E-Journal A COMPARISON OF SOME TRANSLATIONS OF GOOGLE AND THOSE OF TIM KUNCI CULTURAL STUDIES CENTER Asril Marjohan English Education Department Faculty of Languges and Arts Education University of Ganesha e-mail: asrilmarjohan@gmail.com Prof. Dr. Drs. Ida Bagus Putra Yadnya, M.A. Linguistics Department, Postgraduate Program Udayana University Jl. Nias 13 Denpasar, Telp. (0361) 224121 Prof. Dr. Aron Meko Mbete Linguistics Department, Postgraduate Program Udayana University Jl. Nias 13 Denpasar, Telp. (0361) 224121 Prof. Dr. I Ketut Darma Laksana M.Hum. Linguistics Department, Postgraduate Program Udayana University Jl. Nias 13 Denpasar, Telp. (0361) 224121 ABSTRACT This study was aimed at describing Google translations by comparing them with those which were carried out by humans. The human translations used in this study were done by Tim Kunci Cultural Studies Center. This study used qualitative-descriptive method. The sample was purposively selected and consisted of 24 sentences from Chris Barker, Cultural Studies and their two translated versions. The evaluation was done by using Sical’s approach. The results showed that Google translations contained more errors than those made by Tim Kunci Cultural Studies Center. Google made errors that fell into the category of violations of linguistic and sociocultural rules or norms. This was caused by the design of Google Translate that was not based on linguistic and sociocultural analyses. Meanwhile Tim Kunci Cultural Studies Center made errors that indicated reasoning by guessing meanings from context, as what humans are doing in communication. Google translations needed revising in the forms of translating words which had been left untranslated, retranslating words which had been wrongly translated because of not considering context, retranslating phrases which had been wrongly translated because of not analyzing them into their constituents and propositions, retranslating words which had been wrongly translated because of not considering sociocultural norms. In doing the revisions the 2 editor had to read the original text because the translations tended not to give clues to the original patterns of the sentences for their highly varied forms. Meanwhile, the translations of Tim Kunci Cultural Studies Center needed minor editing to make them more effective. There was no need for an intensive reading of the original for their more systematic characteristics. Keywords: Google, human, translation 1. Introduction This study compared Google translations and human translations. Today, translating activities are no longer solely carried out by humans and it seems that machine translation has increasingly become more popular with the consequence of making translating a much easier activity. However, this can threaten professional translators. They are facing a stronger competitor since translational technology has come to a stage in which translational profession has become obsolete. The public will switch to translation machines such as Google Translate which does translating process far much faster than a human translator. Why bother yourself waiting for a professional translator who charges you at a higher rate and who does the job at a slower rate than Google Translate. However, this opinion is not completely correct. Another thing that highly determines the choice is quality. If the quality of Google translations is the same as that of professional translator’s translations, people will choose Google. Some translators state that Google translations can be used for finding information or browsing while for getting a more accurate result one needs the help of a professional translator. Vitek and Cavelier have come to this conclusion when they say that machine translations are a very good tool that a researcher can use when searching relevant patent documents. However, a professional translator is needed to get an official translation of a document (the-future-of-patent translations-human-or-machine). Google Translate does not use context, grammatical and literal models. Instead, it uses a statistics-based approach. Statistical Analysis is applied to a bilingual parallel text with at least one million words in each of the respective languages and 2 sets of monolingual texts with at least one billion words. Statistical models and allogarithms that are derived from the data are used to compile translations of new texts between the two languages without understanding each language or its unique rules (Och, 2006). The probability of an equivalent is calculated by considering thousands of words in the parallel text and the statistical probability is used to determine the choice of lexical equivalents (Baker, 1997:148). Semantic learning technology is used to enhance machine translation by detecting patterns in the documents that have been translated by translators (the-future-of-patent translations- human-or-machine). This study compared the quality of Google Translate’s translations and that of those done by Tim Kunci Cultural Studies Center. The questions posed in this study are (1) What do the errors in Google Translate’s translations look like?; (2) Can we, after considering the errors, conclude that Google Translate’s translations are unpublishable? (3) If they are unpublishable, what an editor has to do to improve the translations. 4) How Google Translate’s translations differ from those performed by Tim Kunci Cultural Studies Center who translated Chris Barker’s Cultural Studies professionally in terms of the errors found in both translations?; and (5) why do they differ? 