Gold standard of the paper titled "LEAD: LLM-enhanced Engine for Author Disambiguation"

Dataset containing the gold standard developed for our "LEAD: LLM-enhanced Engine for Author Disambiguation" paper. Abstract Author Name Disambiguation (AND) is a long-standing challenge in bibliometrics and scientometrics, as name ambiguity undermines the accuracy of bibliographic databases and the reliability of research evaluation. This study addresses the problem of cross-source disambiguation by linking academic career records from CercaUniversità, the official registry of Italian academics, with author profiles in Scopus. We introduce LEAD (LLM-enhanced Engine for Author Disambiguation), a novel hybrid framework that combines semantic features extracted through Large Language Models (LLMs) with structural evidence derived from co-authorship and citation networks. To evaluate the effectiveness of this approach, we constructed a manually validated gold standard of 606 ambiguous cases, representing particularly hard scenarios such as homonyms working at the same institution or in closely related fields. We compare five methods: Label Spreading on co-authorship networks, Bibliographic Coupling on citation networks, a standalone LLM-based approach, an LLM-enriched configuration, and the proposed hybrid pipeline. Results show that LEAD consistently outperforms the other methods, achieving an F1-score of 96.7\% and an accuracy of 95.7\%, while significantly reducing computational cost compared to full LLM solutions. Bibliographic Coupling emerges as the fastest and strongest single-source method. %, whereas Label Spreading provides an efficient baseline for rapid disambiguation. These findings demonstrate that integrating semantic and structural signals within a selective hybrid strategy offers a robust and scalable solution to cross-database author identification. Beyond the specific Italian case, this work contributes to the broader debate on data quality in research information systems, highlighting the potential of hybrid LLM-based approaches for enhancing the reliability of scientometric analyses. Dataset description The study relies on a benchmark dataset specifically designed for hard cases in Author Name Disambiguation (AND). The dataset combines information from two sources: Scopus, providing bibliographic and citation metadata, and CercaUniversità, the official registry of Italian academics and their career records. The Italian academic system is structured into 14 Scientific Areas (SAs), subdivided into Recruitment Field Groups (RFGs), Recruitment Fields (RFs), and Academic Disciplines (ADs). This taxonomy was used to align institutional information with publication data. Although CercaUniversità has begun including Scopus Author IDs (AUIDs) since 2022 - covering around one-third of Italian academics - most records remain without identifiers. To bridge this gap, the authors implemented a two-step matching pipeline. First, a string-matching procedure via the Scopus API retrieved candidate author profiles using combinations of first name, last name, affiliation city, and publication year. This initial search often returned multiple candidate profiles due to homonymy and metadata inconsistencies. Second, a manual annotation phase was conducted to validate correspondences, producing a high-quality ground truth. The final gold standard consists of 606 candidate associations, of which 394 are correct (65%) and 212 incorrect (35%). Each entry contains both institutional attributes (name, RF, AD, university) and a candidate Scopus AUID, plus a binary label for validation. The dataset is unbalanced across fields: Medicine and Industrial/Information Engineering dominate, while humanities and social sciences are underrepresented. This reflects both Scopus coverage and disciplinary publication patterns. Overall, the dataset provides a rigorous testbed for evaluating disambiguation methods under highly ambiguous, cross-source conditions. Field description id: unique record identifier. name: the given name of an Italian academic, as reported in CercaUniversità. surname: the family name of an Italian academic, as reported in CercaUniversità. RF: the Recruitment Field (RF) of an Italian academic, as reported in CercaUniversità. AD: the Academic Discipline (AD) of an Italian academic, as reported in CercaUniversità. university: the institutional affiliation of an Italian academic, as reported in CercaUniversità. AUID: the Scopus Author ID (AUID) of a candidate profile corresponding to the academic under verification. class: binary variable, set to 1 if the AUID was manually validated as correct, and 0 otherwise.