Towards an understanding of the influence of social media on public healthcare by mining health information on the Social Web

Social Media Sites are playing an increasingly important role in the generation and sharing of health information. Research has shown that a significant and increasing population is seeking and utilising health advice found on SMS1. Despite the opportunities, this creates for individual healthcare, their impact on personal health improvement is unclear. Existing research is limited and has found positive results but also that misunderstanding and misinformation are prevalent. A systematic analysis of the content of SMS is non-trivial, as such data sources are unstructured, heterogeneous, often redundant, and can be contradictory. Using the online diabetes community as a case study, we will employ Natural Language Processing techniques to automatically extract concepts, named entities, and relations to identify facts (e.g., diabetes causes dry itchy skin) mentioned from heterogeneous SMS, and integrate and link them in a structured Knowledge Base. The facts will also be quantified based on their frequency of mentions across disparate sources. The KB will capture the health information on SMS in a structured way, facilitating complex querying such as 'how many diabetes patients suffer from itchy skins and what remedies are they using'. This will enable the first crucial steps towards a quantitative and qualitative analysis of health information from SMS.