PyTorch implementation of Video Summarization on Twitch (LOL) dataset
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


PyTorch implementation and dataset of Video Highlight Prediction Using Audience Chat Reactions, 2017 EMNLP.

Add the original chat files.

In the Google Drive, there are two new files and one new directory. is the script I used to cut the twitch chats to separate games.

 python (will read cut_video3.txt to generate chat file for each video and store in final_data)

Library Requirement

Apt-get install

  • ffmpeg


Dataset Download - Google Drive

cd EMNLP17_Twitch_LOL
python (This requires around 378 GB)

Run Training

  • Check the file. This script contains all the configurations of the experiments.
  • If you want to evaluate the trained model, simply use the same command used in training and add -e.


Please cite this paper in your publications if it helps your research:

  title = {Video Highlight Prediction Using Audience Chat Reactions},
  author = {Cheng-Yang Fu, Joon Lee, Mohit Bansal and Alexander C. Berg},
  booktitle = {EMNLP},
  year = {2017}