效用视角下核心引文识别方法初探  

A Preliminary Exploration of Core Citation Recognition Method from the Perspective of Utility

作　　者：李凌[1] 张若楠 崔佳楠 李鑫鑫 Li Ling;Zhang Ruonan;Cui Jianan;Li Xinxin

机构地区：[1]西安交通大学图书馆,陕西西安710049

出　　处：《大学图书馆学报》2023年第6期98-106,127,共10页Journal of Academic Libraries

基　　金：高校图书馆数字资源采购联盟(DRAA)课题“效用视角下核心引文识别与数字资源评估”(课题编号:2023DRAA03);西安交通大学基本科研业务费专项科研项目(人文社科类)“西安交通大学医学部学科及人才竞争情报分析研究”(课题编号:SK2018012)的研究成果之一。

摘　　要：基于引用频次开展的引文分析忽略了引文对施引文献的效用差异,依此展开资源评估、学术影响力判断难免受到“无效”引用干扰。为排除“无效”引文,提升测度及评价数据质量,文章从引文效用的角度,选择引文属性、功能、对象、情感等典型特征,建立引文标注框架。测试逻辑回归与支持向量机等方式对引文自动分类的效果,择优构建“效用视角下核心引文识别模型”。其中,为探索自动化分类方式,在数据标注阶段,除人工标引,还使用大规模预训练语言模型的对话系统ChatGPT进行自动分类标注,并进行建模效果测试,以期为引文自动分类方法及应用实践提供新思路。The citation analysis based on the citation frequency ignores the difference in the utility of citations to the cited documents,and it is inevitable that the resource evaluation and academic influence judgment based on this will be interfered by“invalid”citations.To eliminate“invalid”citations and improve the quality of measurement and evaluation data,the article selects typical features such as citation attributes,functions,objects,and emotions from the perspective of citation utility,and establishes a citation annotation framework,in order to test the effectiveness of logistic regression and support vector machine methods on automatic citation classification in annotated datasets,and to construct a“core citation recognition model from the perspective of utility”.Specifically,in the data annotation stage,in addition to manual indexing,a large-scale pre trained language model dialogue system ChatGPT is also used for automatic classification and modeling performance testing,in order to provide new ideas for automatic citation classification methods and application practices.

关 键 词：引文效用 引文内容分析 引文标注框架 支持向量机 引文自动分类 ChatGPT 

分 类 号：G353.1[文化科学—情报学]