An Annotation Method and Application for Video Contents Based on a Semantic Graph Tomohisa Akafuji 1,3 , Kaname Harumoto 2 , Keishi Kandori 3 , Kˆ oiti Hasida 4 , and Shinji Shimojo 5 1 Graduate School of Information Science and Technology, Osaka University, 5-1 Mihogaoka, Ibaragi, Osaka, 567-0047, Japan , aka@ais.cmc.osaka-u.ac.jp 2 Graduate School of Engineering, Osaka University, Osaka, Japan 3 Information Technology Center, Asahi Broadcasting Corporation, Osaka, Japan 4 Cyber Assist Research Center, National Institute of Advanced Industrial Science and Technology; CREST, Japan Science and Technology Agency, Tokyo, Japan 5 Cybermedia Center, Osaka University, Osaka, Japan Abstract. The progress of digitization at broadcasting stations enables the intellectual production of high-quality contents from video mate- rials. Metadata plays an important role in describing the semantics of multimedia contents by aiding in the semantic structuralization of video material. In this paper, we propose an annotation method for video contents based on the concept of a semantic graph, which allows seman- tic queries on video contents. This method incorporates three ideas: an annotation for metadata on video scenes; time-series links between video scenes; and semantic links to external data. In the proposed annotation method, metadata are provided as human annotations practical for TV productions. We also present the application of a semantic graph to real baseball game contents and evaluate this application to show the usefulness of the proposed annotation method. Keywords: Metadata, semantics, video contents, retrieval 1 Introduction The digitization of video material in broadcasting stations makes storing and processing video contents by computer easy. In addition, both the reuse of con- tents and the production of high-quality contents can be expected. Metadata is essential for realizing video archiving with an intelligent video retrieval function, which aids in the production of high-quality contents. Many video materials are needed for high-quality contents, and some retrieval system is many important factor as an editing support. In this paper, we propose an annotation method including an input support, in which the concept of a semantic graph is introduced. Moreover, broadcasting stations often have valuable external data already attached to video contents. Our annotation method can utilize the external data for annotation support or as the complement of semantically structured annotation data. K. Aizawa, Y. Nakamura, and S. Satoh (Eds.): PCM 2004, LNCS 3331, pp. 478–486, 2004. c  Springer-Verlag Berlin Heidelberg 2004