Machine Translation and Computer-Assisted Translation: a New Way of Translating? Author: Olivia Craciunescu E-mail: [email protected]
Author: Constanza Gerding-Salas E-mail: [email protected]
Author: Susan Stringer-O’Keeffe E-mail: [email protected] Source: http://www.translationdirectory.com
The Need for Translation IT has produced a screen culture that Technology tends to replace the print culture, with printed documents being dispensed Advances in information technology (IT) with and information being accessed have combined with modern communication and relayed directly through computers requirements to foster translation automation. (e-mail, databases and other stored The history of the relationship between information). These computer documents technology and translation goes back to are instantly available and can be opened the beginnings of the Cold War, as in the and processed with far greater fl exibility 1950s competition between the United than printed matter, with the result that the States and the Soviet Union was so intensive status of information itself has changed, at every level that thousands of documents becoming either temporary or permanent were translated from Russian to English and according to need. Over the last two vice versa. However, such high demand decades we have witnessed the enormous revealed the ineffi ciency of the translation growth of information technology with process, above all in specialized areas of the accompanying advantages of speed, knowledge, increasing interest in the idea visual impact, ease of use, convenience, of a translation machine. Although the and cost-effectiveness. At the same time, Cold War has now ended, and despite the with the development of the global importance of globalization, which tends to market, industry and commerce function break down cultural, economic and linguistic more than ever on an international scale, barriers, translation has not become obsolete, with increasing freedom and fl exibility because of the desire on the part of nations in terms of exchange of products and to retain their independence and cultural services. The nature and function of identity, especially as expressed through translation is inevitably affected by these their own language. This phenomenon can changes. There is the need for countries clearly be seen within the European Union, to cooperate in many spheres, such as where translation remains a crucial activity. ecological (Greenpeace), economic The Internet with its universal access to (free trade agreements) humanitarian information and instant communication (Doctors without Borders) and between users has created a physical and educational (exchange programs), etc. geographical freedom for translators that Despite the importance of English, there was inconceivable in the past. is the commonly-held belief that people have • Do the new technologies threaten the the right to use their own language, yet the livelihood of the translator? diversity of languages should not be an • Does automation imply the obstacle to mutual understanding. Solutions disappearance of translation as we to linguistic problems must be found in know it? order to allow information to circulate freely A Short History of Machine and to facilitate bilateral and multilateral Translation relationships. Thus different aspects of modern life have It was not until the twentieth century led to the need for more effi cient methods of that the idea of creating automatic translation. At the present time the demand dictionaries appeared as a solution for translations is not satisfi ed because to the problem of linguistic barriers. there are not enough human translators, In the 1930s two researchers worked or because individuals and organizations independently towards the same goal: do not recognize translation as a complex the Franco-Armenian George Artsrouni activity requiring a high level of skill, and and the Russian Petr Smirnov-Troyanskii. are therefore not prepared to pay what it is The latter was the more important of the worth. In other words, translation is sometimes two because he developed the idea that avoided because it is considered to be too three stages are necessary for a system of expensive. In part, human translation is automatic translation: fi rst an editor who expensive because the productivity of a knows the source language analyzes the human being is essentially limited. Statistics words and converts them into base forms vary, but in general to produce a good according to their syntactic functions; translation of a diffi cult text a translator then a machine organizes the base forms cannot process more than 4-6 pages or into equivalent sequences in the target 2,000 words per day. The economic language; fi nally, this rough version is necessity of fi nding a cheaper solution to corrected by a second editor, familiar international exchange has resulted in with the target language. Despite the continuing technological progress in terms signifi cance of Troyanskii’s work, it of translation tools designed to respond remained generally unknown until the to the translator’s need for immediately- late 1950s. available information and nonsequential The invention of the computer led access to extensive databases. very quickly to attempts to use it for This paper aims at examining the new the translation of natural languages. technologies (machine translation, A letter from Warren Weaver to the electronic dictionaries, terminology computer specialist Norbert Wiener in databases, bilingual texts, grammatical March 1947 is considered to mark the concordances, and translation memories) in beginning of this process. Two years order to determine whether they change the later, in July 1949, Weaver publicized relationship between the translator and the his ideas on the applications of the texts, and if so, then in what way. We will computer to translation and shortly try to answer the following questions: afterwards a number of universities in the United States initiated research • Which computer tools are genuinely into the fi eld of machine translation. In useful to translators? 