Skip to content

Dataset · CanCLID

Cantonese audiobooks and transcriptions

CanCLID/zoengjyutgaai

Recordings by famous storyteller Zhang Yuekai with verified texts for Cantonese TTS and ASR.

What's inside

The corpus contains between 100 thousand and 1 million Cantonese (yue) speech fragments performed by famous radio host Zhang Yuekai (four classical works including Romance of the Three Kingdoms). The audio is segmented and saved in opus format (sources stored separately), and the texts are transcribed in Chinese characters following strict orthography rules without numbers or Latin characters.

What you can do with it

  • Train Cantonese text-to-speech (TTS) models with expressive artistic intonation
  • Train and test Cantonese automatic speech recognition (ASR) systems
  • Linguistic research on spoken Cantonese speech and folklore
Tasks
Automatic speech recognition, Text → speech, Text generation, Embeddings, Audio → audio, Audio classification
Languages
Cantonese
Size
100K–1M rows
Downloads this month
10.5K
Likes
35

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("CanCLID/zoengjyutgaai")