Ontological video annotation and querying system for soccer games

This paper describes a video annotation and querying system which is capable of semi-automatic annotation of videos from text. The extracted metadata is aligned with the corresponding video segments. This allows users to query videos according to their semantic content. We have chosen soccer domain to demonstrate the use of the system. The soccer videos are very suitable for our framework, since it is easy to find web-cast match reports for soccer games. The annotated videos are stored in MPEG-7 format in an object-oriented database. The keyword-based indexing allows fast retrieval of video segments. The system accepts match reports in Turkish which makes querying in Turkish possible.