AZILE: an intelligent verbal interface for the retrieval of pictorial architectural information
Alexander Koutamanis · 2002
The proliferation of on-line image databases and the utility of these images have triggered the development of content-based techniques for indexing and retrieval. Most techniques are characterized by a verbal interpretation of visual patterns for query formulation. The paper describes the integration of such verbal terms for architectural pictorial information in AZILE, a natural language interface that operates through a dialogue with the user. In this dialogue the user expresses queries as normal everyday utterances. These are parsed and matched to a thesaurus of architectural terms and concepts. The meaning and associations of these terms result into a preliminary fuzzy classification of available pictorial information. The purpose of AZILE is three-fold. Firstly, it serves as an incremental refinement of the query. Secondly, it facilitates direct retrieval of suitable information in a browsing fashion. Thirdly, it supports machine learning by automatically indexing of the images with the terms identified in the user's utterances.