Copyright©2019 P-ISSN: 1978-8118 E-ISSN: 2460-710X 289 Lingua Cultura, 13(4), November 2019, 289-296 DOI: 10.21512/lc.v13i4.6067 THE TRANSLATION OF LEXICAL COLLOCATIONS IN UNDERGRADUATE STUDENTS’ THESES’ ABSTRACT: STUDENTS VERSUS GOOGLE TRANSLATE Clara Herlina Karjo1; Ecclesia Metta2 1,2English Department, Faculty of Humanities, Bina Nusantara University Jl. Kemanggisan Illir III No. 45, Palmerah, Jakarta 11480, Indonesia 1claraherlina@yahoo.com; 2mettaokey@gmail.com Received: 28th October 2019/Revised: 11th November 2019/Accepted: 18th November 2019 How to Cite: Karjo, C. H., & Metta, E. (2019). The translation of lexical collocations in undergraduate students’ theses’ abstract: Students versus Google Translate. Lingua Cultura, 13(4), 289-296. https://doi.org/10.21512/lc.v13i4.6067 ABSTRACT This research intended to compare the translations of lexical collocations found in the abstract section of students’ theses. The purposes were to find out the errors in translating lexical collocation either by Google Translate or student translator. The data were taken from twenty working papers of English Literature students at Binus University. The abstracts of these theses (in English and Indonesian) were then processed with Google Translate. Thus, there were four sets of data to analyze: (1) Students’ Text in Indonesian (STI), (2) Google Translate of STI in English (GTE), (3) Students’ Text in English (STE), and (4) Google Translate of STE in Indonesian (GTI). From the data, samples of collocations were taken and categorized based on Hill’s classification of lexical collocations. The lexical collocations found in the four sets of data were scrutinized, compared, and analyzed to find the errors in forms and meaning as well as in the translation. The results reveal that errors in translating collocations are mostly made by Google Translate rather than the students. This research implies that Google Translate still needs improvement in translating collocations, but it is also possible that translation errors occur because of students’ misuse of collocation. Keywords: Google Translate, lexical collocations, translation error, students’ translation INTRODUCTION Recently, more and more people resort to machine translation for translating text from one language to many other languages since the translation can be done within minutes. Besides the speed, machine translation is chosen because this system also pays attention to the grammatical rules in each respective language to ensure that the results will not produce an ambiguous translation or mistranslation from the original. Machine translation keeps improving its system to preserve the originality of the source language. One of the underlying principles of machine translation is the ability to identify statistically significant patterns. So if one translation has been discovered, this method can be applied to similar texts in the future (Medvedev, 2016). There are many kinds of machine translation tools that can be found at present, such as Babble, Bing, Microsoft Translate, Systran, and Google Translate, each of which has its advantages and disadvantages. However, among all the available machine translation tools, it seems that Google Translate (hereinafter is also abbreviated as GT) is the one that is frequently used by people, especially in Indonesia. Google Translate is a free translation machine that supports 103 languages in the world, and the most requested translations are from English to Arabic, Russian, Portuguese, and Indonesian (Aiken, 2019). Google Translate was launched in April 2006 as a statistical translation service. In November 2016, Google announced that Google Translate would become Google Neural Machine Translation (GNMT), which means whole sentences at a time, rather than just piece by piece. In other words, Google will produce more relevant translation (Turovsky, 2016). As a free translation machine, Google Translate has become a necessity for people who need to read and write in English in their work or education. The users include office workers who have to write business letters, university lecturers who have to write research papers or college students who have to write academic papers or thesis in English. If they are not confident with their English, then Google Translate can be a very valuable resource. Even though Google Translate does not always give good translation results (Ambawani, 2014), some people still use 290 LINGUA CULTURA, Vol. 13 No. 4, November 2019, 289-296 it as long as the sentences still make sense and deliver the same meaning as the original ones. From the researchers’ observation, college students, especially those majoring in English language or literature, are the prolific users of Google Translate. They use it to find difficult words until writing the whole essay. The researchers suspect that they write their essay in the source language (Indonesian) first, and then translate it using Google Translate into English, or vice versa. This suspicion is more or less attested when the students have to write an essay in two languages, English and Indonesian. For example, when they write abstracts for their theses. In Binus University, the English Department students have to write their theses in English, but the abstracts should be written in both Indonesian and English. The researchers’ indication is confirmed by Napitupulu’s (2017) research, which has found that many students often use Google Translate to make abstracts because it can translate within a second without taking too much time and effort. The researchers assume that the translation of abstracts is made using Google Translate because many translation errors occurred in their abstracts, particularly in translating lexical collocations. Collocation is the way two words are connected in order to create the natural meaning of the words. Collocation