IRIT at e-Risk

Idriss Abdou Malam (1) , Mohamed Arziki (1) , Mohammed Nezar Bellazrak (1) , Farah Benamara (2, 3) , Assafa El Kaidi (1) , Bouchra Es-Saghir (1) , Zhaolong He (2) , Mouad Housni (1) , Véronique Moriceau (2, 4) , Josiane Mothe (5, 6) , Faneva Ramiandrisoa (5)

1 IRIT - Institut de recherche en informatique de Toulouse
2 IRIT-MELODI - MEthodes et ingénierie des Langues, des Ontologies et du DIscours
3 UT3 - Université Toulouse III - Paul Sabatier
4 LIMSI - Laboratoire d'Informatique pour la Mécanique et les Sciences de l'Ingénieur
5 IRIT-SIG - Systèmes d’Informations Généralisées
6 UT2J - Université Toulouse - Jean Jaurès

In this paper, we present the method we developed when participating to the e-Risk pilot task. We use machine learning in order to solve the problem of early detection of depressive users in social media relying on various features that we detail in this paper. We submitted 4 models which differences are also detailed in this paper. Best results were obtained when using a combination of lexical and statistical features.

Mots clés: e-Risk

Domaines: Théorie de l'information [cs.IT]; Recherche d'information [cs.IR]

Idriss Abdou Malam, Mohamed Arziki, Mohammed Nezar Bellazrak, Farah Benamara, Assafa El Kaidi, et al.. IRIT at e-Risk. 8th International Conference of the CLEF Association (CLEF 2017), Sep 2017, Dublin, Ireland. pp.1-7. ⟨hal-01912779⟩