Learning to Infer Program Sketches

TL;DR

This paper introduces a hybrid program synthesis system that combines pattern recognition and symbolic reasoning, enabling automatic code generation from natural language and examples, with state-of-the-art results on simple English-to-code tasks.

Contribution

It presents a novel intermediate representation and training algorithm that dynamically balances pattern recognition and explicit reasoning without direct supervision.

Findings

Matches neural synthesis memorization performance

Matches symbolic search generalization performance

Achieves state-of-the-art on English description-to-code dataset

Abstract

Our goal is to build systems which write code automatically from the kinds of specifications humans can most easily provide, such as examples and natural language instruction. The key idea of this work is that a flexible combination of pattern recognition and explicit reasoning can be used to solve these complex programming problems. We propose a method for dynamically integrating these types of information. Our novel intermediate representation and training algorithm allow a program synthesis system to learn, without direct supervision, when to rely on pattern recognition and when to perform symbolic search. Our model matches the memorization and generalization performance of neural synthesis and symbolic search, respectively, and achieves state-of-the-art performance on a dataset of simple English description-to-code programming problems.