has been found to cause problems for EFL students all over the world, especially when they have to translate English collocations into their native language collocations, or from their native language collocations into English. In research for Arabic students, Jabak, Abdullah, and Mustapha (2016) have found that 71,86 % of the students could not translate Arabic collocations into equivalent collocations in English. Similarly, Rumanian students are also not able to translate English collocations into Rumanian and vice versa (Pârlog & Punga, 2017). Iranian EFL students are also found to have difficulty in translating English collocations (Ghasemi & Hashemian, 2016). Translating collocations is difficult because it involves assessing the suitability of word collocates. Thus, inappropriate translation of collocations will produce unnatural words and meanings, for example, the phrase ‘dark blue’ should not be translated into biru gelap, but biru tua. Indonesian students also face similar problems in translating collocations, in particular when the collocations occur in academic papers. They often resort to Google Translate, which might not be producing good results either. Therefore, the present research is done to identify the areas of difficulty in translating lexical collocations and being able to find the solution to this problem. Thus, this research aims to analyze the possibility of translation errors made either by the students (ST) or Google Translate in translating lexical collocations. To that end, the researchers propose two research questions: (1) who (ST or Google Translate) can provide a better translation of lexical collocations? Moreover, (2) what kinds of translation errors are made by ST or Google Translate regarding lexical collocations? House (2014) has explained that translation is a linguistic-textual system that changes certain sentences or words from one language into another language. Besides, translation also means delivering in the target language with the same meaning and context as the original language, which the author wants it to be (Newmark, 1988). Similarly, Baker (2018) has also stated that translation is a replacement of the text into something equivalent to the original language. The main point in these definitions is that the translation result should maintain the original message. Thus, if the translation result has a different message from the original one, then it fails to be called translation. Translation can be done manually by a human translator, using printed or offline dictionaries, or it can be done automatically using machine translation. According to Dhakar, Sinha, and Pandey (2013), machine translation is a technology-based system that is able to translate from one language into another. In doing the translation, Dhakar, Sinha, and Pandey (2013) have explained that machine translation follows several steps. The first is ‘analyze’, which means the machine translation will analyze the grammar and vocabulary of the source text. Second is ‘conversion’, which is the process of converting the source text language to the chosen language. The third is ‘synthesis’, which means creating the chosen text language based on its syntactical and morphological rules. Because of these automated systems, machine translation can provide more accuracy compared to human translation, as it also makes sure to pay attention to punctuation and spelling accurately. As one of the most widely used machine translation tools, Google Translate has been the subject of research for many years since its appearance in 2006. Aiken and Balan (2011) first have conducted their research to assess the translation quality of Google Translate for 50 different languages, not just a pair of languages. They have found out that Google Translate translates a European language into another European language much better than those pairs of language involving the Asian language. However, Karami (2014) has claimed that Google Translate could translate either by words, phrases, certain parts of the text, or even a Webpage better for the pair of languages. Meanwhile, Karjo (2016) and Karjo (2015) have found that Google Translate can produce better translation for phrases rather than full sentences since it requires less editing time. Meanwhile, after 10 years of research, Aiken (2019) has found that Google Translate accuracy has increased about 34% for 51 languages since his research in 2011. Despite being the preferable machine translator, many researchers have still found many translation problems with Google Translate, and thus it needs to be improved. Ismail and Hartono (2016) have found that despite being able to translate from many languages, Google Translate still has a problem when translating a complex structure of the sentence, which leads to mistranslation errors. In her research of English into Spanish translation, Allue (2017) has found that Google Translate still makes errors in lexicon grammar, syntax, pragmatic, and punctuation. Similarly, Ambawani (2014) has also discovered that Google Tranlate seems unable to differentiate between the grammatical structure of Indonesian and English, which leads to an inaccurate grammatical structure. Moreover, Medvedev (2016) has also proved in his research that Google Translate tends to ignore the grammatical structure and the text’s context, which leads to ambiguity. As for translation, Google Translate can be allowed for students for quick and easy translation tool. However, lecturers or instructors should exercise caution and guidance on its appropriate use to avoid students’ addiction to this tool (Raza & Nor, 2018). Researches regarding the quality of Google Translate are usually done by doing the Error Analysis (hereinafter EA) on the results of translation (Koponen & Salmi, 2015). EA is first conducted for teaching and learning a foreign