NN-grams: Unifying neural network and n-gram language models for Speech Recognition

TL;DR

NN-grams is a hybrid language model that combines n-grams and neural networks, improving speech recognition performance by leveraging both memorization and generalization capabilities.

Contribution

The paper introduces NN-grams, a novel hybrid model that unifies n-gram counts with neural networks for efficient and effective speech recognition.

Findings

NN-grams outperforms traditional n-gram models on an Italian speech recognition task.

The model is trained efficiently using noise contrastive estimation without an output soft-max layer.

NN-grams effectively combine the strengths of n-grams and neural networks for language modeling.

Abstract

We present NN-grams, a novel, hybrid language model integrating n-grams and neural networks (NN) for speech recognition. The model takes as input both word histories as well as n-gram counts. Thus, it combines the memorization capacity and scalability of an n-gram model with the generalization ability of neural networks. We report experiments where the model is trained on 26B words. NN-grams are efficient at run-time since they do not include an output soft-max layer. The model is trained using noise contrastive estimation (NCE), an approach that transforms the estimation problem of neural networks into one of binary classification between data samples and noise samples. We present results with noise samples derived from either an n-gram distribution or from speech recognition lattices. NN-grams outperforms an n-gram model on an Italian speech recognition dictation task.