Constrained Language Models Yield Few-Shot Semantic Parsers
This provides a rapid bootstrapping method for semantic parsers, addressing the data scarcity problem in natural language processing for tasks requiring structured outputs.
The paper tackles the problem of semantic parsing with limited training data by using large pretrained language models to paraphrase inputs into a controlled English-like sublanguage that can be automatically mapped to meaning representations, achieving surprisingly effective performance that greatly exceeds baseline methods on multiple community tasks.
We explore the use of large pretrained language models as few-shot semantic parsers. The goal in semantic parsing is to generate a structured meaning representation given a natural language input. However, language models are trained to generate natural language. To bridge the gap, we use language models to paraphrase inputs into a controlled sublanguage resembling English that can be automatically mapped to a target meaning representation. Our results demonstrate that with only a small amount of data and very little code to convert into English-like representations, our blueprint for rapidly bootstrapping semantic parsers leads to surprisingly effective performance on multiple community tasks, greatly exceeding baseline methods also trained on the same limited data.