摘要： 传统自适应话题追踪用向量空间模型表示一个话题模型，通常会对话题模型更新带来错误的反馈。针对传统自适应话题追踪中话题模型的不足，提出基于K-Modes聚类的自适应话题追踪方法(K-MATT方法)，用话题类中心代替话题模型，把命名实体向量空间模型作为话题类中心，在追踪过程中不断迭代更新话题类中心，直到话题类中心稳定。实验证明K-MATT方法是有效的。

关键词: 话题追踪, K-MATT方法, 话题类中心

Abstract: Traditional Adaptive Topic Tracking(ATT) uses VSM to express a topic model and bring mistaken feedback to topic model updating. This paper presents an Adaptive Tracking Technique based on K-Modes clustering(K-MATT) to solve the problems caused by traditional topic model expression. This method uses Topic Kind Center(TKC) to substitute topic model and uses named entities VSM to express TKC, updates TKC in topic tracking until TKC is stable. Experiments prove K-MATT method is effective.

Key words: Topic Tracking(TT), K-MATT method, Topic Kind Center(TKC)

中图分类号: 

• TP391