Word Taxonomy for On-line Visual Asset Management and Mining Osmar

We have designed and implemented MultiMediaMiner, a system prototype to mine high-level multimedia information and knowledge from large multimedia repositories like the WWW. WordNet, a semantic network for English, was used to clean and transform sets of keywords extracted from Web pages to index multimedia objects contained in these pages. WordNet was also enriched and used to generate concept hierarchies necessary for interactive information retrieval and the construction of multi-dimensional data cubes for multimedia data mining with