Nos partenaires

CNRS

Rechercher





Accueil du site > Français > Evénements > Séminaires

Séminaires

 

L’IRIT étant localisé sur plusieurs sites, ses séminaires sont organisés et ont lieu soit à l’Université Toulouse 3 Paul Sabatier (UT3), l’Université Toulouse 1 Capitole (UT1), l’INP-ENSEEIHT ou l’Université Toulouse 2 Jean Jaurès (UT2J).

 

Automatic Role Detection in Online Forums

Alberto LUMBRERAS - Equipe SC - IRIT (France)

Jeudi 27 Octobre 2016, 16h00 - 17h30
ENSEEIHT, salle A001
Version PDF :

Abstract

We address the problem of detecting user roles in online discussion forums. A role may be defined as the set of behaviors characteristic of a person. In discussion forums, behaviors are primarily observed through conversations. Hence, we focus our attention on how users discuss. We propose three methods to detect groups of users with similar conversational behaviors.

Our first method for the detection of roles is based on conversational structures. We apply different notions of neighborhood for posts in tree graphs and compare the conversational patterns that they detect as well as the clusters of users with similar conversational patterns.

Our second method is based on stochastic models of growth for conversation threads. Building upon these models we propose a method to find groups of users that tend to reply to the same type of posts. We show that some groups of users (roles) with extreme behaviors have some predictive power, which allow us to make predictions of their future behaviors.

We finally present a "dual-view mixture model" that integrates the type of data used in the two previous methods (feature-based and behavioral or functional-based). The model exploits the idea that users with similar features have similar behaviors. We show that, when this is hypothesis holds, we can find clusters using fewer examples. Besides, the model is non-parametric and uses a Dirichlet Process prior to automatically assess the number of clusters.

 

Retour