language. EA in language learning is a technique for identifying, classifying, and systematically interpreting the unacceptable forms that are produced by learners in learning a foreign language by using any of the principles and procedures provided by linguistics. EA is a set of procedures 291The Translation of Lexical.... (Clara Herlina Karjo; Ecclesia Metta) for identifying, describing, and explaining learners’ errors, which are not only about identifying and detecting errors but also trying to explain why they are made. According to Longman Dictionary of Language Teaching & Applied Linguistics (2010), EA exists for several reasons. The first is to teach English learners easier based on the type of errors they make. The second is to determine the causes of error. Furthermore, the last is to help the beginner to learn English by classifying the common errors in the text. Thus, the method for conducting EA typically follows Gass and Selinker’s method (as cited in Al-Khresheh, 2016), which consists of collecting data, identifying errors, classifying errors, quantifying errors, analyzing sources of error, and remediating for errors. Similar methods of EA are also done in translation studies. Setiawan (2014), in his research on English translation errors in abstracts of educational administration students of the post-graduate school of State University of Medan, has classified errors made in the translations into two types, which are grammatical and context errors. Moreover, Utami (2017), in her study of UIN Sultan Syarif Kasim’s students’ translation, has found that there are three types of grammatical errors made by the students, i.e., global errors, local errors, and other errors. The source of errors, according to Utami (2017), is because most of the students have difficulties in translating the sentences from Indonesian into English due to their insufficient structural knowledge in the target language. As a result, the students tend to translate the sentences literally, which leads them to make some errors. One of the most frequent errors made in the translation is the translation of collocations. According to the Online Oxford Collocation Dictionary, collocation is the way that two words are meant to be with each other in order to make the words sound more natural and appropriate to be delivered. Baker (2018) has stated that words rarely occur on their own; they almost always occur in the company of other words. There are many words in English that cannot be separated from each other as they produce a natural and correct meaning of the context. Basically, collocation does not only apply for a certain language or culture, but the fixated words apply to all languages in the world. However, words are not strung together in random order. There are certain rules of lexical patterning or the ‘likelihood’ of certain words occurring with other words and the typicality of the likely combination of its component. Based on the possible combination of the words, Hill (2000) has divided collocations into four groups. First is unique collocation, in which a particular word can only be collocated with one word. It cannot have paired with another word, for example ‘to shrug your shoulder’. The verb ‘shrug’ cannot be used other than with ‘shoulder’. Second is strong collocation, which is the word combination that commonly used with each other, for example, ‘dogs bark’. The word ‘bark’ usually comes with ‘dogs’. The third is medium-strength collocation, which is a combination that is neither strong nor weak, for example, to ‘hold a meeting’. The word ‘hold’ is commonly used with ‘meeting’, but it also can be used with a limited number of other words such as ‘hand’ or ‘performance’. Fourth is weak collocation, in which the word can be used with many other words, for example, ‘beautiful girl’. The word ‘beautiful’ can also collocate with many other words such as ‘hair’, ‘scenery’, or ‘cake’. Meanwhile, based on the word class of its components, Hill (2000) has classified lexical collocations into eight categories. The first is the adjective + noun. It is a combination of adjective and noun, for example, ‘tall building’. The word ‘building’ is more appropriate to collocate with ‘tall’ rather than ‘high’. The word ‘tall’ can be considered as a medium-strength collocation because it can also combine with other words such as ‘man’ or ‘people’. Second is noun + noun. It is a combination of noun and noun, for example, ‘driving license’. This collocation is considered as a medium-strength collocation as ‘driving’ can collocate with other words such as ‘range’ and ‘school’, while ‘license’ can collocate with ‘marriage’ or ‘pilot’. Third is Verb + (Adjective) + Noun. It is a combination between a verb, adjective, and noun, for example, ‘making a good paper’. However, using adjective is optional, which means it can also be considered as a verb + noun, for example, ‘compose music’. The example can be considered as weak collocation as ‘compose’ can collocate with ‘poem’ or ‘paper’. Fourth is Verb + Adverb. It is a combination of verb and adverb, for example, ‘sing happily’. The example can be considered as weak collocation as ‘sing’ can collocate with ‘loudly’ or ‘cheerfully’. On the other hand, the word ‘happily’ can also collocate with ‘smile’ or ‘laugh’. Fifth is adverb + verb. It is a combination of adverb and verb, for example, ‘half understand’. The word ‘half’ can collocate with ‘awake’ or ‘asleep’; thus, it can be considered as weak collocation. Sixth is Adverb + Adjective. It is a combination of adjective and adverb, for example, ‘completely tired’. It can be considered as weak collocation as ‘completely’ can also collocate with ‘full’ or ‘baked’. Seventh is noun + verb. It is a combination of noun and verb, for example, ‘lions roar’. It can be considered as strong collocation as ‘lions’ naturally collocate with ‘roar’ not ‘shout’. Eighth is verb + expression of the preposition. It is a combination of verb and preposition, for example, ‘escape from’. It can be considered as a unique collocation as ‘escape’ can only collocate with ‘from’. Collocation has become a problem for non-English speakers such as Indonesian because they do not have any guide on which words are meant to each other. For example, some Indonesian people tend to say ‘make a conclusion’, which is literally translated from Indonesian ‘membuat kesimpulan’. However, in English, the correct collocation is ‘draw a conclusion’. The word ‘draw’, according to KBBI, means ‘menggambar’ instead of ‘menarik’. Thus, Indonesian speaker rarely uses ‘draw a conclusion’ because of the assumption that it means ‘menggambar kesimpulan’, which is sound illogical. Researches in the translation of collocation have revealed that the translators will be confronted by numerous problems of various sorts in rendering collocations. The problems can be classified into two broad types; intralingual and interlingual problems (Agah & Soori, 2015). Intralingual problems relate to the problems of identifying collocations and establishing collocability within a language. Meanwhile, interlingual problems revolve around questions of collocability across languages, which constitute the translator’s concern to find the ‘acceptable’ collocations in the target language. METHODS The data for this research are the abstracts from twenty (20) theses of English Department students of Bina Nusantara University from 2013 to 2016 batch. Each 292 LINGUA CULTURA, Vol. 13 No. 4, November 2019, 289-296 thesis contains one abstract in English and one abstract in Indonesian. All the abstracts are then processed using Google Translate, which generates translations of the students’ abstracts. Therefore, there are four data sets obtained for this research; the first set consists of 20 Students’ Original Abstracts in English, abbreviated as STE. The second set consists of 20 Google Translation of STE in Indonesian, abbreviated as GTI. The third set comprises of Students’ Original Abstracts in Indonesian, abbreviated as STI. The last set comprises of Google Translation of STI in English, abbreviated as GTE. From those data, samples of lexical collocations are taken to be analyzed. The samples are chosen based on Hill’s (2000) taxonomy of collocation, consisting of eight categories. The collocations in the students’ texts and their translations are compared and scrutinized. The comparison will show how students and Google form the collocations and what kind of errors they have made in the process of translating the collocations. Errors in translation will be measured based on the formation and meaning equivalence between the source and target language. RESULTS AND DISCUSSIONS Hill (2000) has outlined eight categories of collocation, as stated in the Literature Review section. However, from the data, the researchers only find five categories, which contain translation errors. It means that there are no errors in the other three categories. They are (1) category 1: adjective + noun, (2) category 2: Noun + Noun, (3) Category 3: Verb + (Adjective) + Noun, (4) Category 7: Noun + Verb and (5) Category 8: Verb + expression of preposition. Hence, we did not find translation errors in category 4: Verb + Adverb, category 5: Adverb + Verb, and category 6: Adverb + Adjective. The researchers will explain the description of findings for the translation of collocations of categories 1, 2, 3, 7, and 8. For easier reference, the errors will be marked differently. The underlined words mean that the colocations are not found either in English or in Indonesia. On the other hand, the bolded words mean that the collocations contain translation errors either in structure or meaning. The first category of the lexical collocation that will be analyzed is Adj. + Noun. There are 20 samples taken from the abstracts, which belong to the first category. However, after being analyzed, only 14 samples of the collocations contain errors. Of those 14 samples, five samples are taken and analyzed regarding their mistakes and the possible corrections. Table 1 shows the four versions of abstracts, consisting of the Students’ original works (both in English and Indonesian) and the Google Translate works (both also in English and Indonesian). Table 1 Translations of the First Category Collocations No STE GTI STI GTE 7. the extrinsic approach pendekat-an ekstrinsik pendekatan unsur ekstrinsik the extra- curricu-lar element 10 The first method Metode pertama Cara pertama The first way 12 High class Kelas tinggi Orang kaya The rich Table 1 Translations of the First Category Collocations (Continued) No STE GTI STI GTE 14 Natural identity Identitas alami Identitas asli Original identity 18 Infelicitous utterances Ucapan yang tidak benar Ucapan- ucapan yang infelicitous Infelicitous sayings The words ‘the extrinsic approach’ is translated into ‘pendekatan unsur ekstrinsik’ by ST in abstract 7; thus, there is an addition of the word ‘unsur’ to clarify the meaning of ‘ekstrinsik’. However, this addition causes Google Translate to render this phrase into ‘extracurricular element’. In this case, Google Translate makes two errors. First, it omits the word ‘approach’, and secondly, it translated ‘extrinsic’ into ‘extracurricular’, which has a totally different meaning with ‘extrinsic’. Moreover, the word ‘element’ does not collocate with ‘extracurricular’; thus, it cannot form a collocation. Yet, the word ‘element’ can collocate with other adjectives such as ‘basic’, ‘critical’, ‘essential’, ‘dominant’, etc., so it can form weak collocations. In abstract 10, ‘the first method’ is translated by ST into ‘cara pertama’. In general, ‘method’ has the same sense as ‘cara’, but the usage is different. ‘Method’ means ‘a systematic procedure for accomplishing