Web Context Analysis Based on Generic Ontology

Abstract

Now, the popular search engine system does not take into account the context information of search word, so the returned list of web pages contain a large number of irrelevant web pages. One word can reflect its true semantics only in context. This paper proposed the concepts of “contextual word”, “web page parsing with contextualinformation”, and “context representation” for the first time. Based on general ontology, we use the techniques of the word sense disambiguation to determine the context of the word to realize the web-page parsing in the level of word sense and sentence semantics according to the background of the word in the web-page. First, we transformthe web page into DOM tree, do the web parsing in the tradition method to remove the noise in the web page, extract the main body of the web page, and then use the real time search technology to get the last-modified-time of the web page. Second, we do the Lexical analysis on the body of the web page. Based on general ontology and natural language processing techniques, we mark the word or terms, and get the interpretation corresponding to the context. Third, we use the Named Entity Recognition Technology to get the time and the location information in the web pages,then we organize the information which we obtained into a structure called web context representation that we proposed. Based on the above theoretical basis, the author implements a complete set of web page contextual parse tool—JLUCAS. After a large number analysis of comparative experiments, JLUCAS achieves excellent results in everyaspect. This fully demonstrates thatthe theory and algorithm proposed in this paper can solve the problem of automatic web page contextual parsing, and lay a good basis for ultimately realizing the web contextual search engine.