Meta-Learning a Cross-lingual Manifold for Semantic Parsing

TL;DR

This paper proposes a meta-learning approach to improve cross-lingual semantic parsing, enabling effective transfer with minimal annotated data in new languages, demonstrated on ATIS and Spider datasets.

Contribution

Introduces a first-order meta-learning algorithm that enhances sample efficiency and cross-lingual generalization for semantic parsing in low-resource languages.

Findings

Achieves accurate semantic parsing with ≤10% of training data in new languages.

Effective transfer from high-resource to low-resource languages.

Competitive results on ATIS and Spider datasets.

Abstract

Localizing a semantic parser to support new languages requires effective cross-lingual generalization. Recent work has found success with machine-translation or zero-shot methods although these approaches can struggle to model how native speakers ask questions. We consider how to effectively leverage minimal annotated examples in new languages for few-shot cross-lingual semantic parsing. We introduce a first-order meta-learning algorithm to train a semantic parser with maximal sample efficiency during cross-lingual transfer. Our algorithm uses high-resource languages to train the parser and simultaneously optimizes for cross-lingual generalization for lower-resource languages. Results across six languages on ATIS demonstrate that our combination of generalization steps yields accurate semantic parsers sampling $\le$10% of source training data in each new language. Our approach also trains a competitive model on Spider using English with generalization to Chinese similarly sampling $\le$10% of training data.