Multi-Source News Oriented Domain Event Analysis

Public emergency is a serious threat to people’s lives and property safety, so prevention and rapid disposal of public emergency has become a key area in national science and technology development. In the network environment, virtual network world and real world reflect each other. Events occurred in real world will be expressed by different media through different sides. Currently, analyzing the data on the Internet to sense the real world timely has become a major technical means to effectively respond to emergencies.This paper focuses on the real-time analysis of cross-source news data. Based on in-depth study of the representation model of event, data from different sources on the Internet are collected and analyzed real-timely to recognize news events. The main parts of this paper are listed as follows:Firstly, this paper presents domain event model to describe the emergency classification, temporal and spatial characteristics, and the interactions between theme hazard factor and emergency subject, which supports subsequent event recognition and multidimensional search effectively.Secondly, this paper focuses on cross-source news structured analysis and event detection technology. For cross-source news reprint issue, using the hash means to achieve reprint discrimination reduces redundant computation. For topics diversity issue of open domain news, using domain rules validation to achieve news categorization reduces the amount of calculation and increases precision. For the issue of events constrained by elements, all kinds of elements are extracted from news according to domain event model. Under the constraint of the temporal and spatial elements extracted by heuristic rules, multi-feature similarity strategy is used to aggregate news in form of events. Experiments show the effectiveness of the method proposed in this paper.Thirdly, a prototype system is designed and implemented in this paper based on the requirement of distributed and real-time computing. The results from the food safety domain show that the system runs well and it can basically meet the application requirements.