基于上下文的领域本体概念和关系的提取

Study on context-based domain ontology concept extraction and relation extraction

目前本体学习的研究重点在于概念及关系的提取,概念提取领域一致度与领域相关度相结合的方法取得了比较好的效果,而关系提取则主要采用基于关联规则的方法。这种本体概念、关系学习方法由于只考虑词频,提取结果准确性欠缺。针对这种缺陷,在统计的基础上考虑了语义因素,利用词汇上下文计算概念的语义相似度并将其应用到概念与关系提取中。实验结果表明,词汇上下文与传统统计相结合的方法能够有效改进概念和关系提取的准确度。

Recently, ontology, learning focuses on concept extraction and conceptual relation extraction. For concept extraction, domain relevance combined with domain consistent had yielded better results, and the algorithm of association rules was mainly adopted for relation extraction. Since the traditional methods only considered the word frequency, there were many substantial inaccuracies in learning results. To overcome these shortcomings, this paper proposed a new learning method based on context. In this way, represented semantic similarity between words and could overcome these shortcomings. The experimental results show that this method can effectively improve the performance of ontology, learning system.