Research On Method And Application Of Public Opinion Text Oriented Event Semantic Aggregating Fusion And Spreading Activation

Recent years have witnessed an explosive growth of various public opinion texts such as news, blogs and forums. With the large scale, rapid increasing, loosely distributed and sparsely associated public opinion texts, users have the desire to discover event core semantic information, which helps them make fast and comprehensive understanding of the event developing process. At the meanwhile, as events happen in real-time and their semantics evolve with the time, users are also willing to find relevant event knowledge of their interests, which offers them personalized event knowledge service.For the specific requirements of public opinion text processing, this paper proposes the public opinion text oriented event semantic aggregating fusion and spreading activation method and its theory, and studies related applications. The main contents include:1. Based on Association Link Network(ALN), this paper proposes an event semantic aggregating fusion method. Firstly, we aggregate event semantics from public opinion texts via event semantic community discovery method. Secondly, we assign texts to event semantic communities and reconstruct the event semantic communities to enrich events’ semantics, which realizes the fusion of event semantics.Finally, we aggregate and fuse the event semantics through the splitting iteration of ALN, and discover the event semantics, which helps users having a fast understanding of the knowledge of the events happened in real-time.2. Through simulating the semantic retrieval process of human associative memory spreading activation model, this paper proposes semantic spreading activation process of human associative memory, which expand relevant semantics according to users’ demands. And on this basis, this paper proposes an event semantic spreading activation method. Given the event semantic demands from users, based on the semantic spreading activation process, it finds multi-granularity event knowledge,including event key terms, event knowledge flows and event semantic communities,which offers users the personalized and comprehensive event knowledge service.3. This paper applies the proposed event semantic aggregating fusion method into event discovery, and proposes an event discovery algorithm. We first extract text semantics. Then we find the texts that report the following stories of recent events for event tracking. Finally, we discover the newly happened events using the proposed event semantic aggregating fusion method. Based on six event datasets and five evaluation metrics for event discovery, we conduct event discovery accuracy experiments, event discovery comparative experiments, online event discovery accuracy experiments, online event discovery comparative experiments and performance analysis, to evaluate the accuracy and effectiveness of our proposed event discovery algorithm.4. This paper applies the proposed event semantic spreading activation method into event knowledge recommendation, and proposes an event knowledge recommendation algorithm. Given users’ knowledge demands, it recommends multigranularity event knowledge using the proposed event semantic spreading activation method. Besides, texts that are related to the recommended event knowledge are also offered as the event backgrounds for users’ further understanding. At the same time, through perceiving users’ preferences, we enhance and inhibit user preferences, which make the following recommendation more personalized to meet the user demand. Based on the six event datasets and four evaluation metrics for knowledge recommendation, we conducts recommendation performance comparison experiments, interactive recommendation performance experiments, case studies of recommended event knowledge and system performance analysis, to evaluate the accuracy and effectiveness of event knowledge recommendation algorithm.Research results in this paper provide the theoretical supports and specific methods for text-oriented event semantics study, which can be used in text-oriented event semantic organization and representation, events semantic annotation, event detection and tracking, event semantic search, event knowledge extraction, etc., and solve the problem of web event information explosion without knowledge to some extent.