|
The rise of social networking drives a new wave of revolution in the Internet word, and a large number of users instant message and fast status updating are its characteristic. Thus, in recent years, many studies on social networks have emerged, such as earthquake detection, tracing climate change, sports video highlight detection and so on.
This paper, we propose a new framework for sports game highlights detecting and annotation extraction. In the highlight detection section, we use only the social network text messages, unlike other research using the method with sound / images, not only reduce the computing resources greatly, but also become faster.
We propose a novel framework – A HITS-based Semantic Highlight Detection Framework for Live Sports Games using Chinese Social Media (HITS-SHiDF). In our research, user and highlights are seen as a complete bipartite graph, and using HITS used widely in information retrieval algorithms to do the highlights search.
In the section of highlight semantic annotation, in order to improve the performance of event annotation, we create own segmentation dictionary, using Wikipedia、historical pages of target social media as corpus. In addition, we also propose a new method, trying to enhance the effect on kekphrase extraction.
|