A Hidden Markov Model Approach to Keyword-Based Search over Relational Databases

We present a novel method for translating keyword queries over relationaldatabases into SQL queries with the same intended semantic meaning. Incontrast to the majority of the existing keyword-based techniques, our approachdoes not require any a-priori knowledge of the data instance. It follows a probabilisticapproach based on a Hidden Markov Model for computing the top-K bestmappings of the query keywords into the database terms, i.e., tables, attributesand values. The mappings are then used to generate the SQL queries that areexecuted to produce the answer to the keyword query. The method has been implementedinto a system called KEYRY (from KEYword to queRY).