Web Search Engines A Basis For Term Evaluation

We present an approach for the evaluation of a term significance model, Gain Of Words (GOW) using two different web search engines (WSEs) : Google1 and Yahoo2 . Our term association model extracts significant terms from a single document without using a corpus. We extract the significant terms from the Gain of Words (GOW) parameters’ values based on a sentence level analysis of the document. The experiment is modeled by three sets of queries for the search engines, as single queries, paired queries and triplet queries. The evaluation is based on the documents retrieved from the WSEs using these queries. The results show that the terms or words of higher rank according to this model extract more relevant documents from the WSE as shown through the similarity results with the source documents from where the terms were extracted. We have shown that our model works better than the term frequency inverse sentence frequency (TFISF) model using WSE rankings.