Personal Health Record feeding via medical forums

The huge amount of textual data uploaded in virtual social networks can be effectively exploited to provide (possibly accidental) information in many scenarios. In healthcare discussion forums, users interact to educate themselves about treatments, symptoms, diseases, therapies, usually providing personal health-related information when posting questions or answers. Medical Forum data can be used for feeding Personal Health Record (PHR), the part of EHR where the person can directly manage his personal medical-related information. In this work, we propose an agent-based architecture to first extract textual data from html pages in a set of medical oriented italian social forum. Then, text mining techniques allow to discover health related information coming from the same user in different threads and/or different forums, finally feeding the PHR with such information. The architecture is presented together with a first implementation where 10 medical discussion forums are analyzed, showing promising results when extracting medical information from a subset of 3 over 10 forums.