Doc2Vec based Question and Answer Search System

Abstract

E-learning interaction acts as a positive factor, such as improving learning commitment and learning effect and reducing the dropout rate. As an important function of e-learning interaction, if a learner queries a content that is difficult to understand during learning, a question-and-answer bulletin board that responds to the question is provided by a professor. In the way that the instructor directly answers the learner's questions, real-time feedback is difficult, and the instructor's fatigue increases. The purpose of this study is to achieve the goal of reducing answering time and reducing answering costs by developing a question-and-answer search system that automatically searches for and provides answers to questions created by learners during learning. To this end, this study designed and implemented a question-and-answer search system that provides the most similar query answers to learners by analyzing questions and answers based on Doc2Vec, one of the word embedding technologies, which is a natural language processing technology.   By applying the results of this study to the question-and-answer system, it is expected that the learning effect can be enhanced by providing an immediate answer to the learner's question. In addition, organizations that pay response fees through the national budget, such as the Korea Educational Broadcasting Corporation, will be able to focus more on investments such as improving content quality through budget reduction.