Korean Audio Speech Sample with 516 Hours and 1,077 Speakers

Sample of 516 hours of Korean audio data, recorded by 1,077 speakers with a near-equal gender distribution. The recordings include daily language, interactive sentences, and command phrases, with each speaker contributing approximately half an hour of audio.

Use Cases

• Train an automatic speech recognition model on 516 hours of Korean audio data covering daily language and command phrases.
• Analyze speaker characteristics and gender distribution across the 1,077 speaker corpus for voice technology research.
• Develop and benchmark models for command recognition using the dataset's home and on-board command audio samples.

Strengths

• Contains 516 hours of Korean audio data, providing a substantial sample for speech tasks.
• Includes recordings from 1,077 speakers with a balanced gender distribution of 49% male and 51% female.
• Covers diverse speech contexts including daily language, interactive sentences, and specific commands.

Limitations

• The provided data is only a sample of a larger paid dataset, limiting its scope for full-scale model training.
• The dataset's temporal and geographic coverage is not specified, which may limit its applicability for certain analyses.
• The audio duration per speaker is limited to approximately 30 minutes, which may not capture extensive speaker variability.

Provenance

Source

Nexdata

Collection Method

Recorded audio data from speakers, including daily language, interactive sentences, and commands.

Freshness

Last updated on 2025-04-24.