Combining Text and Image for Content-based Information Retrieval

摘要

In this paper, we face the problem of most Information Retrieval Systems that search documents (e.g. web pages) either by their texts or images content. However, in such complex documents, texts and images don't refer necessarily to the exactly same information, thus, both informational contents should be taken into account in the indexing process. We present our system that provides a conceptual and automatic indexing of documents by considering their heterogeneous content using a domain specific knowledge unit (like ontology). Preliminary results obtained on a specialized corpus of web pages are reported.