Investigation of Effectiveness of Relevant Sentence Search for Web Information Access

Abstract

This paper proposes the approach of supporting Web information access based on relevant sentence search. A typical information retrieval using words as queries tends to let users focus on only top-ranked documents or fail to find relevant words as queries, which would lead to the bias in information collected by users. To mitigate such problems, the proposed approach extracts relevant words from the retrieved result and generates abstracts based on the extracted relevant words. The experimental result using the prototype interface shows the possibility that the proposed relevant sentence search can reduce the difference in the scope of exploration among the test participants.