Meta-Learning for Domain Generalization in Semantic Parsing
arXiv:2010.11988
Abstract
The importance of building semantic parsers which can be applied to new domains and generate programs unseen at training has long been acknowledged, and datasets testing out-of-domain performance are becoming increasingly available. However, little or no attention has been devoted to learning algorithms or objectives which promote domain generalization, with virtually all existing approaches relying on standard supervised learning. In this work, we use a meta-learning framework which targets zero-shot domain generalization for semantic parsing. We apply a model-agnostic training algorithm that simulates zero-shot parsing by constructing virtual train and test sets from disjoint domains. The learning objective capitalizes on the intuition that gradient steps that improve source-domain performance should also improve target-domain performance, thus encouraging a parser to generalize to unseen target domains. Experimental results on the (English) Spider and Chinese Spider datasets show that the meta-learning objective significantly boosts the performance of a baseline parser.
arXiv commentNAACL2021 Camera Ready
References in corpus (9)
- PyTorch: An Imperative Style, High-Performance Deep Learning Library
- Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
- Seq2SQL: Generating Structured Queries from Natural Language using Reinforcement Learning
- TAPAS: Weakly Supervised Table Parsing via Pre-training
- On First-Order Meta-Learning Algorithms
- Improving Text-to-SQL Evaluation Methodology
- Differentiable Dynamic Programming for Structured Prediction and Attention
- La-MAML: Look-ahead Meta Learning for Continual Learning
- Neural Semantic Parsing in Low-Resource Settings with Back-Translation and Meta-Learning