Fact · Speech recognition
Jurafsky and Martin's draft says encoder-decoder speech models implicitly learn a language model from their training data.
Checked against its source on October 17, 2026
In the source's own words
“encoder-decoders implicitly learn a language model for the output domain of letters from their training data”
Speech and Language Processing, Chapter 16: Automatic Speech Recognition (draft), Daniel Jurafsky and James H. Martin, Stanford University, online manuscript of the third edition, 2026, Chapter 16, automatic speech recognition.
Open the original- What it measures
- The draft explains that this implicit model is often not enough, because speech training data contains limited text, so an external language model can be added.
- What it does not show
- Describes the model type in general. Whether a given product adds its own language model stage is a separate question.