GOOGLE TRANSLATE'S PROBLEMS AND ITS POST-EDITING TECHNIQUES OF
AN ENGLISH SHORT STORY TRANSLATION INTO BAHASA INDONESIA
A Research Paper
Submitted as a partial fulfillment of the requirements for Sarjana Sastra degree
by Andi Nurroni
0608877
DEPARTMENT OF ENGLISH LANGUAGE EDUCATION FACULTY OF LANGUAGE AND ARTS EDUCATION
Google Translate's Problems and Its Post-Editing Techniques of an English Short Story
Translation into Bahasa Indonesia
Oleh
Andi Nurroni
Sebuah skripsi yang diajukan untuk memenuhi salah satu syarat memperoleh gelar Sarjana pada Fakultas Pendidikan Bahasa dan Seni
© Andi Nurroni 2012
Universitas Pendidikan Indonesia
Januari 2013
Hak Cipta dilindungi undang-undang.
Skripsi ini tidak boleh diperbanyak seluruhya atau sebagian,
Google Translate's Problems and Its Post-Editing Techniques of an English Short Story Translation into Bahasa Indonesia
Andi Nurroni
Department of English Language Education Indonesia University of Education
2013
andiforshort@gmail.com
Abstract
This study aims 1) to identify the types of translation problem encountered by the most popular internet machine translation Google Translate (GT) in translating an English short story into Bahasa Indonesia, 2) to identify the strategies applied by human translator (HT) in solving the same problems, and 3) to formulate the post-editing techniques which can be formulated by referring to strategies applied by the HT. In identifying the problems, this study focuses on three of six kinds of translation problem in text organization, namely ‘at word level’, ‘above word level’, and ‘at grammatical level’, as proposed by Baker (1992). This study revealed some findings in line with aims of the research. At word level, GT is not sensitive with ‘cultural words’, ‘slang words/spelling’, ‘abbreviation’, and tends to translate words with their general equivalents which causes the translation sounds unnatural. Above word level, GT mostly translates idioms and fixed expression literally, in which sometimes it misinterprets them and translates them in the wrong direction. At grammatical level, the main problem encountered by GT is in translating ‘person system’ as well as ‘information on countability’, ‘information on gender’, and ‘tense and aspect’. Based on the analysis on GT’s problems and HT’s strategies, this study offers 14 general formulas to revise GT’s translation, especially short story. Among others are ‘cultural-words check’, ‘words-with-wide range-of-sense check’, and ‘slang-words check’.
Keywords:
Masalah-masalah Penerjemahan yang Dihadapi Google Translate dalam Menerjemahkan Cerita Pendek Berbahasa Inggris ke Bahasa Indonesia
Beserta Teknik Penyuntingan Akhirnya
Andi Nurroni
Jurusan Pendidikan Bahasa Inggris Universitas Pendidikan Indonesia
2013
andiforshort@gmail.com
Abstrak
Penelitian ini memiliki sejumlah tujuan, yakni 1) mengidentifikasi jenis-jenis masalah penerjemahan yang dihadapi oleh mesin penerjemah Google Translate
(GT) dalam menerjemahkan cerita pendek berbahasa Inggris ke dalam Bahasa Indonesia, 2) untuk mengidentifikasi strategi-strategi penerjemahan yang diterapkan penerjemah manusia dalam menyelesaikan masalah-masalah penerjemahan tersebut, 3) untuk merumuskan teknik-teknik penyuntingan akhir, dengan mengacu pada strategi-strategi yang digunakan penerjemah manusia. Dalam mengidentifikasi masalah-masalah penerjemahan, penelitian ini berfokus pada tiga dari enam jenis masalah penerjemahan di dalam organisasi naskah, yakni ‘pada tingkat kata’, ‘di atas tingkat kata’, dan ‘pada tingkat tatabahasa’, sebagaimana yang dikemukakan oleh Baker (1992). Penelitian ini menghasilkan sejumlah temuan sejalan dengan rumusan masalah yang melandasinya. Pada tingkat kata, GT tidak sensitif terhadap ‘kata budaya’, ‘kata/ejaan slang’,
‘abreviasi’, dan cenderung menerjemahkan kata dengan padanan yang sangat umum yang menyebabkan hasil terjemahan terdegar kurang alami. Di atas tingkat kata, GT sering kali menerjemahkan idiom dan ungkapan baku secara harfiah, yang mana terkadang terjadi salah pemaknaan, lantas diterjemahkan secara keliru.
