面向突发事件的基于UGQA的快速知识供给研究

Because of the unexpectedness and the hazard of emergent events, in order to optimize the decision making and reduce the hazard, it is necessary to provide the related knowledge to decision makers and public quickly, comprehensively and accurately. UGQA (User-Generated Questions and Answers) is a kind of UGC (User-Generated Content) with questions and answers. Based on the huge amount of UGQA in the Internet, supplying the knowledge quickly is important to fulfill the knowledge needs. The decision makers and public are always not familiar with the knowledge which is related to the emergent event. Moreover, the knowledge needs for the emergent event is urgent. Besides the integration of the knowledge needs and knowledge contents, the UGQA has the characteristics which are the huge quantity and the fragmental content. Considering these characteristics, firstly, the research is put on the identification and quality evaluation of the UGQAs which are related to the emergent event. In the research of the quick passive knowledge supply, the knowledge map and expert yellow pages which include the knowledge needs dimension and knowledge content dimension will be construed. Moreover, the representative UGQAs of each cluster will be identified. In the research of the quick active knowledge supply, the personalized knowledge needs will be modeled quickly. Then the UGQA recommendation method which includes two stages will be constructed. In the method, the selected relevant UGQAs are recommended in the first stage and the complementary UGQAs for the viewed UGQA are recommended in the second stage. Moreover, the minimum excellent experts will be recommended according the personalized knowledge needs. Finally, the prototype of the quick knowledge supply system will be constructed and the case study will be conducted with the system. It will be used to verify the research results. The results of this research have great theoretical value and practical significance in making the management of emergent events more scientific and knowledge-driven.

由于突发事件的不可预知性和危害性，发生时需要快速、全面和准确的向决策者和公众供给知识，以提高决策质量、减少危害发生。UGQA（User-Generated Questions and Answers）是问答类用户生成内容。基于互联网上海量的UGQA进行快速知识供给，对及时满足突发事件知识需求有着重要作用。针对突发事件知识陌生且需求紧迫，UGQA数量大、片段化、兼具需求和内容特征的特点，本项目在研究面向突发事件的UGQA识别与质量评价基础上，研究快速被动知识供给，构建面向知识需求和知识内容的二维UGQA知识地图与专家黄页，并识别代表性UGQA。在快速主动知识供给研究中，对个性化知识需求快速建模，然后进行包含相关推荐和互补推荐的两阶段UGQA推荐，并进行最少数量的最优专家推荐。最后通过系统构建与案例应用研究验证理论成果。本项目成果对推动突发事件管理向知识化和科学化发展有重大理论价值和实践意义。

由于突发事件的不可预知性和危害性，发生时需要快速、全面和准确的向决策者和公众供给知识，以提高决策质量、减少危害发生。UGQA（User-Generated Questions and Answers）是问答类用户生成内容。基于互联网上海量的UGQA进行快速知识供给，对及时满足突发事件知识需求有着重要作用。因此本项目进行了面向突发事件的基于UGQA的快速知识供给研究。针对突发事件知识陌生且需求紧迫，UGQA数量大、片段化、兼具需求和内容特征的特点，本项目开展了研究。研究了面向突发事件的UGQA的获取，编写爬虫程序进而获取突发事件相关的UGQA，并对虚拟问答社区、UGQA质量和UGQA中的答案质量以及知识检查服务评价进行了研究。研究了面向突发事件的UGQA的被动知识服务，针对UGQA的特点构建了知识地图和专家黄页。在此基础上，为了进一步提高知识获取的速度，研究了核心问题、核心答案和答案摘要提取方法等。研究了面向突发事件的UGQA的主动知识服务，研究了UGQA的推荐算法，解释性UGQA推荐算法以及专家推荐算法和专家匹配算法。进行了应用与机理研究，开发了原型系统。采集突发事件相关的数据进行实验验证。研究了虚拟问答社区中知识的信任度的影响因素以及突发事件舆情的干预等。在项目研究过程中，积极参加学术会议，进行学术交流。同时，多名博士、硕士研究生参与了项目的研究工作，培养了多名研究生。同时，项目组主要成员曾通过国家“万人计划”青年拔尖人才函评，进入了会评答辩阶段。基于本项目的研究成果，已经发表论文16篇，其中期刊论文12篇（SCIE源期刊10篇，EI源期刊1篇，CSSCI源期刊1篇），会议论文4篇。目前还有7篇论文已投稿到国际SCIE检索期刊和CSSCI检索期刊。还有5篇工作论文待投稿。获得2项科研奖励。本项目成果对推动突发事件管理向知识化和科学化发展有重大理论价值和实践意义。