Development of Analytical Module for Improving the Quality of the Automated Dialogue System

We have developed a system of automated robot to human dialogs. The system allows you to automatically make hundreds of calls per minute. The core of our system uses a full-text search by preliminary specified keys. The developed system showed high work efficiency but required the development of new analytical methods to improve call quality. This paper describes the basis of system architecture and the development of a Hadoop-based analytic module to solve existing problems. It also describes the methods and solutions that we used. It discusses the system requirements for availability and scalability.