Deep Learning was leveraged by the team to construct a parser that achieves 63 percent accuracy on a common benchmark and an error-detecting module that prompts users to clarify ambiguous questions.
Photon was demonstrated by the team at the recent ACL 2020 conference. Team member Victoria Lin described the system in a recent blog post. It is interesting to note that the core of Photon is a neural-network-based semantic parser that has the ability to convert natural language questions from a human user into SQL queries. The parser achieves 63.2% exact-match accuracy on the Spider dataset that is the second highest result achieved to date.
It also incorporates a question corrector that can detect when the human input cannot be translated into SQL; a dialog is initiated by the question corrector with the user to further refine the question using a “chat-bot” style interface.
All in all, Photon deploys a strategy known as semantic parsing that converts the natural-language question into a logical form—essentially translating human language into programming language statements. Photon includes a “human-in-the-loop” question corrector with a purpose of improving the robustness of the system.
A demo version of Photon is available to the public.