Détection des humeurs dépressives sur les réseaux sociaux chinois à partir d'une combinaison de plongements lexicaux et de méthodes textométriques
Résumé
Studies aimed at detecting users at risk of depression on Chinese social media generally rely on metadata or limited textual characteristics of the users, as well as large language models, foregoing any semantic analysis and ignoring the linguistic expression of emotions. We propose an architecture for extracting and analyzing semantic features based on textual statistics to address this issue. Our corpus is derived from the Sina Weibo Depression Dataset. We will use textual statistics methods to analyze the corpus with the goal of constructing an interpretable semantic model, to which we will then combine machine learning and deep learning methods. The results will be analyzed to determine the contribution of the semantic model to detection and its interpretability.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|