Weakly Structured Representations for Flexible Queries
Kenji HANAKAWA, Hideaki TAKEDA, Toyoaki NISHIDA
Most information systems always require us to represent our information in a formal and strict way. This lack of flexibility reduces the usefulness of the information systems. In this paper, we introduce two weakly structured representation techniques for easily constructing and retrieving ill-organized information. The first technique is a word-list-query which is a free ordered sequence of words used to represent user's information needs. The second is an associative graph which is a kind of semantic network consisting of concept nodes, IS-A links and unnamed horizontal links. To parse a word-list-query, we use a concept reduction method which is recursive rewriting of words according to information represented in associative graph. This integration of parsing and information retrieval disambiguates word-list-queries. We implemented an experimental system using these techniques, and show that this system is easy for unskilled users and its accuracy is reasonable.