LLM-SUPPORTED METHOD FOR DEVELOPING DOMAIN- SPECIFIC TAXONOMIES

Authors

  • Cecilia Kotzias Federal University of Santa Catarina
  • Roberto Pacheco Federal University of Santa Catarina
  • Fernando Alvaro Ostuni Gauthier Federal University of Santa Catarina
  • Mônica Carneiro Federal University of Santa Catarina

Keywords:

Design Science Research (DSR), Taxonomy construction, Large Language Models (LLMs), Iterative method

Abstract

The development of taxonomies is a critical process for knowledge organization, yet it is frequently marked by intuitive and non-standardized approaches. This study addresses this gap by proposing the Iterative Hybrid Method, an approach to domain-specific taxonomy construction that integrates conceptual (top-down) and empirical (bottom-up) strategies. Grounded in Design Science Research (DSR), the methodology includes a systematic literature review, iterative expert validation, and the application of Large Language Models (LLMs) to support classification and refinement. The method was applied in constructing a taxonomy for social entities, iterating between empirical observation and theoretical analysis. Results indicate that this structured, adaptable, and replicable approach enhances the accuracy and usability of taxonomies across domains. The study highlights LLMs’ role in automating concept identification and classification while acknowledging transparency and bias limitations. Future work should explore ex-post evaluations and automation integration to further refine taxonomy development processes.

Downloads

Download data is not yet available.

Published

2025-12-05

How to Cite

Kotzias, C., Pacheco, R. ., Alvaro Ostuni Gauthier , F. ., & Carneiro, M. . (2025). LLM-SUPPORTED METHOD FOR DEVELOPING DOMAIN- SPECIFIC TAXONOMIES. International Congress of Knowledge and Innovation - Ciki, 1(1). Retrieved from https://proceeding.ciki.ufsc.br/index.php/ciki/article/view/1734