This page originally authored by Yuki Ichimura (2013). This page covers Machine Translation, including online translating systems in the educational environment.
Machine Translation, referred to as MT, is computer software for automatic natural language translation including free online translators. MT is mostly developed in the form of commercial software and online web translation systems. The systems are designed either for two particular languages (bilingual systems) or for multiple pairs of languages (multilingual systems). Bilingual systems may operate either in one direction only (i.e. from Japanese into English) or in both directions. Although most bilingual systems operate in one direction, multilingual systems usually work for both directions. These days, free online machine translation systems have given rise to its use by helping with multilingual document retrieval and processing. MT can be characterized as the use of the computer in any activity involving language (Somers, 2003).
Machine translation has a long history. It was first conceived in 1949 in seeking the possibilities of using the digital computers for document translation.
By the middle of the 60's, MT research groups had been established in many countries throughout the world, including most European countries, China, Mexico, and Japan.
During the 1980s, MT advanced rapidly, many new operational systems appeared, the commercial market for MT systems of all kinds expanded, and MT research diversified in many directions.
The use of MT accelerated in the 1990s. The commercial agencies, government services and global companies produced a large volume of translation documents.
Since the mid 90s, the Internet has exerted a powerful influence on MT development. MT software products specifically for translating Web pages and electronic mail messages have appeared.
Since about 1994, free online MT systems became available to the general public. CompuServe made a free online MT with Systran first. AltaVista was next in developing Babelfish, another well known website. They were followed by numerous other online services such as Softissimo with online versions of its Reverso systems (Hutchins, 1995).
Types of MT
Hutchins (1995) describes the three main approaches in the classical models of MT.
Direct Translation Approach is referred to as the historically oldest type of MT. This approach uses the system of one particular pair of languages (i.e. Russian as the source language and English as the target language). Translation is directly from the source language text to the target language text. Direct translation systems are normally bilingual and operate only in one direction. The Interlingua approach attempts to convert source language into an artificial language that represents meaning common to more than one language. This artificial language is called "Interlingua". Translation is thus in two stages: from the source language to the Interlingua language and from the Interlingua language to the target language.
In the transfer approach, there are three stages involved. The first stage converts the source language texts into abstract source language-oriented representations; the second stage converts these into equivalent target language-oriented representations; and the third generates the final target language texts. These stages represent the three steps of analysis, transfer, and generation (or synthesis). Rule-Based approach/Knowledge-based approach analyzes various kinds of linguistic rules : rules for syntactic analysis, lexical transfer, syntactic generation, and morphology, etc.
Example-Based Translation and Statistical MT
The Example-Based approach translates a source sentence by matching it with equivalent phrases or word groups from a databank of parallel bilingual texts. After choosing the set of phrases, the system imitates the translation of a similar sentence.
Statistical approach is based on machine-learning technologies and driven by the statistic analysis of a large and structured set of texts from a large volume of parallel human-translated texts (Steding, 2009). Free online translation programs from Google adopt this approach.
Quality of MT Output
The goal of MT systems is to produce high quality and accurate translations, yet their outputs usually require revision and post-editing. Anderson (1995, p. 68) points out that "the current major Machine Translation (MT) evaluation effort, funded by the Advanced Research Projects Agency (ARPA), shows that when compared to expert human translators, MT systems perform only about 65% as well on the average".
Basically there are three types of demand for MT use. The first is the traditional need for translations in ‘publishable’ quality. However, the users' need is often not a perfectly accurate translation but the essence of the original text that is produced quickly (Hutchins, 2010). This second use is called ‘machine translation for assimilation’ in contrast to the production of publishable-quality translations. More recently, the third application has been identified where MT is used in social interaction such as electronic mail and chartrooms, etc. (Hutchins, 2010). Ramm (1994, as cited in McCarthy, B. 2004) address that significant difference between human and machine translation as the linguistic unit on which the translation processes operate, since a human translator hardly translates a single sentence in isolation. Instead, sentences and expressions to be translated are interpreted in the context of other sentences and expressions. The situational background and cultural environment are carefully considered during the human translation process.
MT in Educational Environment
MT is a sub-field of computational linguistics and associated with both Translation and Computer Science academics. Traditionally, MT use in the classroom is associated with translator training programs. Recent development of free online translation systems and global open online educational courses broaden the MT use in educational environments.
