Current Research

Multilingual Intelligence Sharing based on Natural Language Processing

OUTLINE

We are researching natural language processing to share intelligence, especially in Asia. We have so far been developing a Japanese morphological analyzer and a machine translation system from Japanese to Uighur; we have also developed a Japanese-Uighur dictionary. We are currently studying support for translations of Japanese laws and legislation.

TOPICS

(1) Japanese Morphological Analysis

We are studying a new approach to Japanese morphological analysis using the linguistic features of Japanese and developing a simpler morphological analyzer that can deal with other Asian languages. Previous Japanese morphological analyses assumed the conjugation of Japanese verbs. But we simplified the rules for Japanese verbal conjugation by using a derivational grammar that does not assume conjugation. Based on the rules, we developed a simple Japanese morphological analyzer named MAJO.

(2) Japanese-Uighur Machine Translation

Both the Japanese and Uighur languages are agglutinative and have many similarities, such as word order. Therefore, we can get Uighur translation by merely replacing morphemes that are output from a Japanese morphological analyzer with Uighur equivalents. In addition we demonstrated that derivational grammar can deal not only with Japanese but also Uighur. So we realized a Japanese-Uighur machine translation system whose main module is identical to our morphological analyzer MAJO.

(3) Translation of Japanese Statutes

We cooperated on a government translation project of Japanese statutes. This project plans to help foreigners understand Japanese statutes and intends to standardize the translation equivalents of legal terms.
We developed Bilingual KWIC, as shown in the figure, which acquires a bilingual lexicon from a parallel corpus and displays the results in a KWIC format. Display in a KWIC format enables users to easily correct errors of word alignment and compare two or more types of equivalents. In addition, users can easily select an example from the corpus and register it in dictionaries.

FUTURE WORK

The translation project of Japanese statutes plans not only to familiarize foreign residents in Japan with them but also to introduce them to foreign countries for legal assistance. Some of these countries use languages similar to Japanese. Our current translation project only intends to make English translations. But since we think our work on Japanese-Uighur translation is applicable for such languages, we are developing a translation system from Japanese for them.
In addition, we are making an XML format for Japanese statutes and verifying the consolidation of laws to support Japanese legislation.