摘要
提出一种基于语义联系的新闻网页关键词抽取方法,不仅考虑了词语在知识库《知网》中的语义相似度,还考虑词语在具体上下文中的相关性,用词汇链将词语语义联系表示成图形式,在此基础上抽取出新闻网页关键词。对从网易网站选取120篇有核心提示的新闻网页进行测试,实验结果表明,所提出的方法比基于词频的关键词抽取方法和基于《知网》语义相似度构建词汇链的关键词抽取方法,在准确率和召回率上有很大的提高,当抽取关键词个数为3时,比基于词频方法的准确率和召回率分别提高了27.77%和21.38%。
A new keyphrase extraction method based on semantic relations is proposed in this paper. Two kinds of relations are considered one is the semantic similarity between words in HowNet and the other is the word correlation in context. The lexical chains representing the relation graph between phrases are constructed to extract keyphrases. The experimental results show that the proposed method substantially outperforms the method based on term frequency and the method based on lexical chains that are constructed by the semantic similarity based on HowNet,in terms of recall and precision,especially when the number of keyphrases extracted is 3-the accurate rate is increased by 27.77 percent,and the recall rate is increased by 21.38 percent compared with the method based on term frequency.
出处
《广西师范大学学报(自然科学版)》
CAS
北大核心
2009年第1期145-148,共4页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家自然科学基金资助项目(60573174)
中国科学院自动化研究所开放课题“HTML新闻网页过滤与总结系统”资助
关键词
关键词抽取
词汇链
语义联系
keyphrase extraction lexical chain semantic relation