Automatic Question Generation for Hindi and Marathi

Artificial Intelligence has contributed remarkably towards educational purposes in the last few decades. With the rise of online learning, most of the tedious tasks have been transferred from the human hands to machines and programs. Question generation from educational paragraphs has been a challenging activity for educators as it requires significant resources. Although there have been major breakthroughs in the English language towards this goal, there is a need for technology-enabled educational services in Indian regional languages. In this paper, we present automatic question generation for two Indian languages - Hindi and Marathi utilizing Sentence Constituency Parsing with openAI Generative Pre-Trained(GPT) and Text-to-Text Transfer Transformers(T5) models, evaluated by content experts of corresponding languages, generating optimum questions after removing repetitions and inaccuracies in structural constraints on composition.