Transformer-based language modeling and decoding for conversational speech recognition

  • 2020-01-04 23:27:59
  • Kareem Nassar
  • 0

Abstract

We propose a way to use a transformer-based language model in conversationalspeech recognition. Specifically, we focus on decoding efficiently in aweighted finite-state transducer framework. We showcase an approach to latticere-scoring that allows for longer range history captured by a transfomer-basedlanguage model and takes advantage of a transformer's ability to avoidcomputing sequentially.

 

Quick Read (beta)

loading the full paper ...