SBIR Phase I: Democratizing Data Science Through Conversation

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to dramatically improve the human productivity in gathering insights from data, and to democratize data analytics by making it available to a broad class of users within an enterprise. With DataChat, complex analyses can be carried out by simply conversing with a trained chatbot. The proposed approach has the potential to open a new vertical in the analytics market in which chatbots aid humans in carrying out the task of creating, deploying and running complex data science pipelines. This project could lead to the creation of a sub-market in the existing analytic software market, and it could also help improve the productivity of the (non-technology) sectors of the economy that increasingly require high-quality and fast insights from both their archival and real-time datasets. This Small Business Innovation Research (SBIR) Phase I project will take on a number of technical challenges including designing and developing a method to allow programming the underlying program that powers chatbots. Another technical challenge that will be tackled is making it easy to load external data that may not have well-defined schemas. A type inferencing mechanism, and associated set of methods to learn over historical data, will be developed to address this research aspect. Another technical challenge is building good machine learning models, for which a set of mechanisms is proposed that will allow automatic exploration and ranking of machine learning models, aiding the user in picking the right model for the specific task at hand. Overall these technical components will collectively contribute to the different facets of data analysis that are needed to gather insights from data, and will power the overall chatbot approach.