AUTOMATIC CLASSIFICATION OF IFS INNOVATION PROJECTS ACCORDING TO THE 2030 AGENDA USING LARGE LANGUAGE MODELS

The Federal Institutes of Education, Science and Technology (IFs) play a strategic role in promoting innovation and territorial development. The 2030 Agenda recognizes science, technology, and innovation as central instruments for achieving its goals. This article analyzes the titles of innovation projects approved by the Federal Institute of Sergipe (IFS) between 2014 and 2023, classifying them according to the Sustainable Development Goals (SDGs) through a large language model (LLM). The research is exploratory-descriptive in nature and adopts a mixed-methods approach, combining automated text processing with quantitative analysis. Results indicate a strong thematic concentration in SDG 9 (56%), followed by SDG 4 (9%), SDG 8 (7%), and SDG 7 (7%). Together, these four SDGs account for approximately 80% of the initiatives analyzed, reflecting the institutional mission of the IFs to integrate vocational training, applied research, and territorial innovation. The LLM-based classification predominantly yielded medium and low confidence levels, indicating inherent limitations in the exclusive use of project titles as analytical input. The findings suggest that LLMs represent promising tools for the institutional monitoring of innovation policies, although their application requires human validation and an expanded textual corpus for greater interpretive robustness.