Translating Natural Language to SQL using Pointer-Generator Networks and How Decoding Order Matters

TL;DR

This paper extends pointer-generator networks for translating natural language to SQL, investigates the importance of decoding order, and demonstrates improved performance on WikiSQL with insights into multiple decoding paths.

Contribution

It introduces an extension of pointer-generator networks for SQL translation and explores the impact of decoding order and multiple correct decoding paths.

Findings

Outperforms early models on WikiSQL

Decoding order significantly affects translation quality

Reinforce and dynamic oracle methods are explored for decoding

Abstract

Translating natural language to SQL queries for table-based question answering is a challenging problem and has received significant attention from the research community. In this work, we extend a pointer-generator and investigate the order-matters problem in semantic parsing for SQL. Even though our model is a straightforward extension of a general-purpose pointer-generator, it outperforms early works for WikiSQL and remains competitive to concurrently introduced, more complex models. Moreover, we provide a deeper investigation of the potential order-matters problem that could arise due to having multiple correct decoding paths, and investigate the use of REINFORCE as well as a dynamic oracle in this context.