2. Research Method This study used qualitative descriptive method. The data were drawn from Chris Barker’s Cultural Studies textbook which was published by Saga Publications in London in 2000 and consisted of 24 sentences. The sentences were translated by using Google Translate on October 14, 2012. As a comparison, the translations of the same sentences were taken from a translated version of Chris Barker’s textbook by Tim Kunci Cultural Studies Center, Yogyakarta which was published in 2005. To evaluate the translations, attention was focused on the errors in the translated sentences which were then grouped into major (grave or serious) and minor (slight) errors by referring to the Third–generation Sical ( Language Quality Measuring System in Canada) (Williams, 2001: 331). Sical gives the definition of a serious error as follows: Translation: complete failure to render the meaning of a word or passage that contains an essential element of the message: also mistranslation resulting in a contradiction of or significant departure from the meaning of an essential element of the message. Language: incomprehensible, grossly incorrect language or rudimentary error in an essential element of the message. An essential element is related to the conscequence of an error for a client. If the translation is erroneous, this will have a serious effect on the client. Therefore, the translation cannot be used without a revision. As a conscequence, a decision will be made on the acceptability of the translations being compared. If a revision is needed, then it should be specified what types of editing job will be needed to improve the quality of the translations. In brief, the procedures that were followed can be shown in the following figure. Figure 1 Research Procedure 3. Study Results and Discussion Errors found in the Translations performed by using Google Translate (on October14, 2012) and errors made by Tim Kunci Cultural Studies Center can be described as follows. 3.1 Translations Performed by Using Google Translate Viewed from the types of errors found in them Google’s Transalations Translations by Tim Kunci Cultural Studies Center Types and number of Errors Types and number of Errors Types of Revision needed Types of Revision needed 4 1) Error in identifying a dependency relationship (Nababan, 1991:1991-142) Examples of this type of errors are English Expression: Centralized economic forms and determinations Google: Bentuk ekonomi terpusat dan penentuan Its correct translation: Bentuk ekonomi dan penentuan ekonomi terpusat The noun phrase above can be analyzed in terms of its modification structure as follows: Centralized / economic forms and determinations M / H Economic / forms and determinations M / H Note: M = Modifier, H = Head The analysis above shows that economic forms and determinations is modified by centralized and in its turn, forms and determinations is modified by economic. In Indonesian the nominal phrase can be translated into bentuk-bentuk dan determinasi-determinasi ekonomi terpusat. This error is a serious one since it can cause misunderstanding. This error must have been caused by inability to identify immediate constituents. Another example of the same type of error is English expression: Pregiven fixed aim or object Google: Tujuan tetap pra-diberikan atau objek A correct translation: Tujuan atau sasaran tetap yang diasumsikan sudah ada sebelumnya 2) Error due to Missidentification of Figurative Meanings English expression: Sweeping political and economic changes Google: Perubahan politik dan ekonomi yang menyapu A correct translation: Perubahan politik dan ekonomi yang berlangsung dengan cepat dan memberi dampak yang luas.( two among the meanings of the word sweeping according to Webster’s Ninth Collogiate Dictionary are to cover the entire range of dan to move across or along swiftly) The word sweeping in the expression above is used with its figurative meaning. Hence, sweeping should not be translated literally into menyapu. 3) Error due to missidentication of word meanings in context English expression: Argue against in the context Foucault argues against the repressive hypothesis Google: Foucault berpendapat terhadap A correct translation: Foucault memberikan argumen yang menentang hipotesis represif Google mistranslated the preposision against in the above context. Other examples of the error that belongs to this type are (1) For in For Foucault, power is…. translated into Untuk Foucault…. The correct translations are Bagi Foucault and Menurut Foucault…. (2) Ordering in making them grow and ordering them, which was translated into membuat mereka tumbuh dan memerintahkan mereka.The correct translation is membuat … tumbuh dan mengaturnya. (3) Submit in …making them submit, which was translated into ..membuat mereka serahkan.The correct translation is membuat mereka menyerah.The word serahkan is used in an imperative sentence, such as Serahkan uang tebusan itu dengan segera! while --- membuat mereka--- is a declarative sentence that uses a verb with the prefix meN- instead of a verb with suffix –kan. (4) Medicine in …discourses about medicine, which was translated into obat-obatan. The correct translation is kedokteran. (5) Perversity in polymorphous perversity, which was translated into kejahatan polymorphous.The correct translation is penyimpangan berbagai bentuk. (6) Bite in sound-bite which was translated into suara gigitan. The correct translation is efek suara yang tajam. 4) Leaving the English words untranslated as the word polymorphous above. Other examples are (1) Determinate, (2) Reader, and (3) Lacanian. 