A streaming ASR architecture that adds a frozen LLM as the internal language model of a factorized transducer, with vocabulary adaptation and weak-to-strong LM swapping, reports large WER reductions.
RNN- transducer with stateless prediction network,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Transducer-Llama: Integrating LLMs into Streamable Transducer-based Speech Recognition
A streaming ASR architecture that adds a frozen LLM as the internal language model of a factorized transducer, with vocabulary adaptation and weak-to-strong LM swapping, reports large WER reductions.