Pada tingkat tatabahasa, sejumlah masalah utama yang dihadapi GT adalah dalam mengalihbahasakan ‘sistem pronomina’, di smping itu dalam menerjemahkan ‘informasi mengenai jumlah’, ‘informasi mengenai jender’, dan ‘bentuk waktu dan aspek’. Berdasarkan pada analisis atas masalah penerjemahan GT dan strategi yang diterapkan penerjemah manusia, penelitian ini menawarkan 14 kiat umum dalam menyunting hasil penerjemahan GT, khusunya cerita pendek, seperti misalnya ‘melakukan pemeriksaan atas kata budaya’, ‘melakukan pemeriksaan atas kata dengan rentang makna yang luas’, dan ‘melakukan pemeriksaan terhadap kata slang’.
Kata Kunci:
TABLE OF CONTENT
Page of Approval... i
Statement of Authorization ... ii
Preface ... iii
Acknowledgements ... iv
Abstract ... v
Table of Contents ... vi
Chapter I Introduction ... 1
2.1Background ... 2
2.2Research Questions ... 4
2.3Aims of Study ... 5
2.4Organization ... 6
Chapter II Theoretical Foundation ... 8
2.5Introduction ... 8
2.6Definition of Translation ... 8
2.7Significance of Translation ... 9
2.8Machine Translation ... 10
2.9Google Translate ... 13
2.10 Language Functions, Text Categories and Text Types ... 15
2.11 Lexical Meaning ... 17
2.12 Translation Methods ... 20
2.13 Translation Problems ... 22
2.13.1 Problems of Equivalence at Word Level ... 23
2.13.2 Problems of Equivalence above Word Level ... 26
2.13.3 Grammatical Problems ... 34
2.14 Translation Strategies ... 38
2.14.1 Strategies at Word Level ... 38
2.14.3 Strategies at Grammatical Level ... 45
2.15 Translation of Short Story ... 47
Chapter III Research Methodology ... 48
3.1 Formulation of Problems ... 48
3.2 Research Method ... 48
3.3 Data Collection... 50
3.4 Data Analysis ... 52
Chapter IV Findings and Discussions ... 53
4.1 Translation Problems Encountered by Google Translate and The Strategies Applied by Human Translator ... 53
4.1.1 Problems of Equivalence at Word Level ... 53
4.1.2 Problems of Equivalence above Word Level ... 64
4.1.3 Problems of Grammatical Equivalence ... 70
4.2 The General Post-Editing Techniques to Improve the Quality of Google Translate’s Translation of Short Story ... 76
4.2.1 Techniques at Word Level ... 76
4.2.2 Techniques above Word Level... 80
4.2.3 Techniques at Grammatical Level ... 81
Chapter V Conclusions and Suggestions ... 85
5.1 Conclusions ... 85
5.2 Suggestions ... 87
Reference... 87
Appendix ... 89
1
CHAPTER I
INTRODUCTION
1.1Background
Interaction among states and nations throughout the world is in high
traffic, and will always be higher than before. The world’s governments and
citizens meet for the sake of various interests: politics, economy, education,
culture, and so on. Seeing this case further, it is known that language plays a
significant role.
There is not any exchange among civilizations in the world without
language as the bridge. Related to the issue, to use the same language is one of the
methods. In this case, nowadays, English is the first option, apart from French,
Arabic, Chinese, and some other international languages, while another method is
language transfer through translation (and interpretation). According to Newmark
(1988:5), translation is “rendering the meaning of a text into another language in
the way that the author intended the text”.
Communication across nations through translation has existed for a very
long time. As Nida (1991:19-20) says:
“... interlingual communication has been going on since the dawn of
2
The more massive international interaction invites the involved parties to
pay more attention to the role of language, including translation issues. Aside
from the development in theoretical aspects, technology also supports the progress
of translation. Today, technology works in the field of translation to create a
facility named “machine translation” (MT).
MT projects are motivated by the idea that the world needs technology to
support the increasing need of interlingual communication. As an example, at the
beginning of the 1990s, 300 million pages were translated from and into
languages around Europe and United States, and 150 million pages were
translated from and into Japanese for various purposes (Loffler-Laurian, 1996,
cited in Craciunescu, et al., 2004).
Since the idea of MT came up in 1930s (Craciunescu, et al., 2004), the MT
programs have been created and improved to have the characteristics of human
translator. Hutchin (1995, cited in Khosravizadeha & Pashmforoosh, 2011) states
that technically, the main goal of machine translation is to achieve the automated
process and high quality draft of translation. But then, such dream faced various
challenges. One of the practical problems found, as stated by Khosravizadeha &
Pashmforoosh (2011), is that the speaker’s meaning or the pragmatic aspect cannot be easily transferred by such machine.
