知网的形式概念分析及概念相似度研究*

将知网的义项、义原及其关系映射到形式化概念分析的语境中,生成一个基于知网的形式概念格.一方面,提供了一种将知网中概念关系转换为概念格的表征方式,从格中任意一个节点出发,可以很方便地访问到与此相关的各种知识,从而为信息检索和知识推理提供很大方便;另一方面,也提出了一种通过对形式概念格进行分析来计算概念相似度的方法.实验证明该方法克服了以往计算方法的若干不足,并能有效地在相关应用领域如协作学习言论分析中加以应用.

Research on HowNet based on formal concept analysis and concept similarity

A formal concept lattice was implemented which based on HowNet through mapping the concept, sememe and the relationship in HowNet into formal context. On the one hand, the formal concept lattice as a representation transformed all the conceptual relationship in HowNet. In other words, from any node of the lattice, various kinds of related knowledge were accessible. Thus it provided much convenience to information retrieval and knowledge reasoning. On the other hand, a method of calculating the concept similarity was proposed by analyzing the formal concept lattice, which had overcome the some shortcomings of former algorithms and been proved to be effective when applied to analyze speech acts of collaborative learning in experimental result.