@inproceedings{tvqalocalizedcompositionalvideoquestion,
  title = {TVQA: Localized, Compositional Video Question Answering},
  author = {Jie Lei and Licheng Yu and Mohit Bansal and Tamara L. Berg},
  year = {2018},
  booktitle = {EMNLP 2018 10},
  eprint = {1809.01696},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1809.01696v2},
}