Searching for information in long videos can be a time-consuming experience. In order to deal with this problem, we have developed the OnAIR, an ontology-aided information retrieval system, which aims to retrieve clips from video collections. It allows users to enter a query in natural language and searches a digital video database for the clip which best answers the query. Its implementation uses a domain ontology to improve the quality of the retrieved information. The power of description logics and the reasoning mechanisms make it possible to infer new relationships between concepts (at the domain ontology) that would, otherwise, remain unknown. The system uses such information during the search for videos, at a process called query expansion, where the new related concepts are used to enrich the original query.

We have tested the system with a series of queries for the Contemporary Art case study and verified that the use of these semantic web techniques resulted in a relevance improvement of the retrieved documents. We have designed the system to work in a domain-independent way, allowing us to move to other domains by just changing the underlying domain ontology and video collection. OnAIR is maintained and developed by a team of graduate and undergraduate students at IME-USP. We are currently working on a Web 2.0 interface and ways to improve the domain ontology building task.