In the last decades, the existence of MT projects became one of the
burning issues, not only within linguist circles, but also among common society. It
3
benefits: ease of access, high speed and low cost. Even many professional
translators also operate it as a preparation of a rough draft (Champollion, 2003;
Lagoudaki, 2008; O'Hagan & Ashworth, 2002 in Aiken & Balan, 2011) and some
of them use MT to get the gist of foreign text (Altay, 2002 in Aiken & Balan,
2011).
A number of MT programs can be accessed easily as online applications,
some of which are even free of charge. The most well-known MT programs are
Alpaworks, E-lingo, Reverso, Yahoo! Babel Fish, Systran, Transcend, Applied
Language, SDL Automated Translation Solutions, Windows Live Translator, and
Google Translate (Craciunescu, et al., 2004, Aiken & Balan, 2011).
Some studies aimed to compare MT programs were conducted, two of
which were run by Aiken et al. (2009) and “NIST comparison” (2005). The first study, which covered four programs, showed that Google Translate (GT) was the
best, followed by Yahoo!, X10, and Applied Language. The second study, which
covered 22 MT programs, showed that “GT was often first and never lower than third in rankings using text translated from Arabic to English and from Chinese to
English“ (Aiken & Balan, 2011).
Although it is regarded as superior, some studies showed that GT also has
minuses as other MTs do, when it is compared to the work of human translator
(HT). Khosravizadeha and Pashmforoosh (2011) state that GT suffers from a
4
This research is an effort to assess the problems encountered by GT when
translating a short story, and aims to offer general post-editing techniques which
are useful to improve the quality of GT’s text.
In this research, the main theory of translation problems is taken from
Mona Baker’s In Other Words (1992). The English short story used in this research is Shirley Jackson’s Charles, and its translation in Bahasa Indonesia, which was used as comparator, was translated by Syafruddin Hasani.
1.2Research Questions
Based on the background, this research was conducted to answer three
research questions as follows:
1. What problems are encountered by Google Translate in translating
Charles by Shirley Jackson into Bahasa Indonesia?
2. What strategies are applied by the human translator in translating Charles
by Shirley Jackson into Bahasa Indonesia?
3. Based on the problems identified and strategies applied by the human
translator, what post-editing techniques can be formulated to improve the
5
1.3Aims of Study
In line with the research questions, the study was conducted with the
following aims:
1. to identify problems encountered by Google Translate in translating
Charles by Shirley Jackson into Bahasa Indonesia;
2. to identify strategies applied by the human translator in translating Charles
by Shirley Jackson into Bahasa Indonesia;
3. to formulate post-editing techniques to improve the quality of Google
Translate’s translation of short story.
1.4Research Methods
This research is constructed with a descriptive qualitative research design.
Frankle and Wallen (1993:389) suggest that descriptive qualitative is a method of
research to investigate the quality of relationship, activities, situations, or
materials, then to synthesize the information obtained from various sources (e.g.
observations, interviews, and document analysis) into coherent description of
what the researcher has discovered.
1.4.1 Data Collection
1. Identifying and classifying translation problems encountered by GT.
2. Identifying and classifying translation strategies applied by human
6
1.4.2 Data Analysis
1. Elaborating each identified translation problem and relating it to the theory
of translation problem proposed by Baker.
2. Elaborating each identified human translator’s strategy and relating it to the theory of translation strategy proposed by Baker.
1.5Organization
This paper is organized by following the guideline written in the book of
Pedoman Penyusunan Skripsi by Tim Jurusan Bahasa Inggris UPI (2010). This
paper consists of five chapters, each of which has its own role in delivering
contains in detail. Chapter I is Introduction, which aims to introduce the paper in
brief, and which is divided into background, statement of problems, aims of study,
research methods, and organization.
Chapter II is entitled Theoretical Foundation, which is responsible to
elaborate the theories and the previous studies related to the research as an effort
of giving reasonable arguments.
Chapter III is Research Methods. This part is an elaboration of technical
issues in the research, which describes research design, assumptions in the
research, clarification of key terms, steps of finding data, and steps of analyzing
7
Chapter IV is Findings and Discussions presenting the collected data, the
analysis process of the data, and the interpretation and discussion of the findings.
