Ugur Güdükbay,.pdf

1070-986X/07/$25.00 ? 2007 IEEE Published by the IEEE Computer Society 83 We have recently developed a video databasesystem, BilVideo, which provides integrated support for spatiotemporal, semantic, and low-level feature queries.1 As a further development for this system, we present a natural language-based inter- face for query specification. This natural language processing (NLP)-based interface lets users formu- late queries as sentences in English by using a part- of-speech (POS) tagging algorithm. The system then groups the specified queries as object-appear- ance, spatial, and similarity-based object trajecto- ry queries by using POS tagging information. It sends the queries constructed in the form of Prolog facts to the query processing engine, which interacts with both the knowledge base and object-relational database to respond to user queries that contain a combination of spa- tiotemporal, semantic, color, shape, and texture video queries. The query processor seamlessly integrates the intermediate query results returned from these two system components. The system sends the final results to Web clients. What motivates our work is the need for a convenient and flexible natural language-based interface to complement the text-based query interface and the visual query interface of the BilVideo system, because specification of spatial queries using text or visual interfaces is not very easy for novice users. (For examples of how oth- ers have attempted to handle these issues, see the “Related Work” sidebar, next page.) Thus, we developed a natural language-based query interface that’s convenient and offers greater flexibility when specifying queries. The POS-based pattern-matching approach we use in identifying queries helps users specify queries without conforming to strict rules. This approach also lets us adjust our query interface easily as we add new query types to the BilVideo system. BilVideo BilVideo’s spatiotemporal queries contain any combination o