# Learning a Neural Semantic Parser from User Feedback

## TL;DR

This paper introduces a neural semantic parser that learns from user feedback to improve SQL query generation for new domains, enabling rapid deployment and continuous enhancement without complex intermediate steps.

## Contribution

It presents a neural sequence model that directly maps natural language to SQL and leverages user feedback and crowd annotations for iterative improvement.

## Key findings

- Model can be deployed quickly for new domains
- Performance improves over time with user feedback
- No intermediate representations needed

## Abstract

We present an approach to rapidly and easily build natural language interfaces to databases for new domains, whose performance improves over time based on user feedback, and requires minimal intervention. To achieve this, we adapt neural sequence models to map utterances directly to SQL with its full expressivity, bypassing any intermediate meaning representations. These models are immediately deployed online to solicit feedback from real users to flag incorrect queries. Finally, the popularity of SQL facilitates gathering annotations for incorrect predictions using the crowd, which is directly used to improve our models. This complete feedback loop, without intermediate representations or database specific engineering, opens up new ways of building high quality semantic parsers. Experiments suggest that this approach can be deployed quickly for any new target domain, as we show by learning a semantic parser for an online academic database from scratch.