Social media sites such as microblogs and discussion board forums have the potential to be rich source of information about human health. Going beyond simple keyword search and harnessing the data for insights that can benefit public health presents both opportunities and challenges to natural language processing (NLP). In this talk I survey the progress made using NLP methods, e.g. for adverse drug reaction profiling, flu surveillance and the study of depressive disorders. I will then look at

more »

... he technical challenges in understanding such messages, in particular how NLP can automatically encode/normalise laymen's language to the formal terminologies of healthcare professionals. To this end I present state of the art results from our recent work on using deep neural networks to de-conflate word senses as well as 'translating' from social media messages to SNOMED-CT. I will finish by briefly reflecting on the practical and ethical challenges that lie ahead.