Towards fusion of collective knowledge and audio-visual content features for annotating broadcast video

摘要: Broadcasters produce vast collections of video content. However, the lack fine-grained annotations makes it difficult to retrieve fragments interest from these collections. Indeed, manual annotation content is labour-intensive and time-consuming. Moreover, applicability algorithms for automatic limited, given that too many prerequisites need be fulfilled a lot concepts are unidentifiable. At same time, people using social media share their thoughts about they view on television. Therefore, in this Ph.D. research, we plan investigate novel machine learning-based approaches towards task broadcast content, fusing collective knowledge present with output audio-visual analysis algorithms.