A Discriminative Entity-Aware Language Model for Virtual Assistants

TL;DR

This paper introduces a discriminative, entity-aware language model that leverages real-world knowledge from a Knowledge Graph to improve speech recognition accuracy for virtual assistants, especially on less common entities.

Contribution

It extends discriminative n-gram models by integrating KG-based features, enhancing recognition of named entities with minimal impact on overall performance.

Findings

Achieved over 25% relative error rate reduction on test sets with less popular entities.

Maintained performance on standard VA test sets.

Demonstrated effective incorporation of real-world knowledge into language modeling.

Abstract

High-quality automatic speech recognition (ASR) is essential for virtual assistants (VAs) to work well. However, ASR often performs poorly on VA requests containing named entities. In this work, we start from the observation that many ASR errors on named entities are inconsistent with real-world knowledge. We extend previous discriminative n-gram language modeling approaches to incorporate real-world knowledge from a Knowledge Graph (KG), using features that capture entity type-entity and entity-entity relationships. We apply our model through an efficient lattice rescoring process, achieving relative sentence error rate reductions of more than 25% on some synthesized test sets covering less popular entities, with minimal degradation on a uniformly sampled VA test set.