Leveraging Code to Improve In-context Learning for Semantic Parsing

TL;DR

This paper enhances in-context learning for semantic parsing by using programming languages like Python and structured domain descriptions, significantly improving accuracy and reducing demonstration requirements.

Contribution

It introduces a novel approach of leveraging general-purpose programming languages and structured prompts to improve semantic parsing with in-context learning.

Findings

Significant accuracy improvements across datasets

Reduced need for numerous demonstrations

Resemblance to general-purpose code is crucial

Abstract

In-context learning (ICL) is an appealing approach for semantic parsing due to its few-shot nature and improved generalization. However, learning to parse to rare domain-specific languages (DSLs) from just a few demonstrations is challenging, limiting the performance of even the most capable LLMs. In this work, we improve the effectiveness of ICL for semantic parsing by (1) using general-purpose programming languages such as Python instead of DSLs, and (2) augmenting prompts with a structured domain description that includes, e.g., the available classes and functions. We show that both these changes significantly improve accuracy across three popular datasets. Combined, they lead to dramatic improvements (e.g. 7.9% to 66.5% on SMCalFlow compositional split), nearly closing the performance gap between easier i.i.d.\ and harder compositional splits when used with a strong model, and reducing the need for a large number of demonstrations. We find that the resemblance of the target parse language to general-purpose code is a more important factor than the language's popularity in pre-training corpora. Our findings provide an improved methodology for building semantic parsers in the modern context of ICL with LLMs.