The last part, Chapter V is Conclusion and Suggestion. The first part is the
report on the results of the study to give statements as the answers to the research
problems. The second part is on the important points related to the research which
can be offered to support other similar research in the future and any parties who
48
CHAPTER III
RESEARCH METHODOLOGY
This chapter presents information related to some technical issues of the
research. This chapter consists of three parts, those are “Research Method”, “Data
Collection”, and “Data Analysis”.
3.1Formulation of Problems
This research conducted to answer three research questions as follows:
1. What problems which are encountered by Google Translate in
translating Charles by Shirley Jackson into Bahasa Indonesia?
2. What strategies which are applied by the human translator in
translating Charles by Shirley Jackson into Bahasa Indonesia?
3. Based on the problems identified and strategies applied by the
human translator, what post-editing techniques can be formulated to
improve the quality of Google Translate’s translation of short story?
3.2Research Method
This study is qualitative research with a descriptive method. According to
Denzin & Lincoln (2000:4-5):
49
attempting to make sense of or to interpret, phenomena in terms of the meanings people bring to them.
While Burns and Grove (2003:201) states that descriptive method "is
designed to provide a picture of a situation as it naturally happens".
Based on two arguments above, it can be concluded that qualitative
research with descriptive method is the research where the researcher make a
report based on his or her interpretation about the object by following the rules.
In this study, the object is Google Translate, the most popular web-based
machine translation. The study assesses the Google Translate’s problems in
translating Charles short story by Shirley Jackson into Bahasa Indonesia, and
analyzes the strategies applied by the human translator in translating the same
text. Finally, based on the assessments of the Google Translate’s problems and the
strategies applied by the human translator, this study offers a general post-editing
techniques to improve the quality of a short story which is translated from English
into Bahasa Indonesia by Google Translate.
The theory of “translation problems” and “translation strategies” as the
main theories in this research are adopted from Mona Baker’s ideas, taken from
50
3.3Data Collection
The data of this research are the fragments of Charles short story by
Shirley Jackson taken from the original text (English) and the translated version
(Bahasa Indonesia) by Google Translate and Syafruddin Hasani, as the
comparator translator, which are organized by the category of translation
problems proposed by Baker (1992): 1) problems of equivalence at word level, 2)
problems of equivalence above word level, and 3) problems of grammatical
equivalence. Apart from Baker’s category, in this study, there are some new
categories to classify some problems which were not covered by Baker’s
category. Here are the Baker’s categories in pointers:
1. Problems of Equivalence at Word Level:
a. culture-specific concept
b. The source-language concept is not lexicalized in the target language
c. The source-language word is semantically complex
d. The source and target language make different distinction in meaning
e. The target language lacks a superordinate
f. The target language lacks a specific term (hyponym)
g. Differences in physical or interpersonal perspective
h. Differences in expressive meaning
i. Differences in form
51
2. Problems of Equivalence above Word Level:
a. Collocation:
1) The engrossing effect of source text patterning
2) Misinterpreting the meaning of source-language collocation
3) The tension between accuracy and naturalness
4) Culture-specific collocation
5) Marked collocation in the source text
b. Idioms and Fixed Expressions:
1) An idiom or fixed expression may have no equivalent in the target
language
2) An idiom of fixed expression may have a similar counterpart in the
target language, but its context of use may different
3) An idiom may be used in the source text in both its literal and
idiomatic senses at the same time
4) The very convention of using idioms in written discourse, the
context in which they can be used, and their frequency of use may
be different in the source and target language
3. Problems of Grammatical Equivalence:
a. Number
52
d. Tense and Aspect
e. Voice
The next phase in collecting data is categorizing the strategies applied by
the human translator in translating the text, by following Baker’s categories,
including her suggestions and notes. To accommodate strategies which do not
belong to the Baker’s category, some new additional categories needed are
created.
3.2 Data Analysis
After categorizing the short story fragments, the next step is analyzing the
data by elaborating each case of translation problem encountered by Google
Translate and elaborating each case of translation strategy applied by the human
translator in dealing with each translation problem. Based on the analysis, the last
phase is formulating the post-editing techniques to improve the Google
85
CHAPTER V
CONCLUSIONS AND SUGGESTIONS
This chapter provides the conclusions and the suggestions of the research.
The conclusions part states in brief the answers to the research questions, while
the suggestions part offers things which may be taken as benefits from the
research by any party which may concern.
5.1 Conclusions
This study arrives at some findings in line with the proposed research
questions. It can be concluded that GT faces various problems in translating the
short story Charles by Shirley Jackson.
