64 5 acts act VERB fcis-7983 64 6 as as ADP fcis-7983 64 7 a a DET fcis-7983 64 8 feature feature NOUN fcis-7983 64 9 extractor extractor NOUN fcis-7983 64 10 that that PRON fcis-7983 64 11 captures capture VERB fcis-7983 64 12 context context NOUN fcis-7983 64 13 information information NOUN fcis-7983 64 14 , , PUNCT fcis-7983 64 15 including include VERB fcis-7983 64 16 the the DET fcis-7983 64 17 preceding precede VERB fcis-7983 64 18 and and CCONJ fcis-7983 64 19 following follow VERB fcis-7983 64 20 words word NOUN fcis-7983 64 21 in in ADP fcis-7983 64 22 a a DET fcis-7983 64 23 sentence sentence NOUN fcis-7983 64 24 . . PUNCT fcis-7983 65 1 the the DET fcis-7983 65 2 crf crf PROPN fcis-7983 65 3 layer layer NOUN fcis-7983 65 4 considers consider VERB fcis-7983 65 5 dependencies dependency NOUN fcis-7983 65 6 between between ADP fcis-7983 65 7 labels label NOUN fcis-7983 65 8 , , PUNCT fcis-7983 65 9 leading lead VERB fcis-7983 65 10 to to ADP fcis-7983 65 11 more more ADV fcis-7983 65 12 accurate accurate ADJ fcis-7983 65 13 label label NOUN fcis-7983 65 14 predictions prediction NOUN fcis-7983 65 15 . . PUNCT fcis-7983 66 1 compared compare VERB fcis-7983 66 2 to to ADP fcis-7983 66 3 other other ADJ fcis-7983 66 4 text text NOUN fcis-7983 66 5 sequence sequence NOUN fcis-7983 66 6 labeling labeling NOUN fcis-7983 66 7 models model NOUN fcis-7983 66 8 , , PUNCT fcis-7983 66 9 the the DET fcis-7983 66 10 bert bert NOUN fcis-7983 66 11 - - PUNCT fcis-7983 66 12 bi bi ADJ fcis-7983 66 13 - - ADJ fcis-7983 66 14 lstm lstm ADJ fcis-7983 66 15 - - PUNCT fcis-7983 66 16 crf crf NOUN fcis-7983 66 17 model model NOUN fcis-7983 66 18 effectively effectively ADV fcis-7983 66 19 utilizes utilize VERB fcis-7983 66 20 contextual contextual ADJ fcis-7983 66 21 information information NOUN fcis-7983 66 22 and and CCONJ fcis-7983 66 23 label label NOUN fcis-7983 66 24 dependencies dependency NOUN fcis-7983 66 25 , , PUNCT fcis-7983 66 26 resulting result VERB fcis-7983 66 27 in in ADP fcis-7983 66 28 improved improve VERB fcis-7983 66 29 predictive predictive ADJ fcis-7983 66 30 ability ability NOUN fcis-7983 66 31 and and CCONJ fcis-7983 66 32 accuracy accuracy NOUN fcis-7983 66 33 . . PUNCT fcis-7983 67 1 as as ADP fcis-7983 67 2 a a DET fcis-7983 67 3 result result NOUN fcis-7983 67 4 , , PUNCT fcis-7983 67 5 this this DET fcis-7983 67 6 model model NOUN fcis-7983 67 7 has have AUX fcis-7983 67 8 achieved achieve VERB fcis-7983 67 9 good good ADJ fcis-7983 67 10 performance performance NOUN fcis-7983 67 11 in in ADP fcis-7983 67 12 ner ner NOUN fcis-7983 67 13 tasks task NOUN fcis-7983 67 14 and and CCONJ fcis-7983 67 15 has have AUX fcis-7983 67 16 been be AUX fcis-7983 67 17 widely widely ADV fcis-7983 67 18 applied apply VERB fcis-7983 67 19 to to ADP fcis-7983 67 20 various various ADJ fcis-7983 67 21 natural natural ADJ fcis-7983 67 22 language language NOUN fcis-7983 67 23 processing processing NOUN fcis-7983 67 24 tasks task NOUN fcis-7983 67 25 . . PUNCT fcis-7983 68 1 3.1 3.1 NUM fcis-7983 68 2 . . PUNCT fcis-7983 69 1 bert bert PROPN fcis-7983 69 2 pre pre ADJ fcis-7983 69 3 - - ADJ fcis-7983 69 4 trained train VERB fcis-7983 69 5 language language NOUN fcis-7983 69 6 model model NOUN fcis-7983 69 7 bert bert PROPN fcis-7983 69 8 ( ( PUNCT fcis-7983 69 9 bidirectional bidirectional ADJ fcis-7983 69 10 encoder encoder NOUN fcis-7983 69 11 representations representation VERB fcis-7983 69 12 from from ADP fcis-7983 69 13 transformers transformer NOUN fcis-7983 69 14 ) ) PUNCT fcis-7983 69 15 is be AUX fcis-7983 69 16 a a DET fcis-7983 69 17 pre pre ADJ fcis-7983 69 18 - - ADJ fcis-7983 69 19 trained train VERB fcis-7983 69 20 language language NOUN fcis-7983 69 21 model model NOUN fcis-7983 69 22 proposed propose VERB fcis-7983 69 23 by by ADP fcis-7983 69 24 google google PROPN fcis-7983 69 25 in in ADP fcis-7983 69 26 2018[14 2018[14 NUM fcis-7983 69 27 ] ] PUNCT fcis-7983 69 28 . . PUNCT fcis-7983 70 1 it it PRON fcis-7983 70 2 is be AUX fcis-7983 70 3 designed design VERB fcis-7983 70 4 based base VERB fcis-7983 70 5 on on ADP fcis-7983 70 6 the the DET fcis-7983 70 7 transformer transformer NOUN fcis-7983 70 8 model model NOUN fcis-7983 70 9 architecture architecture NOUN fcis-7983 70 10 and and CCONJ fcis-7983 70 11 can can AUX fcis-7983 70 12 use use VERB fcis-7983 70 13 massive massive ADJ fcis-7983 70 14 amounts amount NOUN fcis-7983 70 15 of of ADP fcis-7983 70 16 unlabeled unlabeled ADJ fcis-7983 70 17 text text NOUN fcis-7983 70 18 data datum NOUN fcis-7983 70 19 for for ADP fcis-7983 70 20 pre pre ADJ fcis-7983 70 21 - - NOUN fcis-7983 70 22 training training NOUN fcis-7983 70 23 , , PUNCT fcis-7983 70 24 followed follow VERB fcis-7983 70 25 by by ADP fcis-7983 70 26 fine fine ADV fcis-7983 70 27 - - PUNCT fcis-7983 70 28 tuning tuning NOUN fcis-7983 70 29 for for ADP fcis-7983 70 30 various various ADJ fcis-7983 70 31 natural natural ADJ fcis-7983 70 32 language language NOUN fcis-7983 70 33 processing processing NOUN fcis-7983 70 34 tasks task NOUN fcis-7983 70 35 . . PUNCT fcis-7983 71 1 the the DET fcis-7983 71 2 main main ADJ fcis-7983 71 3 contribution contribution NOUN fcis-7983 71 4 of of ADP fcis-7983 71 5 the the DET fcis-7983 71 6 bert bert PROPN fcis-7983 71 7 model model NOUN fcis-7983 71 8 is be AUX fcis-7983 71 9 its its PRON fcis-7983 71 10 ability ability NOUN fcis-7983 71 11 to to PART fcis-7983 71 12 utilize utilize VERB fcis-7983 71 13 both both DET fcis-7983 71 14 context context NOUN fcis-7983 71 15 information information NOUN fcis-7983 71 16 and and CCONJ fcis-7983 71 17 bidirectional bidirectional ADJ fcis-7983 71 18 processing processing NOUN fcis-7983 71 19 of of ADP fcis-7983 71 20 text text NOUN fcis-7983 71 21 , , PUNCT fcis-7983 71 22 which which PRON fcis-7983 71 23 makes make VERB fcis-7983 71 24 it it PRON fcis-7983 71 25 perform perform VERB fcis-7983 71 26 well well ADV fcis-7983 71 27 in in ADP fcis-7983 71 28 various various ADJ fcis-7983 71 29 natural natural ADJ fcis-7983 71 30 language language NOUN fcis-7983 71 31 processing processing NOUN fcis-7983 71 32 tasks task NOUN fcis-7983 71 33 [ [ X fcis-7983 71 34 15 15 NUM fcis-7983 71 35 ] ] PUNCT fcis-7983 71 36 . . PUNCT fcis-7983 72 1 the the DET fcis-7983 72 2 bert bert PROPN fcis-7983 72 3 model model NOUN fcis-7983 72 4 adopts adopt VERB fcis-7983 72 5 two two NUM fcis-7983 72 6 pre pre ADJ fcis-7983 72 7 - - ADJ fcis-7983 72 8 training training ADJ fcis-7983 72 9 tasks task NOUN fcis-7983 72 10 , , PUNCT fcis-7983 72 11 masked mask VERB fcis-7983 72 12 language language NOUN fcis-7983 72 13 model model NOUN fcis-7983 72 14 ( ( PUNCT fcis-7983 72 15 mlm mlm PROPN fcis-7983 72 16 ) ) PUNCT fcis-7983 72 17 and and CCONJ fcis-7983 72 18 next next ADJ fcis-7983 72 19 sentence sentence NOUN fcis-7983 72 20 prediction prediction NOUN fcis-7983 72 21 ( ( PUNCT fcis-7983 72 22 nsp nsp PROPN fcis-7983 72 23 ) ) PUNCT fcis-7983 72 24 . . PUNCT fcis-7983 73 1 the the DET fcis-7983 73 2 mlm mlm PROPN fcis-7983 73 3 task task PROPN fcis-7983 73 4 randomly randomly ADV fcis-7983 73 5 replaces replace VERB fcis-7983 73 6 some some DET fcis-7983 73 7 words word NOUN fcis-7983 73 8 in in ADP fcis-7983 73 9 the the DET fcis-7983 73 10 input input NOUN fcis-7983 73 11 text text NOUN fcis-7983 73 12 with with ADP fcis-7983 73 13 the the DET fcis-7983 73 14 [ [ X fcis-7983 73 15 mask mask NOUN fcis-7983 73 16 ] ] PUNCT fcis-7983 73 17 symbol symbol NOUN fcis-7983 73 18 , , PUNCT fcis-7983 73 19 and and CCONJ fcis-7983 73 20 then then ADV fcis-7983 73 21 trains train VERB fcis-7983 73 22 the the DET fcis-7983 73 23 model model NOUN fcis-7983 73 24 to to PART fcis-7983 73 25 predict predict VERB fcis-7983 73 26 the the DET fcis-7983 73 27 replaced replace VERB fcis-7983 73 28 words word NOUN fcis-7983 73 29 . . PUNCT fcis-7983 74 1 the the DET fcis-7983 74 2 nsp nsp PROPN fcis-7983 74 3 task task NOUN fcis-7983 74 4 is be AUX fcis-7983 74 5 to to PART fcis-7983 74 6 train train VERB fcis-7983 74 7 the the DET fcis-7983 74 8 model model NOUN fcis-7983 74 9 to to PART fcis-7983 74 10 understand understand VERB fcis-7983 74 11 the the DET fcis-7983 74 12 relationship relationship NOUN fcis-7983 74 13 between between ADP fcis-7983 74 14 sentences sentence NOUN fcis-7983 74 15 in in ADP fcis-7983 74 16 the the DET fcis-7983 74 17 input input NOUN fcis-7983 74 18 text text NOUN fcis-7983 74 19 , , PUNCT fcis-7983 74 20 by by ADP fcis-7983 74 21 determining determine VERB fcis-7983 74 22 whether whether SCONJ fcis-7983 74 23 two two NUM fcis-7983 74 24 sentences sentence NOUN fcis-7983 74 25 are be AUX fcis-7983 74 26 consecutive consecutive ADJ fcis-7983 74 27 . . PUNCT fcis-7983 75 1 the the DET fcis-7983 75 2 bert bert PROPN fcis-7983 75 3 model model NOUN fcis-7983 75 4 has have AUX fcis-7983 75 5 shown show VERB fcis-7983 75 6 good good ADJ fcis-7983 75 7 performance performance NOUN fcis-7983 75 8 in in ADP fcis-7983 75 9 various various ADJ fcis-7983 75 10 natural natural ADJ fcis-7983 75 11 language language NOUN fcis-7983 75 12 processing processing NOUN fcis-7983 75 13 tasks task NOUN fcis-7983 75 14 , , PUNCT fcis-7983 75 15 and and CCONJ fcis-7983 75 16 in in ADP fcis-7983 75 17 the the DET fcis-7983 75 18 medical medical ADJ fcis-7983 75 19 field field NOUN fcis-7983 75 20 , , PUNCT fcis-7983 75 21 it it PRON fcis-7983 75 22 performs perform VERB fcis-7983 75 23 very very ADV fcis-7983 75 24 well well ADV fcis-7983 75 25 in in ADP fcis-7983 75 26 electronic electronic ADJ fcis-7983 75 27 medical medical ADJ fcis-7983 75 28 record record NOUN fcis-7983 75 29 named name VERB fcis-7983 75 30 entity entity NOUN fcis-7983 75 31 recognition recognition NOUN fcis-7983 75 32 . . PUNCT fcis-7983 76 1 nowadays nowadays ADV fcis-7983 76 2 , , PUNCT fcis-7983 76 3 most most ADJ fcis-7983 76 4 studies study NOUN fcis-7983 76 5 on on ADP fcis-7983 76 6 named name VERB fcis-7983 76 7 entity entity NOUN fcis-7983 76 8 recognition recognition NOUN fcis-7983 76 9 in in ADP fcis-7983 76 10 electronic electronic ADJ fcis-7983 76 11 medical medical ADJ fcis-7983 76 12 records record NOUN fcis-7983 76 13 use use VERB fcis-7983 76 14 bert bert NOUN fcis-7983 76 15 model model NOUN fcis-7983 76 16 for for ADP fcis-7983 76 17 word word NOUN fcis-7983 76 18 vector vector NOUN fcis-7983 76 19 representation representation NOUN fcis-7983 76 20 . . PUNCT fcis-7983 77 1 the the DET fcis-7983 77 2 diagram diagram NOUN fcis-7983 77 3 of of ADP fcis-7983 77 4 the the DET fcis-7983 77 5 bert bert PROPN fcis-7983 77 6 model model NOUN fcis-7983 77 7 is be AUX fcis-7983 77 8 shown show VERB fcis-7983 77 9 in in ADP fcis-7983 77 10 figure figure NOUN fcis-7983 77 11 2 2 