1954 the fi rst feasibility trial was carried objectivity and vision, the ALPAC report out as a joint project between IBM and led to a freeze on research into machine the University of Georgetown. Although translation in the US for more than a very limited in scope, the demonstration decade. However, research continued in was considered a success, leading to the Canada, France and Germany and two fi nancing of other projects, both in the US machine translation systems came into and the rest of the world. The fi rst versions being several years later: Systran, used of machine translation programs were by the European Union Commission and based on detailed bilingual dictionaries Taummétéo, created by the University of that offered a number of equivalent words Montreal to translate weather forecasts in the target language for each word listed from French to English. in the source language, as well as a series Important advances occurred during of rules on word order. The complexity of the 1980s. The administrative and the task made it necessary for developers commercial needs of multilingual to continue improving the programs communities stimulated the demand for because of the need for a more systematic translation, leading to the development syntactical focus. Projects were based on in countries such as France, Germany, advances in linguistics, especially on the Canada and Japan of new machine development of transformational generative translation systems such as Logos (from grammar models that appeared to offer German to French and vice versa) and new possibilities for machine translation. the internal system created by the Pan- However, initial optimism soon disappeared. American Health Organization (from Researchers began to think that the semantic Spanish to English and vice versa), as barriers were insurmountable and no longer well as a number of systems produced by saw a solution on the near horizon to the Japanese computer companies. Research problem of machine translation. IBM and also revived in the 1980s because large- the University of Washington produced an scale access to personal computers and operating system called Mark II, but the word-processing programs produced results were disappointing. By 1964 the US a market for less expensive machine government was becoming so concerned translation systems. Companies such about the ineffi ciency of machine translation as ALPS, Weidner, Globalink (North programs that it created the ALPAC America and Europe), (Automatic Language Processing Advisory Sharp, NEC, Mitsubishi, Sanyo (Japan) Committee) to evaluate them. In 1966 this needed these programs. Some of the most committee produced a highly critical report important projects were GETA-Ariane that claimed that machine translation was (Grenoble), SUSY (Saarbrücken), MU slow, ineffi cient and twice as expensive as (Kyoto), and Eurotra (the European Union) human translation, concluding that it was The beginning of the 1990s saw vital not worth investing money in research in developments in machine translation this fi eld. Nevertheless, the report stressed with a radical change in strategy from the need to encourage the development translation based on grammatical rules to of tools to assist the translation process, that based on bodies of texts and examples such as computer dictionaries, databases (for example, the Reverso Program). etc. Although criticized for its lack of Language was no longer perceived as a static entity governed by fi xed rules, but as At the beginning of the 1990s the a dynamic corpus that changes according translation market was as follows (Loffl er- to use and users, evolving through time and Laurian, 1996): adapting to social and cultural realities. To this day machine translation continues to HUMAN MACHINE progress. Large companies are now using TRANSLATION TRANSLATION it more, which also increases software Europe & 300 million 2.5 million the United sales to the general public. This situation pages pages has led to the creation of on-line machine States 150 million 3.5 million translation services such as Altavista, which Japan offer rapid email services, web pages, etc. pages pages in the desired language, as well as to the availability of multilingual dictionaries, It can be seen that only 6 million pages encyclopaedias, and free, direct-access were translated through machine terminology databases. translation, compared with 450 million through human translation, i.e. MT The Translation Market represented only 1.3% of the total. Market The development of machine translation is analysts predict that this percentage will based on supply and demand. On the one not change radically by 2007. They hand, there is new technology available, say that machine translation will remain and on the other, political, social and only about 1% of an over US $10 billion economic need for change. Yet, despite the translation marketplace (Oren, 2004). advances, machine translation still represents The languages for which there was most only a tiny percentage of the market. translation demand in 1991 were:
