An evidential reasoning based LSA approach to document classification for knowledge acquisition

Web is one of major information sources. Failure in proper management of knowledge leads to incorrect results returned by search engines. Therefore, the web should have an effective information retrieval system to improve the correctness of retrieval results. This study provides a method to assign a new document to the fittest category out of predefined categories, where latent semantic analysis (LSA) is used to evaluate each term in documents, the similarity between terms and documents as well as the one between terms and categories. The objective of our method is to fuse evidential reasoning method with LSA which can assign a new document to a predefined category. The method provides better results in performance of classification comparing to the fusion of an evidential reasoning approach with term frequency inverse document frequency (TFIDF).