Contextual weighting approach to compute term weight in layered vector space model

Abstract

The World Wide Web (WWW) is the largest available repository of information. This huge amount of information put forward the challenges of retrieval of trustworthy information from WWW. It defies researchers with new issues of diversity and complexity while retrieving the web information. Information retrieval from the web demands approaches that span beyond conventional information retrieval. Heterogeneity, complexity and the huge volume of web information requires a unique approach to retrieve information. Besides, end-users introduce some difficulties in the retrieval process. Sometimes queries submitted by the user are subtle and ambiguous. The primary concern in information retrieval is the issue of predicting the relevance of documents. In this article, a new approach is proposed that rationally separates web document into five layers, namely, title, header, hyperlink, meta tag and body layer. The proposed method effectively combines the textual information and structural evidence of web document for retrieving information from Web. In the proposed layered vector space model, each layer has an allocated priority which is used to compute weight factor for these layers. The proposed method deduces equation that effectively combines priority of the layer and length of the layer to calculate the weight of the layer.