Unverified50% confidenceFactExact time
The study evaluates on the Whisper model family across three datasets: Fair-Speech, Common Voice 25, and AfriSpeech-200.
1
Sources
50%
Confidence
Long-term
Relevance
9/25/2026
First Seen
Sources
Related Entities
Related Claims
Unverified标准Whisper模型在处理歌声时准确率会下降，社区开发了whisper-timestamped等变体实现逐词级别时间轴对齐72% similarUnverified端侧模型的典型代表包括 Whisper 系列轻量版本（tiny/base/small）、Meta 的 MMS（Massively Multilingual Speech）以及各厂商自研的压缩模型69% similarUnverifiedMainstream ASR systems (such as Whisper and Google Speech-to-Text) are primarily trained on adult speech datasets, and their word error rates on children's speech are typically 30%–50% higher than on adult speech.68% similarUnverifiedWhisper 提供从 tiny 到 large 的多个规格版本，在准确率与计算资源消耗之间提供不同权衡68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/949154API
curl https://kongchang.com/api/v1/knowledge/claims/949154MCP
get_claim(id=949154)