摘要
文章应用自然语言处理,以Sci2所识别的研究热点为效标,分别对基于题名、摘要、关键词和全文探测到的研究前沿进行了预测效度检验。研究发现:首先,无论基于标题、摘要、关键词还是全文等何种分析单元,基于爆发词进行研究前沿的预测都存在一定的效度风险;其次,相对而言,全文在研究前沿预测中的效度最高,而题名、摘要和关键词的预测效度则受到语料数量的影响;第三,如果以不同单项指标(算法)所识别的研究热点为效标,基于不同分析单元所探测的研究前沿存在着效度不一致的情况。这一发现对科学计量领域对研究前沿进行科学化预测具有参考价值。
Present study conducted a Natural Language Processing(NLP)and selected the hot research issues identified by Sci2 as validity criterion to explore the forecast validity of research frontier identified by titles,abstracts,keywords and full text of academic papers.Findings of this research include:firstly,no matter which unit involved into the analysis Processing,the forecasting of research frontiers do existed validity risks.Secondly,full texts,as the analysis corpus,has highest validity comparing to titles,abstracts and keywords which affected by the total amount of the selected papers.Thirdly,if single indexes were taken as the standard to identify the hot research issues,the explored research frontier predicted via different unit of academic papers are various.The findings of present study provide a valuable reference to those researches which focusing on the exploration of research frontier among scientometrics related researches.
出处
《图书与情报》
CSSCI
北大核心
2018年第1期1-7,共7页
Library & Information
基金
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
关键词
研究前沿
共词分析
预测效度
自然语言处理
research frontier
co-words analysis
predictive validity
Natural Language Processing