The history of the relationship between translation and technology goes back to 1950s, when the Cold War urged the
United States and the Soviet Union to translate thousands of documents from Russian to English and vice versa.
Today, machine translation (MT) is a reality of social, commercial and scientific importance. The social importance of MT is realized mainly in communities where more than one language is spoken, such as the European Union. In such communities MT allows for the cultural identity of all states, as expressed through their languages, to be retained. Commercially, MT is important as it reduces delays and provides a, cheaper alternative to the payment of high-skilled human translators. Finally, the scientific importance of MT is obvious, as it provides the testing ground for many new ideas in Artificial Intelligence and Computer Science.
There are several misconceptions surrounding machine translation such as that MT system are useless since the quality of translation provided is very low, or that MT systems threaten the job of human translators. Both misconceptions are because very few of those who are involved in producing translations are fully aware of what is involved in MT systems at a practical level or what is technically feasible today. Therefore, it is considered necessary to discuss the role of MT systems in the translation process.
A MT system or software translates text in one natural language (source language) into another natural language (target language), taking into account the grammatical structure, vocabulary and terminology of both languages and using rules to transfer the information of the source language accurately in the target language. Even the most reliable MT system, however, cannot yet produce immediately usable target texts, as it lacks the knowledge of socio-cultural customs and conventions of the source and the target language. This is the reason why the automatic translation output needs to be revised by a human translator. In such a case, the term Human Assisted Machine Translation (HAMT) is appropriate. Human post-editing of machine translation ensures that the TL content is unambiguous, relevant to the target culture and free of any errors.
On the other hand, MT systems support the human translator with useful tools, such as electronic dictionaries, glossaries, and translation memories, which enable him/her to provide high-quality translations in less time and with less effort than before. In such cases, the term Machine Assisted Human Translation (MAHT) is appropriate.
Therefore, MT plays a particular role in the translation process, either by providing a preliminary version of the target text or by providing the tools that help a human translator refine his/her translation output. Thus, it is obvious that a successful translation product usually comes from the collaboration of both human and machine translators and, therefore, the misconceptions set before are not plausible.
The MT field is a mixture of practice and research. Practical operational systems apply well tested methods both in large scale and smaller systems. Three main approaches have been used in MT systems: the direct strategy, the transfer strategy and the interlingua strategy. Each strategy has provided the basis on which various modern MT systems function.
The direct strategy involves a minimum of linguistic theory. Each word of the source language is linked to a corresponding unit in the target language with a unidirectional correlation, but not the other way round. Companies that develop direct MT systems claim that these systems are designed to assist the translator’s work in terms of performance and efficiency and not to provide final translations.
The transfer strategy is concentrated to the level of representation and it involves three stages. In the linguistic analysis stage a source language dictionary isused to describe the source document. In the transfer stage the linguistic and structural equivalents between the two languages are established. Finally, in the generation stage a target language text is produced with the help of a target language dictionary.
The interlingua strategy involves the analysis of the source text and the development of a representation, which serves as a projection of the source text and a basis for the generation of the target text. This representation is dissimilar from both source and target language. The MT systems that have been developed based on this strategy do not aim at direct translation, but rather reformulate the source language text from its essential information.
In Europe, there are several computer-assisted translation tools, providing free public access or having restricted-access custom-built software, that work with specific language pairs. Such tools include electronic dictionaries, on-line bilingual texts, translation memories and terminology databases. MT Systems and tools assisting in translation of many European languages are presented below:
1. MT Systems: Systran, which was adopted by the European Community in 1976, can be used to translate from/to 7 European languages
2. Real-time Translation Services: AltaVista''s Babel Fish can be used for translations from/to 8 European languages as well as other languages
3. Terminology Databases: CELEX, produced by the European Commission, can be used to translate from/to 11 European languages
4. Translation memories: Trados’ Translation Workbench can be used to translate from/to 58 languages, including all European ones. Déjà vu can be used to translate almost from/to all languages, including bi-directional ones (Arabic and Hebrew) and East Asian double-byte ones. Eurolang’s Optimizer supports at the moment, English, French, German, and Italian as source languages and twenty languages for target output.