Automatic Question Answering System Development

Abstract

A question answering system could read a huge amount of given text and return answers based on what are asked. In this project, the system is comprised of deep neural network which could improve its 'intelligence' when the model is fed a bunch of text with specific domains. It could infer across sentences to find the most suitable answer with the attention mechanism which works like human being during searching answers from texts. The core achievement of this project is that the trained model with 1,000,000 training iterations of dataset in natural language helps the system to provide answers with accuracy of over 95%. In this report, you will know a state-of-the-art technology of Question answering system and details on how the system intimating human beings to answer questions with the Dynamic Memory network.