[目的/意义] 施引作者在引用过程中会概括、提炼被引论文成果价值点，发现此类信息有助于全面、深入地了解被引论文的学术价值。[方法/过程] 提出一种无监督多特征加权的价值点识别方法，进一步发现原文中未提及或未显著提及的价值点；对Athar引用语料库中高被引的20篇文献进行实验。[结果/结论] 实验结果表明，原文中未显著提及但引用中却强调的价值点可揭示被引论文发表后同行的共识与认可、引导跨库检索应用化成果、更新与补充被引论文关键词、收集被引论文主题缩写词等功能，实现对被引论文动态标引，提高论文显示度、检索效率以及跨库关联能力。由此，基于引用内容发现的价值点可以作为一种描述被引论文价值的动态生成的新型元数据即引用标签，发挥重点提示、检索与推荐等功能，丰富引用内容服务。未来将在更多领域、语种、类型以及更大的论文数据集上验证价值点发现的可行性和实用性。
[Purpose/Significance] The citing authors summarize and refine the value points of the cited documents’ achievements during the citation process. Discovering value points is helpful for a comprehensive and indepth understanding of the academic values of the cited documents. [Method/Process] This paper proposed an unsupervised multi-feature weighted method for recognizing the value points, and further discovered the value points that were not mentioned or not significantly mentioned in the cited documents. Experiments were conducted on 20 highly cited documents in Athar’s citation corpus. [Result/Conclusion] The experimental results show that the discovered value points that are not significantly mentioned in the cited documents but are highlighted in the citation can reveal the consensus and recognition of peers after the publication of cited documents, guide the cross-database retrieval of applied outputs, update and supplement the keywords of the cited documents, collect the acronyms of cited documents’ topics, which realize the dynamic indexing of cited documents, improve the prominence and retrieval performance of cited documents, and the ability of cross-database association. The value points discovered in citation content can be used as a new type of dynamically generated metadata named citation tags describing the value of cited documents, which provides the services of highlight, retrieval and recommendation and enriches citation content services. In the future, the feasibility and practicability of the discovery of the value points will be verify in more fields, languages, types and larger paper datasets.
[1] 李洁,孟烨,金佳丽,等.新兴科学引文索引数据库的比较研究[J].大学图书馆学报, 2021, 39(6):48-55, 77.[2] Semantic Scholar[EB/OL].[2022-12-18]. https://www.semanticsholar.org/.[3] Scite[EB/OL].[2022-12-18]. https://scite.ai/.[4] SMALL H G. Cited documents as concept symbols[J]. Social studies of science, 1978, 8(3):327-340.[5] COZZENS S E. Split citation identity:a case study from economics[J]. Journal of the American Society for Information Science, 1982, 33(4):233-236.[6] 陆伟,孟睿,刘兴帮.面向引用关系的引文内容标注框架研究[J].中国图书馆学报, 2014, 40(6):93-104.[7] 许德山.科技论文引用中的观点倾向分析[D].北京:中国科学院文献情报中心, 2012.[8] 马娜.科技论文引用对象识别方法研究[D].北京:中国科学院文献情报中心, 2020.[9] 马娜,张智雄,吴朋民.基于特征融合的术语型引用对象自动识别方法研究[J].数据分析与知识发现, 2020, 4(1):89-98.[10] O'CONNOR J. Citing statements:computer recognition and use to improve retrieval[J]. Information processing&management, 1982, 18(3):125-131.[11] O'CONNOR J. Biomedical citing statements:computer recognition and use to aid full-text retrieval[J]. Information processing&management, 1985, 19(6):361-368.[12] BRADSHAW S. Reference directed indexing:redeeming relevance for subject search in citation indexes[C]//KOCH T, SøLVBERG I T. Research and advanced technology for digital libraries. Berlin:Springer, 2003:499-510.[13] RITCHIE A, TEUFEL S, ROBERTSON S. How to find better index terms through citations[EB/OL].[2022-05-30]. https://aclanthology.org/W06-0804.pdf.[14] RITCHIE A. Citation context analysis for information retrieval[R]. Cambridge:University of Cambridge, 2009.[15] 刘盛博,丁堃,刘则渊.基于引用内容的引文检索与推荐系统[J].情报学报, 2013, 32(11):1157-1163.[16] CAMPOS R, MANGARAVITE V, PASQUALI A, et al. A text feature based automatic keyword extraction method for single documents[C]//PASI G, PIWOWARSKI B, AZZOPARDI L, et al. Advances in information retrieval. Cham:Springer, 2018:684-691.[17] ATHAR A. Citation sentiment corpus[EB/OL].[2022-12-20]. https://cl.awaisathar.com/citation-sentiment-corpus/[18] 默顿.科学社会学——理论与经验研究[M].鲁旭东,林聚任,译.北京:商务印书馆, 2003.
