AI-Based Questions Generation System

Preparing assessment questions is an important part of teaching and learning, but manual question creation can be timeconsuming and inconsistent across topics and difficulty levels. This paper proposes an AI-Based Questions GenerationSystem that transforms supplied educational text, notes, or topics into draft questions in multiple formats, includingmultiple-choice, short-answer, and descriptive questions. The proposed workflow combines document extraction, textpreprocessing, key-concept identification, question generation, suggested-answer creation, distractor generation, qualitychecks, and a human-review interface. Each generated item should retain a reference to its source passage so that users canverify its correctness. A prototype can begin with transparent NLP rules and templates, then optionally integrate a languagemodel to improve wording and variety. The system is intended as an educational assistant rather than an autonomousassessment authority: generated questions may be ambiguous, inaccurate, repetitive, or poorly aligned with a syllabus.Educators must review questions before formal use. The paper proposes an evaluation plan based on relevance, answercorrectness, answerability, grammaticality, diversity, coverage, and educator acceptance. No experimental results areclaimed because the proposed system has not been evaluated in a controlled study