English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Internal language model (ILM) subtraction has been widely applied to improvethe performance of the RNN-Transducer with external language model (LM) fusionfor speech recognition. In this work, we show that sequence discriminativetraining has a strong correlation with ILM subtraction from both theoreticaland empirical points of view. Theoretically, we derive that the global optimumof maximum mutual information (MMI) training shares a similar formula as ILMsubtraction. Empirically, we show that ILM subtraction and sequencediscriminative training achieve similar performance across a wide range ofexperiments on Librispeech, including both MMI and minimum Bayes risk (MBR)criteria, as well as neural transducers and LMs of both full and limitedcontext. The benefit of ILM subtraction also becomes much smaller aftersequence discriminative training. We also provide an in-depth study to showthat sequence discriminative training has a minimal effect on the commonly usedzero-encoder ILM estimation, but a joint effect on both encoder and prediction+ joint network for posterior probability reshaping including both ILM andblank suppression.

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