Problems worth solving at the intersection of NLP and structured data

Most of the value in data sitting in databases remains locked because people who have the context, curiosity, and specific goals that could be served with that data are not able to get answers to their own data questions with ease and speed. On the other hand, people with the technical expertise who have the tools and data context to be able to answer those questions remain overworked and under-appreciated because their job becomes very transactional; they serve one low-value data request after another instead of a strategic role where they can engage in high-level problem solving with data.

A natural language interface that allows all users to ask their data questions in a conversational format that can be trusted and understood without technical knowledge is an extremely promising proposition. While we have made tremendous progress in this quest in the last ten years, we are perhaps only 2% done. In this talk, I will address the technical challenges in building such a product as well as pragmatic approaches that can be taken to build a useful product despite the myriad of unsolved technical problems.