Semantics of user interface for image retrieval: Possibility theory and learning techniques

摘要

We explain and justify the need of a rich semantics for the user interface in interactive image retrieval involving new media. Two ideas to build such interfaces are presented: possibility theory applied to fuzzy data retrieval (with hazy data and/or requests) and a machine learning technique applied to learning the user's deep need (with chosen images as examples and rejected ones as negative examples). We present two prototypes we have realized or undertaken using videodisks and knowledge-based softwares.