The IRIS system is used as the front end for the current demo. IRIS is an open source conversational agent that was built originally to facilitate tasks in data science for non-programmers. We have modified it to recognize key entities (drugs, genes, diseases) and serve as a prototype conversational interface for allowing users to type in free-text and then be prompted for information needed to complete a query. The results of the query are displayed and stored to aid further analysis and ease of information retrieval. IRIS is a collaboration between the Brown Institute at Stanford with support from the NSF in the Bernstein lab at Stanford Computer Science. The backend is written in Python and the front end is written in Javascript/CSS/HTML. IRIS can be compiled into a single executable that can be downloaded and used by a user; it can make API calls to do computationally intensive tasks. Human queries are recognized using logistic regression (trained on a provided set of ?synonymous queries? in English) to find keywords and map to most likely templated queries. IRIS can gather parameters by conversationally asking for additional information. IRIS can be integrated with any other Python-based tools, and is available for use by all participants in the Biomedical Data Translator project freely.
