Taiwanese Mandarin Speech Recordings From 203 Mobile Phone Users

Comprising Taiwanese Mandarin speech recordings from 203 individuals in Taiwan, including 137 females and 66 males. The data was collected in a quiet indoor environment via mobile phone guiding. It is a sample of a larger paid dataset intended for speech and language model training.

Use Cases

• Train a speech recognition model on Taiwanese Mandarin audio recordings from mobile phones.
• Develop voiceprint recognition algorithms using audio data from 137 female and 66 male speakers.
• Analyze regional speech patterns from participants covering cities like Taipei, Kaohsiung, Taichung, and Tainan.
• Fine-tune machine translation models with paired audio and text data from this Taiwanese Mandarin corpus.

Strengths

• Includes recordings from 203 distinct speakers, providing a variety of voices.
• Demographic balance is specified with 137 female and 66 male participants.
• Data was collected in a controlled, quiet indoor environment to reduce background noise.
• Geographic coverage includes multiple major Taiwanese cities such as Taipei, Kaohsiung, Taichung, and Tainan.

Limitations

• The dataset is explicitly a sample of a larger paid dataset, so its scope is limited.
• Specific details on audio length, sampling rate, total hours of speech, or accompanying text transcripts are not provided.
• The gender distribution is imbalanced, with more than twice as many female as male speakers.

Provenance

Source

Nexdata

Collection Method

Recorded via mobile phone guiding in a quiet indoor environment.

Freshness

Last updated on 2025-04-23.

Geography

Taiwan, covering cities including Taipei, Kaohsiung, Taichung, and Tainan.