Improving Logical-Level Natural Language Generation with Topic-Conditioned Data Augmentation and Logical Form Generation

TL;DR

This paper enhances logical-level natural language generation by using GPT-2 for data augmentation and introducing a dual logical form generation task, significantly improving performance with semi-supervised training.

Contribution

It proposes topic-conditioned data augmentation and a logical form generation task, enabling semi-supervised learning to improve Logic2text models without extensive manual annotations.

Findings

Outperforms supervised baselines on Logic2text dataset

Effective utilization of augmented data improves generation quality

Semi-supervised approach benefits from back-translation signals

Abstract

Logical Natural Language Generation, i.e., generating textual descriptions that can be logically entailed by a structured table, has been a challenge due to the low fidelity of the generation. \citet{chen2020logic2text} have addressed this problem by annotating interim logical programs to control the generation contents and semantics, and presented the task of table-aware logical form to text (Logic2text) generation. However, although table instances are abundant in the real world, logical forms paired with textual descriptions require costly human annotation work, which limits the performance of neural models. To mitigate this, we propose topic-conditioned data augmentation (TopicDA), which utilizes GPT-2 to generate unpaired logical forms and textual descriptions directly from tables. We further introduce logical form generation (LG), a dual task of Logic2text that requires generating a valid logical form based on a text description of a table. We also propose a semi-supervised learning approach to jointly train a Logic2text and an LG model with both labeled and augmented data. The two models benefit from each other by providing extra supervision signals through back-translation. Experimental results on the Logic2text dataset and the LG task demonstrate that our approach can effectively utilize the augmented data and outperform supervised baselines by a substantial margin.