Contextual Adapters for Personalized Speech Recognition in Neural Transducers

TL;DR

This paper introduces neural contextual adapters for personalized speech recognition in neural transducer models, improving recognition of user-specific words without retraining the entire model.

Contribution

The paper presents a novel adapter-based method for personalization in neural transducer ASR models that outperforms shallow fusion and preserves pretrained model weights.

Findings

Adapters improve personalization over shallow fusion.

The method works with pretrained models without retraining.

Adapter training is more effective than full fine-tuning.

Abstract

Personal rare word recognition in end-to-end Automatic Speech Recognition (E2E ASR) models is a challenge due to the lack of training data. A standard way to address this issue is with shallow fusion methods at inference time. However, due to their dependence on external language models and the deterministic approach to weight boosting, their performance is limited. In this paper, we propose training neural contextual adapters for personalization in neural transducer based ASR models. Our approach can not only bias towards user-defined words, but also has the flexibility to work with pretrained ASR models. Using an in-house dataset, we demonstrate that contextual adapters can be applied to any general purpose pretrained ASR model to improve personalization. Our method outperforms shallow fusion, while retaining functionality of the pretrained models by not altering any of the model weights. We further show that the adapter style training is superior to full-fine-tuning of the ASR models on datasets with user-defined content.