Peer-reviewed research
Effects of Language Modeling and its Personalization on Touchscreen Typing Performance
Andrew Fowler et al., Proceedings of the 33rd Annual ACM CHI Conference on Human Factors in Computing Systems, 2015 · DOI 10.1145/2702123.2702503
Read the originalRead on October 16, 2026
- What it is
- A simulation-based evaluation of a touchscreen keyboard that combines a spatial model of key taps with a language model, which is the basis of most autocorrect. It reports word error rates with and without language modeling.
- How much weight it carries
- Peer-reviewed CHI paper from Google Research and collaborators. Typing was simulated with a noise model, not measured with real people, so the error rates describe the simulation.
Facts taken from this source
- In the simulated touchscreen study, combining a spatial model with a language model reduced word error rate from a 38.4 percent baseline to 5.7 percent.
“a combined spatial/language model reduces word error rate from a pre-model baseline of 38.4% down to 5.7%” (Abstract)