Amazon has created a dataset of question-answer pairs to train question-answering models. The dataset contains 20,000 complex questions on music, sports, books, movies, geography, politics, video games, and history as well as answers sourced from Wikidata, an open knowledge database based in Germany. The team also translated each question-answer pair into Arabic, French, German, Hindi, Italian, Japanese, Portuguese, and Spanish.
Image credit: Flickr user Yasmeen
