Grant: Strategic Projects - Group

Social media are increasingly popular. Nowadays, Internet users are posting, blogging, tweeting and so on about almost everything, including health topics. Such communities support active discussions on a plethora of topics and have flourished on the Web over the past decade. In particular, Internet users report their mood, their activities, their social interactions and all the events around them. This information could be used as a vast organic sensor network composed of individuals for modelling public health at a population scale. In this project, we focus on automatic prediction and analysis models for mental health. Recently, researchers have shown that social media can be used to detect affective disorders such as depression. The objective of our project is to develop automatic methods to exploit the massive data issued from social media and to apply social web mining and sentiment analysis methods to detect at-risk people and monitor the mental state of population. We will investigate one application scenario for our predictive model which will be used to identify at-risk individuals in online communities. The model will also be used by psychologists and psychiatrists to identify variables related to major mental illness. The algorithms developed in this project can be adapted for other domains (e.g., to identify at-risk kids or victims of bulling in the context of high-school students). We have organised an international group (Canada/France) which has expertise in natural language processing, data mining, social media processing and medical informatics. This consortium also includes psychologists who will focus on the effects the developed methods can have on health care management. The Canadian company involved in the project, Girih Inc. will contribute with their experience with social media data collection and sampling for this kind of data. The joint call NSERC/ANR represents a perfect opportunity for this project. Furthermore, the two groups provide in a synergistic way computational web social mining expertise for both English and French, which is a rare asset in this type of research, as most current work focuses uniquely on English.