Studies on constructing multilingual resources

The need for multilingual resources is becoming increasingly important for various applications of natural language processing. This paper describes the issues of constructing multilingual resources and the actual experiment of generating them. The author conducted an experiment of generating several cross language databases, especially relating to Japanese, Chinese, Korean and some European languages. The idea is to create a new set of databases combining existing language resources with the aid of linguistic knowledge. The author explains the generation of the three types of databases. The first one is the database for the Japanese Katakana words expressing foreign place names which are linked to Chinese Hanzi characters. This has been realized by using the result of transliteration of English, German and French to Chinese Hanzi characters. The second is the generation of a database which maps Japanese Kanji words to Hankul characters. Any application can search the Japanese language resource database using the Hankul pronunciation as a key. The third is the database that shows the relationship of Japanese Kanji words especially proper nouns to corresponding Chinese Pinyin characters. Regarding the last one the author has already conducted an experiment on a substantial amount of data.