Fact · Speech recognition
He and colleagues report that their RNN transducer can outperform a conventional CTC-based model in both latency and accuracy across a number of evaluation categories.
Checked against its source on October 17, 2026
In the source's own words
“we find that the proposed approach can outperform a conventional CTC-based model in terms of both latency and accuracy in a number of evaluation categories”
Streaming End-to-end Speech Recognition For Mobile Devices, Yanzhang He et al., arXiv (ICASSP 2019), 2019, Abstract.
Open the original- What it measures
- Latency here means how soon words appear while the user is still speaking. That is the reason transducers, which stream, are used for on-device dictation.
- What it does not show
- The comparison is between two designs trained by the same team on the same data. It does not show that every transducer beats every CTC model.