Modeling and Analysis in Technology Problem Answering System

Considering the coming of the big data era, we are going to explore the auto-answering problem in technology's curriculum tests with big data. This kind of test contains the properties in Science and Math tests. In this paper, we describe an alternative approach that operates at three layers of representation and reasoning: information retrieval, corpus statistics, and simple inference over a semi-automatically constructed knowledge base. We evaluate the approach on ten years of exam problems from Wuhan University of Technology Fundamentals of Mono-Chip Computers Exams, and show that our overall system has the score of 62.2%. We apply the Generative Adversarial Networks (GANs) thought in one of the methods, and get the improvement of 2.1% over the original methods. We try first for the problems of courses in Colleges and Universities, and leverage GANs in this Technology Problem Answering System.