NUM fcis-7983 77 12 . . PUNCT fcis-7983 77 13 fig fig NOUN fcis-7983 77 14 . . PUNCT fcis-7983 78 1 2 2 NUM fcis-7983 78 2 structure structure NOUN fcis-7983 78 3 of of ADP fcis-7983 78 4 bert bert PROPN fcis-7983 78 5 3.2 3.2 NUM fcis-7983 78 6 . . PUNCT fcis-7983 79 1 bi bi ADJ fcis-7983 79 2 - - ADJ fcis-7983 79 3 lstm lstm ADJ fcis-7983 79 4 network network NOUN fcis-7983 79 5 long long ADJ fcis-7983 79 6 short short ADJ fcis-7983 79 7 - - PUNCT fcis-7983 79 8 term term NOUN fcis-7983 79 9 memory memory NOUN fcis-7983 79 10 ( ( PUNCT fcis-7983 79 11 lstm lstm NOUN fcis-7983 79 12 ) ) PUNCT fcis-7983 79 13 networks network NOUN fcis-7983 79 14 are be AUX fcis-7983 79 15 an an DET fcis-7983 79 16 improved improved ADJ fcis-7983 79 17 version version NOUN fcis-7983 79 18 of of ADP fcis-7983 79 19 recurrent recurrent ADJ fcis-7983 79 20 neural neural ADJ fcis-7983 79 21 networks network NOUN fcis-7983 79 22 ( ( PUNCT fcis-7983 79 23 rnns rnns PROPN fcis-7983 79 24 ) ) PUNCT fcis-7983 79 25 that that PRON fcis-7983 79 26 were be AUX fcis-7983 79 27 introduced introduce VERB fcis-7983 79 28 in in ADP fcis-7983 79 29 1997 1997 NUM fcis-7983 79 30 by by ADP fcis-7983 79 31 hochreiter hochreiter PROPN fcis-7983 79 32 et et PROPN fcis-7983 79 33 al al PROPN fcis-7983 79 34 . . PUNCT fcis-7983 79 35 to to PART fcis-7983 79 36 address address VERB fcis-7983 79 37 the the DET fcis-7983 79 38 issues issue NOUN fcis-7983 79 39 of of ADP fcis-7983 79 40 vanishing vanish VERB fcis-7983 79 41 and and CCONJ fcis-7983 79 42 exploding explode VERB fcis-7983 79 43 gradients gradient NOUN fcis-7983 79 44 in in ADP fcis-7983 79 45 rnns rnns NOUN fcis-7983 79 46 [ [ X fcis-7983 79 47 16 16 NUM fcis-7983 79 48 ] ] PUNCT fcis-7983 79 49 . . PUNCT fcis-7983 80 1 lstm lstm NOUN fcis-7983 80 2 is be AUX fcis-7983 80 3 a a DET fcis-7983 80 4 type type NOUN fcis-7983 80 5 of of ADP fcis-7983 80 6 time time NOUN fcis-7983 80 7 - - PUNCT fcis-7983 80 8 recursive recursive ADJ fcis-7983 80 9 neural neural ADJ fcis-7983 80 10 network network NOUN fcis-7983 80 11 that that SCONJ fcis-7983 80 12 8 8 NUM fcis-7983 80 13 introduces introduce NOUN fcis-7983 80 14 " " PUNCT fcis-7983 80 15 gates gate NOUN fcis-7983 80 16 , , PUNCT fcis-7983 80 17 " " PUNCT fcis-7983 80 18 including include VERB fcis-7983 80 19 an an DET fcis-7983 80 20 input input NOUN fcis-7983 80 21 gate gate NOUN fcis-7983 80 22 , , PUNCT fcis-7983 80 23 forget forget VERB fcis-7983 80 24 gate gate NOUN fcis-7983 80 25 , , PUNCT fcis-7983 80 26 and and CCONJ fcis-7983 80 27 output output NOUN fcis-7983 80 28 gate gate NOUN fcis-7983 81 1 [ [ X fcis-7983 81 2 17 17 NUM fcis-7983 81 3 ] ] PUNCT fcis-7983 81 4 . . PUNCT fcis-7983 82 1 lstm lstm NOUN fcis-7983 82 2 is be AUX fcis-7983 82 3 an an DET fcis-7983 82 4 effective effective ADJ fcis-7983 82 5 technique technique NOUN fcis-7983 82 6 for for ADP fcis-7983 82 7 solving solve VERB fcis-7983 82 8 the the DET fcis-7983 82 9 problem problem NOUN fcis-7983 82 10 of of ADP fcis-7983 82 11 long long ADJ fcis-7983 82 12 - - PUNCT fcis-7983 82 13 term term NOUN fcis-7983 82 14 dependencies dependency NOUN fcis-7983 82 15 . . PUNCT fcis-7983 83 1 as as ADP fcis-7983 83 2 a a DET fcis-7983 83 3 variant variant NOUN fcis-7983 83 4 of of ADP fcis-7983 83 5 rnn rnn PROPN fcis-7983 83 6 , , PUNCT fcis-7983 83 7 lstm lstm NOUN fcis-7983 83 8 is be AUX fcis-7983 83 9 widely widely ADV fcis-7983 83 10 used use VERB fcis-7983 83 11 in in ADP fcis-7983 83 12 the the DET fcis-7983 83 13 field field NOUN fcis-7983 83 14 of of ADP fcis-7983 83 15 natural natural ADJ fcis-7983 83 16 language language NOUN fcis-7983 83 17 processing processing NOUN fcis-7983 83 18 ( ( PUNCT fcis-7983 83 19 nlp nlp NOUN fcis-7983 83 20 ) ) PUNCT fcis-7983 83 21 . . PUNCT fcis-7983 84 1 in in ADP fcis-7983 84 2 the the DET fcis-7983 84 3 lstm lstm PROPN fcis-7983 84 4 model model NOUN fcis-7983 84 5 , , PUNCT fcis-7983 84 6 only only ADV fcis-7983 84 7 unidirectional unidirectional ADJ fcis-7983 84 8 transmission transmission NOUN fcis-7983 84 9 exists exist VERB fcis-7983 84 10 , , PUNCT fcis-7983 84 11 and and CCONJ fcis-7983 84 12 it it PRON fcis-7983 84 13 only only ADV fcis-7983 84 14 considers consider VERB fcis-7983 84 15 the the DET fcis-7983 84 16 information information NOUN fcis-7983 84 17 content content NOUN fcis-7983 84 18 of of ADP fcis-7983 84 19 the the DET fcis-7983 84 20 " " PUNCT fcis-7983 84 21 previous previous ADJ fcis-7983 84 22 context context NOUN fcis-7983 84 23 , , PUNCT fcis-7983 84 24 " " PUNCT fcis-7983 84 25 but but CCONJ fcis-7983 84 26 not not PART fcis-7983 84 27 the the DET fcis-7983 84 28 " " PUNCT fcis-7983 84 29 next next ADJ fcis-7983 84 30 context context NOUN fcis-7983 84 31 . . PUNCT fcis-7983 84 32 " " PUNCT fcis-7983 85 1 in in ADP fcis-7983 85 2 reality reality NOUN fcis-7983 85 3 , , PUNCT fcis-7983 85 4 entity entity NOUN fcis-7983 85 5 named name VERB fcis-7983 85 6 recognition recognition NOUN fcis-7983 85 7 needs need VERB fcis-7983 85 8 to to PART fcis-7983 85 9 consider consider VERB fcis-7983 85 10 all all DET fcis-7983 85 11 information information NOUN fcis-7983 85 12 content content NOUN fcis-7983 85 13 in in ADP fcis-7983 85 14 all all DET fcis-7983 85 15 input input NOUN fcis-7983 85 16 orders order NOUN fcis-7983 85 17 . . PUNCT fcis-7983 86 1 therefore therefore ADV fcis-7983 86 2 , , PUNCT fcis-7983 86 3 the the DET fcis-7983 86 4 current current ADJ fcis-7983 86 5 approach approach NOUN fcis-7983 86 6 to to ADP fcis-7983 86 7 entity entity NOUN fcis-7983 86 8 named name VERB fcis-7983 86 9 recognition recognition NOUN fcis-7983 86 10 in in ADP fcis-7983 86 11 electronic electronic ADJ fcis-7983 86 12 medical medical ADJ fcis-7983 86 13 records record NOUN fcis-7983 86 14 typically typically ADV fcis-7983 86 15 uses use VERB fcis-7983 86 16 bidirectional bidirectional ADJ fcis-7983 86 17 lstm lstm NOUN fcis-7983 86 18 ( ( PUNCT fcis-7983 86 19 bi bi NOUN fcis-7983 86 20 - - ADJ fcis-7983 86 21 lstm lstm ADJ fcis-7983 86 22 ) ) PUNCT fcis-7983 86 23 networks network NOUN fcis-7983 87 1 [ [ X fcis-7983 87 2 18 18 NUM fcis-7983 87 3 ] ] PUNCT fcis-7983 87 4 , , PUNCT fcis-7983 87 5 as as SCONJ fcis-7983 87 6 shown show VERB fcis-7983 87 7 in in ADP fcis-7983 87 8 figure figure NOUN fcis-7983 87 9 3 3 NUM fcis-7983 87 10 . . PUNCT fcis-7983 87 11 fig fig NOUN fcis-7983 87 12 . . PUNCT fcis-7983 87 13 3 3 NUM fcis-7983 87 14 lstm lstm NOUN fcis-7983 87 15 structure structure NOUN fcis-7983 87 16 in in ADP fcis-7983 87 17 figure figure NOUN fcis-7983 87 18 3 3 NUM fcis-7983 87 19 , , PUNCT fcis-7983 87 20 xt xt X fcis-7983 87 21 is be AUX fcis-7983 87 22 the the DET fcis-7983 87 23 input input NOUN fcis-7983 87 24 at at ADP fcis-7983 87 25 time time NOUN fcis-7983 87 26 t t PROPN fcis-7983 87 27 ; ; PUNCT fcis-7983 87 28 ft ft X fcis-7983 87 29 is be AUX fcis-7983 87 30 the the DET fcis-7983 87 31 output output NOUN fcis-7983 87 32 for for ADP fcis-7983 87 33 t+1 t+1 PRON fcis-7983 87 34 at at ADP fcis-7983 87 35 time time NOUN fcis-7983 87 36 t t PROPN fcis-7983 87 37 ; ; PUNCT fcis-7983 87 38 ot ot X fcis-7983 87 39 is be AUX fcis-7983 87 40 the the DET fcis-7983 87 41 output output NOUN fcis-7983 87 42 at at ADP fcis-7983 87 43 time time NOUN fcis-7983 87 44 t t PROPN fcis-7983 87 45 ; ; PUNCT fcis-7983 87 46 ht ht PROPN fcis-7983 87 47 is be AUX fcis-7983 87 48 the the DET fcis-7983 87 49 hidden hidden ADJ fcis-7983 87 50 layer layer NOUN fcis-7983 87 51 representing represent VERB fcis-7983 87 52 the the DET fcis-7983 87 53 output output NOUN fcis-7983 87 54 at at ADP fcis-7983 87 55 time time NOUN fcis-7983 87 56 t t PROPN fcis-7983 87 57 ; ; PUNCT fcis-7983 87 58 Οƒ Οƒ PROPN fcis-7983 87 59 and and CCONJ fcis-7983 87 60 tanh tanh PROPN fcis-7983 87 61 are be AUX fcis-7983 87 62 the the DET fcis-7983 87 63 sigmoid sigmoid NOUN fcis-7983 87 64 function function NOUN fcis-7983 87 65 ; ; PUNCT fcis-7983 87 66 ct ct PROPN fcis-7983 87 67 is be AUX fcis-7983 87 68 the the DET fcis-7983 87 69 cell cell NOUN fcis-7983 87 70 state state NOUN fcis-7983 87 71 at at ADP fcis-7983 87 72 time time NOUN fcis-7983 87 73 t. t. PROPN fcis-7983 87 74 the the DET fcis-7983 87 75 gate gate PROPN fcis-7983 87 76 unit unit NOUN fcis-7983 87 77 calculation calculation NOUN fcis-7983 87 78 formulas formula NOUN fcis-7983 87 79 in in ADP fcis-7983 87 80 lstm lstm NOUN fcis-7983 87 81 are be AUX fcis-7983 87 82 shown show VERB fcis-7983 87 83 below below ADP fcis-7983 87 84 : : PUNCT fcis-7983 88 1 𝑓𝑑 𝑓𝑑 PROPN fcis-7983 88 2 = = PUNCT fcis-7983 88 3 𝜎(π‘Šπ‘“[β„Žπ‘‘βˆ’1 𝜎(π‘Šπ‘“[β„Žπ‘‘βˆ’1 PROPN fcis-7983 88 4 , , PUNCT fcis-7983 88 5 π‘₯𝑑 π‘₯𝑑 ADP fcis-7983 88 6 ] ] PUNCT fcis-7983 88 7 + + CCONJ fcis-7983 88 8 𝑏𝑓 𝑏𝑓 PROPN fcis-7983 88 9 ) ) PUNCT fcis-7983 88 10 ( ( PUNCT fcis-7983 88 11 1 1 X fcis-7983 88 12 ) ) PUNCT fcis-7983 88 13 𝑖𝑑 𝑖𝑑 NOUN fcis-7983 89 1 = = SYM fcis-7983 89 2 𝜎(π‘Šπ‘–[β„Žπ‘‘βˆ’1 𝜎(π‘Šπ‘–[β„Žπ‘‘βˆ’1 PROPN fcis-7983 89 3 , , PUNCT fcis-7983 89 4 π‘₯𝑑 π‘₯𝑑 ADP fcis-7983 89 5 ] ] X fcis-7983 89 6 + + CCONJ fcis-7983 89 7 𝑏𝑖 𝑏𝑖 X fcis-7983 89 8 ) ) PUNCT fcis-7983 89 9 ( ( PUNCT fcis-7983 89 10 2 2 X fcis-7983 89 11 ) ) PUNCT fcis-7983 89 12 π‘œπ‘‘ π‘œπ‘‘ NOUN fcis-7983 89 13 = = SYM fcis-7983 89 14 𝜎(π‘Šπ‘œ[β„Žπ‘‘βˆ’1 𝜎(π‘Šπ‘œ[β„Žπ‘‘βˆ’1 PROPN fcis-7983 89 15 , , PUNCT fcis-7983 89 16 π‘₯𝑑 π‘₯𝑑 ADP fcis-7983 89 17 ] ] X fcis-7983 90 1 + + CCONJ fcis-7983 90 2 π‘π‘œ π‘π‘œ X fcis-7983 90 3 ) ) PUNCT fcis-7983 90 4 ( ( PUNCT fcis-7983 90 5 3 3 X fcis-7983 90 6 ) ) PUNCT fcis-7983 90 7 οΏ½ οΏ½ PROPN fcis-7983 90 8 Μƒ Μƒ PROPN fcis-7983 90 9 οΏ½ οΏ½ PROPN fcis-7983 90 10 𝑑 𝑑 PROPN fcis-7983 90 11 = = PROPN fcis-7983 90 12 π‘‘π‘Žπ‘›β„Ž(π‘Šπ‘[β„Žπ‘‘βˆ’1 π‘‘π‘Žπ‘›β„Ž(π‘Šπ‘[β„Žπ‘‘βˆ’1 PROPN fcis-7983 90 13 , , PUNCT fcis-7983 90 14 π‘₯𝑑 π‘₯𝑑 ADP fcis-7983 90 15 ] ] X fcis-7983 90 16 + + CCONJ fcis-7983 90 17 𝑏𝑐 𝑏𝑐 X fcis-7983 90 18 ) ) PUNCT fcis-7983 90 19 ( ( PUNCT fcis-7983 90 20 4 4 X fcis-7983 90 21 ) ) PUNCT fcis-7983 90 22 𝐢𝑑 𝐢𝑑 PROPN fcis-7983 90 23 = = SYM fcis-7983 90 24 π‘“π‘‘β¨€πΆπ‘‘βˆ’1 π‘“π‘‘β¨€πΆπ‘‘βˆ’1 PROPN fcis-7983 90 25 + + NUM fcis-7983 90 26 𝑖𝑑⨀ 𝑖𝑑⨀ PROPN fcis-7983 90 27 οΏ½ οΏ½ PROPN fcis-7983 90 28 Μƒ Μƒ PROPN fcis-7983 90 29 οΏ½ οΏ½ PROPN fcis-7983 90 30 𝑑 𝑑 PROPN fcis-7983 90 31 ( ( PUNCT fcis-7983 90 32 5 5 NUM fcis-7983 90 33 ) ) PUNCT fcis-7983 90 34 β„Žπ‘‘ β„Žπ‘‘ NOUN fcis-7983 90 35 = = SYM fcis-7983 90 36 π‘œπ‘‘β¨€π‘‘π‘Žπ‘›β„Ž π‘œπ‘‘β¨€π‘‘π‘Žπ‘›β„Ž NOUN fcis-7983 90 37 ( ( PUNCT fcis-7983 90 38 𝐢𝑑 𝐢𝑑 PROPN fcis-7983 90 39 ) ) PUNCT fcis-7983 90 40 ( ( PUNCT fcis-7983 90 41 6 6 NUM fcis-7983 90 42 ) ) PUNCT fcis-7983 90 43 in in ADP fcis-7983 90 44 the the DET fcis-7983 90 45 formula formula NOUN fcis-7983 