README.md

metadata

license: other
language:
  - zh
language_bcp47:
  - zh-TW
task_categories:
  - automatic-speech-recognition
  - text-to-speech
pretty_name: Hung-yi Lee Lecture ASR/TTS
tags:
  - asr
  - tts
  - speech
  - mandarin
  - taiwan
  - lecture
size_categories:
  - 10K<n<100K

Hung-yi Lee Lecture ASR/TTS

This dataset contains speech segments aligned with subtitles from Hung-yi Lee lecture videos. It is intended for Mandarin ASR and TTS experiments.

Dataset Details

• Speaker: Hung-yi Lee
• Language: Traditional Chinese / Taiwan Mandarin (zh-TW)
• Sampling rate: 16 kHz
• Audio format in the published dataset: embedded FLAC audio in Parquet shards
• Source videos with audio: 15
• Segments: 29043
• Total duration: 16.92 hours
• Courses: 機器學習2026, 生成式人工智慧與機器學習導論2025

Columns

• audio: audio segment
• text: subtitle text for ASR/TTS
• normalized_text: currently identical to text
• language: zh-TW
• speaker_id: hung_yi_lee
• source_course, source_title, source_youtube_id, source_url: source metadata
• start, end, duration: segment timing in seconds
• segment_index: original VTT cue index

Notes

Segments are cut from the available MP4 files using their paired VTT subtitles. Items that only had subtitles/transcripts but no downloaded audio are not included as audio samples. Subtitle quality follows the source captions and may require further cleaning for strict TTS training.