Topic-oriented mining and reasoning
Yuefeng Li, Ning Zhong, Yi-Wei Yao · 2005
The discovery of the association between terms and a specified topic is a difficult task. A new data mining technique, topic-oriented mining and reasoning, is presented for this task. The technique consists of two threads: pattern mining and pattern reasoning. Pattern mining means the automatic discovery of interesting user topic models. A novel topic structure is presented for this thread. Pattern reasoning means the utilization and maintenance of the interesting user topic models to determine if an input data is relevant to the specified topic. The innovation for this thread is the use of meta-knowledge for the discovered knowledge. In this way the system can trace errors to update inadequate subtopics in the user topic model. The experimental results show that all objectives we expect for the topic-oriented model are achievable.