EN JP FR DE RU ES Others
As source lang 48% 32% 8% 5% 2% --- 5%
As target lang 45% 24% 12% --- 5% 10% 4% As expected, English dominates the market. The importance of Japanese refl ects the role of Japan in technology and foreign trade, which accounted for two-thirds of translation volume at the end of the 1990s:
Foreign Business IT Science Teaching Literature Journals Trade Administration
40% 25% 10% 10% 5% 5% 5%
At this stage it is important to make a most important of which are terminology distinction between two terms that are databases and translation memories. In closely related and that tend to confuse effect, the computer offers a new way of non-specialists: machine translation approaching text processing of both the (MT) and computer-assisted translation source and target text. Working with a (CAT). These two technologies are the digital document gives us nonsequential consequence of different approaches. access to information so that we can use it They do not produce the same results, and according to our needs. It becomes easy to are used in distinct contexts. MT aims at analyze the sentences of the source text, to assembling all the information necessary verify the context in which a word or a text for translation in one program so that is used, or to create an inventory of terms, a text can be translated without human for example. Likewise, any part of the intervention. It exploits the computer’s target text can be modifi ed at any moment capacity to calculate in order to analyze and parallel versions can be produced the structure of a statement or sentence for comparison and evaluation. All these in the source language, break it down aspects have profound implications for into easily translatable elements and translation, especially in terms of assessing then create a statement with the same the results, since the translator can work structure in the target language. It uses in a more relaxed way because of the huge plurilingual dictionaries, as well as greater freedom to make changes at any corpora of texts that have already been time while the work is in progress. translated. As mentioned, in the 1980s It is important to stress that automatic MT held great promises, but it has been translation systems are not yet capable of steadily losing ground to computer-assisted producing an immediately useable text, translation because the latter responds as languages are highly dependant on more realistically to actual needs. context and on the different denotations CAT uses a number of tools to help the and connotations of words and word translator work accurately and quickly, the combinations. It is not always possible to provide full could produce a target text of the context within the text itself, so that machine same quality as that of a human being. translation is limited to concrete situations However, it is clear that even a human and is considered to be primarily a means translator is seldom capable of producing of saving time, rather than a replacement a polished translation at fi rst attempt. In for human activity. It requires post-editing in reality the translation process comprises order to yield a quality target text. two stages: fi rst, the production of a rough text or preliminary version in the target Cognitive Processes language, in which most of the translation To understand the essential principles problems are solved but which is far from underlying machine translation it is being perfect; and second, the revision necessary to understand the functioning of stage, varying from merely re-reading the human brain. The fi rst stage in human the text while making minor adjustments translation is complete comprehension of the to the implementation of radical changes. source language text. This comprehension It could therefore be said that MT aims at operates on several levels: performing the fi rst stage of this process • Semantic level: understanding words out in an automatic way, so that the human of context, as in a dictionary. translator can then proceed directly to • Syntactic level: understanding words in a the second, carrying out the meticulous sentence. and demanding task of revision. The • Pragmatic level: understanding words in problem is that the translator now faces situations and context. a text that has not been translated by a human brain but by a machine, which Furthermore, there are at least fi ve types of changes the required approach because knowledge used in the translation process: the errors are different. It becomes necessary to harmonize the machine • Knowledge of the source language, which version with human thought processes, allows us to understand the original text. judgements and experiences. Machine • Knowledge of the target language, which translation is thus both makes it possible to produce a coherent text an aid and a trap for translators: an in that language. aid because it completes the fi rst stage • Knowledge of equivalents between the of translation; a trap because it is not source and target languages. always easy for the translator to keep • Knowledge of the subject fi eld as well the necessary critical distance from a as general knowledge, both of which aid text that, at least in a rudimentary way, comprehension of the source language is already translated, so that mistakes