Adding Connectionist Temporal Summarization into Conformer to Improve Its Decoder Efficiency For Speech Recognition
This work addresses decoding efficiency for speech recognition systems, offering incremental improvements in speed and accuracy for ASR applications.
The paper tackles the problem of improving decoding efficiency in Conformer models for speech recognition by proposing a novel connectionist temporal summarization (CTS) method, which reduces the decoding budget by up to 20% on LibriSpeech and 11% on FluentSpeech data without losing accuracy, and even improves accuracy on the LibriSpeech 'test-other' set with a 6% relative reduction in word error rate at beam width 1.
The Conformer model is an excellent architecture for speech recognition modeling that effectively utilizes the hybrid losses of connectionist temporal classification (CTC) and attention to train model parameters. To improve the decoding efficiency of Conformer, we propose a novel connectionist temporal summarization (CTS) method that reduces the number of frames required for the attention decoder fed from the acoustic sequences generated by the encoder, thus reducing operations. However, to achieve such decoding improvements, we must fine-tune model parameters, as cross-attention observations are changed and thus require corresponding refinements. Our final experiments show that, with a beamwidth of 4, the LibriSpeech's decoding budget can be reduced by up to 20% and for FluentSpeech data it can be reduced by 11%, without losing ASR accuracy. An improvement in accuracy is even found for the LibriSpeech "test-other" set. The word error rate (WER) is reduced by 6\% relative at the beam width of 1 and by 3% relative at the beam width of 4.