Scalable Multi Corpora Neural Language Models for ASR

  • 2019-07-02 23:28:52
  • Anirudh Raju, Denis Filimonov, Gautam Tiwari, Guitang Lan, Ariya Rastrow
  • 4

Abstract

Neural language models (NLM) have been shown to outperform conventionaln-gram language models by a substantial margin in Automatic Speech Recognition(ASR) and other tasks. There are, however, a number of challenges that need tobe addressed for an NLM to be used in a practical large-scale ASR system. Inthis paper, we present solutions to some of the challenges, including trainingNLM from heterogenous corpora, limiting latency impact and handlingpersonalized bias in the second-pass rescorer. Overall, we show that we canachieve a 6.2% relative WER reduction using neural LM in a second-pass n-bestrescoring framework with a minimal increase in latency.

 

Quick Read (beta)

loading the full paper ...