UserLibri: A Dataset for ASR Personalization Using Only Text

TL;DR

This paper introduces UserLibri, a new dataset for on-device speech recognition personalization using only text data, demonstrating improved WER with personalized language models.

Contribution

It presents a novel dataset, UserLibri, and shows that personalized language models trained on text-only data can enhance speech recognition accuracy.

Findings

Lowered average word error rate across users

Achieved 2.5 WER reduction on test-other users with streaming models

Demonstrated effectiveness of text-only personalization for ASR

Abstract

Personalization of speech models on mobile devices (on-device personalization) is an active area of research, but more often than not, mobile devices have more text-only data than paired audio-text data. We explore training a personalized language model on text-only data, used during inference to improve speech recognition performance for that user. We experiment on a user-clustered LibriSpeech corpus, supplemented with personalized text-only data for each user from Project Gutenberg. We release this User-Specific LibriSpeech (UserLibri) dataset to aid future personalization research. LibriSpeech audio-transcript pairs are grouped into 55 users from the test-clean dataset and 52 users from test-other. We are able to lower the average word error rate per user across both sets in streaming and nonstreaming models, including an improvement of 2.5 for the harder set of test-other users when streaming.