Deep learning for automatic dialogue system

Abstract

Dialogue systems, also known as interactive conversations agents, virtual agents and sometimes chatterbots has been widely applied into daily life from entertainment to automation customer services such as personalised medical service and online shopping. The project aims to research and develop the state of the art deep learning technology for various modules of dialogue system. Main work item is to build up query understanding, knowledge database and reply generating modules. Technology wise, LSTM network will be core part of target system. In this project, quality of data input is one of the priority factors that affect output results. Hence, research on current existing dataset of sentence corpus will also be conducted. In the end a demo system should be provided to show dialogue system with daily conversation content. This system should able to generate a reply with a given sentence.