Unverified50% confidenceFactExact time
KeyASRBench is a bilingual benchmark for measuring ASR systems' ability to preserve keywords (person names, product names, technical terms, version numbers, relational phrases) as whole phrases.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTraditional ASR systems handle proper nouns by building domain dictionaries and hotword lists, which require manual maintenance, undergo frequent version updates, and incur high costs for cross-domain migration.64% similarUnverifiedKeyphrase, keyword, and term are often used interchangeably in the literature, but the granularity and linguistic properties they refer to are not the same, and this ambiguity directly affects the comparability of evaluations.63% similarUnverifiedKeyASRBench currently covers Chinese and English and a fixed keyword set, and its control set is 30 minutes in size; its value lies in proposing a reproducible KER framework rather than providing a universally applicable industry threshold.63% similarUnverifiedSpeechFast provides a custom vocabulary feature, where users can add proper nouns, personal names and industry terms to improve recognition accuracy.62% similarUnverifiedIntra-sentential switching poses the greatest challenge to ASR systems, because it requires the model to complete language-identification switching within an extremely short time span while maintaining syntactically coherent understanding.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1153734API
curl https://kongchang.com/api/v1/knowledge/claims/1153734MCP
get_claim(id=1153734)