text. may go undetected. In no sense should • Knowledge of socio-cultural aspects, that a translation produced automatically be is, of the customs and conventions of the considered fi nal, even if it appears on the source and target cultures. surface to be coherent and correct. Machine Translation Strategies Given the complexity of the phenomena that underlie the work of a human translator, it Machine translation is an autonomous would be absurd to claim that a machine operating system with strategies and approaches that can be classifi ed as There are a number of systems that follows: function on the same principle: for example SPANAM, used for Spanish- • the direct strategy English translation since 1980, and • the transfer strategy SYSTRAN, developed in the United • the pivot language strategy States for military purposes to translate Russian into English. After modifi cation The direct strategy, the fi rst to be used in designed to improve its functioning, machine translation systems, involves a SYSTRAN was adopted by the European minimum of linguistic theory. This approach Community in 1976. At present it can be is based on a predefi ned source language- used to translate the following European target language binomial in which each languages: word of the source language syntagm is • Source languages: English, French, directly linked to a corresponding unit in German, Spanish, Italian, Portuguese, the target language with a unidirectional and Greek. correlation, for example from English to • Target languages: English, French, Spanish but not the other way round. The German, Spanish, Italian, Portuguese, best-known representative of this approach Greek, Dutch, Finnish, and Swedish. is the system created by the University of In addition, programs are being created Georgetown, tested for the fi rst time in 1964 for other European languages, such as on translations from Russian to English. The Hungarian, Polish and Serbo-Croatian. Georgetown system, like all existing systems, is based on a direct approach with a strong Apart from being used by the European lexical component. The mechanisms for Commission, SYSTRAN is also used by morphological analysis are highly developed NATO and by Aérospatiale, the French and the dictionaries extremely complex, but aeronautic company, which has played the processes of syntactical analysis and an active part in the development of the disambiguation are limited, so that texts system by contributing its own terminology need a second stage of translation by human bank for French-English and English- translators. The following is an example French translation and by fi nancing the that follows the direct translation model: specialized area related to aviation.
Source language text La jeune fi lle a acheté deux livres Breakdown in source language La jeune fi lle acheter deux livre Lexical Transfer The young girl buy two book Adaptation in target language The young girl bought two books Outside Europe, SYSTRAN is used by de Montréal (University of Montreal The United States Air Force because of its Machine Translation) was created by the interest in Russian-English translation, by Canadian Government in 1965. It has the XEROX Corporation, which adopted been functioning to translate weather machine translation at the end of the 1970s forecasts from English to French since and which is the private company that has 1977 and from French to English since contributed the most to the expansion of 1989. One of the oldest effective systems machine translation, and General Motors, in existence, TAUMMÉTÉO carries out which through a license from Peter Toma is both a syntactic and a semantic analysis allowed to develop and sell the applications and is 80% effective because weather of the system on its own account. It should forecasts are linguistically restricted and be noted that in general the companies that clearly defi ned. It works with only 1,500 develop direct machine translation systems lexical entries, many of which are proper do not claim that they are designed to nouns. In short, it carries out limited produce good fi nal translations, but rather repetitive tasks, translating texts that are to facilitate the translator’s work in terms highly specifi c, with a limited vocabulary of effi ciency and performance (Lab, p.24). (although it uses an exhaustive dictionary) The transfer strategy and stereotyped syntax, and there is focuses on the concept of “level of perfect correspondence from structure to representation” and involves three stages. structure. The analysis stage describes the source The pivot language strategy is based document linguistically and uses a source on the idea of creating a representation language dictionary. The transfer stage of the text independent of any particular transforms the results of the analysis stage language. This representation functions and establishes the linguistic and structural as a neutral, universal central axis equivalents between the two languages. that is distinct from both the source It uses a bilingual dictionary from source language and the target language. In language to target language. The generation theory this method reduces the machine stage produces a document in the target translation process to only two stages: language on the basis of the linguistic data analysis and generation. The analysis of the source language by means of a target of the source text leads to a conceptual language dictionary. representation, the diverse components The transfer