90 46 , , PUNCT fcis-7983 90 47 wi wi PROPN fcis-7983 90 48 , , PUNCT fcis-7983 90 49 wf wf PROPN fcis-7983 90 50 , , PUNCT fcis-7983 90 51 and and CCONJ fcis-7983 90 52 wo will AUX fcis-7983 90 53 are be AUX fcis-7983 90 54 weight weight NOUN fcis-7983 90 55 matrices matrix NOUN fcis-7983 90 56 that that PRON fcis-7983 90 57 connect connect VERB fcis-7983 90 58 to to ADP fcis-7983 90 59 the the DET fcis-7983 90 60 gate gate NOUN fcis-7983 90 61 unit unit NOUN fcis-7983 90 62 ; ; PUNCT fcis-7983 90 63 bi bi NOUN fcis-7983 90 64 , , PUNCT fcis-7983 90 65 bf bf NOUN fcis-7983 90 66 , , PUNCT fcis-7983 90 67 and and CCONJ fcis-7983 90 68 bo bo PROPN fcis-7983 90 69 are be AUX fcis-7983 90 70 the the DET fcis-7983 90 71 bias bias NOUN fcis-7983 90 72 values value NOUN fcis-7983 90 73 , , PUNCT fcis-7983 90 74 represents represent VERB fcis-7983 90 75 element element ADJ fcis-7983 90 76 - - ADJ fcis-7983 90 77 wise wise ADJ fcis-7983 90 78 multiplication multiplication NOUN fcis-7983 90 79 . . PUNCT fcis-7983 91 1 3.3 3.3 NUM fcis-7983 91 2 . . PUNCT fcis-7983 92 1 radical radical ADJ fcis-7983 92 2 embedding embed VERB fcis-7983 92 3 unlike unlike ADP fcis-7983 92 4 english english PROPN fcis-7983 92 5 , , PUNCT fcis-7983 92 6 chinese chinese ADJ fcis-7983 92 7 characters character NOUN fcis-7983 92 8 are be AUX fcis-7983 92 9 ideographic ideographic ADJ fcis-7983 92 10 , , PUNCT fcis-7983 92 11 and and CCONJ fcis-7983 92 12 most most ADJ fcis-7983 92 13 of of ADP fcis-7983 92 14 the the DET fcis-7983 92 15 radical radical ADJ fcis-7983 92 16 components component NOUN fcis-7983 92 17 still still ADV fcis-7983 92 18 retain retain VERB fcis-7983 92 19 their their PRON fcis-7983 92 20 original original ADJ fcis-7983 92 21 meanings meaning NOUN fcis-7983 92 22 , , PUNCT fcis-7983 92 23 with with ADP fcis-7983 92 24 rich rich ADJ fcis-7983 92 25 intrinsic intrinsic ADJ fcis-7983 92 26 features feature NOUN fcis-7983 92 27 . . PUNCT fcis-7983 93 1 for for ADP fcis-7983 93 2 example example NOUN fcis-7983 93 3 , , PUNCT fcis-7983 93 4 " " PUNCT fcis-7983 93 5 口 口 INTJ fcis-7983 93 6 " " PUNCT fcis-7983 93 7 ( ( PUNCT fcis-7983 93 8 mouth mouth NOUN fcis-7983 93 9 ) ) PUNCT fcis-7983 93 10 often often ADV fcis-7983 93 11 appears appear VERB fcis-7983 93 12 in in ADP fcis-7983 93 13 symptom symptom NOUN fcis-7983 93 14 entities entity NOUN fcis-7983 93 15 , , PUNCT fcis-7983 93 16 such such ADJ fcis-7983 93 17 as as ADP fcis-7983 93 18 vomiting vomiting NOUN fcis-7983 93 19 ; ; PUNCT fcis-7983 93 20 " " PUNCT fcis-7983 93 21 月 月 X fcis-7983 93 22 " " PUNCT fcis-7983 93 23 is be AUX fcis-7983 93 24 a a DET fcis-7983 93 25 simplified simplified ADJ fcis-7983 93 26 form form NOUN fcis-7983 93 27 of of ADP fcis-7983 93 28 " " PUNCT fcis-7983 93 29 肉 肉 NOUN fcis-7983 93 30 " " PUNCT fcis-7983 93 31 ( ( PUNCT fcis-7983 93 32 meat meat NOUN fcis-7983 93 33 ) ) PUNCT fcis-7983 93 34 , , PUNCT fcis-7983 93 35 which which PRON fcis-7983 93 36 is be AUX fcis-7983 93 37 often often ADV fcis-7983 93 38 related relate VERB fcis-7983 93 39 to to ADP fcis-7983 93 40 body body NOUN fcis-7983 93 41 parts part NOUN fcis-7983 93 42 such such ADJ fcis-7983 93 43 as as ADP fcis-7983 93 44 the the DET fcis-7983 93 45 heart heart NOUN fcis-7983 93 46 and and CCONJ fcis-7983 93 47 chest chest NOUN fcis-7983 93 48 ; ; PUNCT fcis-7983 93 49 " " PUNCT fcis-7983 93 50 η–’ η–’ NOUN fcis-7983 93 51 " " PUNCT fcis-7983 93 52 often often ADV fcis-7983 93 53 appears appear VERB fcis-7983 93 54 in in ADP fcis-7983 93 55 symptom symptom NOUN fcis-7983 93 56 descriptions description NOUN fcis-7983 93 57 , , PUNCT fcis-7983 93 58 such such ADJ fcis-7983 93 59 as as ADP fcis-7983 93 60 pain pain NOUN fcis-7983 93 61 . . PUNCT fcis-7983 94 1 therefore therefore ADV fcis-7983 94 2 , , PUNCT fcis-7983 94 3 measuring measure VERB fcis-7983 94 4 the the DET fcis-7983 94 5 similarity similarity NOUN fcis-7983 94 6 between between ADP fcis-7983 94 7 entities entity NOUN fcis-7983 94 8 to to ADP fcis-7983 94 9 some some DET fcis-7983 94 10 extent extent NOUN fcis-7983 94 11 through through ADP fcis-7983 94 12 radical radical ADJ fcis-7983 94 13 features feature NOUN fcis-7983 94 14 is be AUX fcis-7983 94 15 feasible feasible ADJ fcis-7983 94 16 [ [ X fcis-7983 94 17 19 19 NUM fcis-7983 94 18 ] ] PUNCT fcis-7983 94 19 . . PUNCT fcis-7983 95 1 in in ADP fcis-7983 95 2 addition addition NOUN fcis-7983 95 3 , , PUNCT fcis-7983 95 4 radical radical ADJ fcis-7983 95 5 embedding embedding NOUN fcis-7983 95 6 can can AUX fcis-7983 95 7 enhance enhance VERB fcis-7983 95 8 the the DET fcis-7983 95 9 semantic semantic ADJ fcis-7983 95 10 information information NOUN fcis-7983 95 11 of of ADP fcis-7983 95 12 characters character NOUN fcis-7983 95 13 , , PUNCT fcis-7983 95 14 such such ADJ fcis-7983 95 15 as as ADP fcis-7983 95 16 improving improve VERB fcis-7983 95 17 the the DET fcis-7983 95 18 model model NOUN fcis-7983 95 19 's 's PART fcis-7983 95 20 generalization generalization NOUN fcis-7983 95 21 ability ability NOUN fcis-7983 95 22 for for ADP fcis-7983 95 23 characters character NOUN fcis-7983 95 24 that that PRON fcis-7983 95 25 appear appear VERB fcis-7983 95 26 only only ADV fcis-7983 95 27 in in ADP fcis-7983 95 28 the the DET fcis-7983 95 29 test test NOUN fcis-7983 95 30 set set VERB fcis-7983 95 31 and and CCONJ fcis-7983 95 32 not not PART fcis-7983 95 33 in in ADP fcis-7983 95 34 the the DET fcis-7983 95 35 training training NOUN fcis-7983 95 36 set set NOUN fcis-7983 95 37 [ [ X fcis-7983 95 38 20,21 20,21 NUM fcis-7983 95 39 ] ] PUNCT fcis-7983 95 40 . . PUNCT fcis-7983 96 1 to to PART fcis-7983 96 2 fully fully ADV fcis-7983 96 3 utilize utilize VERB fcis-7983 96 4 the the DET fcis-7983 96 5 potential potential ADJ fcis-7983 96 6 intrinsic intrinsic ADJ fcis-7983 96 7 features feature NOUN fcis-7983 96 8 in in ADP fcis-7983 96 9 electronic electronic ADJ fcis-7983 96 10 medical medical ADJ fcis-7983 96 11 record record NOUN fcis-7983 96 12 texts text NOUN fcis-7983 96 13 , , PUNCT fcis-7983 96 14 a a DET fcis-7983 96 15 cnn cnn PROPN fcis-7983 96 16 - - PUNCT fcis-7983 96 17 based base VERB fcis-7983 96 18 radical radical ADJ fcis-7983 96 19 extraction extraction NOUN fcis-7983 96 20 framework framework NOUN fcis-7983 96 21 is be AUX fcis-7983 96 22 designed design VERB fcis-7983 96 23 . . PUNCT fcis-7983 97 1 the the DET fcis-7983 97 2 cnn cnn PROPN fcis-7983 97 3 network network NOUN fcis-7983 97 4 is be AUX fcis-7983 97 5 used use VERB fcis-7983 97 6 to to PART fcis-7983 97 7 extract extract VERB fcis-7983 97 8 the the DET fcis-7983 97 9 implicit implicit ADJ fcis-7983 97 10 radical radical ADJ fcis-7983 97 11 information information NOUN fcis-7983 97 12 inside inside ADP fcis-7983 97 13 characters character NOUN fcis-7983 97 14 and and CCONJ fcis-7983 97 15 fine fine ADJ fcis-7983 97 16 - - PUNCT fcis-7983 97 17 tune tune NOUN fcis-7983 97 18 during during ADP fcis-7983 97 19 the the DET fcis-7983 97 20 training training NOUN fcis-7983 97 21 process process NOUN fcis-7983 97 22 . . PUNCT fcis-7983 98 1 the the DET fcis-7983 98 2 network network NOUN fcis-7983 98 3 consists consist VERB fcis-7983 98 4 of of ADP fcis-7983 98 5 three three NUM fcis-7983 98 6 parts part NOUN fcis-7983 98 7 : : PUNCT fcis-7983 98 8 radical radical ADJ fcis-7983 98 9 embedding embed VERB fcis-7983 98 10 layer layer NOUN fcis-7983 98 11 , , PUNCT fcis-7983 98 12 convolutional convolutional ADJ fcis-7983 98 13 layer layer NOUN fcis-7983 98 14 , , PUNCT fcis-7983 98 15 and and CCONJ fcis-7983 98 16 max max PROPN fcis-7983 98 17 - - PUNCT fcis-7983 98 18 pooling pool VERB fcis-7983 98 19 layer layer NOUN fcis-7983 98 20 . . PUNCT fcis-7983 99 1 all all DET fcis-7983 99 2 radicals radical NOUN fcis-7983 99 3 in in ADP fcis-7983 99 4 this this DET fcis-7983 99 5 paper paper NOUN fcis-7983 99 6 are be AUX fcis-7983 99 7 obtained obtain VERB fcis-7983 99 8 from from ADP fcis-7983 99 9 the the DET fcis-7983 99 10 corpus corpus NOUN fcis-7983 99 11 . . PUNCT fcis-7983 100 1 3.4 3.4 NUM fcis-7983 100 2 . . PUNCT fcis-7983 100 3 dictionary dictionary NOUN fcis-7983 100 4 embedding embed VERB fcis-7983 100 5 clinical clinical ADJ fcis-7983 100 6 texts text NOUN fcis-7983 100 7 contain contain VERB fcis-7983 100 8 many many ADJ fcis-7983 100 9 specialized specialized ADJ fcis-7983 100 10 terms term NOUN fcis-7983 100 11 , , PUNCT fcis-7983 100 12 which which PRON fcis-7983 100 13 are be AUX fcis-7983 100 14 constructed construct VERB fcis-7983 100 15 into into ADP fcis-7983 100 16 a a DET fcis-7983 100 17 term term NOUN fcis-7983 100 18 dictionary dictionary NOUN fcis-7983 100 19 as as ADP fcis-7983 100 20 features feature NOUN fcis-7983 100 21 for for ADP fcis-7983 100 22 the the DET fcis-7983 100 23 model model NOUN fcis-7983 100 24 [ [ X fcis-7983 100 25 22 22 NUM fcis-7983 100 26 ] ] PUNCT fcis-7983 100 27 . . PUNCT fcis-7983 101 1 given give VERB fcis-7983 101 2 a a DET fcis-7983 101 3 sentence sentence NOUN fcis-7983 101 4 𝑋 𝑋 NOUN fcis-7983 101 5 = = SYM fcis-7983 101 6 ( ( PUNCT fcis-7983 101 7 π‘₯1 π‘₯1 NOUN fcis-7983 101 8 , , PUNCT fcis-7983 101 9 π‘₯2 π‘₯2 NOUN fcis-7983 101 10 , , PUNCT fcis-7983 101 11 … … PUNCT fcis-7983 101 12 , , PUNCT fcis-7983 101 13 π‘₯𝑛 π‘₯𝑛 NOUN fcis-7983 101 14 ) ) PUNCT fcis-7983 101 15 of of ADP fcis-7983 101 16 length length NOUN fcis-7983 101 17 n n CCONJ fcis-7983 101 18 , , PUNCT fcis-7983 101 19 a a DET fcis-7983 101 20 feature feature NOUN fcis-7983 101 21 vector vector NOUN fcis-7983 101 22 𝑑𝑖is 𝑑𝑖is NOUN fcis-7983 101 23 constructed construct VERB fcis-7983 101 24 for for ADP fcis-7983 101 25 each each DET fcis-7983 101 26 character character NOUN fcis-7983 101 27 π‘₯𝑖based π‘₯𝑖base VERB fcis-7983 101 28 on on ADP fcis-7983 101 29 the the DET fcis-7983 101 30 dictionary dictionary ADJ fcis-7983 101 31 d d PROPN fcis-7983 101 32 and and CCONJ fcis-7983 101 33 the the DET fcis-7983 101 34 context context NOUN fcis-7983 101 35 . . PUNCT fcis-7983 102 1 dictionary dictionary PROPN fcis-7983 102 2 d d PROPN fcis-7983 102 3 is be AUX fcis-7983 102 4 constructed construct VERB fcis-7983 102 5 at at ADP fcis-7983 102 6 the the DET fcis-7983 102 7 entity entity NOUN fcis-7983 102 8 level level NOUN fcis-7983 102 9 , , PUNCT fcis-7983 102 10 while while SCONJ fcis-7983 102 11 the the DET fcis-7983 102 12 text text NOUN fcis-7983 102 13 sequence sequence NOUN fcis-7983 102 14 is be AUX fcis-7983 102 15 based base VERB fcis-7983 102 16 on on ADP fcis-7983 102 17 character character NOUN fcis-7983 102 18 - - PUNCT fcis-7983 102 19 level level NOUN fcis-7983 102 20 tagging tagging NOUN fcis-7983 102 21 , , PUNCT fcis-7983 102 22 so so CCONJ fcis-7983 102 23 different