Enhancing Pragmatic Processing: A Two-Dimension Approach to Detecting Intentions in Spanish

Resumen

Recent advancements in Natural Language Processing (NLP), driven by the impressive performance of Large Language Models (LLMs), enable studies to address more complex linguistic levels such as semantics and pragmatics. However, available resources annotated with pragmatic information remain scarce for most languages. To address this gap, we present a Spanish annotation scheme for communicative intentions comprising two typologies: one for identifying the global intention of a message and another for the intentions of its textual segments. After validating the scheme, we introduce INTENT-ES, the first Spanish corpus of tweets annotated with their global and segment intentions. We leverage this corpus to evaluate the performance of traditional Machine Learning systems and current LLMs on intention classification. Considering the results, we believe these resources will benefit the NLP research community, facilitating the evaluation of LLMs in pragmatic tasks and integrating pragmatic information into NLP systems.