关键词: 全信息理论, 全信息知识库, 层次知识获取, 标量聚类

Abstract:

A new method of automatic entity relation extraction is proposed. Based on the Bootstrapping algorithm, the hierarchy knowledge extraction model can be designed. The inner specific word extraction model and outer pattern extraction model can be nested each other to extract automatically knowledge, so that the specific dictionary and pattern rules used for the entity relation extraction is achieved. Combined with the Comprehensive Information Theory, the semantic and pragmatic information can be added into the relation extraction patterns to generate the Comprehensive Information Knowledge-Base (CIKB). Both the experiments of relation extraction and the evaluation have been done.

Key words: comprehensive information theory, comprehensive information knowledge base, hierarchy knowledge extraction, and scalar cluster