MT for Global Online Academic Courses
The wider supply of Open Educational Resources has increased the learning opportunities for global learners around the world. On the other hand, it calls the need to localize resources among online global education communities (Ling and Chien, 2009). Ling and Chien (2009) examined the MT use for global academic purposes. Their Open courseware at MIT adopted both of machine and natural translations. Post-editing from translators and experts of academic fields were provided. Their main findings of MT translation between English and Chinese were significant time savers for translation. They suggest draft translation with computer assistance for course contents translation on the Open Courseware. In order to make the machine translated text comprehensive, examining the text with understanding of the target language is effective (Lin & Chien, 2009). The more open courseware is available in the rapid pace, the more the needs in localization would increase, in order to reduce linguistic barriers and enable equal accessibility to the academic knowledge in a timely manner. MIT offers translated OpenCourseWare courses in the eight languages (as of March, 2013).
Acikgoz and Sert’s (2006, as cited in Lin & Chien, 2009) states
" MT, having a history almost as old as the modern digital computer, emerged as an attempt to overcome the intricacy of 'being informed' in a group of offers to sustain communication".
Free Online Web Translators
Free online translation system is the most accessible form of MT. It is mostly used to get the gist of what a foreign text says, not to publish the results (Nino, 2009). In other words, free online MT is not suited for publication, because their system cannot be customized according to the language pair or the type of text and purposes (Nino, 2009). In the foreign language classroom, many students still use free online MT outputs as a language resource for their assignments, and it has caused the problems relating to plagiarism. On the other hand, some researchers propose the positive use of MT in the language classrooms.
Positive/Negative Aspects of Free Online Translators
Nino (2009) addresses positive and negative aspects of free online translators.
- Positive Aspects
- Wide Availability
- Good with lexical translation: The fact that free online MT translates short lexical units reasonably well.
- Good with repetitive, simply-structured texts: The free online MT works reasonably well with not so complex structured texts such as weather reports or technical manuals.
- Negative Aspects
- Literal Translation: Outputs of free online translators include many errors and keep original structures of source languages.
- Many grammatical inaccuracies
- Discursive inaccuracies: They produce inaccuracies in connectives and co-reference between the sentences.
- Spelling errors: They present orthographical inaccuracies such as punctuation and capitalization errors, letter omissions or unnecessary letters.
- Unable to account for cultural references
- Unnatural writing
Instructional Drawbacks of Free Online Translators in the Foreign Language Classroom
In the survey from Nino ( 2009 ) among students and language teachers’ perception on online translation systems, while the students accepted free online translator positively, the teachers perceived that in the future, when the quality of the online translation systems' output is better, they can start thinking about incorporating MT into the language class. Students' motivation for using online MT for foreign language assignment is usually due to a lack of time, energy, imagination, linguistic insight or a lack of confidence (McCathy, 2004). Students have access to free translation already. Researchers find that it is not realistic to ignore or merely hope to eliminate online MT use for students' translation assignments (McCarthy, 2004). Accordingly, online translation poses problems for formal assessment or students' language development. The instructors and researchers raise discussion point to be plagiarism in the use of MT for the assignment production. McCathy (2004) argues that in higher education, submitting an online translated work for translation assignment is against university's policy (McCathy, 2004). For prevention of MT use in the writing and translation assignment, Stedeing (2009) proposed a language instructor’s clear policy and demonstration of MT technology in the language classroom. Teachers should also be aware that a large number of online translation services very often use the same third party products, which is a reason why different website produce identical translation mistakes ( Stedeing, 2009).
MT for Language Classroom Applications
Despite the issue around online translating systems and the foreign language classrooms, some models using MT as a tool for language learning are addressed.
MT as a Computer-Assisted Language Learning (CALL)
Some academics discuss that translation is often part of foreign-language learning, and therefore, learning about MT should be part of the curriculum for language learners (Somers, 2003., Lewis, 1997., and DeCesaris, 1995) . On the other hand, MT is not designed for the purpose of language learning, and learners should be wary of using them. From the studies, Somers (2003) illustrates three models using MT for language learning.
