Cantonese Storytelling Audio Collection by Zhang Yuekai

A Cantonese audio dataset features storyteller Zhang Yuekai narrating four classic literary works, including 'Romance of the Three Kingdoms' and 'Water Margin'. It is designed for TTS and ASR model training, as well as linguistic and literary research. The dataset contains audio files and corresponding standardized text transcripts.

Use Cases

• Train a Cantonese TTS model using the standardized text transcripts and corresponding 48kHz opus audio files.
• Develop or benchmark a Cantonese ASR system on the storyteller's audio recordings.
• Analyze linguistic patterns in classic literature using the full-width punctuation and character-based text transcripts.
• Study the prosody and style of a specific Cantonese storytelling artist across four major works.

Strengths

• Audio is provided in a 48kHz opus format suitable for direct training use.
• Text transcripts are standardized using full-width punctuation and character-based writing per specific typographic guidelines.
• Features a culturally significant Cantonese storyteller with recordings spanning from the 1970s.

Limitations

• The dataset is limited to four specific literary works, not a broad sample of Cantonese speech.
• Audio source quality and recording conditions over the decades may be inconsistent.
• Sample data, row counts, and file size are not provided for assessment.

Provenance

Source

CanCLID

Collection Method

Audio collected from Guangdong radio broadcasts and segmented for training.

Time Range

Recordings span from the 1970s onward.

Freshness

null

Geography

Guangdong, China; Cantonese language focus.