Abstract Syntax Networks for Code Generation and Semantic Parsing

TL;DR

This paper introduces abstract syntax networks, a new modeling framework that generates well-structured outputs like code and semantic parses by constructing abstract syntax trees, achieving state-of-the-art results on multiple benchmarks.

Contribution

The paper presents a novel framework for code generation and semantic parsing that constructs outputs as abstract syntax trees using a dynamically structured decoder.

Findings

Achieved 79.2 BLEU and 22.7% exact match on Hearthstone dataset.

Performed competitively on Atis, Jobs, and Geo datasets.

Outperformed previous state-of-the-art methods without task-specific engineering.

Abstract

Tasks like code generation and semantic parsing require mapping unstructured (or partially structured) inputs to well-formed, executable outputs. We introduce abstract syntax networks, a modeling framework for these problems. The outputs are represented as abstract syntax trees (ASTs) and constructed by a decoder with a dynamically-determined modular structure paralleling the structure of the output tree. On the benchmark Hearthstone dataset for code generation, our model obtains 79.2 BLEU and 22.7% exact match accuracy, compared to previous state-of-the-art values of 67.1 and 6.1%. Furthermore, we perform competitively on the Atis, Jobs, and Geo semantic parsing datasets with no task-specific engineering.