UNSL at eRisk 2022: Decision policies with history for early classification

For the 2022 edition of the CLEF eRisk Laboratory, our research group at Universidad Nacional de San Luis (UNSL) added new approaches and improvements concerning our last participation. We proposed two decision policies for EarlyModel that take into account historic information available to the models, and incorporated two score normalization steps into the SS3 model. We also significantly reduced the runtime to process the inputs. Despite not having achieved the best performances, our team obtained the best results for the $\text{ERDE}_{50}$ in tasks T1 and T2. Besides, considering the $F_{\text{latency}}$, we were the third-best team for both tasks. Finally, a couple of our models got some of the best performance for the ranking-based metrics for task T1.