A Neural Question Answering System for Supporting Software Engineering Students

QA (Question Answering) is the task of automatically answer natural language questions posed by humans. Usually, QA approaches use a combination of computational linguistics, information retrieval and knowledge representation to find answers for questions. In a teaching-learning process, it is critical that teachers use a range of teaching strategies to effectively meet the needs of individual learners. Thus, QA approaches can be effectively used to support the teaching-learning process. In this article, we exploit neural networks for QA to support the teaching-learning process. Particularly, we use DMN+ (improved dynamic memory networks) and SeqToSeq (sequence to sequence) with a corpus of SE (software engineering) texts to effectively answer questions commonly posed by SE learners. Experimental results show that DMN+ is more effective than SeqToSeq for this task with up to 77% accuracy.