Article ID: | iaor1997591 |
Country: | Japan |
Volume: | 37 |
Issue: | 6 |
Start Page Number: | 1007 |
End Page Number: | 1016 |
Publication Date: | Jun 1996 |
Journal: | Transactions of the Information Processing Society of Japan |
Authors: | Suganuma Akira, Kasahara Susumu, Ushijima Kazuo |
Keywords: | computers, artificial intelligence, information |
Japanese documents are usually analyzed with the grammatical analysis like the morphological analysis. Since this analysis requires much time, the authors constructed an extraction method with only textual information for the passive voice, the negative expressions, the conjunctive particle ‘GA,’ and so on. And they are developing a system of writing tools SUIKOU with the textual analysis method. This paper describes a method to estimate the conjugation of Japanese verbs and adjectives with a textual analysis. The authors prepare two kinds of tables for a conjugational part and a non-conjugational part of Japanese verbs and adjectives. This method analyzes a Japanese document from tail to head of a sentence with these tables. Using the method, they construct a new extraction method for the conjunctive particle ‘GA’ and the negative expressions. This extraction method guarantees to extract all requested items, but it may extract some unrequested items. The precision of the new extracting method for the conjunctive particle ‘GA’ is above 98%, and is improved by 7% in comparison with that of old one, although the new method is able to extract the conjunctive particle ‘GA’ as fast as old one. [In Japanese.]