Text Classification to Weave Medical Advice with Patient Experiences.

Unlike face-to-face patient support groups, clinicians providing medical advice to peer-patient conversations is challenging in online patient communities due to its large scale. We are developing a system that weaves medical advice into online patient conversations. As an initial step, we are exploring text classification of answers contributed by health professionals and patients. Using unigrams, we were able to predict with 91.9% accuracy whether an answer was written by a health professional or a patient. Introduction Unlike face-to-face support groups, it is challenging for health professionals to provide medical advice in the context of online peer-patient conversations because of time and resource limitations. To address this gap, we are developing a system that weaves health professionals’ medical advice available on the Web into peer-patient conversations in online communities. Health professionals’ blogs or answers posted in online community forums often lack meta-data about the authorship of the information. To explore automatically identifying health professionals’ answers from the Web, we looked at classifying health professionals’ answers versus patients’ answers. Previous studies looked at various ways—semantic, syntactic, and statistical features—to classify medical questions, but no studies have classified authorship of answers in online patient community settings.