something’, and it is usually used in an academic setting. Meanwhile, ‘cara’ or ‘way’ means ‘a style of doing something’, which is used in a broader sense. Therefore, the rendering of ‘first method’ into ‘cara pertama’ is not appropriate. In Indonesian, the word ‘method’ is borrowed into ‘metode’. At abstract 12, the Google Translate has translated ‘high class’ into ‘kelas tinggi’, which does not collocate to each other as ‘kelas’ does not match with ‘tinggi’. It should be ‘kelas atas’. In English, ‘high’ can be used to signify height or level, such as in ‘high mountain’ or ‘high degree’, but in Indonesian, it should be differentiated into ‘tinggi’ or ‘atas’. In addition, this is considered as weak collocation because ‘kelas’ can collocate with many words such as, ‘menengah’, ‘bawah’, or ‘sosial’. While Google Translate makes a collocational error that ST renders ‘high class’ into ‘orang kaya’. In some sense, people who belong to the high class are rich people, but translating ‘high class’ into ‘orang kaya’ is unacceptable because it has a different meaning. Next, at the abstract 14, ST makes the terms ‘natural identity’, which does not collocate to each other. The words that collocate with ‘identity’ include ‘true’, ‘false’, ‘common’, ‘cultural’, and ‘national’. Thus, it leads into collocation error made by Google Translate, which renders this phrase as ‘identitas alami’. In this context, ‘true identity’ might be a better choice for the author. At abstract number 18, the student does not translate the word ‘infelicitous’, so her translation still contains the original word. In Indonesian, this word can be translated as ‘tidak pantas’, or ‘tidak benar’, but this translation does not correct either. The phrase ‘infelicitous utterances’ is a fixed terminology in pragmatics referring to certain conditions that are not fulfilled before saying the utterances. Meanwhile, GTE translates ‘ucapan-ucapan’ into ‘sayings’, which is more related to the proverb, maxim, or expression. Thus, both ST and Google Translate make an error in translating and forming the collocation. According to these results, this analysis might agree 293The Translation of Lexical.... (Clara Herlina Karjo; Ecclesia Metta) with Ambawani (2014) regarding collocation problems for Indonesian people. Apparently, Indonesian people tend to ignore collocation; for example, the author uses ‘natural identity’ rather than ‘true identity’ just because ‘natural’ in KBBI means ‘alami’. The second category of the lexical collocation analyzed is Noun + Noun. Out of the 20 abstracts that are analyzed, there are eight samples of errors relating to the second category. Out of those eight samples, five samples are taken to analyze in detail that can be seen in Table 2. Table 2 Translation of the Second Category Collocation No STE GTI STI GTE 3 the students’ performan ce kinerja siswa hasil murid- murid the results of students 8 the image of a labor citra persalin an citra buruh the image of workers 9 Memory disorder Gangguan memori Gangguan memori Memory impairment 11 Autistic savant Sarjana autistik Autis terpelajar Learned autistic 11 Library research Kajian pustaka Penelitian kepustaka an Literature review Noun + noun collocations in this category have three formats: [noun + noun], [possessive noun + noun], and [noun + of + noun]. In abstract 3, ‘student’s performance’ is rendered as ‘kinerja siswa’ by Google Translate, which is applicable. However, ST renders it as ‘hasil murid-murid’, which does not have the same sense as ‘the students’ performance’ in English. In this case, Google Translate gives a better translation. On the contrary, in abstract 8, Google Translate makes a serious misinterpretation of the phrase ‘the image of labor’. The word ‘labor’ has two meanings; (1) physical workers, or (2) the process of childbirth. The author uses the first sense for ‘labor’, while Google Translate takes the second sense, so it translates the phrase into ‘citra persalinan’. However, Google Translate is correct in rendering ‘citra buruh’ as ‘the image of workers’, since ‘labor’ (buruh) is usually referred to manual workers in collective, so it cannot be written as ‘a labor’. In abstract 9, the phrase ‘memory disorder’ is translated as ‘gangguan memori’ by both ST and Google Translate. However, in back translation, Google Translate renders it as ‘memory impairment’, which has a slight meaning difference with ‘memory disorder’. A disorder is an illness that disrupts normal physical or mental functions, while impairment is a state where a function is weakening or damaging. Thus, ‘disorder’ can best be translated into ‘gangguan’, while ‘impairment’ maybe for ‘pelemahan’. In terms of collocational pair, both words ‘memory’ and ‘disorder’ can collocate with many other words, mostly adjectives. However, ‘memory disorder’ forms a strong collocation, because it is commonly used with each other. ‘Autistic savant’ in abstract 11 is translated as ‘sarjana autis’ by Google Translate and ‘autis terpelajar’ by ST. It is then back-translated into ‘learned autistic’ by Google Translate. The problem here lies in the word ‘savant’, which means ‘a learned person, especially a distinguished scientist’. Thus, the use of ‘sarjana’ or ‘terpelajar’ is not appropriate because they are not the right equivalence for ‘savant’. In the context of the abstract, this ‘savant’ refers to one character of the movie Rain Man, which is not a scientist. The more appropriate collocation is ‘idiot savant’. An ‘autistic savant’ can be considered as a unique colocation. The last sample for this category is found in abstract 11, in which ‘library research’ is translated into ‘kajian pustaka’ by Google Translate, but ‘penelitian kepustakaan’ by ST, and then it is rendered into ‘literature review’ by Google Translate. In this case, the translation chosen by the author (ST) is more appropriate than