strategy, developed by GETA of which are matched by the generation (Groupe d’Etude pour la Traduction module to their equivalents in the target Automatique / Machine Translation Study language. The research on this strategy Group) in Grenoble, France, led by is related to artifi cial intelligence and the B. Vauquois, has stimulated other research representation of knowledge. The systems projects. Some, such as the Canadian based on the idea of a pivot language do TAUMMÉTÉO and the American METAL, not aim at direct translation, but rather are already functioning. Others are still at reformulate the source text from the the experimental stage, for example, SUSY essential information. At the present time in Germany and EUROTRA, which is a the transfer and pivot language strategies joint European project. TAUM, an acronym are generating the most research in for Traduction Automatique de l’Université the fi eld of machine translation. With regard to the pivot language strategy, it is most important aspects of MT, it allows worth mentioning the Dutch DLT (Distributed translators to concentrate on producing Language Translation) project which ran from a high-quality target text. Perhaps 1985 to 1990 and which used Esperanto then “machine translation” is not an as a pivot language in the translation of 12 appropriate term, since the machine only European languages. completes the fi rst stage of the process. It It should be repeated that unless the systems would be more accurate to talk of a tool function within a rigidly defi ned sphere, as that aids the translation process, rather is the case with TAUM-MÉTÉO, machine than an independent translation system. translation in no way offers a fi nished The following is a relatively recent product. As Christian Boitet, director of classifi cation of some MT programs based GETA (Grenoble) says in an interview on the results obtained from a series of tests given to the journal Le français dans le that focused on errors and intelligibility monde Nº314 in which he summarizes the in the target texts (Poudat, p.51):
Translator Address Characteristics
Alphaworks® www.alphaworks.ibm.com Translates English into seven languages; transfer method 3
E-lingo® www.elingo.com Twenty pairs of languages available; transfer method 2 Reverso® www.trans.voila.fr Thirteen pairs of languages available; transfer method 1
Systran® www.systransoft.com Twelve pairs of languages available; direct transfer method 4
Analysis of Some Errors in Machine- SOURCE TEXT: Le Monde Diplomatique, translated Texts September 2002 For the purpose of analyzing errors in Depuis le 11 septembre 2001, l’esprit machine-translated texts, it is revealing to guerrier qui souffl e sur Washington semble compare such a translation with that done avoir balayé ces scrupules. Désormais by a human translator. An article from Le comme l’a dit le président George W. Monde Diplomatique has been chosen, as Bush, “qui n’est pas avec nous est avec les this is a newspaper that is originally written terroristes”. in French but which is then translated into 17 other languages. In this case we will compare the French to English translations produced by Systran, Reverso and a human translator. Systran Reverso Human translation
Since September 11, 2001, the warlike Since September 11, 2001, the warlike Since 11 September 2001 the spirit which blows on Washington seems spirit which blows on Washington warmongering mood in Washington to have swept these scruples. From now seems to have swept (annihilated) seems to have swept away such on, like said it the president George W these scruples. Henceforth, as said it scruples. From that point, as President the president George W. Bush, “which George Bush put it, “either you are Bush, “which is not with us is with the (who) is not with us is with the terrorists”. with us or you are with the terrorists.” terrorists”. (37 words) (35 +2 words) (36words)
The fi rst point to be made is that MT is a Although Reverso’s translation is translation method that focuses on the not completely correct, it translates source language, while human translation comme into “as”, which is the aims at comprehension of the target correct choice for this context. language. Machine translations are therefore often inaccurate because they • Errors in usage take the words from a dictionary and follow the situational limitations set by the The translation is understandable in that program designer. Various types of errors the MT produces the meaning but does can be seen in the above translations. not respect usage:
• Errors that change the meaning Original: semble avoir balayé ces of the lexeme scrupules
1. Words or phrases that are apparently Systran: seems to have swept these correct but which do not translate the scruples meaning in context: Reverso: seems to have swept (annihilated) Original: l’esprit guerrier these scruples
Systran: the warlike spirit HT: seems to have swept away such scruples Reverso: the warlike spirit Original: qui n’est pas avec nous est avec HT: the warmongering mood les terroristes
2. Words without meaning: Systran: which is not with us is with the terrorists Original: comme l’a dit le président Reverso: which (who) is not with us is George W. Bush with the terrorists
Systran: like said it the president George HT: either you are with us or with the W. Bush terrorists
