This project generates multiple choice questions from text by identifying keywords as potential answers, replacing them with blanks to form question bases, transforming sentences into question format, and generating similar-word distractors as incorrect options. It employs spacy for NLP processing, naive bayes for classification, word embeddings and cosine similarity for keyword and distractor identification. The approach has been published at RANLP2021 conference and evolved to use T5 transformers in a related repository.