Automatic question generation based on sentence structure analysis using machine learning approach

Automatic question generation is one of the most challenging tasks Natural Language Processing. It requires "bidirectional" language processing: firstly, system has to understand input text (Natural Understanding) and it then generate questions also in form Generation). In this article, we introduce our framework for generating factual from unstructured English language. uses a combination traditional linguistic approaches based on sentence patterns with several machine learning methods. We firstly obtain lexical, syntactic semantic information an construct hierarchical set each sentence. The features extracted used automated new transformation rules. Our process totally data-driven because rules are obtained initial sentence-question pairs. advantages approach lie simple expansion which allows us various types continuous improvement by reinforcement learning. includes evaluation module estimates quality generated questions. serves as filter selecting best eliminating incorrect ones or duplicates. have performed experiments evaluate correctness compared state-of-the-art systems. results indicate that outperforms systems comparable created humans. published interface all datasets evaluated questions, so possible follow up work.