Fact · Accessibility and health
23.8percent of utterances truncated
In the same study, the baseline endpointer cut off an average of 23.8 percent of utterances from people who stutter.
Checked against its source on October 16, 2026
In the source's own words
“the baseline endpointer model truncates on average 23.8% of utterances”
From User Perceptions to Technical Improvement: Enabling People Who Stutter to Better Use Speech Recognition (full text), Colin Lea et al., Proceedings of the 2023 CHI Conference on Human Factors in Computing Systems (arXiv preprint), 2023, Section 4.2.1 (endpointer).
Open the original (DOI 10.48550/arXiv.2302.09044)- What it measures
- The endpointer decides when a person has finished speaking. The figure is the average across 41 Phase 2 participants.
- What it does not show
- Truncation is cutting the speech off too early. The authors attribute the high rate to the baseline model being trained mostly on general population speech, which is poorly matched to these speakers.
Source
Cited on
- Voice Assistants vs Dictation: Siri, Alexa, GeminiGuide
- Real-Time Speech Recognition: Streaming and LatencyGuide
- AI Voice Agents: How They Work and Where They FailGuide
- Speech Recognition for Stuttering and DysarthriaGuide
- Does speech recognition work if you stutter?Question
- How does speech recognition know when you stop talking?Question
- How fast does a voice AI agent need to respond?Question