Google Translate. ‘Literature review’ only refers to a specific chapter of a research report, while ‘library research’ refers to the method used for doing the research. The analysis of collocation errors for the second category might support Karjo’s (2015) works regarding Google Translate quality. It appears that Google Translate indeed has a problem in choosing the equivalent words when the target language has more than one meaning for the same word. For example, in the third abstract, ‘the image of labor’ becomes ‘citra persalinan’, which shows that Google Translate makes a wrong choice of meaning. The third category of the lexical collocation that is analyzed is Verb + (Adjective) + Noun. Out of the 20 abstracts being analyzed, there are five samples of errors relating to the third category. All of those five samples are analyzed in Table 3. Table 3 Translations of the Third Category Collocation No STE GTI STI GTE 1 Acquire English slang Mengakui sisi bahasa Inggris slang Mempero leh kata gaul get English slang 2 Compo sing the written text. menyusun teks tertulis mengarang sebuah text tertulis Com posing a written text. 7 Compos ing music menulis musik Menggu bah musik Com pose music 13 Compose the prob- lems Menyusun permasala han Menyusun permasala han Arrange the problems 14 Adopt foreign lifestyle Mengadopsi gaya hidup asing Mempela jari gaya hidup Inggris Study ing British lifestyle In the first abstract, ‘mengakuisisi bahasa Inggris slang’ in Indonesian does not collocate with each other. The word ‘acquires’ means ‘to obtain (an object, or property) for oneself’ and ‘to learn or develop’ (a habit, skill, or quality), which in Indonesian it is borrowed into ‘akuisisi’. However, for English slang, the more appropriate word is ‘to obtain’, thus the translation should be ‘mendapatkan/memperoleh Bahasa Inggris slang’. This combination can be considered as weak collocation as ‘memperoleh’ can collocate with many words such as ‘hasil’, ‘alat’, or ‘pendidikan’. The problem with the word ‘compose’ occurs in abstract 2, 7, and 13. The word ‘compose’ means ‘write or create’, and it usually collocates with music or poetry. It also has another meaning of ‘forming and arranging the parts in an artistic way’. In the seventh abstract, the author writes ‘composing music’, but Google Translate translates it 294 LINGUA CULTURA, Vol. 13 No. 4, November 2019, 289-296 into ‘menulis musik’, which does not form a collocation. In Indonesian, there is a specific word for ‘composing music’, i.e., ‘menggubah musik’. In abstract 2, the author translates the word ‘compose’ into ‘mengarang’, which is acceptable in terms of meaning. However, it does not collocate with the word ‘suatu teks’. In the last abstract 13, the word ‘problem’ does not collocate with ‘compose’. The word ‘problem’ can form weak collocate, such as ‘create’, ‘pose’, ‘solve’, ‘present’, etc. So, both Google Translate and ST makes a mistake in forming the collocation. In abstract 14, the phrase ‘adopt foreign style’ is translated into ‘mempelajari gaya hidup Inggris’ by Google Translate. The word ‘adopt’ can mean ‘choose to take up or follow’ (an idea, method, or course of action), or ‘legally take (another’s child) and bring it up as one’s own’. This word is borrowed into Indonesian into ‘mengadopsi’, but the common meaning used is the second one (mengadopsi anak – to adopt a child). Thus, rendering ‘adopt’ into ‘mempelajari’ (studying) is incorrect in terms of meaning. It is better translated as ‘menerapkan gaya hidup asing’. Based on the result, the analysis might disagree with Aiken and Balan’s (2011) research as they have said that Google Translate did a better job only when translating from European to another European language rather than to Asian languages. This analysis shows that Google Translate still does a good job when translating to Indonesian, an Asian country, as there are only a few mistakes in translation. Overall, Google Translate uses the borrowed words if it could not find the Indonesian word equivalence. The seventh category of the lexical collocation that will be analyzed is noun + verb. The seventh category is chosen because there is no mistake found in the abstracts relating to the fourth, fifth, and sixth category. Out of the 20 abstracts scrutinized, there are seventeen samples, which contain errors regarding the seventh category. Five out of those 17 samples will be taken to be analyzed in detail. The focus on this category is the noun and what verb is suitable or collocates with that noun. It can be seen in Table 4. Table 4 Translations of The Seventh Category Collocation No STE GTI STI GTE 2 The writer held (qualitative method) Penulis melaku kan Penulis mengguna kan The writer uses 2 This thesis concerns with Thesis ini berkaitan dengan Skripsi ini berpusat pada This thesis focuses on 6 Sexist lan- guage exist Bahasa sek- sis memang ada Bahasa seksis terbukti benar Sexist language is proven to be true 9 This research examines Penelitian ini meneliti Penelitian ini membahas This research discusses 13 This thesis focuses on Tesis ini berfokus pada Skripsi ini berpusat pada This thesis centered on In abstract 2, the mistake is initiated by the author when she writes, ‘the writer held a qualitative method’. Relating to the object of the verb, ‘qualitative method’, the verb ‘held’ does not fit in this context since ‘held’ in Indonesian is not ‘menggunakan’, but ‘memegang’ or ‘mengadakan’. That is why Google Translate renders it as ‘Penulis melakukan metode’, which produces the awkward translation. However, this mistake is generated from the author herself. In abstract 2 and 13, the problem lies in the word ‘focus on’. The translation for ‘focus on’ is ‘berpusat pada’ or ‘berfokus pada’, which is a borrowed word from English. However, Google Translate translates it into ‘centered on’ in abstract 13. The word ‘centered’ has more spatial sense, which puts something in the center. Thus, the phrase ‘this thesis centered on’ does not form a correct collocation. Meanwhile, in abstract 2, the writer uses the words ‘this thesis concerns with’, which is translated into ‘thesis ini berpusat pada’. Though both phrases form collocations, ‘concerns with’ are not equivalent to ‘berpusat pada’. In abstract 6, the author makes the wrong translation of the word ‘exist’, which she translates into ‘terbukti benar’. ‘Exist’ means ‘have objective reality or being’, which is ‘ada’, ‘mewujud’ in Indonesian. ‘Being exist’ does not mean ‘proven to be true’ or ‘terbukti benar’. In this case, Google Translate correctly translates the phrase. In abstract 9, Google Translate uses redundant words to translate ‘this research examines’. It creates the translation ‘penelitian ini meneliti’, which sounds awkward and does not form a collocation. The collocated words for the research include ‘demonstrate’, ‘indicate’, ‘prove’, ‘reveal’, ‘shows’, etc. Thus, ‘research’ can form weak collocation because it can collocate with many other words. The collocation errors presented in this section shows that students have little knowledge of collocations, so they choose the words that do not collocate with each other. This result is similar to the results of Arabic students’ translation in Jabak, Abdullah, & Mustapha (2016). For example, in the second abstract, the author uses the word ‘held’, which does not collocate with the word ‘method’. It is more natural if she uses ‘apply’ instead of ‘held’. This phrase indicates students’ lack of knowledge of English collocations. The eighth category of the lexical collocation that will be analyzed is verb + expression of preposition. This is the last category of the lexical collocation category, and out of the 20 abstracts that are analyzed, there are only four samples of errors regarding the eighth category. All of those four samples will be taken to be analyzed in detail. It can be seen in Table 5. Table 5 Translations of Eighth Category Translation No STE GTI STI GTE 8 Analyzed with Dianalisa dengan Dianalisa dengan Analyzed by 8 (conclu- sions) are drawn to answer (kesimpulan) ditarik untuk (kesim- pulan) disimpulkan untuk (conclu- sions) are summary zed to 12 measured by several variables diukur dengan beberapa variabel diukur melalui beberapa variable measured by several variables 12 Struggle for principal berjuang un- tuk prinsipal berjuang memperta hankan prinsip hidup strive to maintain the principle of life 295The Translation of Lexical.... (Clara Herlina Karjo; Ecclesia Metta) The discussion of the suitable preposition for some words relates to the expressions before or after the collocation in question. The expression after the words ‘analyzed with’ in the first sample is speech act theory. So, the speech act theory functions as the instrument for analysis. The problem here is that the author makes wrong collocation by using ‘with’. It should be ‘analyzed by (using)’, which can be translated into ‘dianalisa dengan (menggunakan)’. In the second sample of abstract 8, the complete phrase is ‘conclusions are drawn to answer problem formulation’. Thus, the word ‘drawn to’ forms a strong collocation with ‘conclusion’, become ‘to draw a conclusion’. Similarly, in Indonesian, this phrase can be translated into ‘menarik kesimpulan’, which also makes a strong collocation, because literally ‘to draw’ means ‘menggambar’ not ‘menarik’. Thus, the first Google Translate’s translation ‘kesimpulan ditarik untuk’ is suitable. Meanwhile, ST translates it into ‘kesimpulan disimpulkan’, which is redundant and does not form a collocation. A similar problem occurs in abstract 12, where ST uses ‘diukur melalui’, which is a wrong translation of ‘measured by (several variables)’. Google Translate renders it as ‘diukur dengan’. ‘Melalui’ is not translated into ‘by’ in English, but ‘through’. However, ‘measured through’ also does not form a collocation. Thus, in this case, Google Translate makes better translation for this phrase. Still at the twelfth abstract, ‘struggle for’ is not translated as a verb + preposition in Indonesian; the author translates it into ‘berjuang mempertahankan’. It seems that the author eliminates the preposition but added a verb ‘mempertahankan’ to replace the preposition ‘for’. The problem here occurs because the author makes an unintelligible phrase ‘struggle for principal’, which is then translated by Google Translate into ‘berjuang untuk principal’, which also does not make sense. ST might want to write ‘struggle for life principle’. The results in this category might support (Haghighi, 2018) an indication that some collocations are language- specific. For example, western people often say ‘catch a cold’, but in Indonesian, it is said ‘masuk angin’ (literally ‘enter a wind’). Similarly, in English, people say ‘draw a conclusion’, but in Indonesian, it is ‘menarik kesimpulan’ (literally ‘to pull a conclusion’). CONCLUSIONS The findings of this research show that there are quite many errors in terms of lexical collocation, which are made by students or Google Translate. Out of 20 abstracts of the students’ working papers, there are only five categories of lexical collocation errors found in the abstracts and their translations. These categories are Adj. + Noun, Noun + Noun, Verb + (Adj.) + Noun, Noun + Verb, and Verb + expression of preposition. Thus, there are three categories of lexical collocation that do not incite errors, mostly involving the use of adverbs. Overall, Google Translate makes five wrong translation errors, while the students make 14 errors. Google Translate makes more errors in forming the lexical collocations. On the other hand, students make more errors in translating the collocations, i.e., finding the equivalence for the collocations. Because of these, Google Translate makes