Librispeech Transducer Model with Internal Language Model Prior Correction
This work addresses a specific bottleneck in speech recognition for researchers and practitioners, offering an incremental improvement to transducer models.
The authors tackled the problem of integrating external language models into transducer-based speech recognition by proposing a method to subtract an estimated internal language model prior, achieving over 14% relative improvement over standard shallow fusion on Librispeech.
We present our transducer model on Librispeech. We study variants to include an external language model (LM) with shallow fusion and subtract an estimated internal LM. This is justified by a Bayesian interpretation where the transducer model prior is given by the estimated internal LM. The subtraction of the internal LM gives us over 14% relative improvement over normal shallow fusion. Our transducer has a separate probability distribution for the non-blank labels which allows for easier combination with the external LM, and easier estimation of the internal LM. We additionally take care of including the end-of-sentence (EOS) probability of the external LM in the last blank probability which further improves the performance. All our code and setups are published.