Designing Domain Specific Word Embeddings: Applications to Disease Surveillance

Traditional disease surveillance can be augmented with a wide variety of realtime sources such as news and social media. However, these sources are in general unstructured and construction of surveillance tools such as taxonomical correlations and trace mapping involves considerable human supervision. In this paper, we motivate a disease vocabulary driven word2vec model (Dis2Vec) which we use to model diseases and constituent attributes as word embeddings from the HealthMap news corpus. We use these word embeddings to create disease taxonomies and evaluate our model accuracy against human annotated taxonomies. We compare our accuracies against several state-of-the art word2vec methods. Our results demonstrate that Dis2Vec outperforms traditional distributed vector representations in its ability to faithfully capture disease attributes and accurately forecast outbreaks.