Clone from "friedrichor/MSVD".
MSVD contains 1,970 videos, each of which is paired with ~40 captions.
We adopt the official split:
@inproceedings{chen2011collecting,
title={Collecting highly parallel data for paraphrase evaluation},
author={Chen, David and Dolan, William B},
booktitle={Proceedings of the Annual Meeting of the Association for… See the full description on the dataset page:
https://huggingface.co/datasets/VLM2Vec/MSVD.