5) The use of two words that refer to the same concept without putting them side by side and inserting one of them in a bracket to show that the two words are interchangeable. An example is the use of the word kekuasaan and power. From the description above the following can be summed up: Google translations contained many weaknesses and the serious ones among them were (1) leaving English words untranslated (lexical interference), (2) selecting the meanings of words that did not fit into the context, and (3) errors in identifying a dependency relationship. Errors that belong to the first category caused the translations to leave essential meanings that needed to be made explixit to the reader to help him or her understand the source language text, errors of the second category caused inaccuracy since the translations did not communicate the correct meanings of the polysemous words according to the context, and errors of the third category caused misunderstandings of the propositional contents of the source language text that resulted in errors in understanding the text. In addition, there were errors in selecting the lexical meanings such as in the translation of perversity into kejahatan in polymorphous perversity, which is unacceptable to the reader as a member of a community who is fully aware of the norms and law. The reader will not accept such a type of translation since he or she is aware that a baby cannot be blamed for what it does. A criminal is usually a person who is regarded as having been able to differentiate good deeds from bad ones that he or she deserves to get a reward or penalty for his or her action. 6 3.2 Translations Performed by Tim Kunci Cultural Studies Center Viewed from Errors Found in them 1) Errors that result in the use of reasoning to determine meanings in context Example: English expression: Sweeping political and economic changes Tim Kunci Cultural Studies Center (henceforth called Tim): Perubahan politik dan ekonomi penting. The word sweeping in the nominal phrase sweeping political and economic changes was translated into penting. This translation is wrong since the meaning of the word in the above context is ‘extensive’, which is ‘ekstensif’ or ‘meluas’ in Indonesian. The translator might have inferred the meaning in the context. Another example of the errors of this category is as follows. English expression: that subjectivity is a discursive construction Tim bahwa subjektivitas adalah konstruksi sosial The word discursive has been wrongly understood to have the same meaning as the word social. 1) Partial translation Example: English expression: The pioneering work of Saussure Tim Karya Saussure The word pioneering in the noun phrase above was left untranslated and ommited. 2) Leaving English words untranslated (lexical interference) English expression: In Lacanian term Tim : Dalam istilah Lacanian The word Lacanian in the English expression above has been maintained in the translation above. The word should be translated into Lacan Another example of error of this type is: English expression: Sexed subjects Tim: Subjek –subjek terkelaminkan The correct translation is topik-topik yang digolongkan menurut jenis kelamin. Here the translator overgeneralized the equivalents of loan words without looking at the context in which they occur. The English word subject has many meanings and for most of them the translator has to find other words that fit in the context in Indonesian. The word Subject can mean course (mata pelajaran), citizen (warga negara), subject of a sentence (subjek), subject of study (subjek penelitian), etc. 3) Wrongly applied Literal translation Example: English expression: British Asian teenage viewers of soap opera in the UK Tim: Remaja Asia Inggris penonton soap opera sabun di Inggris The translator could have explicitly specified the relationship between the noun Asia and English (Inggris) above. Thus, the translation will read: Remaja Inggris yang berasal dari Asia in which it is clear that the English teenagers come from Asia. In addition, the word penonton can be replaced by yang menonton to make it sound natural to the Indonesian reader. The cause of the error was a wrong application of literal translation and the deviation from the convention or common collocational patterns in the target language. Other examples of this type of error are: (1) Carpet bombing Tim: Pengeboman karpet, which should be pengeboman intensif (2) The message of the Author-God Tim: (4) Pesan –Pengarang – Tuhan, which should be wahyu atau firman Tuhan (5) Human animal Tim: Hewan manusia, which should be manusia dari segi sifat hewaniahnya 3.3 Differences between Errors Found in Translations Performed by Using Google Translate and Those Made by Tim Kunci Cultural Studies Center The errors found in the translations performed by using Google Translate were serious while those made by Tim Kunci Cultural Studies Center were not. Many words were left untranslated such as polymorphous, determinate, reader and Lacanian. In the translations carried out by Tim Kunci Cultural Center, there was only one word left untranslated, namely, the word Lacanian. The translations peformed by using Google Translate tended to contain word meanings that did not fit in the context, such as argue against, which was translated into berpendapat terhadap, for which was translated into untuk in the context for Foucault, power is… ordering that is translated into memerintahkan in the context making them grow and ordering them, medicine which was translated into obat-obatan in the context discourses about medicine, perversity which was translated into kejahatan in the context polymorphous perversity and bite which was translated into gigitan in the context sound bite. Furthermore, Google Translate produced translations containing errors indicating the wrong analyses of phrases into their immediate constituents that caused errors in understanding the propositional contents. As an example, Google translated centralized economic forms and determinations into bentuk ekonomi terpusat dan penentuan. Tim Kunci Cultural Studies Center did not make such types of errors. On the other hand, Tim Kunci Cultural Studies Center made errors indicating that the translator guessed the meanings of words in context through reasoning. Two examples of this type of error are the translation of the word sweeping in sweeping economic and political changes into penting and the 8 translation of discourse in discourse construction into sosial. In addition to the types of errors that have been described so far, Google Translate also made errors that are related to the violation of a grammatical rule in Indonesian. An example of this type of error is the translation of the word submit into serahkan in making them submit. Such a type of error was not found in the translations made by Tim Kunci Cultural Studies Center. Nor was the error made by Google in translating the word perversity into kejahatan. Tim Kunci Cultural Studies Center translated it into kenakalan. 