SINAI at eRisk@CLEF 2022: Approaching Early Detection of Gambling and Eating Disorders with Natural Language Processing

2025-09-18

AI Summary

This study addresses two critical mental health risk identification tasks: early detection of pathological gambling and severity assessment of eating disorders. We propose a multimodal text analysis framework that integrates Transformer-based sentence embeddings, contextualized word vectors, lexical diversity measures, text length and complexity metrics, and fine-grained sentiment features. For gambling risk detection, we develop a multidimensional feature classification model; for eating disorder severity estimation, we design a semantic similarity–based regression approach. In the CLPsych 2023 shared task, our system achieved an F1-score of 0.808 for gambling detection—ranking second among 41 teams—and placed second among three finalist teams for eating disorder severity prediction. To our knowledge, this is the first work to jointly leverage contextualized semantic modeling and multidimensional linguistic features for dual-task mental health risk quantification, significantly enhancing both early detection robustness and clinical interpretability.

Technology Category

• Natural Language Processing: Language Grounding & Multi-modal NLP
• Machine Learning: Multimodal Learning
• Cognitive Modeling & Cognitive Systems: Computational Creativity

Application Category

• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

This paper describes the participation of the SINAI team in the eRisk@CLEF lab. Specifically, two of the proposed tasks have been addressed: i) Task 1 on the early detection of signs of pathological gambling, and ii) Task 3 on measuring the severity of the signs of eating disorders. The approach presented in Task 1 is based on the use of sentence embeddings from Transformers with features related to volumetry, lexical diversity, complexity metrics, and emotion-related scores, while the approach for Task 3 is based on text similarity estimation using contextualized word embeddings from Transformers. In Task 1, our team has been ranked in second position, with an F1 score of 0.808, out of 41 participant submissions. In Task 3, our team also placed second out of a total of 3 participating teams.

Problem

• Early detection of pathological gambling signs
• Measuring severity of eating disorder signs
• Using NLP for mental health assessment

Innovation

• Transformers sentence embeddings with volumetry features
• Text similarity using contextualized word embeddings
• Lexical diversity and emotion-related metrics integration