A Domain-Specific Concept-Based Searching System(Text Mining II)

概要

論文の詳細を見る
Search engines are helpful tools for finding information from a large collection of electronic documents. However, current keyword-based engines, in which searching is basically to match the keywords in queries and those in documents, are not performing well in terms of the precision and recall criteria. The main reason is that, in natural language, one word can have different senses and different words can mean the same thing in particular contexts. This paper is to share our experience in developing a document searching system that aims at answering to "what a user means" rather than "what a user says". Here, queries and documents are abstracted into vectors of pre-defined concepts and, posed a query, the answers will be those documents whose concept vectors and that of the query have the smallest angles. It requires a corpus of specific knowledge about the concepts and their associated terms in a domain of discourse. Our chosen documents are about regulations of financial management in Vietnam. Experiments have shown that the implemented concept-based searching system outperforms traditional keyword-based ones.
一般社団法人情報処理学会の論文
2004-12-04