At the word level, GT is not sensitive to cultural words, slang words or
spelling, and abbreviations. In addition, it tends to translate words with their
general equivalents which causes the translation to sound unnatural, mainly in
words with wide range of senses, including in translating words with
unlexicalized concept and words which lacks of hyponym equivalents.
Above the word level, GT mostly translates idioms and fixed expressions
literally, in which case it sometimes misinterprets them and translate them
86
translating prepositions, terms containing hyphen and terms having complex word
order.
At the grammatical level, the main problem encountered by GT is in
translating person system as well as information on countability, information on
gender, and tense and aspect.
Compared to GT, when translating the same text, human translator
(HT)shows his or her superiority. HT can apply several translation strategies to
solve problems posed by a translation unit by keeping the information meaningful
and the sounds natural in the target text. In translating at the word level, HT
mostly finds the nearest equivalents which can represent the message of the
source-text as completely as possible in the target-text. HT also can omit
unneeded words to prevent redundancy.
Above the word level, apart from keeping the information accurate, HT also
keeps the translation sounding natural. When the ready equivalents do not exist,
HT will paraphrase them with related or unrelated words. On the other hand, at
the grammatical level, HT keeps the translation following the logical order of
words, thus the translation conveys a clear message to the target readers.
Based on the analysis of the problems which are encountered by GT and the
strategies applied by HT to solve the problem in question, this study offers 14
87
especially short story, which can be applied to fix the unsolved problems at the
word level, above word level, and at the grammatical level.
Last but not least, this study suffers from several weaknesses. First, The
assessment made is based only on a single short story, so the cases of problem
may be very limited to be generalized. Second, this study does not include other
translation problems in other text organizations as proposed by Baker, which are
thematic and information structures, cohesion, and pragmatic. Thus, the
assessment is less comprehensive.
5.2 Suggestions
Practically, the results of the study offer a simple help for anyone who may
concern themselves with improving the quality of GT’s translation, especially in
translating English short story into Bahasa Indonesia. For future research related
to the issue, any researcher may focus on the other types of text and may assess
other text organizations. Theoretically, this study may become a reference for any
party which is interested in machine translation study, particularly Google
REFERENCES
Ahmad, M. (2012). Students’ Strategies in Translating an English Text into Bahasa Indonesia: A Case Study at the English Education Department of FPBS UPI Bandung. Unpublished Thesis
Aiken, M. & Balan, S. (2011). An Analysis of Google Translate Accuracy. Retrieved August 14 2013, from http://www.translationdirectory.com
Albir, L. M. and Hurtado, A. (2002). Translation Techniques Revisited: A Dynamic and Functionalis Approach. Translations Journal, volume 47 no. 4. Retrieved August 14 2013, from http://id.erudit.org/iderudit/008033ar
Fraenkle, J. R. & Wallen, N. G. (1993). How to Design Evaluative Research in Education (Second Edition). Singapore: McGraw Hill Book Co.
Baker, M. (1992). In Other Words: a Course Book on Translation. London: Routledge
Carciunecu, O, Gerding-Salas, C, Stringer-O'Keeffe, S. (2004) Machine Translation and Computer-Assisted Translation: a New Way of Translating? Translation Journal, Volume 8, No. 3 2004. Retrieved August 14 2013, from http://www.translationdirectory.com
Denzin, N. K. & Lincoln, Y. S. (2000). Introduction: The discipline and practice of qualitative research. Thousand Oaks, CA: Sage. Retrieved August 14 2013, from http://www.sagepub.com
Ethan, S. (2010, February 10). Comparison of online machine translation tools. Retrieved December 15, 2011, from www.tcworld.info
Khosravizadeha, P. & Pashmforoosh, R. (2011). Google Translation: A Semantic Structure Analysis. Retrieved August 14 2013, from Lingoistica.com
Koponen, M. (2010). Assesing Machine Translation Quality with Error Analysis. Retrieved 14 August 2013, from www.sktl.fi/@Bin/.../Koponen_MikaEL2010.pdf
Larson M. L. (1984). Meaning Based Translation: a Guide to Cross-language Equivalence.
London: University Press of America
Neal, D. (2012, April 27). Google Translate Translets Millions Books Woth a Day. Retrieved August 14, 2013, from http://www.theinquirer.net/inquirer/news
Newmark, P. (1988). A Textbook of Translation. Hertforshire: Prentice Hall
Nida A. E.. (1991). Theories of Translation. Journal TTR: traduction, terminologie, redaction, vol. 4, no. 1
