Abstract
Machine reading comprehension (MRC) has emerged as a pivotal area of research within the realm of natural language processing, aiming to equip machines with the ability to understand and extract information from textual sources. This study investigates diverse approaches employed in the field of MRC, analyzing methodologies, techniques and their respective strengths and limitations. We delve into traditional rule-based systems, statistical models and contemporary deep learning architectures, highlighting advancements in neural network-based approaches such as attention mechanisms and transformer models. Furthermore, we explore the challenges posed by different types of MRC tasks, including span extraction, multiple-choice questions and cloze-style questions, along with benchmark datasets commonly used for evaluation.