一种基于改进似然比的术语自动抽取方法

A Term Extraction Approach Based on Modified Log-likelihood Ratio

术语自动抽取是信息处理领域的基础性课题,日益受到研究者的关注。似然比方法能有效抽取低频词汇,但抽取准确率偏低。为了解决这个问题,将似然比的抽取结果用C-value进行改进。实验证明,两者相结合,在保证似然比方法高召回率的前提下,比单纯依靠似然比方法抽取准确率提高了约8%。

Term extraction is a basic subject in information processing and is attracting more and more attention nowadays. In order to extract low frequency words effectively, Log-likelihood ratio method is used but with a low precision rate. To solve this problem,C-value method is used to deal with the results of Log-likelihood ratio. Experiment results show that by combining the two methods ,the precision is improved in the premise of ensuing high recall rate of Log-likelihood ratio method. The proposed method can improve the precision by about 8; compared with the Log-likelihood ratio method.