Few-Shot Table-to-Text Generation with Prefix-Controlled Generator

TL;DR

This paper introduces Prefix-Controlled Generator (PCG), a prompt-based method that improves few-shot table-to-text generation by controlling content and structure, reducing hallucinations, and better adapting PLMs to table data.

Contribution

The paper proposes a novel prompt-based approach, PCG, that enhances few-shot table-to-text generation by incorporating task-specific and input-specific prefixes, addressing hallucinations and structural differences.

Findings

Substantial improvements over baselines in automatic evaluations.

Effective control of factual content and word order.

Better adaptation to low-resource scenarios.

Abstract

Neural table-to-text generation approaches are data-hungry, limiting their adaptation for low-resource real-world applications. Previous works mostly resort to Pre-trained Language Models (PLMs) to generate fluent summaries of a table. However, they often contain hallucinated contents due to the uncontrolled nature of PLMs. Moreover, the topological differences between tables and sequences are rarely studied. Last but not least, fine-tuning on PLMs with a handful of instances may lead to over-fitting and catastrophic forgetting. To alleviate these problems, we propose a prompt-based approach, Prefix-Controlled Generator (i.e., PCG), for few-shot table-to-text generation. We prepend a task-specific prefix for a PLM to make the table structure better fit the pre-trained input. In addition, we generate an input-specific prefix to control the factual contents and word order of the generated text. Both automatic and human evaluations on different domains (humans, books and songs) of the Wikibio dataset show substantial improvements over baseline approaches.