基于CRFs的中文领域术语自动抽取的研究.pdfVIP

基于CRFs的中文领域术语自动抽取研究* 1 1,2 1,2 2 2 2 王海雄 ,郭剑毅 ,余正涛 ,毛存礼 , 张朝胜 ,雷春雅 1. 昆明理工大学信息工程与自动化学院，昆明 650051 2. 云南省计算机技术应用重点实验室智能信息处理研究所，昆明 650051 E-mail: it_tianlang @163.com 摘 要：本文提出基于条件随机场 （CRFs ）的中文领域术语的自动抽取方法，在给定领域分类语料前提 下，该方法不仅很好解决了未登录词中的领域术语抽取的问题，而且还很好的解决三字以上长字词领域术 语的抽取问题。本文以云南旅游领域为实验对象，并与基于互信息结合T 评价的传统领域术语抽取方法作 比较，人工评测显示基于CRFs 的方法能准确有效的抽取领域术语，并能很好的抓取未登录词中的领域术 语和长字词领域术语，明显提高了领域术语的抽取的准确率和召回率。 关键词：CRFs ；机器学习；互信息；领域术语；特征选择； Automatic Chinese Domain Terms Extraction Based on conditional Random Fields 1 1,2 1,2 2 2 2 HaiXiong Wang , JianYi GUO ,ZhengTao Yu ,CunLi Mao ,ChaoSheng Zhang , ChunYa Lei 1. The School of Information Engineering and Automation, Kunming University of Science and Technology, Kunming 650051 2. The Institute of Intelligent Information Processing , Computer Technology Application Key Laboratory of Yunnan Province, Kunming 650051 E-mail: it_tianlang @163.com Abstract ：This paper proposed for automatic extracting domain terms from domain corpora based on CRFs . Under the premise of a given domain classification of corpus, this method is not only good to solve the problem of extracting domain terms from unknown Chinese words, but also have a very good solution to the problem of more than three characters long domain term extraction . This paper took the field of tourism in YunNan Province as the experimental subjects, furthermore, compared with the traditional method based on Combination of Mutually information and T-Value. Artificial evaluation showed that based on CRFs had a satisfactory results. This method is ve