Few-Shot NLG with Pre-Trained Language Model

TL;DR

This paper introduces a few-shot natural language generation approach using pre-trained language models, achieving strong performance across domains with limited data, and outperforming existing baselines by over 8 BLEU points.

Contribution

It proposes a novel few-shot NLG method leveraging pre-trained models, demonstrating effective content selection and language modeling with minimal training data.

Findings

Achieves over 8 BLEU points improvement with 200 examples

Demonstrates strong cross-domain generalization

Outperforms existing baselines significantly

Abstract

Neural-based end-to-end approaches to natural language generation (NLG) from structured data or knowledge are data-hungry, making their adoption for real-world applications difficult with limited data. In this work, we propose the new task of \textit{few-shot natural language generation}. Motivated by how humans tend to summarize tabular data, we propose a simple yet effective approach and show that it not only demonstrates strong performance but also provides good generalization across domains. The design of the model architecture is based on two aspects: content selection from input data and language modeling to compose coherent sentences, which can be acquired from prior knowledge. With just 200 training examples, across multiple domains, we show that our approach achieves very reasonable performances and outperforms the strongest baseline by an average of over 8.0 BLEU points improvement. Our code and data can be found at \url{https://github.com/czyssrs/Few-Shot-NLG}