SAPPHIRE: Approaches for Enhanced Concept-to-Text Generation

TL;DR

SAPPHIRE introduces simple yet effective techniques to improve concept-to-text generation, significantly enhancing performance and addressing common issues like lack of commonsense, specificity, and fluency in generated text.

Contribution

The paper proposes SAPPHIRE, a set of augmentation and phrase infilling methods, to improve concept-to-text generation models like BART and T5.

Findings

Notable performance improvements on the CommonGen task

Enhanced fluency and commonsense in generated texts

Effective addressing of baseline model issues

Abstract

We motivate and propose a suite of simple but effective improvements for concept-to-text generation called SAPPHIRE: Set Augmentation and Post-hoc PHrase Infilling and REcombination. We demonstrate their effectiveness on generative commonsense reasoning, a.k.a. the CommonGen task, through experiments using both BART and T5 models. Through extensive automatic and human evaluation, we show that SAPPHIRE noticeably improves model performance. An in-depth qualitative analysis illustrates that SAPPHIRE effectively addresses many issues of the baseline model generations, including lack of commonsense, insufficient specificity, and poor fluency.