Reverso: as said it the president George As already mentioned, human translation W. Bush concentrates on the target language, preferring to depart from the source HT: as President George Bush put it language, if necessary, in order to reproduce meaning. For example, the human and transform it according to the needs of translator clearly chose “the warmongering the specifi c task. Thus computer-assisted mood in Washington” as a better contextual translation gives the translator on-the-spot translation of l’esprit guerrier qui souffl e sur fl exibility and freedom of movement, Washington than the more literal versions together with immediate access to seen in the machine translations. an astonishing range of upto-date information. The result is an enormous Because MT aims primarily at comprehension saving of time. and not at the production of a perfect target The following are the most important text, it is important to follow two basic rules computer tools in the translator’s in order to make the best use of programs. workplace, from the most elementary to First, we need to recognize that certain types the most complex: of texts, such as poetry, for example, are Electronic Dictionaries, Glossaries not suitable for MT. Second, it is essential to and Terminology Databases correct the source text, as even one letter can radically change Consulting electronic or digital dictionaries meaning, as in the following example: We on the computer does not at fi rst appear shook hand translates into “Nous avons radically different from using paper secoué la main”; but We shook hands dictionaries. However, the advantages becomes “Nous nous sommes serrés la soon become clear. It takes far less time main”. The omission of an s in the source text to type in a word on the computer and is enough to make the machine translation receive an answer than to look through incomprehensible. It is of additional interest a paper dictionary; there is immediate to note that the fi nal s of serrés access to related data through is a mistake because the MT program does links; and it is possible to use several not take into account the subtleties of French dictionaries simultaneously by working grammar with regard to the agreement of with multiple documents. the past participle. Electronic dictionaries are available in Computer-assisted Translation several forms: as software that can be In practice, computer-assisted translation is installed in the computer; as CD-ROMs a complex process involving specifi c tools and, most importantly, through the Internet. and technology adaptable to the needs The search engine Google, for example, of the translator, who is involved in the gives us access to a huge variety of whole process and not just in the editing monolingual and bilingual dictionaries in stage. The computer becomes a workstation many languages, although it is sometimes where the translator has access to a variety necessary to become on-line subscribers, of texts, tools and programs: for example, as with the Oxford English Dictionary. monolingual and bilingual dictionaries, On-line dictionaries organize material for parallel us from their corpus because they are not texts, translated texts in simply a collection of words in isolation. a variety of source and target languages, For example, we can ask for all words and terminology databases. Each translator related to one key word, or for all words can create a personal work environment that come from a particular language. That is to say, they allow immediate cross- but provide an additional method of access to information. handling texts for translation. They are For help with specifi c terminology there is word-processing programs that produce a wide range of dictionaries, glossaries a list of all the occurrences of a string of and databases on the Internet. Le Nouveau letters within a defi ned corpus with the Grand Dictionnaire Terminologique objective of establishing patterns that are developed in Quebec, Canada contains otherwise not clear. These letters may 3 million terms in French and English form part of a word, such as a prefi x or belonging to 200 fi elds. Another important suffi x for example, or a complete word, resource is EURODICAUTOM, a multilingual or a group of words. Specifi c functions terminology database created by the of concordances include giving statistical European Union in 1973 that covers a variety data about the number of words or of specialized areas, both scientifi c and non- propositions, classifying words etc. scientifi c (the list begins: Agriculture, Arts, in terms of frequency or alphabetical Automation...). In addition, there are web order and, most importantly perhaps, sites that offer information on terminology identifying the exact context in which that is helpful to translators. One such site is the words occur. Information can be that of the TERMISTI research center attached accumulated and stored as more texts to the Higher Institute for Translators and are translated, producing a database Interpreters (ISTI) in Brussels available for consultation at any time in (http://www.termisti.refer.org) a non-sequential way. which provides information on the following: Concordances are particularly valuable for translating specialized texts with fi xed • Dictionaries available on Internet such as vocabulary and expressions that have a those mentioned. clearly defi ned meaning. They ensure • Terminology networks such as terminological consistency, providing the RIFAL (Réseau international francophone translator with more control over the text, d’aménagement linguistique), RITERM irrespective of length and complexity. (Red Iberoamericana de