An approach to automatic answering for English reading comprehension tests

Abstract

This study focuses on the reading comprehension problem with multiple-choice answers, using the BERT model to achieve the highest performance. The ultimate goal is to create a solution to help solve reading comprehension problems without any reasoning or knowledge, suitable for the level of students in grades six and seven. The model will solve factoid questions from a given text. Our research topic will use a deep learning model-based approach to create a model that automatically answers the English reading comprehension question. We obtain promising results to give an accuracy of 78 percent.

Keywords

• bert
• multiple choice
• masked language model (mlm)
• question answering (qa)
• tokenization
• transformer
• word embeddings