共词分析识别研究热点的效标关联效度研究:基于自然语言处理  

The Criterion-related Validity of the Hot Research Issues Identified by Co-words Analysis:a Study based on the Natural Language Processing

作　　者：杨丽 张彤彤 周文杰 

机构地区：[1]西北师范大学商学院甘肃兰州730070

出　　处：《图书与情报》2018年第1期15-19,共5页Library & Information

基　　金：本文系国家自然科学基金项目“基于共词分析的科学计量信效度研究”(项目编号:71563042)研究成果之一.

摘　　要：文章应用自然语言处理的方法,对样本文献中的题名、摘要和全文进行分词,并连同关键词一起,分别提取了四种分析单元下的高频词并应用Pajek和Sci2两个软件工具和常用的八种指标(算法)分别进行了研究热点的识别。然后,以全文为效标,分别运用相关分析和配对样本t检验,对题名、摘要和关键词在研究热点识别上的同时效度进行了检验。研究发现:(1)基于摘要而识别的研究热点同时效度最高,而基于关键词所识别的研究热点同时效度相对较低,具有一定效度风险;(2)在研究热点的识别方面,文本比词的同时效度高,而且文本的长度对于同时效度有着一定影响。Present study conducted a Natural Language Processing(NLP)on the titles,abstracts and whole paper of sample literature,together with keywords,to exact the high frequency words from all of above 4 units to identify the hot research issues via Pajek and Sci2.Moreover,present study performed Criterion-related Validity through correlation analysis and paired samples t-test by setting the hot research issues identified by whole paper as criterion and compared the correlation coefficient and t-value between whole papers and titles,abstracts and keywords to identify the Concurrency Validity of co-words analysis on hot research issues.The findings of this research include that:a)Those hot research issues identified by the abstract has a higher Concurrency Validity keywords.b)Aiming to identify the hot research issues,text is better than words from the perspective of Concurrency Validity,however,Validity is affected by the length of sampled text.

关 键 词：共词分析 研究热点 效标关联效度 自然语言处理 

分 类 号：G254.29[文化科学—图书馆学]