This is the TED Talk dataset for VoiceCoach, CHI 2020, https://arxiv.org/abs/2001.07876.

The datasets contain meta information of 2623 TED Talks in official TED.com website until Jun 7th, 2019.

The meta information includes fields: 'author', 'datefilmed', 'totalviews', 'comments', 'language', 'downloadlink', 'vidlen', 'aws-transcripts', 'datecrawled', 'datepublished', 'title', 'id', 'url', 'keywords', 'videoname', 'ratings', and complete information is stored in the field 'alldata_JSON'.

• url: original video link
• aws-transcripts: Each video in the dataset is transcribed by AWS. It has two fileds, including:
  • transcript: all words in the video
  • words: an array containing detailed information about all words. e.g.,
    • "start_time": "12.94",
    • "end_time": "13.25",
    • "alternatives": [{"confidence": "0.9097", "content": "we"}], "type": "pronunciation"}]

tedvideo_download.py contains the code for downloading ted videos from TED.com

You can use ffmpeg to convert .mp4 to other audio formats (e.g., mp3, wav, etc.)

(updating)

If you use this dataset, please cite our paper

VoiceCoach: Interactive Evidence-based Training for Voice Modulation Skills in Public Speaking

Preprint: https://arxiv.org/abs/2001.07876

Authors: Xingbo Wang, Haipeng Zeng, Yong Wang, Aoyu Wu, Zhida Sun, Xiaojuan Ma, Huamin Qu

The dataset is shared under the Creative Commons license.