Pre-Editing and Post-Editing
Somers (2003) reported that MT was useful for advanced students for the first draft translation into their native language. After the post editing, the students produced the improved edition with comments on the translation process. They discussed the comments to associate the errors and the general problems of MT. Alternatively, classifying of the errors with a linguistic analysis is proposed. Belam (2003, as cited in Nino, 2009) argued that post-editing can make the students focus on new vocabulary, expressions, grammar points and stylistic aspects. Kliffer (2005) pointed the value of MT post-editing for weaker students as it is less stressful than having students do the entire translation themselves.
Using MT as a Bad Model
This model uses MT's weakness and mistakes to bring out language differences between the source language and the target language in order to reinforce learners' correct grammar and style. One of MT's weaknesses and mistakes inhabit in the language differences. A study using English-Hebrew MT system reported that for translation into the learners' native language, it is a useful exercise, since poor quality translations are often close to structure of the source language. By focusing on the construction in the poor translations, students' knowledge of grammatical structure is increased (Lewis, 1997). This model is expected to prevent plagiarism by students’ awareness in the limitation of MT technology (Stendeing, 2009). Nevertheless, this approach is controversial, because it might reinforce incorrect language habits in the learners.
Translation Training Program, "TransIt-Tiger"
TransIt-Tiger is a translation-training program operated on a PC, not a translation tool. This program is designed for higher education to help language learners to accomplish linguistic competence in both their target and native language. It draws learners' attention to the task, by a two -stage activity. First, the program helps learners with their first translation with the use of glossaries, a dictionary, and prepared questions and hints. The hints focus on grammatical points and the questions direct the learner towards an appropriate solution. In the second stage, the learners are provided with two alternative translations. The students edit their first edition considering the provided two versions, and produce a final translation. Although this program was expected to increase the students' autonomy in their learning, the research on the program reported that the students did not use the program out of their own initiative (Fayard, 1999).
MT for Linguistic and Computer Science classes
MT is related to the field of linguistic and computer science, in which the use of computers to translate natural languages is studied; namely, Computational Linguistics (CL). For the students of the area, MT can be used to illustrate problems in language analysis. Exercises can be developed to raise students' awareness in weaknesses and problems of MT. MT output translation can be used for linguistic error analysis and focusing on one particular problem area.
- Anderson, D. (1995). Machine Translation As a Tool in Second Language Learning. CALICO Journal, Volume 13-1.
- Fayard, N.(1999). Integrating TransIt-TIGER French into the Second Year Language Curriculum. CALL-EJ Online 1:1. Retrieved from http://callej.org/journal/1-1/fayard.html
- Google Translate http://translate.google.com/about/
- Nino, A. (2009). Machine translation in foreign language learning: language learners’ and tutors’ perceptions of its advantages and disadvantages. ReCALL, 21, pp241-258. doi:10.1017/S0958344009000172.
- Hutchins, J.(1995). MACHINE TRANSLATION: A BRIEF HISTORY. From: Concise history of the language sciences: from the Sumerians to the cognitivists. Edited by E.F.K.Koerner and R.E.Asher. Oxford: Pergamon Press. Pages 431-445
- Hutchins, J.(2010). Machine translation: a concise history. Journal of Translation Studies, vol.13, 1-2. pp.29-70.
- Lewis, D. (2006). Machine Translation in a Modern Languages Curriculum. Computer Assisted Language Learning, 10:3, 255-271.
- Lin, G.H. &Chien, P.S. (2009). Machine Translation for Academic Purposes. Paper presented at the International Conference on TESOL and Translation. Retrieved from http://www.academia.edu/1753976/Machine_Translation_for_Academic_Purposes
- McCarthy, B. (2004). Does Online Machine Translation Spell the End of Take-Home Translation Assignments?. CALL-EJ OnlineVol. 6, No.1. Retrieved from http://callej.org/journal/6-1/mccarthy.html
- Robberecht,P. ()Critical Review of TransIt-Tiger. CALICO Journal, Volume 18 Number 2. 472-489.
- Somers, H.(2003). Machine Translation in the Classroom. in Computers and Translation. A translator's guide. John Benjamins Publishing Company. Philadelphia, PA.
- Somers, H. (2001).Three Perspectives on MT in the Classroom. Retrieved from http://www.dlsi.ua.es/tmt/docum/TMT4.pdf
- Steding, S. (2009). Machine Translation in the German Classroom: Detection, Reaction, Prevention. Teaching German, 42:2, 178-189.