Improving End-to-end Speech Recognition with Pronunciation-assisted Sub-word Modeling
This addresses a specific bottleneck in speech recognition systems for applications requiring high accuracy, though it is incremental as it builds on existing sub-word methods.
The paper tackles the problem of sub-word segmentation in end-to-end speech recognition by proposing pronunciation-assisted sub-word modeling (PASM), which leverages pronunciation information to improve accuracy, showing it outperforms character-based and byte-pair encoding baselines.
Most end-to-end speech recognition systems model text directly as a sequence of characters or sub-words. Current approaches to sub-word extraction only consider character sequence frequencies, which at times produce inferior sub-word segmentation that might lead to erroneous speech recognition output. We propose pronunciation-assisted sub-word modeling (PASM), a sub-word extraction method that leverages the pronunciation information of a word. Experiments show that the proposed method can greatly improve upon the character-based baseline, and also outperform commonly used byte-pair encoding methods.