different ADJ fcis-7983 102 24 schemes scheme NOUN fcis-7983 102 25 are be AUX fcis-7983 102 26 used use VERB fcis-7983 102 27 to to PART fcis-7983 102 28 represent represent VERB fcis-7983 102 29 dictionary dictionary ADJ fcis-7983 102 30 features feature NOUN fcis-7983 102 31 . . PUNCT fcis-7983 103 1 through through ADP fcis-7983 103 2 the the DET fcis-7983 103 3 construction construction NOUN fcis-7983 103 4 steps step NOUN fcis-7983 103 5 of of ADP fcis-7983 103 6 the the DET fcis-7983 103 7 dictionary dictionary ADJ fcis-7983 103 8 feature feature NOUN fcis-7983 103 9 vector vector NOUN fcis-7983 103 10 , , PUNCT fcis-7983 103 11 given give VERB fcis-7983 103 12 a a DET fcis-7983 103 13 sentence sentence NOUN fcis-7983 103 14 x x NOUN fcis-7983 103 15 , , PUNCT fcis-7983 103 16 the the DET fcis-7983 103 17 character character NOUN fcis-7983 103 18 embedding embed VERB fcis-7983 103 19 𝑒𝑖 𝑒𝑖 NOUN fcis-7983 103 20 and and CCONJ fcis-7983 103 21 the the DET fcis-7983 103 22 feature feature NOUN fcis-7983 103 23 vector vector NOUN fcis-7983 103 24 di di NOUN fcis-7983 103 25 for for ADP fcis-7983 103 26 each each DET fcis-7983 103 27 character character NOUN fcis-7983 103 28 π‘₯𝑖 π‘₯𝑖 PRON fcis-7983 103 29 are be AUX fcis-7983 103 30 obtained obtain VERB fcis-7983 103 31 . . PUNCT fcis-7983 104 1 in in ADP fcis-7983 104 2 common common ADJ fcis-7983 104 3 models model NOUN fcis-7983 104 4 , , PUNCT fcis-7983 104 5 the the DET fcis-7983 104 6 original original ADJ fcis-7983 104 7 bi bi ADJ fcis-7983 104 8 - - ADJ fcis-7983 104 9 lstm lstm ADJ fcis-7983 104 10 - - PUNCT fcis-7983 104 11 crf crf NOUN fcis-7983 104 12 model model NOUN fcis-7983 104 13 only only ADV fcis-7983 104 14 takes take VERB fcis-7983 104 15 𝑒𝑖 𝑒𝑖 PRON fcis-7983 104 16 as as ADP fcis-7983 104 17 input input NOUN fcis-7983 104 18 . . PUNCT fcis-7983 105 1 since since SCONJ fcis-7983 105 2 dictionary dictionary ADJ fcis-7983 105 3 features feature NOUN fcis-7983 105 4 can can AUX fcis-7983 105 5 provide provide VERB fcis-7983 105 6 valuable valuable ADJ fcis-7983 105 7 information information NOUN fcis-7983 105 8 for for ADP fcis-7983 105 9 chinese chinese ADJ fcis-7983 105 10 electronic electronic ADJ fcis-7983 105 11 medical medical ADJ fcis-7983 105 12 record record NOUN fcis-7983 105 13 named name VERB fcis-7983 105 14 entity entity NOUN fcis-7983 105 15 recognition recognition NOUN fcis-7983 105 16 tasks task NOUN fcis-7983 105 17 , , PUNCT fcis-7983 105 18 they they PRON fcis-7983 105 19 are be AUX fcis-7983 105 20 integrated integrate VERB fcis-7983 105 21 into into ADP fcis-7983 105 22 the the DET fcis-7983 105 23 bert bert NOUN fcis-7983 105 24 - - PUNCT fcis-7983 105 25 bi bi ADJ fcis-7983 105 26 - - ADJ fcis-7983 105 27 lstm lstm ADJ fcis-7983 105 28 - - PUNCT fcis-7983 105 29 crf crf NOUN fcis-7983 105 30 model model NOUN fcis-7983 105 31 . . PUNCT fcis-7983 106 1 4 4 X fcis-7983 106 2 . . NUM fcis-7983 106 3 experiments experiment NOUN fcis-7983 106 4 and and CCONJ fcis-7983 106 5 results result VERB fcis-7983 106 6 4.1 4.1 NUM fcis-7983 106 7 . . PUNCT fcis-7983 107 1 experimental experimental ADJ fcis-7983 107 2 environment environment NOUN fcis-7983 107 3 and and CCONJ fcis-7983 107 4 experimental experimental ADJ fcis-7983 107 5 data datum NOUN fcis-7983 107 6 set set VERB fcis-7983 107 7 the the DET fcis-7983 107 8 specific specific ADJ fcis-7983 107 9 experimental experimental ADJ fcis-7983 107 10 environment environment NOUN fcis-7983 107 11 is be AUX fcis-7983 107 12 as as SCONJ fcis-7983 107 13 follows follow VERB fcis-7983 107 14 : : PUNCT fcis-7983 107 15 python python PROPN fcis-7983 107 16 was be AUX fcis-7983 107 17 chosen choose VERB fcis-7983 107 18 as as ADP fcis-7983 107 19 the the DET fcis-7983 107 20 development development NOUN fcis-7983 107 21 language language NOUN fcis-7983 107 22 , , PUNCT fcis-7983 107 23 the the DET fcis-7983 107 24 cpu cpu NOUN fcis-7983 107 25 model model NOUN fcis-7983 107 26 is be AUX fcis-7983 107 27 amd amd ADJ fcis-7983 107 28 ryzen ryzen ADJ fcis-7983 107 29 7 7 NUM fcis-7983 107 30 4800h 4800h NUM fcis-7983 107 31 , , PUNCT fcis-7983 107 32 and and CCONJ fcis-7983 107 33 pycharm pycharm PROPN fcis-7983 107 34 was be AUX fcis-7983 107 35 chosen choose VERB fcis-7983 107 36 as as ADP fcis-7983 107 37 the the DET fcis-7983 107 38 development development NOUN fcis-7983 107 39 tool tool NOUN fcis-7983 107 40 . . PUNCT fcis-7983 108 1 in in ADP fcis-7983 108 2 this this DET fcis-7983 108 3 study study NOUN fcis-7983 108 4 , , PUNCT fcis-7983 108 5 the the DET fcis-7983 108 6 ccks2019 ccks2019 PROPN fcis-7983 108 7 dataset dataset VERB fcis-7983 108 8 was be AUX fcis-7983 108 9 used use VERB fcis-7983 108 10 for for ADP fcis-7983 108 11 experimentation experimentation NOUN fcis-7983 108 12 . . PUNCT fcis-7983 109 1 ccks2019 ccks2019 PROPN fcis-7983 109 2 ( ( PUNCT fcis-7983 109 3 china china PROPN fcis-7983 109 4 conference conference PROPN fcis-7983 109 5 on on ADP fcis-7983 109 6 knowledge knowledge NOUN fcis-7983 109 7 graph graph NOUN fcis-7983 109 8 and and CCONJ fcis-7983 109 9 semantic semantic ADJ fcis-7983 109 10 computing computing NOUN fcis-7983 109 11 ) ) PUNCT fcis-7983 109 12 is be AUX fcis-7983 109 13 an an DET fcis-7983 109 14 academic academic ADJ fcis-7983 109 15 conference conference NOUN fcis-7983 109 16 on on ADP fcis-7983 109 17 knowledge knowledge NOUN fcis-7983 109 18 graphs graph NOUN fcis-7983 109 19 and and CCONJ fcis-7983 109 20 semantic semantic ADJ fcis-7983 109 21 computing computing NOUN fcis-7983 109 22 , , PUNCT fcis-7983 109 23 as as ADV fcis-7983 109 24 well well ADV fcis-7983 109 25 as as ADP fcis-7983 109 26 a a DET fcis-7983 109 27 chinese chinese PROPN fcis-7983 109 28 - - PUNCT fcis-7983 109 29 based base VERB fcis-7983 109 30 natural natural ADJ fcis-7983 109 31 language language NOUN fcis-7983 109 32 processing processing NOUN fcis-7983 109 33 ( ( PUNCT fcis-7983 109 34 nlp nlp NOUN fcis-7983 109 35 ) ) PUNCT fcis-7983 109 36 competition competition NOUN fcis-7983 109 37 . . PUNCT fcis-7983 110 1 the the DET fcis-7983 110 2 named name VERB fcis-7983 110 3 entity entity NOUN fcis-7983 110 4 recognition recognition NOUN fcis-7983 110 5 ( ( PUNCT fcis-7983 110 6 ner ner NOUN fcis-7983 110 7 ) ) PUNCT fcis-7983 110 8 part part NOUN fcis-7983 110 9 of of ADP fcis-7983 110 10 the the DET fcis-7983 110 11 ccks2019 ccks2019 PROPN fcis-7983 110 12 dataset dataset VERB fcis-7983 110 13 mainly mainly ADV fcis-7983 110 14 includes include VERB fcis-7983 110 15 annotated annotated ADJ fcis-7983 110 16 data datum NOUN fcis-7983 110 17 of of ADP fcis-7983 110 18 entities entity NOUN fcis-7983 110 19 related relate VERB fcis-7983 110 20 to to ADP fcis-7983 110 21 medical medical ADJ fcis-7983 110 22 fields field NOUN fcis-7983 110 23 , , PUNCT fcis-7983 110 24 including include VERB fcis-7983 110 25 6 6 NUM fcis-7983 110 26 types type NOUN fcis-7983 110 27 of of ADP fcis-7983 110 28 entities entity NOUN fcis-7983 110 29 , , PUNCT fcis-7983 110 30 namely namely ADV fcis-7983 110 31 diseases disease NOUN fcis-7983 110 32 and and CCONJ fcis-7983 110 33 diagnoses diagnosis NOUN fcis-7983 110 34 , , PUNCT fcis-7983 110 35 surgeries surgery NOUN fcis-7983 110 36 , , PUNCT fcis-7983 110 37 drugs drug NOUN fcis-7983 110 38 , , PUNCT fcis-7983 110 39 anatomical anatomical ADJ fcis-7983 110 40 sites site NOUN fcis-7983 110 41 , , PUNCT fcis-7983 110 42 imaging imaging NOUN fcis-7983 110 43 examinations examination NOUN fcis-7983 110 44 , , PUNCT fcis-7983 110 45 and and CCONJ fcis-7983 110 46 laboratory laboratory NOUN fcis-7983 110 47 tests test NOUN fcis-7983 110 48 , , PUNCT fcis-7983 110 49 with with ADP fcis-7983 110 50 a a DET fcis-7983 110 51 total total NOUN fcis-7983 110 52 of of ADP fcis-7983 110 53 1379 1379 NUM fcis-7983 110 54 data datum NOUN fcis-7983 110 55 . . PUNCT fcis-7983 111 1 the the DET fcis-7983 111 2 specific specific ADJ fcis-7983 111 3 entity entity NOUN fcis-7983 111 4 types type NOUN fcis-7983 111 5 are be AUX fcis-7983 111 6 shown show VERB fcis-7983 111 7 in in ADP fcis-7983 111 8 table table NOUN fcis-7983 111 9 1 1 NUM fcis-7983 111 10 . . PUNCT fcis-7983 111 11 table table NOUN fcis-7983 111 12 1 1 NUM fcis-7983 111 13 . . PUNCT fcis-7983 112 1 ccks2019 ccks2019 PROPN fcis-7983 112 2 dataset dataset VERB fcis-7983 112 3 ccks2019 ccks2019 PROPN fcis-7983 112 4 disease disease NOUN fcis-7983 112 5 and and CCONJ fcis-7983 112 6 diagnosis diagnosis NOUN fcis-7983 112 7 imaging imaging NOUN fcis-7983 112 8 examination examination NOUN fcis-7983 112 9 laboratory laboratory NOUN fcis-7983 112 10 test test NOUN fcis-7983 112 11 operation operation NOUN fcis-7983 112 12 drug drug NOUN fcis-7983 112 13 anatomical anatomical ADJ fcis-7983 112 14 site site NOUN fcis-7983 112 15 train train VERB fcis-7983 112 16 2116 2116 NUM fcis-7983 112 17 222 222 NUM fcis-7983 112 18 318 318 NUM fcis-7983 112 19 765 765 NUM fcis-7983 112 20 456 456 NUM fcis-7983 112 21 1486 1486 NUM fcis-7983 112 22 test test NOUN fcis-7983 112 23 682 682 NUM fcis-7983 112 24 91 91 NUM fcis-7983 112 25 193 193 NUM fcis-7983 112 26 140 140 NUM fcis-7983 112 27 263 263 NUM fcis-7983 112 28 447 447 NUM fcis-7983 112 29 4.2 4.2 NUM fcis-7983 112 30 . . PUNCT fcis-7983 113 1 evaluation evaluation NOUN fcis-7983 113 2 metrics metric NOUN fcis-7983 113 3 the the DET fcis-7983 113 4 precision precision NOUN fcis-7983 113 5 ( ( PUNCT fcis-7983 113 6 p p NOUN fcis-7983 113 7 ) ) PUNCT fcis-7983 113 8 , , PUNCT fcis-7983 113 9 recall recall INTJ fcis-7983 113 10 ( ( PUNCT fcis-7983 113 11 r r NOUN fcis-7983 113 12 ) ) PUNCT fcis-7983 113 13 , , PUNCT fcis-7983 113 14 and and CCONJ fcis-7983 113 15 f1 f1 NOUN fcis-7983 113 16 - - PUNCT fcis-7983 113 17 score score NOUN fcis-7983 113 18 were be AUX fcis-7983 113 19 used use VERB fcis-7983 113 20 as as ADP fcis-7983 113 21 the the DET fcis-7983 113 22 evaluation evaluation NOUN fcis-7983 113 23 metrics metric NOUN fcis-7983 113 24 in in ADP fcis-7983 113 25 the the DET fcis-7983 113 26 experiments experiment NOUN fcis-7983 113 27 , , PUNCT fcis-7983 113 28 and and CCONJ fcis-7983 113 29 their their PRON fcis-7983 113 30 formulas formula NOUN fcis-7983 113 31 are be AUX fcis-7983 113 32 as as SCONJ fcis-7983 113 33 follows follow VERB fcis-7983 113 34 : : PUNCT fcis-7983 113 35 𝑃 𝑃 PROPN fcis-7983 113 36 = = PUNCT fcis-7983 113 37 𝑇𝑃 𝑇𝑃 NOUN fcis-7983 113 38 𝑇𝑃+𝐹𝑃 𝑇𝑃+𝐹𝑃 NUM fcis-7983 113 39 ( ( PUNCT fcis-7983 113 40 7 7 X fcis-7983 113 41 ) ) PUNCT fcis-7983 113 42 𝑅 𝑅 NOUN fcis-7983 113 43 = = PUNCT fcis-7983 113 44 𝑇𝑃 𝑇𝑃 PROPN fcis-7983 113 45 𝑇𝑃+𝐹𝑁 𝑇𝑃+𝐹𝑁 NOUN fcis-7983 113 46 ( ( PUNCT fcis-7983 113 47 8) 8) NUM fcis-7983 113 48 𝐹1 𝐹1 NOUN fcis-7983 113 49 = = SYM fcis-7983 113 50 2×𝑃×𝑅 2×𝑃×𝑅 NOUN fcis-7983 113 51 𝑃+𝑅 𝑃+𝑅 NOUN fcis-7983 113 52 ( ( PUNCT fcis-7983 113 53 9 9 NUM fcis-7983 113 54 ) ) PUNCT fcis-7983 113 55 in in ADP fcis-7983 113 56 the the DET fcis-7983 113 57 formula formula NOUN fcis-7983 113 58 , , PUNCT fcis-7983 113 59 tp tp NOUN fcis-7983 113 60 represents represent VERB fcis-7983 113 61 true true ADJ fcis-7983 113 62 positive positive ADJ fcis-7983 113 63 which which PRON fcis-7983 113 64 is be AUX fcis-7983 113 65 the the DET fcis-7983 113 66 9 9 NUM fcis-7983 113 67 number number NOUN fcis-7983 113 68 of of ADP fcis-7983 113 69 positive positive ADJ fcis-7983 113 70 instances instance NOUN fcis-7983 113 71 that that PRON fcis-7983 113 72 are be AUX fcis-7983 113 73 correctly correctly ADV fcis-7983 113 74 predicted predict VERB fcis-7983 113 75 as as ADP fcis-7983 113 76 positive positive ADJ fcis-7983 113 77 , , PUNCT fcis-7983 113 78 fp fp X fcis-7983 113 79 represents represent VERB fcis-7983 113 80 false false ADV fcis-7983 113 81 positive positive ADJ fcis-7983 113 82 which which PRON fcis-7983 113 83 is be AUX fcis-7983 113 84 the the DET fcis-7983 113 85 number number NOUN fcis-7983 113 86 of of ADP fcis-7983 113 87 negative negative ADJ fcis-7983 113 88 instances instance NOUN fcis-7983 113 89 that that PRON fcis-7983 113 90 are be AUX fcis-7983 113 91 wrongly wrongly ADV fcis-7983 113 92 predicted predict VERB fcis-7983 113 93 as as ADP fcis-7983 113 94 positive positive ADJ fcis-7983 113 95 , , PUNCT fcis-7983 113 96 and and CCONJ fcis-7983 113 97 fn fn NOUN fcis-7983 113 98 represents represent VERB fcis-7983 113 99 false false ADJ fcis-7983 113 100 negative negative ADJ fcis-7983 113 101 which which PRON fcis-7983 113 102 is be AUX fcis-7983 113 103 the the DET fcis-7983 113 104 number number NOUN fcis-7983 113 105 of of ADP fcis-7983 113 106 positive positive ADJ fcis-7983 113 107 instances instance NOUN fcis-7983 113 108 that that PRON fcis-7983 113 109 are be AUX fcis-7983 113 110 wrongly wrongly ADV fcis-7983 113 111 predicted predict VERB fcis-7983 113 112 as as ADP fcis-7983 113 113 negative negative ADJ fcis-7983 113 114 . . PUNCT fcis-7983 114 1 f1 f1 ADJ fcis-7983 114 2 - - PUNCT fcis-7983 114 3 score score NOUN fcis-7983 114 4 takes take VERB fcis-7983 114 5 into into ADP fcis-7983 114 6 account account NOUN fcis-7983 114 7 both both DET fcis-7983 114 8 precision precision NOUN fcis-7983 114 9 and and CCONJ fcis-7983 114 10 recall recall NOUN fcis-7983 114 11 of of ADP fcis-7983 114 12 a a DET fcis-7983 114 13 classification classification NOUN fcis-7983 114 14 model model NOUN fcis-7983 114 15 , , PUNCT fcis-7983 114 16 and and CCONJ fcis-7983 114 17 is be AUX fcis-7983 114 18 the the DET fcis-7983 114 19 harmonic harmonic ADJ fcis-7983 114 20 mean mean NOUN fcis-7983 114 21 of of ADP fcis-7983 114 22 precision precision NOUN fcis-7983 114 23 and and CCONJ fcis-7983 114 24 recall recall NOUN fcis-7983 114 25 . . PUNCT fcis-7983 115 1 4.3 4.3 NUM fcis-7983 115 2 . . PUNCT fcis-7983 116 1 experimental experimental ADJ fcis-7983 116 2 result result NOUN fcis-7983 116 3 an an DET fcis-7983 116 4 improved improved ADJ fcis-7983 116 5 feature feature NOUN fcis-7983 116 6 fusion fusion NOUN fcis-7983 116 7 method method NOUN fcis-7983 116 8 for for ADP fcis-7983 116 9 electronic electronic ADJ fcis-7983 116 10 medical medical ADJ fcis-7983 116 11 record record NOUN fcis-7983 116 12 named name VERB fcis-7983 116 13 entity entity NOUN fcis-7983 116 14 recognition recognition NOUN fcis-7983 116 15 based base VERB fcis-7983 116 16 on on ADP fcis-7983 116 17 bert bert PROPN fcis-7983 116 18 - - PUNCT fcis-7983 116 19 bi bi NOUN fcis-7983 116 20 - - ADJ fcis-7983 116 21 lstmcrf lstmcrf PROPN fcis-7983 116 22 is be AUX fcis-7983 116 23 proposed propose VERB fcis-7983 116 24 by by ADP fcis-7983 116 25 combining combine VERB fcis-7983 116 26 chinese chinese ADJ fcis-7983 116 27 radical radical ADJ fcis-7983 116 28 and and CCONJ fcis-7983 116 29 domain domain NOUN fcis-7983 116 30 dictionary dictionary NOUN fcis-7983 116 31 features feature VERB fcis-7983 116 32 with with ADP fcis-7983 116 33 deep deep ADJ fcis-7983 116 34 learning learning NOUN fcis-7983 116 35 models model NOUN fcis-7983 116 36 . . PUNCT fcis-7983 117 1 by by ADP fcis-7983 117 2 studying study VERB fcis-7983 117 3 the the DET fcis-7983 117 4 distribution distribution NOUN fcis-7983 117 5 of of ADP fcis-7983 117 6 radicals radical NOUN fcis-7983 117 7 for for ADP fcis-7983 117 8 different different ADJ fcis-7983 117 9 entities entity NOUN fcis-7983 117 10 in in ADP fcis-7983 117 11 the the DET fcis-7983 117 12 dataset dataset NOUN fcis-7983 117 13 , , PUNCT fcis-7983 117 14 the the DET fcis-7983 117 15 basic basic ADJ fcis-7983 117 16 features feature NOUN fcis-7983 117 17 of of ADP fcis-7983 117 18 chinese chinese ADJ fcis-7983 117 19 characters character NOUN fcis-7983 117 20 are be AUX fcis-7983 117 21 embedded embed VERB fcis-7983 117 22 into into ADP fcis-7983 117 23 the the DET fcis-7983 117 24 model model NOUN fcis-7983 117 25 to to PART fcis-7983 117 26 enrich enrich VERB fcis-7983 117 27 the the DET fcis-7983 117 28 semantic semantic ADJ fcis-7983 117 29 features feature NOUN fcis-7983 117 30 . . PUNCT fcis-7983 118 1 three three NUM fcis-7983 118 2 models model NOUN fcis-7983 118 3 are be AUX fcis-7983 118 4 compared compare VERB fcis-7983 118 5 , , PUNCT fcis-7983 118 6 and and CCONJ fcis-7983 118 7 the the DET fcis-7983 118 8 bert bert NOUN fcis-7983 118 9 - - PUNCT fcis-7983 118 10 bi bi ADJ fcis-7983 118 11 - - ADJ fcis-7983 118 12 lstm lstm ADJ fcis-7983 118 13 - - PUNCT fcis-7983 118 14 crf crf NOUN fcis-7983 118 15 model model NOUN fcis-7983 118 16 with with ADP fcis-7983 118 17 fusion fusion NOUN fcis-7983 118 18 of of ADP fcis-7983 118 19 radical radical ADJ fcis-7983 118 20 embedding embedding NOUN fcis-7983 118 21 and and CCONJ fcis-7983 118 22 domain domain NOUN fcis-7983 118 23 dictionary dictionary NOUN fcis-7983 118 24 has have AUX fcis-7983 118 25 improved improve VERB fcis-7983 118 26 the the DET fcis-7983 118 27 performance performance NOUN fcis-7983 118 28 of of ADP fcis-7983 118 29 chinese chinese ADJ fcis-7983 118 30 electronic electronic ADJ fcis-7983 118 31 medical medical ADJ fcis-7983 118 32 record record NOUN fcis-7983 118 33 named name VERB fcis-7983 118 34 entity entity NOUN fcis-7983 118 35 recognition recognition NOUN fcis-7983 118 36 . . PUNCT fcis-7983 119 1 on on ADP fcis-7983 119 2 the the DET fcis-7983 119 3 ccks2019 ccks2019 PROPN fcis-7983 119 4 dataset dataset VERB fcis-7983 119 5 , , PUNCT fcis-7983 119 6 the the DET fcis-7983 119 7 f1 f1 PROPN fcis-7983 119 8 scores score NOUN fcis-7983 119 9 based base VERB fcis-7983 119 10 on on ADP fcis-7983 119 11 radical radical ADJ fcis-7983 119 12 embedding embedding NOUN fcis-7983 119 13 and and CCONJ fcis-7983 119 14 domain domain NOUN fcis-7983 119 15 dictionary dictionary NOUN fcis-7983 119 16 features feature NOUN fcis-7983 119 17 have have AUX fcis-7983 119 18 increased increase VERB fcis-7983 119 19 by by ADP fcis-7983 119 20 4.92 4.92 NUM fcis-7983 119 21 % % NOUN fcis-7983 119 22 and and CCONJ fcis-7983 119 23 6.90 6.90 NUM fcis-7983 119 24 % % NOUN fcis-7983 119 25 , , PUNCT fcis-7983 119 26 respectively respectively ADV fcis-7983 119 27 , , PUNCT fcis-7983 119 28 compared compare VERB fcis-7983 119 29 to to ADP fcis-7983 119 30 the the DET fcis-7983 119 31 bert bert PROPN fcis-7983 119 32 - - PUNCT fcis-7983 119 33 bi bi ADJ fcis-7983 119 34 - - ADJ fcis-7983 119 35 lstm lstm ADJ fcis-7983 119 36 - - PUNCT fcis-7983 119 37 crf crf NOUN fcis-7983 119 38 model model NOUN fcis-7983 119 39 , , PUNCT fcis-7983 119 40 indicating indicate VERB fcis-7983 119 41 that that SCONJ fcis-7983 119 42 adding add VERB fcis-7983 119 43 radical radical ADJ fcis-7983 119 44 embedding embedding NOUN fcis-7983 119 45 and and CCONJ fcis-7983 119 46 dictionary dictionary ADJ fcis-7983 119 47 features feature NOUN fcis-7983 119 48 can can AUX fcis-7983 119 49 effectively effectively ADV fcis-7983 119 50 improve improve VERB fcis-7983 119 51 the the DET fcis-7983 119 52 performance performance NOUN fcis-7983 119 53 of of ADP fcis-7983 119 54 the the DET fcis-7983 119 55 model model NOUN fcis-7983 119 56 . . PUNCT fcis-7983 120 1 table table NOUN fcis-7983 120 2 2 2 NUM fcis-7983 120 3 . . PUNCT fcis-7983 120 4 comparison comparison NOUN fcis-7983 120 5 of of ADP fcis-7983 120 6 experimental experimental ADJ fcis-7983 120 7 results result NOUN fcis-7983 120 8 of of ADP fcis-7983 120 9 different different ADJ fcis-7983 120 10 models model NOUN fcis-7983 120 11 model model VERB fcis-7983 120 12 p p NOUN fcis-7983 120 13 r r NOUN fcis-7983 120 14 f1 f1 NOUN fcis-7983 120 15 bert bert PROPN fcis-7983 120 16 - - PUNCT fcis-7983 120 17 bi bi NOUN fcis-7983 120 18 - - ADJ fcis-7983 120 19 lstm lstm ADJ fcis-7983 120 20 - - PUNCT fcis-7983 120 21 cr cr NOUN fcis-7983 120 22 72.25 72.25 NUM fcis-7983 120 23 % % NOUN fcis-7983 120 24 71.69 71.69 NUM fcis-7983 120 25 % % NOUN fcis-7983 120 26 71.62 71.62 NUM fcis-7983 120 27 % % NOUN fcis-7983 120 28 radical radical ADJ fcis-7983 120 29 embedding embed VERB fcis-7983 120 30 78.36 78.36 NUM fcis-7983 120 31 % % NOUN fcis-7983 120 32 76.88 76.88 NUM fcis-7983 120 33 % % NOUN fcis-7983 120 34 76.54 76.54 NUM fcis-7983 120 35 % % NOUN fcis-7983 120 36 dictionary dictionary NOUN fcis-7983 120 37 embedding embed VERB fcis-7983 120 38 81.28 81.28 NUM fcis-7983 120 39 % % NOUN fcis-7983 120 40 78.34 78.34 NUM fcis-7983 120 41 % % NOUN fcis-7983 120 42 78.52 78.52 NUM fcis-7983 120 43 % % NOUN fcis-7983 120 44 5 5 NUM fcis-7983 120 45 . . PUNCT fcis-7983 121 1 conclusion conclusion NOUN fcis-7983 121 2 based base VERB fcis-7983 121 3 on on ADP fcis-7983 121 4 the the DET fcis-7983 121 5 named name VERB fcis-7983 121 6 entity entity NOUN fcis-7983 121 7 recognition recognition NOUN fcis-7983 121 8 data datum NOUN fcis-7983 121 9 set set VERB fcis-7983 121 10 provided provide VERB fcis-7983 121 11 by by ADP fcis-7983 121 12 ccks2019 ccks2019 PROPN fcis-7983 121 13 ( ( PUNCT fcis-7983 121 14 chinese chinese ADJ fcis-7983 121 15 conference conference NOUN fcis-7983 121 16 on on ADP fcis-7983 121 17 knowledge knowledge NOUN fcis-7983 121 18 graph graph NOUN fcis-7983 121 19 and and CCONJ fcis-7983 121 20 semantic semantic ADJ fcis-7983 121 21 computing computing NOUN fcis-7983 121 22 ) ) PUNCT fcis-7983 121 23 , , PUNCT fcis-7983 121 24 this this DET fcis-7983 121 25 paper paper NOUN fcis-7983 121 26 proposes propose VERB fcis-7983 121 27 a a DET fcis-7983 121 28 chinese chinese PROPN fcis-7983 121 29 named name VERB fcis-7983 121 30 entity entity NOUN fcis-7983 121 31 recognition