Effective internal language model training and fusion for factorized transducer model

arXiv:2404.01716v1 Announce Type: cross
Abstract: The internal language model (ILM) of the neural transducer has been widely studied. In most prior work, it is mainly used for estimating the ILM score and is subsequently subtracted during inference to facilitate improved integration with external language models. Recently, various of factorized transducer models have been proposed, which explicitly embrace a standalone internal language model for non-blank token prediction. However, even with the adoption of factorized transducer models, limited improvement has been observed …

abstract arxiv cs.ai cs.cl cs.lg eess.as fusion inference integration language language model language models language model training prior training type work