Query Ambiguity Indication Using Infrequent Term Cooccurrences

Conventional search engines are designed mainly for general keyword search. Therefore, in many cases, we can find no appropriate combination of query terms. In this paper, we present a query disambiguation method by using infrequent term cooccurrences. This strategy comes from the following idea: terms appearing with a wide variety of terms cannot establish an independent topic. Based on this hypothesis, terms are weighted. The experimental results show that the terms ranked higher by our method can improve the average precision of Web search when added to the original query terms. As compared with other term ranking methods, our method gives higher ranks to the terms denoting more particular and adequate stuff and referring to more specific concepts.