Unverified50% confidenceOpinionExact time
WER cannot fully represent the user experience of voice interaction: the handling of proper nouns, accents, numbers and homophones, as well as the restoration of punctuation and sentence segmentation, all significantly affect downstream understanding.
1
Sources
50%
Confidence
Long-term
Relevance
11/2/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedIn traditional voice interaction systems, paralinguistic information carried by speech—such as emotion, tone, and speech rate—is completely lost when converted to plain text during the ASR stage.73% similarUnverifiedIn voice agent scenarios, neither WER nor MOS is sufficient: WER only covers comprehension and does not reflect intent understanding; MOS only measures the naturalness of single-sentence synthesis and does not measure multi-turn dialogue context consistency or emotional expression.71% similarUnverifiedCurrent speech interaction models can say the right words but ignore non-verbal vocalizations, which imposes an inherent limitation on naturalness.71% similarUnverifiedThe experience quality of a voice product depends largely on the naturalness of the conversation and the coverage of scenarios.70% similarUnverifiedAutomatic speech recognition can still make errors in scenarios involving technical terminology, multiple speakers, and accents, directly affecting retrieval quality.69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1128118API
curl https://kongchang.com/api/v1/knowledge/claims/1128118MCP
get_claim(id=1128118)