Application of Abstractive Summarization in Multiple Choice Question Generation
Ayushi Mathur, M. Suchithra · 2022 International Conference on Computational Intelligence and Sustainable Engineering Solutions (CISES) · 2022
This research aims to provide a solution to automate multiple choice question generation using natural language processing supported by abstractive summarization. The text or chapter imported by the user is transformed with the help of PEGASUS for Abstractive Summarization developed by Google AI in 2020 enabling us to get important in a paraphrased manner. These sentences will be used as questions. The key-value or word which will be the answer and will be removed from the sentence is determined with the help of KeyBERT which uses Bidirectional Encoder Representations from Transformers (BERT) embeddings and was developed by Maarten Grootendorst. KeyBERT considers the semantics of a word while performing keyword extraction. To generate the incorrect options pertaining to the respective question, we use sense2vec which was trained on Reddit comments and returns distractors (word similar to our keyword/answer). We generate questions along with the options based on the text received from the user at the end of the process.