[Session C4. 일정계획(2); 인공지능] 디렉토리형 검색엔진의 문서분류시스템에 관한 연구

In this study, we first propose the overall structure of knowledge-based directory service system and identify the detailed function of each subsystem. And then we focus on the classification methods which can be adopted in the web document classification subsystem. Feature weighting TFIDF, one of extended TFIDF which is the traditional vector model in information retrieval technique, is devised to overcome the limitations of existing methods in terms of computing resources and classification accuracy. To demonstrate the performance of the Feature weighting TFIDF, we compare our method with the traditional TFIDF and k-Nearest Neighbor method using case-based reasoning. The domain area we applied to is the web-based agricultural technology information service for domestic farmers. And the sample documents we use were classified by human experts in Rural Development Administration.