Savant: video annotation support for GSS meeting memory

• Main Author: Rathie, Susan E.
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/8233

Much research effort has been expended in the areas of facilitation and Group Support Systems, with a view to improving the efficiency and effectiveness of meetings. One area involves the development of database systems to capture information in support of meeting and organizational memory. Uses of such information include updating members of a group on issues discussed at meetings from which they were absent, and passing on information to a new individual taking on an existing role. These memory systems have to date been . somewhat limited in their expressiveness as they are based in textual data. The rich record video provides makes it a likely medium to more effectively support meeting memory. However, raw video is not easily searchable, and some form of indexing is needed to facilitate later retrieval. These indices are provided through an annotation process. Since manual annotation of video is a long, difficult and potentially error-prone task even when using a computerized annotation tool, automated support for annotation is desirable, and forms the subject of this work. Derived from the facilitation and multimedia literatures, a model based on the concept of automatic video annotation is proposed. This model facilitates the indexing and retrieval of multimedia data from meetings conducted with Group Support System (GSS). A prototype system, Semi-Automatic Video Annotation Tool (SAVANT), was developed in conformance with this model. SAVANT uses domain knowledge of meetings and video to provide first-level annotations of GSS meetings based on the computer logs generated by the GSS. These annotations can be used on their own or to facilitate further manual annotation. The research reported in this thesis (1) develops a model for semi-automatic annotation support that extends the concept of meeting memory to provide knowledge-based support for multimedia information; (2) proposes a channel-based approach to annotation automation and demonstrates this approach through the implementation of a prototype of the GSS Log channel; and (3) demonstrates the feasibility of automating annotation support for GSS meetings by conducting a field test. The field test involved real-time multimedia data capture from a real-life meeting situation using an existing GSS (MeetingPlace), integrating that information with the log captured from that GSS meeting, and showing how different researchers or annotators could use that log in conjunction with a semi-automated annotation tool to aid in their further analysis of the events from such a meeting.