Strategies for Training Large Vocabulary Neural Language Models

Abstract: Training neural network LLMs over large vocabularies is still computationally very costly compared to count-based models such as Kneser-Ney. At the same time, neural LLMs are gaining popularity for many applications such as speech recognition and machine translation whose success depends on scalability. We present a systematic comparison of strategies to represent and train large vocabularies, including softmax, hierarchical softmax, target sampling, noise contrastive estimation and self normalization. We further extend self normalization to be a proper estimator of likelihood and introduce an efficient variant of softmax. We evaluate each method on three popular benchmarks, examining performance on rare words, the speed/accuracy trade-off and complementarity to Kneser-Ney.