recognition NOUN fcis-7983 121 32 method method NOUN fcis-7983 121 33 based base VERB fcis-7983 121 34 on on ADP fcis-7983 121 35 bert bert PROPN fcis-7983 121 36 - - PUNCT fcis-7983 121 37 bi bi ADJ fcis-7983 121 38 - - ADJ fcis-7983 121 39 lstm lstm ADJ fcis-7983 121 40 - - PUNCT fcis-7983 121 41 crf crf NOUN fcis-7983 121 42 model model NOUN fcis-7983 121 43 , , PUNCT fcis-7983 121 44 and and CCONJ fcis-7983 121 45 introduces introduce VERB fcis-7983 121 46 the the DET fcis-7983 121 47 function function NOUN fcis-7983 121 48 of of ADP fcis-7983 121 49 side side NOUN fcis-7983 121 50 radical radical ADJ fcis-7983 121 51 and and CCONJ fcis-7983 121 52 dictionary dictionary ADJ fcis-7983 121 53 to to PART fcis-7983 121 54 improve improve VERB fcis-7983 121 55 the the DET fcis-7983 121 56 recognition recognition NOUN fcis-7983 121 57 ability ability NOUN fcis-7983 121 58 of of ADP fcis-7983 121 59 the the DET fcis-7983 121 60 model model NOUN fcis-7983 121 61 . . PUNCT fcis-7983 122 1 specifically specifically ADV fcis-7983 122 2 , , PUNCT fcis-7983 122 3 the the DET fcis-7983 122 4 author author NOUN fcis-7983 122 5 adds add VERB fcis-7983 122 6 the the DET fcis-7983 122 7 partial partial ADJ fcis-7983 122 8 radicals radical NOUN fcis-7983 122 9 and and CCONJ fcis-7983 122 10 dictionary dictionary ADJ fcis-7983 122 11 information information NOUN fcis-7983 122 12 to to ADP fcis-7983 122 13 the the DET fcis-7983 122 14 input input NOUN fcis-7983 122 15 layer layer NOUN fcis-7983 122 16 , , PUNCT fcis-7983 122 17 and and CCONJ fcis-7983 122 18 inputs input VERB fcis-7983 122 19 them they PRON fcis-7983 122 20 together together ADV fcis-7983 122 21 with with ADP fcis-7983 122 22 the the DET fcis-7983 122 23 bert bert PROPN fcis-7983 122 24 representation representation PROPN fcis-7983 122 25 vector vector NOUN fcis-7983 122 26 into into ADP fcis-7983 122 27 the the DET fcis-7983 122 28 bi bi ADJ fcis-7983 122 29 - - ADJ fcis-7983 122 30 lstm lstm ADJ fcis-7983 122 31 layer layer NOUN fcis-7983 122 32 for for ADP fcis-7983 122 33 feature feature NOUN fcis-7983 122 34 extraction extraction NOUN fcis-7983 122 35 . . PUNCT fcis-7983 123 1 then then ADV fcis-7983 123 2 , , PUNCT fcis-7983 123 3 the the DET fcis-7983 123 4 model model NOUN fcis-7983 123 5 is be AUX fcis-7983 123 6 trained train VERB fcis-7983 123 7 and and CCONJ fcis-7983 123 8 predicted predict VERB fcis-7983 123 9 using use VERB fcis-7983 123 10 crf crf NOUN fcis-7983 123 11 layer layer NOUN fcis-7983 123 12 , , PUNCT fcis-7983 123 13 and and CCONJ fcis-7983 123 14 the the DET fcis-7983 123 15 model model NOUN fcis-7983 123 16 is be AUX fcis-7983 123 17 tested test VERB fcis-7983 123 18 and and CCONJ fcis-7983 123 19 evaluated evaluate VERB fcis-7983 123 20 in in ADP fcis-7983 123 21 detail detail NOUN fcis-7983 123 22 . . PUNCT fcis-7983 124 1 the the DET fcis-7983 124 2 results result NOUN fcis-7983 124 3 show show VERB fcis-7983 124 4 that that SCONJ fcis-7983 124 5 the the DET fcis-7983 124 6 introduction introduction NOUN fcis-7983 124 7 of of ADP fcis-7983 124 8 partial partial ADJ fcis-7983 124 9 radicals radical NOUN fcis-7983 124 10 and and CCONJ fcis-7983 124 11 dictionary dictionary ADJ fcis-7983 124 12 information information NOUN fcis-7983 124 13 can can AUX fcis-7983 124 14 significantly significantly ADV fcis-7983 124 15 improve improve VERB fcis-7983 124 16 the the DET fcis-7983 124 17 recognition recognition NOUN fcis-7983 124 18 ability ability NOUN fcis-7983 124 19 of of ADP fcis-7983 124 20 the the DET fcis-7983 124 21 model model NOUN fcis-7983 124 22 , , PUNCT fcis-7983 124 23 and and CCONJ fcis-7983 124 24 the the DET fcis-7983 124 25 combination combination NOUN fcis-7983 124 26 of of ADP fcis-7983 124 27 bert bert NOUN fcis-7983 124 28 - - PUNCT fcis-7983 124 29 bi bi ADJ fcis-7983 124 30 - - ADJ fcis-7983 124 31 lstm lstm ADJ fcis-7983 124 32 - - PUNCT fcis-7983 124 33 crf crf NOUN fcis-7983 124 34 model model NOUN fcis-7983 124 35 can can AUX fcis-7983 124 36 further far ADV fcis-7983 124 37 improve improve VERB fcis-7983 124 38 the the DET fcis-7983 124 39 recognition recognition NOUN fcis-7983 124 40 accuracy accuracy NOUN fcis-7983 124 41 . . PUNCT fcis-7983 125 1 in in ADP fcis-7983 125 2 conclusion conclusion NOUN fcis-7983 125 3 , , PUNCT fcis-7983 125 4 the the DET fcis-7983 125 5 proposed propose VERB fcis-7983 125 6 chinese chinese PROPN fcis-7983 125 7 named name VERB fcis-7983 125 8 entity entity NOUN fcis-7983 125 9 recognition recognition NOUN fcis-7983 125 10 method method NOUN fcis-7983 125 11 based base VERB fcis-7983 125 12 on on ADP fcis-7983 125 13 bert bert PROPN fcis-7983 125 14 - - PUNCT fcis-7983 125 15 bi bi ADJ fcis-7983 125 16 - - ADJ fcis-7983 125 17 lstm lstm ADJ fcis-7983 125 18 - - PUNCT fcis-7983 125 19 crf crf NOUN fcis-7983 125 20 model model NOUN fcis-7983 125 21 introduces introduce VERB fcis-7983 125 22 the the DET fcis-7983 125 23 functions function NOUN fcis-7983 125 24 of of ADP fcis-7983 125 25 side side NOUN fcis-7983 125 26 radicals radical NOUN fcis-7983 125 27 and and CCONJ fcis-7983 125 28 dictionaries dictionary NOUN fcis-7983 125 29 at at ADP fcis-7983 125 30 the the DET fcis-7983 125 31 same same ADJ fcis-7983 125 32 time time NOUN fcis-7983 125 33 , , PUNCT fcis-7983 125 34 which which PRON fcis-7983 125 35 can can AUX fcis-7983 125 36 significantly significantly ADV fcis-7983 125 37 improve improve VERB fcis-7983 125 38 the the DET fcis-7983 125 39 recognition recognition NOUN fcis-7983 125 40 ability ability NOUN fcis-7983 125 41 of of ADP fcis-7983 125 42 the the DET fcis-7983 125 43 model model NOUN fcis-7983 125 44 and and CCONJ fcis-7983 125 45 has have VERB fcis-7983 125 46 good good ADJ fcis-7983 125 47 practical practical ADJ fcis-7983 125 48 value value NOUN fcis-7983 125 49 . . PUNCT fcis-7983 126 1 future future ADJ fcis-7983 126 2 research research NOUN fcis-7983 126 3 direction direction NOUN fcis-7983 126 4 can can AUX fcis-7983 126 5 further far ADV fcis-7983 126 6 explore explore VERB fcis-7983 126 7 how how SCONJ fcis-7983 126 8 to to PART fcis-7983 126 9 apply apply VERB fcis-7983 126 10 this this DET fcis-7983 126 11 method method NOUN fcis-7983 126 12 to to ADP fcis-7983 126 13 more more ADV fcis-7983 126 14 complex complex ADJ fcis-7983 126 15 natural natural ADJ fcis-7983 126 16 language language NOUN fcis-7983 126 17 processing processing NOUN fcis-7983 126 18 tasks task NOUN fcis-7983 126 19 , , PUNCT fcis-7983 126 20 such such ADJ fcis-7983 126 21 as as ADP fcis-7983 126 22 text text NOUN fcis-7983 126 23 classification classification NOUN fcis-7983 126 24 , , PUNCT fcis-7983 126 25 relation relation NOUN fcis-7983 126 26 extraction extraction NOUN fcis-7983 126 27 , , PUNCT fcis-7983 126 28 etc etc X fcis-7983 126 29 . . X fcis-7983 127 1 in in ADP fcis-7983 127 2 addition addition NOUN fcis-7983 127 3 , , PUNCT fcis-7983 127 4 other other ADJ fcis-7983 127 5 techniques technique NOUN fcis-7983 127 6 such such ADJ fcis-7983 127 7 as as ADP fcis-7983 127 8 transfer transfer NOUN fcis-7983 127 9 learning learning NOUN fcis-7983 127 10 and and CCONJ fcis-7983 127 11 confrontation confrontation NOUN fcis-7983 127 12 training training NOUN fcis-7983 127 13 can can AUX fcis-7983 127 14 be be AUX fcis-7983 127 15 combined combine VERB fcis-7983 127 16 to to PART fcis-7983 127 17 improve improve VERB fcis-7983 127 18 the the DET fcis-7983 127 19 performance performance NOUN fcis-7983 127 20 and and CCONJ fcis-7983 127 21 efficiency efficiency NOUN fcis-7983 127 22 of of ADP fcis-7983 127 23 the the DET fcis-7983 127 24 model model NOUN fcis-7983 127 25 . . PUNCT fcis-7983 128 1 references reference NOUN fcis-7983 128 2 [ [ X fcis-7983 128 3 1 1 X fcis-7983 128 4 ] ] PUNCT fcis-7983 128 5 chinchor chinchor NOUN fcis-7983 128 6 n.muc-6 n.muc-6 PROPN fcis-7983 128 7 named name VERB fcis-7983 128 8 entity entity NOUN fcis-7983 128 9 task task NOUN fcis-7983 128 10 definition definition NOUN fcis-7983 128 11 ( ( PUNCT fcis-7983 128 12 version version NOUN fcis-7983 128 13 2.1 2.1 NUM fcis-7983 128 14 ) ) PUNCT fcis-7983 129 1 [ [ X fcis-7983 129 2 c c X fcis-7983 129 3 ] ] PUNCT fcis-7983 129 4 . . PUNCT fcis-7983 130 1 proceedings proceeding NOUN fcis-7983 130 2 of of ADP fcis-7983 130 3 the the DET fcis-7983 130 4 6th 6th ADJ fcis-7983 130 5 conference conference NOUN fcis-7983 130 6 on on ADP fcis-7983 130 7 message message NOUN fcis-7983 130 8 understanding understanding NOUN fcis-7983 130 9 , , PUNCT fcis-7983 130 10 columbia columbia PROPN fcis-7983 130 11 , , PUNCT fcis-7983 130 12 maryland maryland PROPN fcis-7983 130 13 , , PUNCT fcis-7983 130 14 1995:142 1995:142 NUM fcis-7983 130 15 - - SYM fcis-7983 130 16 194 194 NUM fcis-7983 130 17 . . PUNCT fcis-7983 131 1 [ [ X fcis-7983 131 2 2 2 X fcis-7983 131 3 ] ] X fcis-7983 131 4 xiang xiang PROPN fcis-7983 131 5 xiaowen xiaowen PROPN fcis-7983 131 6 , , PUNCT fcis-7983 131 7 shi shi PROPN fcis-7983 131 8 xiaodong xiaodong PROPN fcis-7983 131 9 , , PUNCT fcis-7983 131 10 zeng zeng PROPN fcis-7983 131 11 hualin hualin PROPN fcis-7983 131 12 a a DET fcis-7983 131 13 chinese chinese PROPN fcis-7983 131 14 named name VERB fcis-7983 131 15 entity entity NOUN fcis-7983 131 16 recognition recognition NOUN fcis-7983 131 17 system system NOUN fcis-7983 131 18 that that PRON fcis-7983 131 19 combines combine VERB fcis-7983 131 20 statistics statistic NOUN fcis-7983 131 21 and and CCONJ fcis-7983 131 22 rules rule VERB fcis-7983 131 23 [ [ X fcis-7983 131 24 j j X fcis-7983 131 25 ] ] X fcis-7983 131 26 . . PUNCT fcis-7983 132 1 computer computer NOUN fcis-7983 132 2 applications application NOUN fcis-7983 132 3 , , PUNCT fcis-7983 132 4 2005 2005 NUM fcis-7983 132 5 , , PUNCT fcis-7983 132 6 25 25 NUM fcis-7983 132 7 ( ( PUNCT fcis-7983 132 8 10 10 NUM fcis-7983 132 9 ): ): PUNCT fcis-7983 132 10 3 3 NUM fcis-7983 132 11 . . PUNCT fcis-7983 133 1 [ [ X fcis-7983 133 2 3 3 X fcis-7983 133 3 ] ] X fcis-7983 133 4 zhang zhang PROPN fcis-7983 133 5 chuanyan chuanyan PROPN fcis-7983 133 6 , , PUNCT fcis-7983 133 7 hong hong PROPN fcis-7983 133 8 xiaoguang xiaoguang PROPN fcis-7983 133 9 , , PUNCT fcis-7983 133 10 peng peng PROPN fcis-7983 133 11 chaohui chaohui PROPN fcis-7983 133 12 , , PUNCT fcis-7983 133 13 et et PROPN fcis-7983 133 14 al al PROPN fcis-7983 133 15 mesh mesh NOUN fcis-7983 133 16 entity entity NOUN fcis-7983 133 17 activity activity NOUN fcis-7983 133 18 extraction extraction NOUN fcis-7983 133 19 based base VERB fcis-7983 133 20 on on ADP fcis-7983 133 21 support support NOUN fcis-7983 133 22 vector vector NOUN fcis-7983 133 23 machine machine NOUN fcis-7983 133 24 and and CCONJ fcis-7983 133 25 extended extend VERB