New generation Systran translation system

MT Summit VIII Santiago de Compostela (Spain), 18 - 22 September 2001 New Generation Systran Translation System Jean Senellart, Péter Dienes, Tamás Váradi SYSTRAN, HIS 1, rue du Cimetière - BP.7 95237, Soisy-sous-Montmorency Cedex France senellart@systran.fr, dienes@nytud.hu, varadi@nytud.hu Abstract In this paper, we present the design of the new generation Systran translation systems, currently utilized in the development of English-Hungarian, English-Polish, English-Arabic, French-Arabic, Hungarian-French and Polish-French language pairs. The new design, based on the traditional Systran machine translation expertise and the existing linguistic resources, addresses the following aspects: efficiency, modularity, declarativity, reusability, and maintainability. Technically, the new systems rely on intensive use of state-of-the-art finite automaton and formal grammar implementation. The finite automata provide the essential lookup facilities and the natural capacity of factorizing intuitive linguistic sets. Linguistically, we have introduced a full monolingual description of linguistic information and the concept of implicit transfer. Finally, we present some by-products that are directly derived from the new architecture: intuitive coding tools, spell checker and syntactic tagger. Keywords: machine translation architecture, implicit transfer, declarativity, management of linguistic resources Introduction The Systran machine translation (MT) system today, generally classified as a transfer translation system, is the result of 30-years worth of accumulation of linguistic and technical expertise. The tremendous amount of development resulted in very large multilingual linguistic resources. During this time, the system has been rewritten several times, progressively integrating the latest MT technology, for example, backtracking capacity, introduction of a fine-grained semantic description, etc. The evolution of the systems shows that the major problem inherent in Systran is not so much a question of technical or linguistic capacity, but more a question of a) maintainability of linguistic resources, b) reusability of the resources, c) global mastering of the linguistic process during translation, and d) opening of the system to the very demanding market. In today’s market, MT is facing more demanding • requirements than ever: To rapidly develop new languages and cross- language translation systems; • To provide personalized translation service, e.g., personalized dictionaries, customerizable format filters and specific grammars; • To expand the translation service to more general natural language processing, e.g., multilingual indexing, document authoring, intuitive integration with translation memory; • To provide easy access for direct terminology exchange, e.g., exchange between different systems, integration of linguistic acquisition tools. In response to the above requirements, we have initiated a thorough redesign of the Systran translation engine from the ground up. The development work is carried out in the framework of the MATCHPAD project, a 5th Framework EU funded project, which is aimed at extending the suit of Systran MT systems into Hungarian and Polish. So far, we have developed new language pairs involving Hungarian and Polish (i.e. English to Hungarian and Polish, and Hungarian and Polish to French), while the Arabic (i.e. English and French to Arabic) and new cross-language systems are in progress. Although the new technical structure is quite different from the traditional systems, the most important feature of the redesign lies in the connection of the …

Content truncated to 500 words