Ontology building for concept indexing on multimedia information retrieval system

Abstract

With the rising trend of distance education and open source concept, multime- dia lecture video has become increasingly popular for self-learning and knowl- edge sharing with large demand of source. However, huge amount of video sources with broad range of topics make it difficult for users to access the par- ticular lecture video or the specific position of a lecture video that meets his/her requirement. Therefore, a system enabling users to browse lecture videos, search and retrieve required information of their interest is critical and necessary. In this project, we build a system with model-view-controller framework architec- ture. With this structure design, the system is able to add new scenario func- tions, change interface view format or design easily and flexibly, which means a system with high modifiability, reusability and maintainability. To make the indexing more domain-oriented with less ambiguity and provide more high-level structural knowledge representation of a collection domain, after realizing in- dexing on keywords, we also experiment to build a concept tree for one collec- tion of videos under the same topic, containing summarized essential concepts in the lecture contents to allow users to index based on concepts rather than only indexing on the exact words.