Blank Collapse: Compressing CTC emission for the faster decoding

TL;DR

This paper introduces a simple method to compress CTC emission calculations, significantly accelerating beam search decoding in speech recognition with minimal accuracy loss, validated through experiments and theoretical analysis.

Contribution

The paper proposes a novel, straightforward approach to reduce CTC emission calculations, enabling up to 78% faster decoding in speech recognition tasks.

Findings

Achieves up to 78% faster decoding speed

Minimal accuracy loss with the proposed method

More effective with higher model accuracy

Abstract

Connectionist Temporal Classification (CTC) model is a very efficient method for modeling sequences, especially for speech data. In order to use CTC model as an Automatic Speech Recognition (ASR) task, the beam search decoding with an external language model like n-gram LM is necessary to obtain reasonable results. In this paper we analyze the blank label in CTC beam search deeply and propose a very simple method to reduce the amount of calculation resulting in faster beam search decoding speed. With this method, we can get up to 78% faster decoding speed than ordinary beam search decoding with a very small loss of accuracy in LibriSpeech datasets. We prove this method is effective not only practically by experiments but also theoretically by mathematical reasoning. We also observe that this reduction is more obvious if the accuracy of the model is higher.