AUTOMATIC INDEXING IN THE AGE OF LANGUAGE MODELS

Authors

Keywords:

automatic indexing, indexing evaluation, language models

Abstract

Bibliographic databases and institutional repositories incorporate millions of documents annually, making exclusively manual processing unfeasible. In this context, indexing is essential for information organization and retrieval. This research compares three paradigms of automatic indexing under controlled vocabulary: a heuristic system based on rules and terminological recognition, a supervised machine learning approach for multilabel classification, and a large language model approach implemented in a hybrid pipeline combining lexical and semantic retrieval. A corpus of more than 9,500 documents in Agriculture was constructed, including 8,000 documents for training and three test sets with 500 documents each. The evaluation used reference standards derived from the AGRICOLA database, obtained by replicating search strategies and manually aligning the indexing. Indexing was restricted to the controlled vocabulary of the National Agricultural Library, previously normalized and represented in textual and vector formats. The results show that comparison among the approaches cannot be reduced to technological substitution: each paradigm responds differently to the requirements of automatic indexing. In the investigated corpus, hybrid solutions combining terminological control, textual retrieval, and semantic inference appear more suitable for exploring the potential of language models than using them as self-sufficient indexers.

Author Biographies

  • Marco Antônio Gusmão Carvalho, Unesp de Marília

    Holds a Master's degree in Information Science from the Paulista State University (UNESP) and a specialization in Technologies for Web Applications. He holds a bachelor's and teaching degree in Systems Analysis and Information Technologies from the Faculdade de Tecnologia de Ourinhos. He works as a lecturer of technical subjects related to software development in technical courses and high school at ETEC Professor Mário Antônio Verza, where he also coordinates the Internet Computing and High School courses. He also teaches in the Higher Education course in Intelligent Systems Technology at FATEC Shunji Nishimura and in the Software Engineering course at UNIFIO University Center..

  • Mariângela Spotti Lopes Fujita, Unesp de Marília

    PhD in Communication Sciences from the University of São Paulo (1992). She is currently a Senior Lecturer at the Faculty of Philosophy and Sciences at UNESP, Marília Campus, where she provides voluntary services in graduate teaching, research, and extension. She is a permanent faculty member in the Graduate Program in the research line of Information Production and Organization at the UNESP Information Science Program. As a researcher, she participates in the Research Groups on Thematic Representation of Information (leader) and Reading, Organization, Representation, Production, and Use of Information at UFPB (member). She conducts research activities at UNESP with a CNPq Research Productivity Fellowship, level 1B.

  • Isidoro Gil Leiva, University of Murcia

    He holds a PhD in Hispanic Philology from the University of Murcia (1997) and a degree in Library and Information Science from the University of Murcia (1992), as well as a degree in Geography and History (1994) from the University of Granada. He is currently a Full Professor at the University of Murcia. He is a member of the "Documentary Analysis" Research Group at UNESP – Marília Campus, where he has maintained a scientific cooperation plan since 2007. From 1998 to the present, he has taught in various master's and doctoral programs at the Polytechnic University of Valencia and the University of Murcia. Since 2007, he has also participated as a faculty member in PPGCI doctoral programs: at UNESP; DINTER (UNESP and UFCe); and PROCAD-CAPES (UNESP and UFPB).

Downloads

Published

2026-11-13

Conference Proceedings Volume

2026: XXVI ENCONTRO NACIONAL DE PESQUISA E PÓS-GRADUAÇÃO EM CIÊNCIA DA INFORMAÇÃO

Section

GT 2 – Organização e Representação do Conhecimento

Categories