Exploiting neural networks in question answering to support the teaching-learning process

Abstract

Question Answering (QA) systems brought a fresh perspective to the Information Retrieval (IR) research area, enabling humans to ask natural language question to a computational system which can retrieve a single and precise answer. In front of this nature, we understand that this kind of approach could bring a valuable contribution to a Teaching- Learning process. This study aims to analyze how question answering algorithms performs when applied to an educational environment. To achieve this goal, we developed a Systematic Literature Review (SLR) which enlighten the concepts, definitions and patterns of QA field, guiding us on which algorithm, approach and paradigm were more suitable for our needs. After that we created our own educational corpus and tested two approaches with it. As a result, we concluded that, QA systems can be used as a important tool on a teaching-learning process as we could reach a 77% match on factoid answers.