Fouille de forums de discussion en santé : Identification des caractéristiques des contributeurs.

Health online discussion fora are precious sources of information from the therapeutic education. They permit to study the behaviours of the participants but also their understanding of their disorder and their needs when facing health problems. Yet, the analysis of the exchanges between users and their behaviour requires to collect demographic information on the internet users posting messages. We present an approach combining lexica and terminological resources with supervised classifiers for the identification of age and gender of users, their disorder and relation between the user and the patient. We apply our approch on diabetes mellitus discussion fora in French. Results vary according to the extracted information. Gender and age range detection achieves a good quality while disorders and relation between the author of the message and the patient are more difficult to identify. [ABSTRACT FROM AUTHOR]