Terminología) However, they are not so helpful to literary • European terminology projects such as translators, who are constantly faced Human Language Technologies, with problems relating to the polysemic Information Society Technologies. and metaphorical use of language. • Translation Schools Nevertheless, some literary translators • Forums and diffusion/discussio n lists use concordances as they clearly have a • Conferences potential role in all kinds of translation. • Journals such as the International On-line Bilingual Texts Journal of Lexicography, La banque des mots, L’actualité terminologique, A bilingual corpus normally consists of a Méta, Terminogramme, Terminologies source text plus its translation, previously nouvelles, Terminology, Terminometro, carried out by human translators. This Translation Journal, Apuntes. type of document, which is stored Concordances electronically, is called a bi-text. It facilitates later translations by supplying Computer concordances do not replace ready solutions to fi xed expressions, tools such as dictionaries and glossaries, thus automating part of the process. The growth of the translation market has A translation memory can be used in two led to increased interest on the part of ways: companies and international organizations in collections of texts or corpora in different 1. In interactive mode: The text to languages stored systematically on-line and be translated is on the computer screen available for immediate consultation. and the translator selects the segments one by one to translate them. After Translation Memories each selection the program searches its Translation memories represent one of memory for identical or similar segments the most important applications of on-line and produces possible translations in a bilingual texts, going back to the beginning separate window. The translator accepts, of the 1980s with the pioneering TSS system modifi es or rejects the suggestions. of ALPS, later Alpnet. This was succeeded 2. In automatic mode: The program at the beginning of the 90s by programs automatically processes the whole such as Translator Manager, Translator’s source-language text and inserts into Workbench, Optimizer, Déjà Vu, Trados the target-language text the translations and Eurolang, among others. In its simplest it fi nds in the memory. This is a more form, a translation memory is a database useful mode if there is a lot of repetition in which a translator stores translations because it avoids treating each segment for future reuse, either in the same text or in a separate operation. other texts. Basically the program records bilingual pairs: a source-language segment A translation memory program is normally (usually a sentence) combined with a made up of the following elements: target-language segment. If an identical or similar source-language segment comes a. A translation editor, which protects up later, the translation memory program the target text format. will fi nd the previously-translated segment and automatically suggest it for the new b. A text segment localizer. translation. The translator is free to accept it without change, or edit it to fi t the c. A terminological tool for dictionary current context, or reject it altogether. Most management. programs fi nd not only perfect matches but also partially-matching segments. This d. An automatic system of analysis for computer-assisted translation tool is most new texts. useful with texts possessing the following characteristics: e. A statistical tool that indicates the • Terminological homogeneity: The number of words translated and to be meaning of terms does not vary. translated, the language, etc. • Phraseological homogeneity: Ideas or actions are expressed or Thus translation memory programs are described with the same words based on the accumulation and storing of • Short, simple sentences: These knowledge that is recycled according to increase the probability of repetition need, automating the use of terminology and reduce ambiguity. and access to dictionaries. When translation tasks are repeated, memories a decline in the quality of production. save the translator valuable time and even Some translators totally reject machine physical effort: for example, keyboard use translation because they associate it with can be reduced by as much as 70% with the point of view that translation is merely some texts. Memories also simplify project one more marketable product based management and team translation by on a calculation of investment versus ensuring consistency. However, translation profi ts. They defi ne translation as an art memories can only deal with a text that possesses its own aesthetic criteria simplistically in terms of linguistic segments; that have nothing to do with profi t and they cannot, unlike the human translator, loss, but are rather related to creativity have a vision of the text as a whole with and the power of the imagination. This regard to ideas and concepts or overall applies mostly, however, to specifi c kinds message. A human translator may choose to of translation, such as that of literary rearrange or redistribute the information in texts, where polysemy, connotation and the source text because the target language style play a crucial role. It is clear that and culture demand a different content computers could not even begin to replace relationship to create coherence or facilitate human translators with such texts. Even comprehension. Another disadvantage of with other