Unverified50% confidenceFactExact time
通用ASR模型的训练数据大多来自播客、有声书、YouTube视频等宽带高质量音频源，与呼叫中心的窄带嘈杂环境存在显著差异
1
Sources
50%
Confidence
Long-term
Relevance
8/5/2026
First Seen
Sources
Related Claims
UnverifiedMainstream 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.71% similarUnverified紧急报警场景中的语音特征（高语速、气喘、哭泣、语句中断重叠）在主流ASR训练数据集中严重不足，形成显著的领域偏移问题69% similarUnverifiedCurrent mainstream ASR systems such as OpenAI's Whisper and Google Speech-to-Text still experience significantly higher error rates when dealing with accents, variations in speech rate, background noise, and specialized terminology68% similarVerifiedModels such as Whisper and Wav2Vec 2.0 learn general acoustic representations on large amounts of unlabeled audio through self-supervised pre-training, reducing reliance on manually annotated data.68% similarUnverifiedAudio tokens are trained with the objective of reconstructing high-fidelity audio, retaining all acoustic information including timbre, emotion, and ambient sound, rather than preserving only linguistic content as ASR output does.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/692312API
curl https://kongchang.com/api/v1/knowledge/claims/692312MCP
get_claim(id=692312)