Searching the Web: a Semantics-Based Approach
Current search engines such as Google are mainly keyword-based, whereby a query is represented by a set of keywords. Such a query language is not expressive enough to allow users to present the subject of the web documents that they want to find. Consequently, one often receives many useless results when searching the web. This paper proposes a semantics-based approach to web search engines in order to increase their precision. Our assumption is that the subjects of the documents that one wants to search for can be expressed by a set of concepts and relations between them. We propose to use conceptual graphs to represent both user queries and document descriptions, on the basis of an ontology built up for a particular domain. In order to reduce the computational cost, documents are first filtered to provide only those that contain the concepts in the query. Graph matching is then performed to return relevant documents. A prototype of the proposed system is also presented for demonstration.