Cognitive bootstrapping for knowledge
transfer in automatic sports video annotation

The conventional approach to designing automatic video annotation systems involves the development of application specific solutions based on computer vision and machine learning techniques. Any changes to the application will normally require a careful redesign and retraining. In this talk we explore cognitive bootstrapping as a mechanism for knowledge transfer between application scenarios. The problem of cognitive bootstrapping will be introduced and the key enabling components of cognitive bootstrapping will identified and discussed. The ideas and some early results will be illustrated on the problem of automatic sports video annotation.



Short bio