Automated question generation and question answering from Turkish texts

While exam-style questions are a fundamental educational tool serving variety of purposes, manual construction is complex process that requires training, experience and resources. Automatic question generation (QG) techniques can be utilized to satisfy the need for continuous supply new by streamlining their generation. However, compared automatic answering (QA), QG more challenging task. In this work, we fine-tune multilingual T5 (mT5) transformer in multitask setting QA, answer extraction tasks using Turkish QA datasets. To best our knowledge, first academic work performs automated text-to-text from texts. Experimental evaluations show proposed achieves state-of-the-art performance on TQuADv1, TQuADv2 datasets XQuAD split. The source code pretrained models available at https://github.com/obss/turkish question-generation.