Development of a Multilingual Information Retrieval and Check System Based on Database Semantic

Technical documents for multilateral agreements or international business transactions are normally produced in a bilingual or multilingual form. Being mostly of legal nature, these documents require especially accurate and speedy translations by expert translators. In order to aid these experts, automatic ways of checking translation results (such as a spelling checker) would be highly desirable. This paper describes the MIRAC system for Multilingual Information Retrieval And Checking. It is designed to find translation errors in multilingual documents, and to evaluate the overall results of translation. Unlike a machine translation or a translation memory system [Volk98, Webb98], the primary function of the MIRAC system is to evaluate previously translated and aligned documents in source and target languages, while dynamically building a database that consists of aligned multilingual texts carrying semantically equivalent content.