eRisk: Early Risk Prediction on the Internet
eRisk explores the evaluation methodology, effectiveness metrics and practical applications (particularly those related to health and safety) of early risk detection on the Internet.
Tasks:
-
Task 1: Search for Symptoms of Depression
Rank sentences from users according to their relevance to each of the 21 symptoms of the BDI-II questionnaire. Training data consists of sentence-tagged datasets from 2023 and 2024, with new test data including contextual information (previous and next sentences).
-
Task 2: Contextualized Early Detection of Depression
Participants analyze full conversational interactions to classify users with signs of depression, considering the conversational context beyond isolated user writings. The test phase includes writings with full conversational dynamics, while the training phase uses isolated user submissions.
-
Pilot Task: Conversational Depression Detection via LLMs
Participants interact with a persona powered by a large language model (LLM) that is fine-tuned using types of depressive and non depressive users. The objective is to detect signs of depression, with participants limited to a specified number of messages to engage with the LLM.
Organizers
- Javier Parapar, University of A Coruña
- Anxo Perez, University of A Coruña
- Xi Wang, University of Sheffield
- Fabio Crestani, Università della Svizzera Italiana
Contact
- anxo.pvila@udc.es
- javierparapar@udc.es
- xi.wang@sheffield.ac.uk
- fabio.crestani@usi.ch
