摘要
以全文索引为基础的网页搜索引擎检索相关度偏低。针对这一问题,本文提出了一种基于查询日志分析的中文网页关键词抽取方法。该方法利用用户对网页与查询词的相关性判断来选择关键词。为了量化用户的相关性判断,提出了单位篇幅停留时间、逆向点击率、排名补偿因子3个指标,并对其进行综合加权。在查询串分词、同义词识别及多义词消歧、关键短语组配方面,也做了特殊处理。实验结果表明:抽取关键词的准确率较高,综合性能也高于TF.IDF和SVM方法。该方法能得到较满意的关键词抽取效果。
The webpage search engine based on the full-text index provides low correlation. To solve this problem, this paper proposes a keyword extraction method for Chinese pages based on query log analysis. The method selects keywords according to users＇ judgment of relevance on the page and query words. In order to quantify the relevance judgment, three indexes, such as residence time per unit length, inverted click rate and rank compensation factor, are proposed of which are then comprehensively weighted. In this paper, these processes, such as query string segmentation, synonym recognition, polysemy disambiguation, keyphrase matching, are specially treated. The experiment results show that the precision rate is high, and the comprehensive performance is better than that of the TF.IDF method and the SVM method. The proposed method can obtain satisfactory effect of the keyword extraction.
出处
《广西师范大学学报（自然科学版）》
CAS
北大核心
2015年第2期42-48,共7页
Journal of Guangxi Normal University:Natural Science Edition
基金
国家社会科学基金资助项目(14CJL001)
关键词
查询日志
关键词抽取
关键短语组配
同义词识别
多义词消歧
query log
keyword extraction
keyphrase matching
synonym recognition
polysemy disambiguation
