Decoupling Structure and Lexicon for Zero-Shot Semantic Parsing

TL;DR

This paper presents a zero-shot semantic parsing method that generalizes to new domains by decoupling structure and lexicon, achieving high accuracy without domain-specific training data.

Contribution

It introduces a novel approach that maps utterances to domain-independent logical forms and then fills in domain-specific details, enabling zero-shot generalization.

Findings

Achieves 53.4% accuracy on 7 domains in Overnight dataset.

Outperforms other zero-shot baselines significantly.

Matches performance of parsers trained on over 30% of target domain data.

Abstract

Building a semantic parser quickly in a new domain is a fundamental challenge for conversational interfaces, as current semantic parsers require expensive supervision and lack the ability to generalize to new domains. In this paper, we introduce a zero-shot approach to semantic parsing that can parse utterances in unseen domains while only being trained on examples in other source domains. First, we map an utterance to an abstract, domain-independent, logical form that represents the structure of the logical form, but contains slots instead of KB constants. Then, we replace slots with KB constants via lexical alignment scores and global inference. Our model reaches an average accuracy of 53.4% on 7 domains in the Overnight dataset, substantially better than other zero-shot baselines, and performs as good as a parser trained on over 30% of the target domain examples.