Current Research

Integration and Application of Knowledge and Language Information

OUTLINE

We are promoting research on knowledge and language processing for integration and applications to realize advanced information systems. So far we have developed fundamental models for representation and reasoning of knowledge with exceptions, and integration of knowledge bases, and also applied technologies for morphological analysis and machine translation based on phonology. We are now engaged in the research and development of systems that support humans by combining logical structures and meaning included in linguistic information with the knowledge that experts have acquired experientially.

TOPICS

(1) Translation Support System for Japanese Statutes

It is becoming necessary to translate Japanese statutes into foreign languages due to social and economic globalization, etc. To continuously perform translations at low cost and with high quality, we are developing a support system for translators. In particular, since a standard bilingual dictionary must be constructed that contains technical terms and wordings of statutes. we are developing technologies that automatically extract bilingual candidates from a bilingual statute corpus. We are also developing a support system for dictionary registration utilizing Bilingual KWIC, which visually shows translation correspondence between the source and object languages in KeyWord In Context (KWIC) form. Furthermore, we are developing a quality check system of translation based on standard dictionaries.

(2) E-legislation

Legislation is concerned with the drafting and management of statute documents. Although it has been carried out manually by human experts with rich knowledge about the document format and the expressions of sentences based on custom, the number of statutes is huge, and their burden is increasing. To solve the problem, we are developing a system that supports legislation, which we call e-legislation. In particular, to support works by considering the logical structures of a statute such as articles and items as units, we are constructing a XML database of structured statute documents. Moreover, since there is a non-retroactive principle in laws, not only the current versions but also the old ones have to be stored in the database. To restore the old versions, we are developing an automated consolidation system that generates a new version from the old one and an amendment statute based on analysis of amendment sentences.

(3) Construction of a Multilingual Thesaurus

We can easily find synonyms by using a multilingual thesaurus, which is helpful in various natural language processing tasks including translation. In the automated construction of a thesaurus, similarity between words is generally calculated using surface information acquired from the corpora such as the cooccurence of words.

Against it, we are developing a method that calculates similarity using probability distribution of the latent meaning of a word from the cooccurrence relation between words through such grammatical relations as between verbs and nouns.

FUTURE WORK

By further developing the above technologies, we will establish an integrated environment where all works are consistently concerned with statute translation, legislation, and other legal information processing. We seek to construct an advanced information system as social infrastructure by developing a system for the international sharing of legal information.