Unverified50% confidenceFactExact time
LibriSpeech test-clean contains 40 speakers and 2620 utterances, from which the author drew 50 stratified samples (10 speakers, 5 utterances each), about 1002 reference words and 356 seconds of audio.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
Unverified语音数据通常被编码为离散的音频token，每秒语音大约对应25-50个token70% similarUnverifiedLibriSpeech、CommonVoice 等主流语音识别训练集大多来自朗读录音或受控环境，信噪比高、语速均匀、句子完整68% similarUnverifiedThe researchers chose the acoustic organization of vowels as the testing platform, covering fifteen languages and twenty-seven representations.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1013777API
curl https://kongchang.com/api/v1/knowledge/claims/1013777MCP
get_claim(id=1013777)