Grammatical Analysis of Japanese KANA-Sentences in Old Style by Computer

Dictionaries with concept data and corpora with part-of-speech tags are important tools in natural language processing. However, there is not any such data tool for Japanese old style language. Therefore we studied a method for preparing such corpora and realized an old-Japanese corpus with 150,000 words. We performed an experiment of lexical analysis on old Japanese by statistically learning of the part-of-speech tagged corpus. In addition, we studied a method of acquiring a common concept by finding a portion where concepts concerning the words concentrated on a thesaurus. In order to estimate the method, we applied it to English-Japanese Dictionary, which is a sequence of records each having an English headword and Japanese equivalent words. The result shows some usefulness. However, it is not sufficient to get the concepts of old-Japanese words using old Japanese dictionary. This is a future study.