Does the Order of Training Samples Matter? Improving Neural Data-to-Text Generation with Curriculum Learning

TL;DR

This paper investigates how the order of training samples affects neural data-to-text generation, demonstrating that curriculum learning with a novel difficulty metric accelerates training and improves output quality.

Contribution

It introduces a soft edit distance difficulty metric for curriculum learning in data-to-text generation, leading to faster convergence and higher BLEU scores.

Findings

Training time reduced by 38.7%

BLEU score increased by 4.84 points

Faster convergence with curriculum learning

Abstract

Recent advancements in data-to-text generation largely take on the form of neural end-to-end systems. Efforts have been dedicated to improving text generation systems by changing the order of training samples in a process known as curriculum learning. Past research on sequence-to-sequence learning showed that curriculum learning helps to improve both the performance and convergence speed. In this work, we delve into the same idea surrounding the training samples consisting of structured data and text pairs, where at each update, the curriculum framework selects training samples based on the model's competence. Specifically, we experiment with various difficulty metrics and put forward a soft edit distance metric for ranking training samples. Our benchmarks show faster convergence speed where training time is reduced by 38.7% and performance is boosted by 4.84 BLEU.