摘要
[目的/意义]引文情感识别是全文本计量时代引文内容分析的重要研究议题之一,它与引文动机/功能识别、引文主题分析、引文摘要自动生成等存在较强的关联性,可为学术评价、知识图谱构建/绘制等问题的解决提供有效的研究支撑,具有较高研究价值。[方法/过程]通过文献调研分析,从引文语料集创建、情感词典使用、情感识别算法应用及存在问题4个方面,对国内外引文情感识别的研究进展进行全面梳理和分析评述。[结果/结论]引文情感识别已从早期的基于情感词典方法发展到当前基于机器学习算法的新阶段,并正由传统机器学习进一步向深度学习推进。亟待解决的主要问题有:(1)缺乏大规模高质量的引文语料集,对引文语料蕴含的特有价值(引文特征)的挖掘利用严重不足;(2)情感词典方法严重依赖情感词典自身的完备性,机器学习算法(分类模型)的参数优化及识别效果仍有提升空间,对两类方法的有机融合利用尚需深入探索;(3)更细粒度和更多维度的引文情感识别研究及相关应用实践有待进一步拓展和深化。
[Purpose/significance]Citation sentiment identification(CSI)is one of the important research tasks of Citation Context Analysis(CCA)in full-text era,and it has strong correlation with other CCA tasks,such as citation motive/function iden-tification,citation theme analysis,and automatic citation abstracting,etc.CSI can provide valuable research support for the solu-tion of academic evaluation and knowledge graph construction/mapping.[Method/process]Based on literature investigation,this paper comprehensively discusses the research progress of CSI at home and abroad from four aspects,such as citation corpus crea-tion,the use of sentiment lexicon,sentiment identification algorithm application and existing some problems.[Result/conclusion]CSI has progressed from the early sentiment lexicon-based methods to a new stage based on machine learning algorithms,and is be-ing further pushed from traditional machine learning to deep learning.The main problems to be concerned/solved are as follows:①One of the main obstacles for future research on CSI is lack of citation corpus with bigger size&higher quality,and making good use of more citation features extracting from corpus also has many things to do.②All methods based on sentiment lexicon rely heavi-ly on the completeness of the lexicon itself,various machine learning algorithms(classification models)also need to be improved,further experimental research is needed to optimize model parameters and effectively integrate with sentiment lexicon methods.③More fine-grained and multi-dimensional sentiment identification(including sentiment polarity&its intensity)are some valuable as-pects for the following research of CSI.
出处
《情报理论与实践》
CSSCI
北大核心
2024年第1期173-181,189,共10页
Information Studies:Theory & Application
关键词
引文情感识别
引文情感分析
引文内容分析
情感词典
机器学习
citation sentiment identification
citation sentiment analysis
citation context analysis
sentiment lexicon
machine learning
作者简介
王心玥(ORCID:0009-0001-0257-8610),女,2000年生,博士生。研究方向:科学计量学;通信作者:赵丹群,ORCID:0000-0003-0685-6689,Email:zdq@pku.edu.cn,女,1966年生,博士,教授,博士生导师。研究方向:科学计量学,引文分析与学术评价,信息检索与信息组织等。
