Team 9: A Comparison of Simple vs. Complex Models for Suicide Risk Assessment

AbstractThis work presents the systems explored as part of the CLPsych 2021 Shared Task. More specifically, this work explores the relative performance of models trained on social me- dia data for suicide risk assessment. For this task, we aim to investigate whether or not simple traditional models can outperform more complex fine-tuned deep learning mod- els. Specifically, we build and compare a range of models including simple baseline models, feature-engineered machine learning models, and lastly, fine-tuned deep learning models. We find that simple more traditional machine learning models are more suited for this task and highlight the challenges faced when trying to leverage more sophisticated deep learning models.