PsTuts-VQA Dataset

What is PsTuts-VQA?

PsTuts-VQA is a video question answering dataset on the narrated instructional videos for Adobe Photoshop. All the videos and their manual transcripts in English were obtained from the official website of the software. To collect a high-quality question answering dataset, three image editing experts using the software were hired and asked to watch the videos in sequence and generate a question and answer pair at a moment in the video. Each answer is linked to an entity or an option in an existing knowledge base for the software. The dataset includes the following resources: 76 videos (5.6 hours in total), 17,768 question-answer pairs, and a domain knowledge-base with 1,236 entities and 2,196 options.

Downloads

https://github.com/adobe-research/PsTuts-VQA-Dataset

Contact

Wentian Zhao

Paper

Video Question Answering on Screencast Tutorials

Wentian Zhao, Seokhwan Kim, Ning Xu, Hailin Jin

To appear in Proceedings of the 29th International Joint Conference on Artificial Intelligence and the 17th Pacific Rim International Conference on Artificial Intelligence (IJCAI-PRICAI 2020), Yokohama, Jul 2020.