Slovak Dataset for Multilingual Question Answering

SK-QuAD is the first manually annotated dataset of questions and answers in Slovak. It consists more than 91k factual from various fields. Each question has an answer marked corresponding paragraph. also contains negative examples form "unanswered questions" "plausible answers". The published free charge for scientific use. We aim to contribute creation Slovak or multilingual systems generating a natural language. paper provides overview existing datasets answering. describes annotation process statistically analyzes created content. expands possibilities training evaluation language models. Experiments show that achieves state-of-the-art results improves answering other languages zero-shot learning. compare effect machine-translated data with annotated. Additional improve modeling low-resourced languages.