Indonesian Speech Recordings From 496 Native Speakers

Comprising 360 hours of Indonesian speech data collected from 496 native speakers. Each speaker recorded approximately 400 sentences covering categories like economics, entertainment, news, figures, letters, and oral content in a quiet environment.

Use Cases

• Train an automatic speech recognition (ASR) model on 360 hours of Indonesian audio covering multiple content categories.
• Analyze speaker variation and pronunciation patterns across a corpus of 496 native Indonesian speakers.
• Develop text-to-speech (TTS) models using the paired audio and text data from the reading tasks.

Strengths

• 360 hours of valid audio data provides a substantial resource for speech model training.
• Data collected from 496 native Indonesian speakers supports diversity in speaker characteristics.
• Recordings made in a quiet environment help ensure audio clarity and reduce background noise.

Limitations

• The dataset is described as a sample of a larger paid dataset, so the available scope may be limited.
• Specific details on audio formats, sampling rates, and text transcriptions are not provided in the input.
• Potential demographic or regional bias exists as the input does not specify the geographic distribution of the 496 speakers.

Provenance

Source

Nexdata

Collection Method

Recorded via mobile phone reading by native speakers in a quiet environment.

Time Range

null

Freshness

null

Geography

Indonesia (inferred from data collection from Indonesian native speakers)