Fact · Voice and AI
The 2024 paper says end-to-end speech recognition systems still have a significant error rate for entity names that appear rarely in their training data.
Checked against its source on October 17, 2026
In the source's own words
“these systems still have a significant error rate for entity names which appear infrequently in their training data”
Retrieval Augmented Correction of Named Entity Speech Recognition Errors, Ernest Pusateri and et al. (nine authors in total), arXiv, 2024, Abstract.
Open the original- What it measures
- Entity names are people, places, products and similar names. The paper proposes correcting them with retrieved names from a list.
- What it does not show
- The paper does not give one number for all calls. Its results are on synthetic voice assistant audio.