Pilot qualitative analysis dataset on the contribution of Metascience Observatory articles to policy documents

This dataset is part of an ongoing study aimed at identifying the specific types of contributions made by papers published by Metascience Observatories (MOs), retrieved from OpenAlex, to policy documents indexed in Overton that mention those papers. It is part of an article currently being written by Cesar Pereira, Sergio Salles-Filho, and Cassidy Sugimoto, within the project Investigating the relationship between science diplomacy and global DGT: the role of inclusive metascience observatories, coordinated by Prof. Cassidy Sugimoto at the Georgia Institute of Technology, United States. This spreadsheet contains the results of a pilot qualitative content analysis conducted to examine how articles published by MOs were mobilized in policy documents. For this pilot, four observatories were selected: BELSPO (Belgium), CONACYT (Paraguay), NISTEP (Japan), and OCyT (Colombia). The selection sought to ensure diversity among observatories while maintaining a manageable corpus of publications and policy documents for methodological validation. The pilot was carried out in two analytical stages using Gemini 2.5 Pro via API. In the first stage, selected Overton fields - especially highlight doi, highlight text, and highlight context - were analyzed in order to classify the type of contribution made by the cited article to the policy document. In the second stage, the interpretation was deepened by combining the article abstract, the description of the policy document, and the results of the first classification stage, producing an analytical synthesis of the article’s contribution and its thematic alignment with a predefined set of axes. The prompts used in these stages were previously deposited in Zenodo and constitute the methodological basis of this dataset. This spreadsheet corresponds to the structured results produced through the application of those prompts in the pilot stage. The prompts are available at: Pereira, C. A. (2026). Prompt to Gemini 2.5 to match articles with policy documents. Zenodo. https://doi.org/10.5281/zenodo.19255312. Each row in the spreadsheet represents one analytical instance of a cited article in relation to a specific policy document. Therefore, the unit of analysis is not the article alone, but the pair cited article X citing policy document. For this reason, the same value in the doi_cited column may appear more than once whenever the same scientific article was cited in different policy documents identified by distinct policy_document_id values. The spreadsheet includes bibliographic and institutional identification fields, textual inputs used in the LLM-assisted analysis, the exact prompts applied in each stage, and the structured outputs generated by Gemini 2.5 Pro. The main columns are: mo: abbreviated name of the Metascience Observatory. name: full name of the Metascience Observatory. ror_id: ROR identifier of the observatory. doi_cited: DOI of the scientific article cited in the policy document. article_title: title of the cited scientific article. article_abstract: abstract of the cited scientific article. policy_document_id: unique identifier of the citing policy document in Overton. policy_document_title: title of the citing policy document. llm_document_description: description or summary of the policy document used as analytical input. highlight_text: bibliographic reference or cited text segment extracted from Overton. highlight_context: excerpt of the policy document showing the context in which the article was mentioned. prompt_a: full prompt used in the first analytical stage. contribution_category_(gemini_prompt_a): classification generated in the first stage regarding the type of contribution of the article to the policy document. prompt_b: full prompt used in the second analytical stage. contribution_assessment_(gemini_prompt_b): synthetic qualitative assessment generated in the second stage on the substantive contribution of the article to the policy document. adherence_overall_(gemini_prompt_b): overall indication of whether the contribution was considered adherent to at least one predefined thematic axis. adherence_terms_(gemini_prompt_b): list of thematic axes identified as relevant in the assessment. adherence_explanation_(gemini_prompt_b): short explanation justifying the thematic adherence assessment. This dataset was organized to support transparency, traceability, and reproducibility of the pilot procedure, allowing readers to examine how the qualitative analysis was operationalized and how the outputs generated by the language model were structured for subsequent interpretation. This record should be read in conjunction with the Zenodo records containing the prompts, since the spreadsheet stores the empirical results derived from their application.