Temporal fine-tuning for early risk detection

Resumen

Early Risk Detection (ERD) on the Web aims to identify promptly users facing social and health issues. Users are analyzed post[1]by-post, and it is necessary to guarantee correct and quick answers, which is particularly challenging in critical scenarios. ERD involves opti[1]mizing classiőcation precision and minimizing detection delay. Standard classiőcation metrics may not suffice, resorting to speciőc metrics such as ERDEθ that explicitly consider precision and delay. The current re[1]search focuses on applying a multi-objective approach, prioritizing clas[1]siőcation performance and establishing a separate criterion for decision time. In this work, we propose a completely different strategy, temporal őne-tuning, which allows tuning transformer-based models by explicitly incorporating time within the learning process. Our method allows us to analyze complete user post histories, tune models considering differ[1]ent contexts, and evaluate training performance using temporal metrics. We evaluated our proposal in the depression and eating disorders tasks for the Spanish language, achieving competitive results compared to the best models of MentalRiskES 2023. We found that temporal őne-tuning optimized decisions considering context and time progress. In this way, by properly taking advantage of the power of transformers, it is possible to address ERD by combining precision and speed as a single objective.