面向维基百科的领域知识演化关系抽取

Evolutionary Relation Extraction for Domain Knowledge in Wikipedia

互联网下同一领域中不同知识概念间存在多种关系,其中演化关系对于用户学习和理解领域知识,梳理领域知识的前序和后续逻辑关系具有重要意义,然而网络数据的多样和无序使用户难以准确有序地获取领域知识关系.针对该问题,提出一种面向中文维基百科领域知识的演化关系抽取方法,利用语法分析特征,挖掘演化关系模式,构建演化关系推理模型,采用基于句子层面的关系抽取算法识别领域知识演化关系,最后在真实的维基百科数据集上对该文方法进行了性能评测.实验表明,该方法具有较高的关系抽取准确率和召回率,能有效地抽取出维基百科中领域知识的演化关系.同时,基于实验抽取结果构建知识图谱,能有效挖掘领域学科下知识集合的演化体系,识别重难点知识,对学科建设以及相关课程教学具有一定的指导意义.

There are many relations among different concepts in the same field of knowledge from Internet. The evolutionary relation is very important for users to learn and understand the domain knowledge as well as sorting out the successive logic relationship between two different concepts. However, the diversity and disorder of the network data made it difficult for users to obtain relations among domain knowledge accurately and orderly. Aiming at this issue, we propose an evolutionary relation extraction method for domain knowledge in Chinese Wikipedia. Firstly, we construct a reasoning model for evolutionary relation using different syntax features as well as patterns discovery for evolutionary relation, and then detect the evolutionary relation for domain knowledge taking advantage of a sentence level relation extraction algorithm. Finally, we evaluated the method on the real Wikipedia data set in this paper. Experiments show that our method has higher precision and recall than the existing models and our proposal is effective against evolutionary relation extraction for domain knowledge in Wikipedia. Moreover, based on the experimental results, we construct a knowledge map in one field, which can represent the evolution structure of the knowledge-object collection and identify important and difficult points of knowledge effectively. Therefore, this method has a certain guiding significance to the subject construction and the related course teaching.