Analysis of Machine Reading Comprehension Problem Using Machine Learning Techniques

Abstract

Machine reading Comprehension is a significant challenge in the field of natural language programming. In this problem, the objective is to read and grasp a given text passage before responding to questions that are dependent on the material. The most modern machine reading comprehension systems have accuracy levels that are superior to those of humans. On the other hand, when domains are switched, the majority of machine reading comprehension systems see a considerable drop in performance. However, certain machine reading comprehension systems have previously outperformed humans on a range of standard datasets, despite the evident and vast disparity among them. This is the case even though MRC models are not designed to read like humans. This demonstrates the need for enhancing the currently available datasets, assessment criteria, and models to progress the machine reading comprehension models toward "actual" comprehension. In this work, the analysis of the machine reading comprehension problem performed by using logistic regression, K- nearest neighbor, and random forest. This strategy will include perspectives that are topic-oriented, concept-oriented, and time-oriented, and it will provide support for the summary of multilingual texts with the assistance of several machine reading comprehension models that are currently in development.