iSearch: an interpretation based framework for keyword search in relational databases

Keyword search has become an effective information retrieval method for structured data. Existing works in relational database keyword search have addressed the problems of finding and evaluating candidate results. However, given that keyword queries are inherently ambiguous, it is often the case that candidate results do not match users’ search intention. In this paper, we analyze the limitations of current keyword search techniques and introduce the problem of generating and ranking keyword query interpretations. We propose a novel 3-phase keyword search paradigm which consists of: (1) the ability to predict query interpretations; (2) incorporate user feedback to to remove keyword ambiguities; (3) a ranking model to evaluate a query interpretation.