Error-driven Pruning of Language Models for Virtual Assistants

TL;DR

This paper introduces a customized entropy pruning technique for language models in virtual assistants, balancing model size and accuracy by selectively retaining infrequent n-grams, leading to improved WER performance.

Contribution

It proposes a novel keep list construction method for entropy pruning and discriminative techniques to reduce model size while maintaining accuracy gains.

Findings

8% average WER reduction on targeted test set

Best model is 3 times larger than baseline

Discriminative methods effectively reduce model size

Abstract

Language models (LMs) for virtual assistants (VAs) are typically trained on large amounts of data, resulting in prohibitively large models which require excessive memory and/or cannot be used to serve user requests in real-time. Entropy pruning results in smaller models but with significant degradation of effectiveness in the tail of the user request distribution. We customize entropy pruning by allowing for a keep list of infrequent n-grams that require a more relaxed pruning threshold, and propose three methods to construct the keep list. Each method has its own advantages and disadvantages with respect to LM size, ASR accuracy and cost of constructing the keep list. Our best LM gives 8% average Word Error Rate (WER) reduction on a targeted test set, but is 3 times larger than the baseline. We also propose discriminative methods to reduce the size of the LM while retaining the majority of the WER gains achieved by the largest LM.