Accelerated Reinforcement Learning for Sentence Generation by Vocabulary Prediction

TL;DR

This paper introduces a dynamic vocabulary prediction method to reduce action space in reinforcement learning for sentence generation, resulting in faster training, less memory use, and improved BLEU scores.

Contribution

The paper proposes a novel dynamic vocabulary prediction approach that significantly improves reinforcement learning efficiency and performance in sentence generation tasks.

Findings

Achieves approximately 2.7x faster reinforcement learning

Uses about 2.3x less GPU memory

Attains equal or better BLEU scores with faster decoding

Abstract

A major obstacle in reinforcement learning-based sentence generation is the large action space whose size is equal to the vocabulary size of the target-side language. To improve the efficiency of reinforcement learning, we present a novel approach for reducing the action space based on dynamic vocabulary prediction. Our method first predicts a fixed-size small vocabulary for each input to generate its target sentence. The input-specific vocabularies are then used at supervised and reinforcement learning steps, and also at test time. In our experiments on six machine translation and two image captioning datasets, our method achieves faster reinforcement learning ($\sim$2.7x faster) with less GPU memory ($\sim$2.3x less) than the full-vocabulary counterpart. The reinforcement learning with our method consistently leads to significant improvement of BLEU scores, and the scores are equal to or better than those of baselines using the full vocabularies, with faster decoding time ($\sim$3x faster) on CPUs.