Accelerated Reinforcement Learning for Sentence Generation by Vocabulary Prediction
This addresses efficiency bottlenecks in reinforcement learning for NLP tasks like machine translation and image captioning, offering practical speed and memory gains, though it is incremental as it builds on existing methods.
The paper tackles the large action space problem in reinforcement learning for sentence generation by introducing dynamic vocabulary prediction to reduce the vocabulary size per input, resulting in faster reinforcement learning (~2.7x), less GPU memory usage (~2.3x less), and improved or equal BLEU scores with faster decoding (~3x faster on CPUs).
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.