CTCBERT: Advancing Hidden-unit BERT with CTC Objectives

TL;DR

This paper introduces CTCBERT, a training method for hidden-unit BERT that uses CTC objectives to improve alignment flexibility and overall speech recognition performance, showing consistent WER improvements on Librispeech.

Contribution

The paper proposes CTCBERT, a novel training approach that replaces CE loss with CTC loss for hidden-unit BERT, enhancing alignment learning and speech recognition accuracy.

Findings

CTCBERT outperforms HuBERT with 2%-11% relative WER reduction.

Using CTC objectives improves alignment flexibility in training.

Slight finetuning benefits observed with blank-related parameters.

Abstract

In this work, we present a simple but effective method, CTCBERT, for advancing hidden-unit BERT (HuBERT). HuBERT applies a frame-level cross-entropy (CE) loss, which is similar to most acoustic model training. However, CTCBERT performs the model training with the Connectionist Temporal Classification (CTC) objective after removing duplicated IDs in each masked region. The idea stems from the observation that there can be significant errors in alignments when using clustered or aligned IDs. CTC learns alignments implicitly, indicating that learning with CTC can be more flexible when misalignment exists. We examine CTCBERT on IDs from HuBERT Iter1, HuBERT Iter2, and PBERT. The CTC training brings consistent improvements compared to the CE training. Furthermore, when loading blank-related parameters during finetuning, slight improvements are observed. Evaluated on the Librispeech 960-100h setting, the relative WER improvements of CTCBERT are 2%-11% over HuBERT and PERT on test-other data.