Skip to content

Nexdata-AI/3881-Hours-Mandarin-Spontaneous-Speech-Data

Repository files navigation

3881-Hours-Mandarin-Spontaneous-Speech-Data

Description

3,881 hours - Mandarin Spontaneous Speech Data, the content covering multiple subjects. All the speech audio was manually transcribed into text content; speaker identity, gender, and other attribution are also annotated. This dataset can be used for voiceprint recognition model training, corpus construction for machine translation, and algorithm research introduction, etc.

For more details, please refer to the link: https://www.nexdata.ai/datasets/speechrecog/1024?source=Github

Specifications

Format

16kHz, 16bit, wav, mono channel;

Content Category

Interview; Sports; Variety; Course; Entertainment, Service, etc.

Annotation

annotation for the transcription text, speaker identification, gender

Language

Mandarin

Accuracy

at a Sentence Accuracy Rate (SAR) of being no less than 95%

Application scenarios

speech recognition, video caption generation and video content review

Licensing Information

Commercial License