On the Relation between Internal Language Model and Sequence Discriminative Training for Neural Transducers

arXiv:2309.14130v2 Announce Type: replace-cross
Abstract: Internal language model (ILM) subtraction has been widely applied to improve the performance of the RNN-Transducer with external language model (LM) fusion for speech recognition. In this work, we show that sequence discriminative training has a strong correlation with ILM subtraction from both theoretical and empirical points of view. Theoretically, we derive that the global optimum of maximum mutual information (MMI) training shares a similar formula as ILM subtraction. Empirically, we show that ILM subtraction …

abstract arxiv correlation cs.cl cs.lg cs.sd eess.as fusion language language model performance recognition rnn show speech speech recognition training type work