Web-Document Retrieval by Genetic Learning of Importance Factors for HTML Tags

In contrast to conventional documents, a Web document consists of a number of tags which provide hints on the structure of the documents. In this paper, we propose a Web-document retrieval method using the characteristics of HTML tags. This method learns the importance of tags from a training text set. We use a genetic algorithm for learning the importance weights. We also present a modi ed similarity measure which uses the tag information. Experiments have been performed on the TREC document collection consisting of 247,491 documents. Compared to the traditional IR method, the proposed method has achieved 15% improvement in average precision.