Unmet Need:
With the rise of accessible healthcare information on the Internet, it is increasingly essential for individuals to obtain relevant and accurate information for their questions. There is particular emphasis for this in light of the COVID-19 pandemic and potential future viral outbreaks and health crises.
Technology Overview:
Inventors at Johns Hopkins have developed a natural language interface in the form of a health chat bot that may be integrated with existing or new mobile health applications. The culmination of several techniques including Latent Dirichlet Allocation with topic overgeneration, BM25 similarity ranking, Discriminative Information retrieval model, Standard term frequency model, BERT deep learning model, and Polyencoder model resulted in a COVID-19 chat bot that was accurate in understanding the intent of questions in. natural language 74.4% of the time.
Stage of Development:
Pilot phase with functional product.
Publications:
N/A
