Abstract: Ontology learning has become a hot research in the field of computer. Ontology learning currently focuses on concept extraction and conceptual relation extraction. Concerning the poor accuracy of existing learning methods, a new learning method based on Log-Likelihood Ratio（LLR） is proposed. In this way, the concepts and concept pairs can be extracted from Chinese text for estimating the significance of concepts and conceptual relations. The experimental results show that the method can effectively improve the performance of concept extraction and relation extraction.

Key words: ontology learning, concept extraction, relation extraction, Log-Likelihood Ratio（LLR）

摘要： 本体学习已成为计算机领域的一个研究热点，目前本体学习的研究重点在于概念及关系的提取。针对现有学习方法准确率不高，提出一种结合对数似然比（Log-Likelihood Ratio，LLR）的本体学习方法，采用对数似然比计算概念与领域及概念与概念之间的相关性，将其应用到概念与关系提取中。实验结果表明，结合对数似然比的学习方法能够有效改进概念和关系提取的准确度。

关键词: 本体学习, 概念提取, 关系提取, 对数似然比