Exploiting Captions for Access to Multimedia Databases

摘要

Descriptive captions help organize noncompetitive media. But automated use ofcaptions in retrieval from computerized multimedia databases has not been much examined because it would seem to require significant natural language processing. We argue that captions can be naturally expressed in a restricted language whose interpretations is easier than general natural-language understanding. We describe a multimedia database system that stores interpreted captions in predicate calculus for each media datum; it then interprets restricted-language queries, and finds matching media objects.