Semantic-based enhanced logistic regression model with dual mode classification for accurate MCQs generation

The developments in language models and architectures, such as the transformer, allow us to investigate how teachers benefit from these endeavors. In this research, neural language models are employed to generate Multiple Choice Questions (MCQs) automatically from Wikipedia. MCQs are a common form of assessment. Despite the advantages, creating MCQs manually is difficult, time-consuming, and prone to mistakes. This is because the development of multiple-choice questions requires knowledge from the MCQ developers, as each MCQ consists of a question, a correct answer, and other possibilities. This research presents a method that combines semantic and machine learning techniques to create technically accurate MCQ stems of varying sorts. This paper proposes a Semantic-based Enhanced Logistic Regression Model with Dual Mode Classification (SbELR-DMC) for accurate MCQs and Odd one-out model questions for assessment of students. The proposed model when contrasted with the traditional models exhibits better accuracy.