errors in back translation as it tends to translate exactly based on the source texts. It seems that if students do not make wrong translation errors, Google Translate might be able to avoid collocation errors. In conclusion, Google Translate still needs improvement regarding the translation of lexical collocations. However, some of the errors in translation happen because of the inaccuracy of the source text. In this case, if the students know how to form collocation in Indonesian or in English, the translation error by Google Translate can be minimized. Errors in Google Translate might occur because people do not write accurately in the source language. As a result, Google Translate will not give a good translation. Thus, when the source language text is more correctly structured, it is more likely that Google Translate will provide better translation in the target language. The results of this research imply that students, especially those who need to write their final papers, should be given more knowledge about the collocations in English so that they can write in better English. As for translation, Google Translate can be allowed for students for quick and easy translation tool. However, lecturers or instructors should exercise caution and guidance on its appropriate use to avoid students’ addiction to this tool. For further research, the researchers might compare the results of Google Translate for two types of text, one being the original students’ writing, and the other one is the edited version of the students’ writing. This can be done to prove that well- written source text will produce a better translation in GT. REFERENCES Agah, M., & Soori, A. (2015). Comparative study of collocation among the languages. Language in India, 15(1), 527–537. Aiken, M. (2019). An updated evaluation of Google Translate accuracy. Studies in Linguistics and Literature, 3(3), 253–260. http://doi.org/10.22158/sll.v3n3p253. Aiken, M., & Balan, S. (2011). An analysis of Google Translate accuracy. Translation Journal, 16(2), 25– 32. Al-Khresheh, M. H. (2016). A review study of contrastive analysis theory. Journal of Advances in Humanities and Social Sciences, 2(6), 49–59. http://doi. org/10.20474/jahss-2.6.5. Allue, B. R. (2017). The reliability and limitations of Google Translate: A bilingual, bidirectional, and genre-based evaluation. Entreculturas, 9, 67–80. Ambawani, S. (2014). Grammatical errors on Indonesian- English translation by Google Translate. In Prosiding Seminar Nasional Aplikasi Sains & Teknologi. Yogyakarta, Indonesia. pp. 333–338. Baker, M. (2018). In other words: A coursebook on translatoin (Third Edition). London: Routledge. Dhakar, B. S., Sinha, S. K., & Pandey, K. K. (2013). A survey of translation quality of English to Hindi online translation systems (Google and Bing). International Journal of Scientific and Research Publications, 3(1), 2250–3153. Ghasemi, H., & Hashemian, M. (2016). A comparative study of Google Translate translations: An error analysis of English-to-Persian and Persian-to- English translations. English Language Teaching, 9(3), 13–17. http://doi.org/10.5539/elt.v9n3p13. Haghighi, H. (2018). A multifaceted approach to the translation of collocations from English to Persian. 296 LINGUA CULTURA, Vol. 13 No. 4, November 2019, 289-296 Applied Linguistics Research Journal, 2(2), 8–25. http://doi.org/10.14744/alrj.2018.03511. Hill, J. (2000). Revising priorities: From grammatical failure to collocational success. In Teaching Collocation: Further Developments in the Lexical Approach (p. 245). London: LTP. House, J. (2014). Translation quality assessment: Past and present. In Translation: A Multidisciplinary Approach. London: Palgrave Macmillan UK. pp. 241– 264. http://doi.org/10.1057/9781137025487_13. Ismail, A., & Hartono, R. (2016). Errors made in Google Translate in the Indonesian to English translation of news items texts. Journal of English Language Teaching, 5(2), 1–6. doi: https://doi.org/10.15294/ elt.v5i2.11228. Jabak, O. O., Abdullah, S., & Mustapha, N. F. (2016). The difficulty of translating collocations from Arabic into English encountered by a sample of Arab students. SSRN, 3(1), 1-10. Karami, O. (2014). The brief view on Google Translate machine. In Artificial Intelligence on Natural Language. Vienna: Vienna University of Technology. Karjo, C. H. (2015). Toward a better use of Google Translate. Jakarta: Atma Jaya University. Karjo, C. H. (2016). IT-based translation: How accurate are they? In KIMLI. Denpasar: MLI. pp. 128–131. Koponen, M., & Salmi, L. (2015). On the correctness of machine translation: A machine translation post- editing task. The Journal of Specialised Translation, 23(23), 118–136. Medvedev, G. (2016). Google Translate in teaching English. The Journal of Teaching English for Specific and Academic Purposes, 4(1), 181–193. Napitupulu, S. (2017). Analysing Indonesian-English abstract translation errors by Google Translate. International Journal of English Language and Linguistics Research, 5(2), 15–23. Newmark, P. (1988). A textbook of translation. New York: Prentice Hall. Pârlog, H., & Punga, L. (2017). Difficulties of translating English collocations into Romanian. B.A.S. British and American Studies, 23, 255–274. Raza, M. A., & Nor, F. M. (2018). Google Translate in EFL classroom. International Journal of Translation, 30(1), 7–21. Setiawan, Y. (2014). English Translation errors in abstract of educational administration student of post graduate school of State University in Medan. Journal of Education, 7(1), 67–71. Turovsky, B. (2016). Found in translation: More accurate, fluent sentences in Google Translate. Retrieved October 23, 2019 from https://blog.google/products/ translate/found-translation-more-accurate-fluent- sentences-google-translate/. Utami, S. (2017). The source of errors in Indonesian-English translation. Jurnal KATA, 1(2), 192–202.