An Intelligent Method For Extracting Hotspot Events in News Bulletin

Journalists must understand the news bulletins accurately before they can get their hands on hotspot events. Identifying key points in a news bulletin is the cornerstone for accurately extracting hotpot events. However, the manual mining of hotspot events containing key points from massive news bulletins is a time-consuming and laborious work, and now we relies mainly on keyword search and rule matching. In order to accomplish the task, which is labor-intensive and material, journalists must have a good prior knowledge of all kinds of news bulletins on various topics. We tried to apply event extraction techniques to capture key points in the news bulletins. In news bulletins, however, the type of hotspot event that contains the key points is not properly defined. Existing event extraction methods are difficult to solve the problem of multiple hotspot events sharing the same parameters or trigger words in a sentence, which is common in news bulletins. In this paper, we proposed a mechanism to define hotspot events and a two-level label mapping method based on BERT and GloVe, which can better solve the above problems and facilitate the automatic extraction of hotspot events from multi-source massive news bulletin. Experimental results show that this method can accurately obtain the key points in news bulletin.