Fact · Speech recognition
The Conformer paper introduces the convolution-augmented transformer for speech recognition, named Conformer, which combines convolution neural networks with transformers.
Checked against its source on October 17, 2026
In the source's own words
“we propose the convolution-augmented transformer for speech recognition, named Conformer”
Conformer: Convolution-augmented Transformer for Speech Recognition, Anmol Gulati et al., arXiv (Interspeech 2020), 2020, Abstract.
Open the original- What it measures
- The convolution layers capture local patterns in the audio, and the attention layers capture long-range context. Most speech encoders since 2020 have used this combination or a close variant.
- What it does not show
- The claim that most later encoders use it is a summary of the field, not a figure from the paper. Conformer results are on LibriSpeech, an English read-speech benchmark.