fcis-7983 133 26 conditional conditional ADJ fcis-7983 133 27 random random ADJ fcis-7983 133 28 field field NOUN fcis-7983 134 1 [ [ X fcis-7983 134 2 j j X fcis-7983 134 3 ] ] X fcis-7983 134 4 . . PUNCT fcis-7983 135 1 journal journal PROPN fcis-7983 135 2 of of ADP fcis-7983 135 3 software software NOUN fcis-7983 135 4 science science NOUN fcis-7983 135 5 , , PUNCT fcis-7983 135 6 2012 2012 NUM fcis-7983 135 7 , , PUNCT fcis-7983 135 8 23 23 NUM fcis-7983 135 9 ( ( PUNCT fcis-7983 135 10 10 10 NUM fcis-7983 135 11 ): ): PUNCT fcis-7983 135 12 16 16 NUM fcis-7983 135 13 . . PUNCT fcis-7983 136 1 [ [ X fcis-7983 136 2 4 4 X fcis-7983 136 3 ] ] X fcis-7983 136 4 huang huang PROPN fcis-7983 136 5 z z PROPN fcis-7983 136 6 , , PUNCT fcis-7983 136 7 wei wei PROPN fcis-7983 136 8 x x PROPN fcis-7983 136 9 , , PUNCT fcis-7983 136 10 kai kai PROPN fcis-7983 136 11 y. y. PROPN fcis-7983 136 12 a a DET fcis-7983 136 13 bidirectional bidirectional ADJ fcis-7983 136 14 lstm lstm PROPN fcis-7983 136 15 - - PUNCT fcis-7983 136 16 crf crf NOUN fcis-7983 136 17 model model NOUN fcis-7983 136 18 for for ADP fcis-7983 136 19 sequence sequence NOUN fcis-7983 136 20 labeling labeling NOUN fcis-7983 137 1 [ [ X fcis-7983 137 2 j j X fcis-7983 137 3 ] ] X fcis-7983 137 4 . . PUNCT fcis-7983 138 1 computer computer NOUN fcis-7983 138 2 science science NOUN fcis-7983 138 3 , , PUNCT fcis-7983 138 4 2015 2015 NUM fcis-7983 138 5 . . PUNCT fcis-7983 139 1 [ [ X fcis-7983 139 2 5 5 X fcis-7983 139 3 ] ] PUNCT fcis-7983 139 4 strubelle strubelle NOUN fcis-7983 139 5 e e NOUN fcis-7983 139 6 , , PUNCT fcis-7983 139 7 verga verga NOUN fcis-7983 139 8 p p NOUN fcis-7983 139 9 , , PUNCT fcis-7983 139 10 belanger belanger PROPN fcis-7983 139 11 d d PROPN fcis-7983 139 12 , , PUNCT fcis-7983 139 13 etc etc X fcis-7983 139 14 . . X fcis-7983 139 15 fast fast ADJ fcis-7983 139 16 and and CCONJ fcis-7983 139 17 accurate accurate ADJ fcis-7983 139 18 entity entity NOUN fcis-7983 139 19 recognition recognition NOUN fcis-7983 139 20 using use VERB fcis-7983 139 21 iterative iterative ADJ fcis-7983 139 22 expansion expansion NOUN fcis-7983 139 23 convolution convolution NOUN fcis-7983 139 24 [ [ X fcis-7983 139 25 j j X fcis-7983 139 26 ] ] X fcis-7983 139 27 . . PUNCT fcis-7983 140 1 two two NUM fcis-7983 140 2 thousand thousand NUM fcis-7983 140 3 and and CCONJ fcis-7983 140 4 seventeen seventeen NUM fcis-7983 140 5 . . PUNCT fcis-7983 141 1 [ [ X fcis-7983 141 2 6 6 NUM fcis-7983 141 3 ] ] PUNCT fcis-7983 141 4 yan yan PROPN fcis-7983 141 5 h h PROPN fcis-7983 141 6 , , PUNCT fcis-7983 141 7 deng deng PROPN fcis-7983 141 8 b b PROPN fcis-7983 141 9 , , PUNCT fcis-7983 141 10 li li PROPN fcis-7983 141 11 x x PROPN fcis-7983 141 12 , , PUNCT fcis-7983 141 13 et et PROPN fcis-7983 141 14 al al PROPN fcis-7983 141 15 . . PROPN fcis-7983 141 16 tener tener PROPN fcis-7983 141 17 : : PUNCT fcis-7983 141 18 adaptive adaptive ADJ fcis-7983 141 19 transformer transformer NOUN fcis-7983 141 20 encoder encoder NOUN fcis-7983 141 21 for for ADP fcis-7983 141 22 named name VERB fcis-7983 141 23 entity entity NOUN fcis-7983 141 24 recognition recognition NOUN fcis-7983 142 1 [ [ X fcis-7983 142 2 j j X fcis-7983 142 3 ] ] X fcis-7983 142 4 . . PUNCT fcis-7983 143 1 two two NUM fcis-7983 143 2 thousand thousand NUM fcis-7983 143 3 and and CCONJ fcis-7983 143 4 nineteen nineteen NUM fcis-7983 143 5 . . PUNCT fcis-7983 144 1 [ [ X fcis-7983 144 2 7 7 X fcis-7983 144 3 ] ] X fcis-7983 144 4 yang yang PROPN fcis-7983 144 5 j j PROPN fcis-7983 144 6 , , PUNCT fcis-7983 144 7 teng teng PROPN fcis-7983 144 8 z z PROPN fcis-7983 144 9 , , PUNCT fcis-7983 144 10 zhang zhang PROPN fcis-7983 144 11 m m PROPN fcis-7983 144 12 , , PUNCT fcis-7983 144 13 et et PROPN fcis-7983 144 14 al al PROPN fcis-7983 144 15 . . PUNCT fcis-7983 145 1 combining combine VERB fcis-7983 145 2 discrete discrete ADJ fcis-7983 145 3 and and CCONJ fcis-7983 145 4 neural neural ADJ fcis-7983 145 5 features feature NOUN fcis-7983 145 6 for for ADP fcis-7983 145 7 sequence sequence NOUN fcis-7983 145 8 labeling labeling NOUN fcis-7983 145 9 [ [ X fcis-7983 145 10 j j X fcis-7983 145 11 ] ] X fcis-7983 145 12 . . PUNCT fcis-7983 146 1 two two NUM fcis-7983 146 2 thousand thousand NUM fcis-7983 146 3 and and CCONJ fcis-7983 146 4 seventeen seventeen NUM fcis-7983 146 5 . . PUNCT fcis-7983 147 1 [ [ X fcis-7983 147 2 8 8 NUM fcis-7983 147 3 ] ] X fcis-7983 147 4 wu wu PROPN fcis-7983 147 5 s s PROPN fcis-7983 147 6 , , PUNCT fcis-7983 147 7 song song NOUN fcis-7983 147 8 x x NOUN fcis-7983 147 9 , , PUNCT fcis-7983 147 10 feng feng PROPN fcis-7983 147 11 z. z. PROPN fcis-7983 147 12 mect mect PROPN fcis-7983 147 13 : : PUNCT fcis-7983 147 14 a a DET fcis-7983 147 15 cross cross NOUN fcis-7983 147 16 transform transform NOUN fcis-7983 147 17 based base VERB fcis-7983 147 18 on on ADP fcis-7983 147 19 multivariate multivariate NOUN fcis-7983 147 20 data datum NOUN fcis-7983 147 21 embedding embed VERB fcis-7983 147 22 for for ADP fcis-7983 147 23 chinese chinese ADJ fcis-7983 147 24 named name VERB fcis-7983 147 25 entity entity NOUN fcis-7983 147 26 recognition recognition NOUN fcis-7983 148 1 [ [ X fcis-7983 148 2 j j X fcis-7983 148 3 ] ] X fcis-7983 148 4 . . PUNCT fcis-7983 149 1 two two NUM fcis-7983 149 2 thousand thousand NUM fcis-7983 149 3 and and CCONJ fcis-7983 149 4 twenty twenty NUM fcis-7983 149 5 - - PUNCT fcis-7983 149 6 one one NUM fcis-7983 149 7 . . PUNCT fcis-7983 150 1 [ [ X fcis-7983 150 2 9 9 NUM fcis-7983 150 3 ] ] X fcis-7983 150 4 liu liu PROPN fcis-7983 150 5 z z PROPN fcis-7983 150 6 , , PUNCT fcis-7983 150 7 zhu zhu PROPN fcis-7983 150 8 c c X fcis-7983 150 9 , , PUNCT fcis-7983 150 10 zhao zhao PROPN fcis-7983 150 11 t. t. PROPN fcis-7983 150 12 chinese chinese PROPN fcis-7983 150 13 named name VERB fcis-7983 150 14 entity entity NOUN fcis-7983 150 15 recognition recognition NOUN fcis-7983 150 16 based base VERB fcis-7983 150 17 on on ADP fcis-7983 150 18 sequence sequence NOUN fcis-7983 150 19 marking mark VERB fcis-7983 150 20 method method NOUN fcis-7983 150 21 : : PUNCT fcis-7983 150 22 based base VERB fcis-7983 150 23 on on ADP fcis-7983 150 24 characters character NOUN fcis-7983 150 25 or or CCONJ fcis-7983 150 26 words word NOUN fcis-7983 150 27 ? ? PUNCT fcis-7983 151 1 [ [ X fcis-7983 151 2 j j X fcis-7983 151 3 ] ] X fcis-7983 151 4 . . PUNCT fcis-7983 152 1 springer springer NOUN fcis-7983 152 2 press press PROPN fcis-7983 152 3 , , PUNCT fcis-7983 152 4 2010 2010 NUM fcis-7983 152 5 . . PUNCT fcis-7983 153 1 [ [ X fcis-7983 153 2 10 10 NUM fcis-7983 153 3 ] ] X fcis-7983 153 4 yang yang PROPN fcis-7983 153 5 x x PROPN fcis-7983 153 6 , , PUNCT fcis-7983 153 7 mao mao PROPN fcis-7983 153 8 k. k. PROPN fcis-7983 153 9 using use VERB fcis-7983 153 10 comprehensive comprehensive ADJ fcis-7983 153 11 knowledge knowledge NOUN fcis-7983 153 12 to to PART fcis-7983 153 13 learn learn VERB fcis-7983 153 14 multi multi ADJ fcis-7983 153 15 prototype prototype NOUN fcis-7983 153 16 word word NOUN fcis-7983 153 17 embedding embed VERB fcis-7983 153 18 from from ADP fcis-7983 153 19 single single ADJ fcis-7983 153 20 prototype prototype NOUN fcis-7983 153 21 word word NOUN fcis-7983 153 22 embedding embed VERB fcis-7983 153 23 [ [ X fcis-7983 153 24 j j X fcis-7983 153 25 ] ] X fcis-7983 153 26 . . PUNCT fcis-7983 154 1 expert expert NOUN fcis-7983 154 2 systems system NOUN fcis-7983 154 3 and and CCONJ fcis-7983 154 4 applications application NOUN fcis-7983 154 5 , , PUNCT fcis-7983 154 6 2016 2016 NUM fcis-7983 154 7 , , PUNCT fcis-7983 154 8 56 56 NUM fcis-7983 154 9 ( ( PUNCT fcis-7983 154 10 september september PROPN fcis-7983 154 11 ): ): PUNCT fcis-7983 154 12 291 291 NUM fcis-7983 154 13 - - SYM fcis-7983 154 14 299 299 NUM fcis-7983 154 15 . . PUNCT fcis-7983 155 1 [ [ X fcis-7983 155 2 11 11 NUM fcis-7983 155 3 ] ] X fcis-7983 155 4 zhang zhang PROPN fcis-7983 155 5 y y PROPN fcis-7983 155 6 , , PUNCT fcis-7983 155 7 yang yang PROPN fcis-7983 155 8 j. j. PROPN fcis-7983 155 9 net net PROPN fcis-7983 155 10 enrollment enrollment NOUN fcis-7983 155 11 rate rate NOUN fcis-7983 155 12 in in ADP fcis-7983 155 13 china china PROPN fcis-7983 155 14 using use VERB fcis-7983 155 15 lattice lattice NOUN fcis-7983 155 16 lstm lstm NOUN fcis-7983 155 17 [ [ X fcis-7983 155 18 j j X fcis-7983 155 19 ] ] X fcis-7983 155 20 . . PROPN fcis-7983 155 21 2018 2018 NUM fcis-7983 155 22 . . PUNCT fcis-7983 156 1 chung chung PROPN fcis-7983 156 2 j j PROPN fcis-7983 156 3 , , PUNCT fcis-7983 156 4 gulcehre gulcehre PROPN fcis-7983 156 5 c c PROPN fcis-7983 156 6 , , PUNCT fcis-7983 156 7 cho cho PROPN fcis-7983 156 8 k k PROPN fcis-7983 156 9 h h PROPN fcis-7983 156 10 , , PUNCT fcis-7983 156 11 et et PROPN fcis-7983 156 12 al al PROPN fcis-7983 156 13 . . PROPN fcis-7983 157 1 empirical empirical ADJ fcis-7983 157 2 evaluation evaluation NOUN fcis-7983 157 3 of of ADP fcis-7983 157 4 gated gate VERB fcis-7983 157 5 recurrent recurrent ADJ fcis-7983 157 6 neural neural ADJ fcis-7983 157 7 networks network NOUN fcis-7983 157 8 for for ADP fcis-7983 157 9 sequence sequence NOUN fcis-7983 157 10 modeling modeling NOUN fcis-7983 158 1 [ [ X fcis-7983 158 2 j]//arxiv j]//arxiv PROPN fcis-7983 158 3 preprint preprint NOUN fcis-7983 158 4 , , PUNCT fcis-7983 158 5 2014 2014 NUM fcis-7983 158 6 : : PUNCT fcis-7983 158 7 arxiv arxiv NOUN fcis-7983 158 8 : : PUNCT fcis-7983 158 9 1412.3555 1412.3555 NUM fcis-7983 158 10 . . PUNCT fcis-7983 159 1 [ [ X fcis-7983 159 2 12 12 NUM fcis-7983 159 3 ] ] X fcis-7983 159 4 li li PROPN fcis-7983 159 5 jie jie PROPN fcis-7983 159 6 , , PUNCT fcis-7983 159 7 meng meng PROPN fcis-7983 159 8 ke ke PROPN fcis-7983 159 9 . . PUNCT fcis-7983 160 1 mfe mfe PROPN fcis-7983 160 2 - - PROPN fcis-7983 160 3 ner ner NOUN fcis-7983 160 4 : : PUNCT fcis-7983 160 5 multi multi ADJ fcis-7983 160 6 feature feature NOUN fcis-7983 160 7 fusion fusion NOUN fcis-7983 160 8 embedding embed VERB fcis-7983 160 9 for for ADP fcis-7983 160 10 chinese chinese ADJ fcis-7983 160 11 named name VERB fcis-7983 160 12 entity entity NOUN fcis-7983 160 13 recognition recognition NOUN fcis-7983 161 1 [ [ X fcis-7983 161 2 j j X fcis-7983 161 3 ] ] X fcis-7983 161 4 . . PUNCT fcis-7983 162 1 two two NUM fcis-7983 162 2 thousand thousand NUM fcis-7983 162 3 and and CCONJ fcis-7983 162 4 twenty twenty NUM fcis-7983 162 5 - - PUNCT fcis-7983 162 6 one one NUM fcis-7983 162 7 . . PUNCT fcis-7983 163 1 [ [ X fcis-7983 163 2 13 13 NUM fcis-7983 163 3 ] ] X fcis-7983 163 4 devlin devlin PROPN fcis-7983 163 5 j j PROPN fcis-7983 163 6 , , PUNCT fcis-7983 163 7 chang chang PROPN fcis-7983 163 8 m m PROPN fcis-7983 