| Summary: | 碩士 === 中原大學 === 資訊工程學系 === 88 === Information retrievals, such as Boolean, vector and probability models, use index terms to get the documents of related contents. However, the relationships among index terms are usually ignored. Such relationships, no matter whether semantic or quantitative, may be valuable within retrievals.
This paper utilized the quantitative associations of index terms to form a weighted undirected graph, called TAG (term association graph). A TAG can be obtained from the association rules of a structured document. A structured document consisting of structures such as chapter, section, heading and hyperlink can also be represented as an undirected graph, called Structure Document Graph (SDG). For any document, the index term content can be obtained from its SDG. Then the corresponding TAG for this document can be calculated under the criterion of minimal accumulated association. When applied to hypertext information retrieval, TAG can be used to find the similarities of index terms, then the similarity between query and document.
A term association-based hypertext information retrieval system is established to implement our idea. The web documents are translated into SDGs; subsequently, the corresponding TAGs can be found and then clustered in several groups. Real examples prove that such system can retrieve related documents successfully.
|
|---|
