Scholarlink

<p>This study evaluates the use of Author Bibliographic Coupling Analysis (ABCA) as an additional criterion for the disambiguation of academic researcher names in databases. As a proof of concept, the dataset consists of the 32 recipients of the Derek de Solla Price Medal, with their scholarly outputs indexed in the Web of Science (WoS) database. The method aims to determine whether similarity (presence of coupling) or dissimilarity (absence of coupling) in cited references (authors) between two researchers sharing similar names is more effective. The methodological procedures involved three stages: (I) search strategies using “Last Name, First Name” and “Last Name and the first name's initial” if the former criterion was not met; (II) upon locating records for a given researcher, extracting the list of all cited authors for each identified profile (medal-winning researchers) and performing bibliographic coupling in two ways: (a) coupling the official profile with ambiguous profiles if an official profile existed; (b) coupling all profiles when no official profile was available, as it was not possible to determine the correct one; and (III) evaluating the coupling results to ascertain whether the identified profiles—ambiguous or non-official—corresponded to the same person. A qualitative (manual) analysis of the metadata for each identified profile was conducted, along with a quantitative evaluation using validation metrics, including precision, [positive predictive value (PPV)], negative predictive value (NPV), recall, <i>F</i>1-score, and accuracy. The coupling process resulted in a total of 243 calculations. The results indicated: 26 couplings with values ≥ 1 that corresponded to the same author; 3 null couplings that also corresponded to the same author; 158 null couplings where the profiles consistently belonged to different authors; and 56 couplings with values ≥ 1 where the profiles did not belong to the same researcher. The study concludes that the method is more effective in identifying dissimilarity (null coupling), as indicated by its high NPV (98.14%). In 158 out of 161 cases of null coupling, the method correctly determined that the profiles represented different individuals. This outcome is due to the absence of bibliographic coupling, reflecting a clear lack of similarity between the analyzed profiles. On the other hand, the method's precision was lower (31.71%), demonstrating limitations in identifying connections in cases of non-null coupling. Additionally, it achieved a recall of 89.65%, indicating a strong ability to correctly identify profiles belonging to the same author, while the <i>F</i>1-Score of 46.85% reflects a moderate balance between precision and recall. The accuracy of 75.72% shows that, overall, the method correctly classified most of the analyzed cases. Thus, the method proves to be more reliable for excluding false combinations than for suggesting profile merges.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Author Bibliographic Coupling Analysis as an additional criterion for Author Name Disambiguation: initial approaches

  • 期刊论文

  • 发表日期：

    2025年12月7日

查看全文
  • Rafael Gutierres Castanha,
  • Raidan Cruz Silveira

摘要

This study evaluates the use of Author Bibliographic Coupling Analysis (ABCA) as an additional criterion for the disambiguation of academic researcher names in databases. As a proof of concept, the dataset consists of the 32 recipients of the Derek de Solla Price Medal, with their scholarly outputs indexed in the Web of Science (WoS) database. The method aims to determine whether similarity (presence of coupling) or dissimilarity (absence of coupling) in cited references (authors) between two researchers sharing similar names is more effective. The methodological procedures involved three stages: (I) search strategies using “Last Name, First Name” and “Last Name and the first name's initial” if the former criterion was not met; (II) upon locating records for a given researcher, extracting the list of all cited authors for each identified profile (medal-winning researchers) and performing bibliographic coupling in two ways: (a) coupling the official profile with ambiguous profiles if an official profile existed; (b) coupling all profiles when no official profile was available, as it was not possible to determine the correct one; and (III) evaluating the coupling results to ascertain whether the identified profiles—ambiguous or non-official—corresponded to the same person. A qualitative (manual) analysis of the metadata for each identified profile was conducted, along with a quantitative evaluation using validation metrics, including precision, [positive predictive value (PPV)], negative predictive value (NPV), recall, F1-score, and accuracy. The coupling process resulted in a total of 243 calculations. The results indicated: 26 couplings with values ≥ 1 that corresponded to the same author; 3 null couplings that also corresponded to the same author; 158 null couplings where the profiles consistently belonged to different authors; and 56 couplings with values ≥ 1 where the profiles did not belong to the same researcher. The study concludes that the method is more effective in identifying dissimilarity (null coupling), as indicated by its high NPV (98.14%). In 158 out of 161 cases of null coupling, the method correctly determined that the profiles represented different individuals. This outcome is due to the absence of bibliographic coupling, reflecting a clear lack of similarity between the analyzed profiles. On the other hand, the method's precision was lower (31.71%), demonstrating limitations in identifying connections in cases of non-null coupling. Additionally, it achieved a recall of 89.65%, indicating a strong ability to correctly identify profiles belonging to the same author, while the F1-Score of 46.85% reflects a moderate balance between precision and recall. The accuracy of 75.72% shows that, overall, the method correctly classified most of the analyzed cases. Thus, the method proves to be more reliable for excluding false combinations than for suggesting profile merges.