Keyword search on relational databases using keyword query interpretation

A user who wants to get information from a relational database needs to know database schema and structured query languages like SQL. The ordinary users are not familiar to those things, so searching information from relational databases is hard to them. Keyword search is a solution of the problem, where a keyword query is a simple and user-friendly search model that can be issued by writing a list of keywords. Because a keyword query can be interpreted variously, a large number of answers are returned. Thus, we need to handle the ambiguity problem. In this paper, we propose a keyword search method on relational databases. This method finds joined tuples as answers, partitions them by interpretations of the query, and ranks those groups of answers. Our ranking method focuses on finding answers derived from interpretations of the query that are similar to the interpretation in minds of ordinary users. An experimental evaluation using real data shows the performance of our ranking method.