摘要： 传统的话题检测方法仅通过最初几篇话题相关报道的特征来表示话题，不能适应话题动态变化的特点。为此，提出一种话题案例知识库的动态模型。采用资源描述框架，实现话题案例知识表示，并在层次语义树基础上，利用案例融合策略实现话题案例知识库的动态更新，使用最大容忍优化策略解决话题质心漂移问题。实验结果表明，通过提高话题案例知识的全面性和内聚性，该模型能够改进话题检测和追踪的效果。

关键词: 话题检测, 案例知识库, 动态模型, 优化策略, 层次语义树, 话题质心

Abstract: As topic dynamically change over time, it is difficult to stand for all related story with feature of the first several stories. A dynamic model of topic case knowledge base is proposed in order to solve this problem. The model adopts Resource Description Framework(RDF) to represent case knowledge base, and then realizes dynamically updating of topic feature by merging case based on Hierarchical Semantic Tree(HST). In addition, it solves the problem of topic drift cased by the model with maximum enduring optimization strategy. Experimental results show that the model can greatly increase accuracy of topic detection through enhancing completeness and cohesiveness of case knowledge.

Key words: topic detection, case knowledge base, dynamic model, optimization strategy, Hierarchical Semantic Tree(HST), topic centroid

中图分类号: 

• TP301.6