2014 – present
Diesner’s team is developing a natural-language processing solution for probabilistic entity detection and classification in the domain of healthcare. The core of the solution are prediction models built by using supervised and/or semi-supervised machine learning techniques. The resulting models can be used to annotate natural language text data documents for entity classes. The team will perform fact extraction from medical text data documents as well as map tokens to predefined medical codes. Both tasks involve the same steps: 1) building and evaluating prediction models, 2) helping to integrate the prediction models into IMO’s workflow, 3) building an inference engine for practical applications, 4) building a technical solution with which IMO can update the prediction models, and 5) help to integrate the update solution into their workflow.