FAIRView: Visual Explanations for Video Summaries

FAIRView Corpus: Visual Explanations for Video Summaries

This repository contains:

 200 videos machine annotated with key concepts extracted from video subtitles and video frames;

800 video summaries (four summaries per video) machine annotated with concepts extracted from video subtitles and frames;

3200 visual explanations (four explanations per video summary with different levels of transparency):

semantic coverage

semantic prominence

quantity coverage

distance

human validated utility of each visual explanation type (mTurk crowdsourcing study_1 with 20 videos and single summary)

human validated representativeness of video summaries with regard to their original videos (mTurk crowdsourcing study_2 with 18 videos with 2 summaries per video). In total 36 video-summary pairs.

All results are published in the following paper:

 Oana Inel, Nava Tintarev and Lora Aroyo: **Eliciting User Preferences for Personalized Explanations for Video Summaries**. UMAP 2020.

If you find the paper and the data useful in your research, please consider citing:

@inproceedings{inel2020eliciting,
  title={Eliciting User Preferences for Personalized Explanations for Video Summaries},
  author={Inel, Oana and Tintarev, Nava and Aroyo, Lora},
  booktitle={To Appear in the Proceedings of the 28th Conference on User Modeling, Adaptation and Personalization (UMAP)},
  year={2020},
  organization={ACM}
}