kinds of texts, our analysis memories is that training time is essential of the roles and capabilities of both MT for effi cient use and even then it takes time and CAT shows that neither is effi cient to build up an extensive database and accurate enough to eliminate the i.e. they are not immediate time-savers necessity for human translators. In fact, straight out of the box. Finally, it should be so-called machine translation would be stressed that translation memory programs more accurately described as computer- are designed to increase the quality assisted translation too. Translators and effi ciency of the translation process, should recognize and learn to exploit the particularly with regard to specialized texts potential of the new technologies to help with non-fi gurative language and fi xed them to be more rigorous, consistent and grammatical constructions, but they are not productive without feeling threatened. designed to replace the human translator. Some people ask if the new technologies have created a new profession. It could Conclusion: The Impact of the New be claimed that the resources available Technologies on Translators to the translator through information technology imply a change in the It has long been a subject of discussion relationship between the translator and whether machine translation and computer- the text, that is to say, a new way of assisted translation could convert translators translating, but this does not mean that into mere editors, making them less the result is a new profession. However, important than the computer programs. The there is clearly the development of new fear of this happening has led to a certain capabilities, which leads us to point out a rejection of the new technologies on the number of essential aspects of the current part of translators, not only because of a situation. Translating with the help of possible loss of work and professional the computer is defi nitely not the same prestige, but also because of concern about as working exclusively on paper and with paper products such as conventional dictionaries, because computer tools provide DISTRIBUTED LANGUAGE us with a relationship to the text which is TRANSLATION (DLT) much more fl exible than a purely lineal www.fb10.uni-bremen.de/linguistik/ reading. Furthermore, the Internet with its khwagner/mt /ppt/DLT.ppt universal access to information and instant (consulted 19 Oct. 2003). communication between users has created a physical and geographical freedom for HUTCHINS, W. John. Translation. translators that was inconceivable in the Technology and the Translator. past. We share the conviction that translation URL: www.bcs.org.uk/siggroup/nalatran/ has not become a new profession, but the nalamtr7/nalamt76.htm changes are here to stay and will continue (consulted 9 May 2002). to evolve. Translators need to accept the HUTCHINS, W. John. Computer-based new technologies and learn how to use Translation Systems and Tools. them to their maximum potential as a URL: www.eamt.org/archive/hutchins_intro. means to increased productivity and quality html (consulted 8 July 2002). improvement. HUTCHINS, W. John. The State of Machine References Translation in Europe and Future Prospects. ARNOLD, Doug, BALKAN, Lorna et al. URL: www.hltcentral.org/page- Machine Translation: An Introductory Guide. RICO PEREZ, Celia. New Trends in URL: http://clwww.essex.ac.uk.MTbook/HTML/ Machine Translation, Meta, XLII, 4, 1997. (consulted 13 May 2002). URL: www.erudit.org/erudit/meta/v42n 04 (consulted 9 May 2002). BEDARD, Claude. Ce qu’il faut savoir sur les mémoires de traduction. RICO PEREZ, Celia. From Novelty to URL: www.terminotix.com/Terminotixfr/info/ Ubiquity: Computers and Translation at art_5html (consulted 17 June 2002). the Close of the Industrial Age. URL: http://accurapid.com/journal/15mt2.htm BERTRAND-GASTALDY, Suzanne. (consulted 9 May, 2002). La traduction automatique ou assistée par ordinateur. RICO PEREZ, Celia. Universals of URL: www.esi.umontreal.ca/~moalv/blt 6134/ Language, Sublanguage and Artifi cial index.html (consulted 6 June 2002). Languages: May the Machine Compete with the Translator? CHAMPOLLION, Yves. Machine URL: www.ccl.umist.ac.uk/events/ Translation (MT and Future of the conference/multimedia1.htm Translation Industry. (consulted 18 May, 2002). URL: http://accurapid.com/journal/15mt.htm (consulted 2 May 2002). ROSEN, Mariam et TSAO, Claude. Demain, la traduction automatique. DELISLE, Cynthia. La traduction URL: www.mondediplomatique.fr/mav/ automatique: espoirs ou déboires? Internet/ROSEN/traduction.html URL: www.cafi .org/colloques/compterendu- (consulted 19 July, 2002). mard-2000.html (consulted 14 June 2002). SAMUELSON-BROWN, G. A Practical Guide for Translators (2nd Ed.). London: Multilingual Matters Ltd, 1995.
SEVERINI, Alfi ero. Mise au point d’un système de traduction automatique italien français. Université Paris 13, 2001. Traduction—terminotique -Lexicographie. URL: http://www.owil.org/traduc.htm (consulted 28 July, 2002).
TRUJILLO, A. Translation engines: techniques for machine translation. London: Springer, 1999. URL: http://www.ccl.umist.ac.uk/staff/iat/ transeng (consulted 30 Sept.2003).
TRUJILLO, A. Estrategias de traducción automática. URL: http://www.imim.es/quark/ num19/019053.htm (consulted 13 Oct. 2003).
ZIMINA, Maria. Alignement de textes bilingues par classifi cation ascendante hiérarchique. URL: http://www.cavi.univ-paris3.fr/ lexicometrica/jadt/jadt20 00/pdf/77/77.pdf (consulted 3 August 2002).