4. Conclusion and Suggestions 4.1 Conclusion Viewed from the types of errors, the translations performed by Google Translate contained far more errors than those by Tim Kunci Cultural Studies Center. Google Translate made errors related to the violation of grammatical rules and norms in the community since Google Translate was not designed based on a linguistic analysis and analysis of sociocultural aspects of the language. Meanwhile, Tim Kunci Cultural Studies Center made errors indicating the use of reasoning by guessing word meanings in context, as what is commonly done by a human being in communication. This is done by the human when he or she does not know the meaning of a word. There are two things that he or she may do in such a situation. First, he or she may analyze the word into its morphemes and secondly, he or she may relate the word to other words found together with the new word in the context. The translations produced by Google Translate need revising, by translating English words that were left untranslated, retranslating the words wrongly translated since they were translated without considering the context, translating phrases by analyzing them first into their constituents to get the propositional contents before finding their equivalents to maintain their propositions in the target language and retranslating the words wrongly translated since they were translated without considering the sociocultural norms in the community. All of these jobs need to be done by reading the original text intensively due to the varied and unsystematic patterns of the sentences. Meanwhile, the translations done by Tim Kunci Cultural Studies Center only need to be edited in one place or another to get a more effective result. Unlike the translations performed by Google Translate, the translations done by Tim Kunci Cultural Studies Center can be edited without reading the original text intensively due to their more systematic characteristics. Google Translate made errors related to grammatical rules since it was not based on a linguistic analysis but on statistical analysis applied to a parallel bilingual text and two sets of monolingual texts. The word submit was translated into the Indonesian word serahkan by referring to a parallel English-Indonesian text, rather than by looking at the linguistic context in which the word occurs. Google Translate is not a human being that can differentiate the context in which the word kejahatan (crime) was used from the context in which the word penyimpangan (deviation) was used. Thus perversity was translated into kejahatan in the context of a baby prior to the resolution of the Oedipus Complex. On the other hand, Tim Kunci Cultural Studies Center made errors indicating the use of reasoning by guessing word meanings from context, a practice that is commonly done by a human being in communication. The guess can be right and it can be wrong. Thus it shows that the translation has been done by a human being rather than a machine. 4.2 Suggestions From the comparison between the translations performed by using Google Translate and the ones performed by professional translators (Tim Kunci Cultural Studies Center), Google Translate can be used for searching information and in using the results of such browsing process the user of Google Translate should be careful of its contents. He or she should be aware that they have been produced by a machine rather by a human being. Thus the results tend to contain errors which a human editor can correct. After finding the information, the user needs a professional translator to do the translation of the document(s) that he or she has selected from the translations performed by Google Translate. In this manner, the time and cost can be used more economically. A professional translator can use the service given by Google Translate in finding the meaning of a word faster than by looking it up in a manual dictionary. 5. Acknowledgements In this opportunity, I would like to express my appreciation to Prof. Dr. Drs. Ida Bagus Putra Yadnya, M.A., as supervisor, Prof. Dr. Aron Meko Mbete, as co-supervisor 1,and Prof. Dr. I Ketut Darma Laksana, M.Hum, as co-supervisor 2, and the board of examiners: Prof. Dr. I Gusti Made Sutjaja, M.A,. Prof. Dr. I Wayan Pastika, M.S., Prof. Dr. I Nengah Sudipa, M.A, Prof. Dr. I Ketut Seken, M.A., and Dr. Frans I Made Brata, M.Mum for their comments and suggestions. References ----------The Future of Patent Translations: Human or Machine? http//intellogist. Craciumescu, O., Salas, C.G, O’Keeffe S., S., 2004. Machine Translation and Computer- Assisted Translation: A New Way of Translating? Translation Journal and the Author 2004. URL: htttp://accurapid.com/journal/29bias.htm 12/20/2010. Baker, M. (ed.).1997. Routledge Encyclopedia of Translation Studies. London: Routledge. Barker, C., 2000. Cultural Sudies: Theory and Practice. Sage Publication: London. Barker, C. 2005. Cultural Studies: Teori dan Praktik (Tim Cultural Studies Center, Penerj.). Yogyakarta: PT. Bentang Pustaka. Nababan, M.R. 1999. Teori Menerjemahkan Bahasa Inggris. Yogyakarta: Pustaka Pelajar 10 Och,F.J. 2006. The Allignment Template Approach to Statistical Machine Translation. MIT Press Journal, 30 940, 417-449. Williams, M. 2001. The Application of Argumentation Theory to Translation Quality Assessment. Meta XLVI, 2, 2001 326- 344. .