163 9 w w PROPN fcis-7983 163 10 , , PUNCT fcis-7983 163 11 lee lee PROPN fcis-7983 163 12 k k PROPN fcis-7983 163 13 , , PUNCT fcis-7983 163 14 et et PROPN fcis-7983 163 15 al al PROPN fcis-7983 163 16 . . PUNCT fcis-7983 163 17 bert bert PROPN fcis-7983 163 18 : : PUNCT fcis-7983 163 19 pre pre VERB fcis-7983 163 20 training training NOUN fcis-7983 163 21 of of ADP fcis-7983 163 22 deep deep ADJ fcis-7983 163 23 bidirectional bidirectional ADJ fcis-7983 163 24 converters converter NOUN fcis-7983 163 25 for for ADP fcis-7983 163 26 language language NOUN fcis-7983 163 27 comprehension comprehension NOUN fcis-7983 163 28 [ [ X fcis-7983 163 29 j j X fcis-7983 163 30 ] ] X fcis-7983 163 31 . . PUNCT fcis-7983 164 1 two two NUM fcis-7983 164 2 thousand thousand NUM fcis-7983 164 3 and and CCONJ fcis-7983 164 4 eighteen eighteen NUM fcis-7983 164 5 . . PUNCT fcis-7983 165 1 [ [ X fcis-7983 165 2 14 14 NUM fcis-7983 165 3 ] ] PUNCT fcis-7983 165 4 yu yu PROPN fcis-7983 165 5 tongrui tongrui PROPN fcis-7983 165 6 , , PUNCT fcis-7983 165 7 jin jin PROPN fcis-7983 165 8 ran run VERB fcis-7983 165 9 , , PUNCT fcis-7983 165 10 han han PROPN fcis-7983 165 11 xiaozhen xiaozhen PROPN fcis-7983 165 12 , , PUNCT fcis-7983 165 13 li li PROPN fcis-7983 165 14 jiahui jiahui PROPN fcis-7983 165 15 , , PUNCT fcis-7983 165 16 yu yu PROPN fcis-7983 165 17 ting ting PROPN fcis-7983 165 18 . . PUNCT fcis-7983 166 1 review review NOUN fcis-7983 166 2 of of ADP fcis-7983 166 3 research research NOUN fcis-7983 166 4 on on ADP fcis-7983 166 5 natural natural ADJ fcis-7983 166 6 language language NOUN fcis-7983 166 7 processing processing NOUN fcis-7983 166 8 pre pre NOUN fcis-7983 166 9 training training NOUN fcis-7983 166 10 models model NOUN fcis-7983 167 1 [ [ X fcis-7983 167 2 j j X fcis-7983 167 3 ] ] X fcis-7983 167 4 . . PUNCT fcis-7983 168 1 computer computer NOUN fcis-7983 168 2 engineering engineering NOUN fcis-7983 168 3 and and CCONJ fcis-7983 168 4 applications application NOUN fcis-7983 168 5 , , PUNCT fcis-7983 168 6 2020,56 2020,56 PROPN fcis-7983 168 7 ( ( PUNCT fcis-7983 168 8 23 23 NUM fcis-7983 168 9 ): ): PUNCT fcis-7983 168 10 12 12 NUM fcis-7983 168 11 - - SYM fcis-7983 168 12 22 22 NUM fcis-7983 168 13 . . PUNCT fcis-7983 169 1 [ [ X fcis-7983 169 2 15 15 NUM fcis-7983 169 3 ] ] X fcis-7983 169 4 hochreiter hochreiter PROPN fcis-7983 169 5 s s PROPN fcis-7983 169 6 , , PUNCT fcis-7983 169 7 schmidhuber schmidhuber PROPN fcis-7983 169 8 j j PROPN fcis-7983 169 9 , , PUNCT fcis-7983 169 10 long long ADJ fcis-7983 169 11 term term NOUN fcis-7983 169 12 and and CCONJ fcis-7983 169 13 short short ADJ fcis-7983 169 14 - - PUNCT fcis-7983 169 15 term term NOUN fcis-7983 169 16 memory memory NOUN fcis-7983 169 17 [ [ X fcis-7983 169 18 j j X fcis-7983 169 19 ] ] X fcis-7983 169 20 . . PUNCT fcis-7983 170 1 neurocomputing neurocomputing NOUN fcis-7983 170 2 , , PUNCT fcis-7983 170 3 1997 1997 NUM fcis-7983 170 4 , , PUNCT fcis-7983 170 5 9 9 NUM fcis-7983 170 6 ( ( PUNCT fcis-7983 170 7 8) 8) NUM fcis-7983 170 8 : : PUNCT fcis-7983 170 9 1735 1735 NUM fcis-7983 170 10 - - SYM fcis-7983 170 11 1780 1780 NUM fcis-7983 170 12 . . PUNCT fcis-7983 171 1 [ [ X fcis-7983 171 2 16 16 NUM fcis-7983 171 3 ] ] X fcis-7983 171 4 wu wu PROPN fcis-7983 171 5 zongyou zongyou PROPN fcis-7983 171 6 , , PUNCT fcis-7983 171 7 bai bai PROPN fcis-7983 171 8 kunlong kunlong PROPN fcis-7983 171 9 , , PUNCT fcis-7983 171 10 yang yang PROPN fcis-7983 171 11 linrui linrui PROPN fcis-7983 171 12 , , PUNCT fcis-7983 171 13 et et PROPN fcis-7983 171 14 al al PROPN fcis-7983 171 15 a a DET fcis-7983 171 16 review review NOUN fcis-7983 171 17 of of ADP fcis-7983 171 18 research research NOUN fcis-7983 171 19 on on ADP fcis-7983 171 20 text text NOUN fcis-7983 171 21 mining mining NOUN fcis-7983 171 22 in in ADP fcis-7983 171 23 electronic electronic ADJ fcis-7983 171 24 medical medical ADJ fcis-7983 171 25 records record NOUN fcis-7983 171 26 [ [ X fcis-7983 171 27 j j X fcis-7983 171 28 ] ] X fcis-7983 171 29 . . PUNCT fcis-7983 172 1 computer computer NOUN fcis-7983 172 2 research research NOUN fcis-7983 172 3 and and CCONJ fcis-7983 172 4 development development NOUN fcis-7983 172 5 , , PUNCT fcis-7983 172 6 2021 2021 NUM fcis-7983 172 7 , , PUNCT fcis-7983 172 8 58 58 NUM fcis-7983 172 9 ( ( PUNCT fcis-7983 172 10 3 3 NUM fcis-7983 172 11 ): ): PUNCT fcis-7983 172 12 15 15 NUM fcis-7983 172 13 . . PUNCT fcis-7983 173 1 [ [ X fcis-7983 173 2 17 17 NUM fcis-7983 173 3 ] ] X fcis-7983 173 4 zhao zhao PROPN fcis-7983 173 5 r r PROPN fcis-7983 173 6 , , PUNCT fcis-7983 173 7 wang wang PROPN fcis-7983 173 8 d d PROPN fcis-7983 173 9 , , PUNCT fcis-7983 173 10 yan yan PROPN fcis-7983 173 11 r r NOUN fcis-7983 173 12 , , PUNCT fcis-7983 173 13 et et PROPN fcis-7983 173 14 al al PROPN fcis-7983 173 15 . . PROPN fcis-7983 173 16 machine machine NOUN fcis-7983 173 17 health health NOUN fcis-7983 173 18 monitoring monitoring NOUN fcis-7983 173 19 based base VERB fcis-7983 173 20 on on ADP fcis-7983 173 21 local local ADJ fcis-7983 173 22 feature feature NOUN fcis-7983 173 23 gated gate VERB fcis-7983 173 24 recursive recursive ADJ fcis-7983 173 25 unit unit NOUN fcis-7983 173 26 networks network NOUN fcis-7983 173 27 [ [ X fcis-7983 173 28 j j X fcis-7983 173 29 ] ] X fcis-7983 173 30 . . PUNCT fcis-7983 174 1 ieee ieee PROPN fcis-7983 174 2 industrial industrial ADJ fcis-7983 174 3 electronic electronic ADJ fcis-7983 174 4 trading trading NOUN fcis-7983 174 5 , , PUNCT fcis-7983 174 6 2018 2018 NUM fcis-7983 174 7 . . PUNCT fcis-7983 175 1 [ [ X fcis-7983 175 2 18 18 NUM fcis-7983 175 3 ] ] X fcis-7983 175 4 wu wu PROPN fcis-7983 175 5 s s PROPN fcis-7983 175 6 , , PUNCT fcis-7983 175 7 song song NOUN fcis-7983 175 8 x x NOUN fcis-7983 175 9 , , PUNCT fcis-7983 175 10 feng feng PROPN fcis-7983 175 11 z. z. PROPN fcis-7983 175 12 mect mect PROPN fcis-7983 175 13 : : PUNCT fcis-7983 175 14 a a DET fcis-7983 175 15 cross cross NOUN fcis-7983 175 16 transform transform NOUN fcis-7983 175 17 based base VERB fcis-7983 175 18 on on ADP fcis-7983 175 19 multivariate multivariate NOUN fcis-7983 175 20 data datum NOUN fcis-7983 175 21 embedding embed VERB fcis-7983 175 22 for for ADP fcis-7983 175 23 chinese chinese ADJ fcis-7983 175 24 named name VERB fcis-7983 175 25 entity entity NOUN fcis-7983 175 26 recognition recognition NOUN fcis-7983 176 1 [ [ X fcis-7983 176 2 j j X fcis-7983 176 3 ] ] X fcis-7983 176 4 . . PUNCT fcis-7983 177 1 two two NUM fcis-7983 177 2 thousand thousand NUM fcis-7983 177 3 and and CCONJ fcis-7983 177 4 twenty twenty NUM fcis-7983 177 5 - - PUNCT fcis-7983 177 6 one one NUM fcis-7983 177 7 . . PUNCT fcis-7983 178 1 [ [ X fcis-7983 178 2 19 19 NUM fcis-7983 178 3 ] ] PUNCT fcis-7983 178 4 xu xu PROPN fcis-7983 179 1 c c X fcis-7983 179 2 , , PUNCT fcis-7983 179 3 wang wang PROPN fcis-7983 179 4 f f PROPN fcis-7983 179 5 , , PUNCT fcis-7983 179 6 han han PROPN fcis-7983 179 7 j j PROPN fcis-7983 179 8 , , PUNCT fcis-7983 179 9 et et PROPN fcis-7983 179 10 al al PROPN fcis-7983 179 11 . . PUNCT fcis-7983 180 1 using use VERB fcis-7983 180 2 multiple multiple ADJ fcis-7983 180 3 embedding embed VERB fcis-7983 180 4 technology technology NOUN fcis-7983 180 5 for for ADP fcis-7983 180 6 chinese chinese ADJ fcis-7983 180 7 named name VERB fcis-7983 180 8 entity entity NOUN fcis-7983 180 9 recognition recognition NOUN fcis-7983 180 10 : : PUNCT fcis-7983 180 11 10.1145/33577384.3358117 10.1145/33577384.3358117 NUM fcis-7983 181 1 [ [ X fcis-7983 181 2 p p X fcis-7983 181 3 ] ] X fcis-7983 181 4 . . PUNCT fcis-7983 182 1 2019 2019 NUM fcis-7983 182 2 . . PUNCT fcis-7983 183 1 10 10 NUM fcis-7983 184 1 [ [ SYM fcis-7983 184 2 20 20 NUM fcis-7983 184 3 ] ] X fcis-7983 184 4 zhang zhang PROPN fcis-7983 184 5 yunqiu yunqiu PROPN fcis-7983 184 6 , , PUNCT fcis-7983 184 7 wang wang PROPN fcis-7983 184 8 yang yang PROPN fcis-7983 184 9 , , PUNCT fcis-7983 184 10 li li PROPN fcis-7983 184 11 bocheng bocheng PROPN fcis-7983 184 12 . . PUNCT fcis-7983 185 1 chinese chinese ADJ fcis-7983 185 2 electronic electronic ADJ fcis-7983 185 3 medical medical ADJ fcis-7983 185 4 record record NOUN fcis-7983 185 5 named name VERB fcis-7983 185 6 entity entity NOUN fcis-7983 185 7 recognition recognition NOUN fcis-7983 185 8 based base VERB fcis-7983 185 9 on on ADP fcis-7983 185 10 roberta roberta PROPN fcis-7983 185 11 wwm wwm PROPN fcis-7983 185 12 dynamic dynamic ADJ fcis-7983 185 13 fusion fusion NOUN fcis-7983 185 14 model model NOUN fcis-7983 186 1 [ [ X fcis-7983 186 2 j j X fcis-7983 186 3 ] ] X fcis-7983 186 4 . . PUNCT fcis-7983 187 1 data datum NOUN fcis-7983 187 2 analysis analysis NOUN fcis-7983 187 3 and and CCONJ fcis-7983 187 4 knowledge knowledge NOUN fcis-7983 187 5 discovery discovery NOUN fcis-7983 187 6 , , PUNCT fcis-7983 187 7 222,6 222,6 NUM fcis-7983 187 8 ( ( PUNCT fcis-7983 187 9 z1 z1 NOUN fcis-7983 187 10 ): ): PUNCT fcis-7983 187 11 242 242 NUM fcis-7983 187 12 - - SYM fcis-7983 187 13 250 250 NUM fcis-7983 187 14 . . PUNCT fcis-7983 188 1 [ [ X fcis-7983 188 2 21 21 NUM fcis-7983 188 3 ] ] X fcis-7983 188 4 zhang zhang PROPN fcis-7983 188 5 y y PROPN fcis-7983 188 6 , , PUNCT fcis-7983 188 7 yang yang PROPN fcis-7983 188 8 j. j. PROPN fcis-7983 188 9 net net PROPN fcis-7983 188 10 enrollment enrollment NOUN fcis-7983 188 11 rate rate NOUN fcis-7983 188 12 in in ADP fcis-7983 188 13 china china PROPN fcis-7983 188 14 using use VERB fcis-7983 188 15 lattice lattice NOUN fcis-7983 188 16 lstm lstm NOUN fcis-7983 188 17 [ [ X fcis-7983 188 18 j j X fcis-7983 188 19 ] ] X fcis-7983 188 20 . . PUNCT fcis-7983 189 1 two two NUM fcis-7983 189 2 thousand thousand NUM fcis-7983 189 3 and and CCONJ fcis-7983 189 4 eighteen eighteen NUM fcis-7983 189 5 . . PUNCT fcis-7983 190 1 [ [ X fcis-7983 190 2 22 22 NUM fcis-7983 190 3 ] ] X fcis-7983 190 4 peng peng PROPN fcis-7983 190 5 m m PROPN fcis-7983 190 6 , , PUNCT fcis-7983 190 7 ma ma PROPN fcis-7983 190 8 r r PROPN fcis-7983 190 9 , , PUNCT fcis-7983 190 10 zhang zhang PROPN fcis-7983 190 11 q q PROPN fcis-7983 190 12 , , PUNCT fcis-7983 190 13 et et PROPN fcis-7983 190 14 al al PROPN fcis-7983 190 15 . . PROPN fcis-7983 190 16 simplify simplify VERB fcis-7983 190 17 the the DET fcis-7983 190 18 use use NOUN fcis-7983 190 19 of of ADP fcis-7983 190 20 vocabulary vocabulary NOUN fcis-7983 190 21 in in ADP fcis-7983 190 22 chinese chinese ADJ fcis-7983 190 23 ner ner NOUN fcis-7983 191 1 [ [ X fcis-7983 191 2 j j X fcis-7983 191 3 ] ] X fcis-7983 191 4 . . PUNCT fcis-7983 192 1 two two NUM fcis-7983 192 2 thousand thousand NUM fcis-7983 192 3 and and CCONJ fcis-7983 192 4 nineteen nineteen NUM fcis-7983 192 5 . . PUNCT