CEX Project - GoldStandard Fulltext

This gold standard collection comprises 107 scholarly articles, from 27 disciplines (according to the Scimago subject categories), each represented in multiple structured formats to support research on document processing, citation analysis, and linked data modeling. The list of all articles is provided in the GS_info.txt file. If an entry is marked as “Restricted”, it means that the corresponding PDF is not Open Access. As a result, these PDFs are not included in the dataset and cannot be found in the GoldStandard_PDF folder. On the other hand, all articles that are Open Access are shared in this folder in their original PDF format. For each article, we provide four complementary files, designed to capture the document’s content, structure, and citation information comprehensively: The original PDF, available for Open Access papers, preserves the published layout and formatting of the article. A TEI XML file, which is a full-text, machine-readable version of the article encoded using the Text Encoding Initiative (TEI) guidelines. This XML representation includes the article’s title, authors, affiliations, and publication metadata. It also provides a detailed structure of the text, segmented into sections, paragraphs, and subsections, and marks up figures, tables, and formulas. In addition, in-text citations are linked to their respective bibliographic references, and persistent identifiers such as DOIs and ORCIDs are embedded when available. A JSON file containing citation annotations, which maps each in-text citation pointer to its context within the article. Specifically, for every citation, this file identifies the section where it appears (first-level section if numbered), the complete sentence in which the citation is found, and the corresponding bibliographic reference. This format serves as a ground truth for tasks such as citation extraction and reference linking. A JSON-LD file, which provides a Linked Data representation of the article. This file is compliant with the OCDM v2.0.1 specification and is intended to facilitate the integration of scholarly content into semantic web applications and knowledge graphs. All TEI XML files were generated using Grobid (version 0.8.2), while the JSON and JSON-LD files were produced using the CEX tool. Importantly, each of these files has been carefully manually corrected and refined to ensure accuracy and reliability.