Development and Performance Evaluation of Chatbot System for Office Automation

Abstract

Many employees perform repetitive office tasks inefficiently, hence, spend considerable time on them. Furthermore, existing tools for saving time on repetitive tasks were developed for large-scale services such as banks and messengers. Accordingly, for small groups of workers, repetitive tasks continue to be time-consuming. This study developed a user-friendly chatbot system based on KakaoTalk and evaluated its performance in automating repetitive tasks in a small workforce. First, functional and performance requirements were identified on the basis of the KakaoTalk chatbot policy to construct a practical chatbot system. Subsequently, a KakaoTalk channel was established and a chatbot server was set up to operate the chatbot system. The first chatbot was developed in a simple manner for minimizing difficulties in its development and for user convenience. In order to meet the system requirements, the chatbot performance was improved by writing the source code of programs used by backend chatbot servers in different programming languages. Furthermore, the use of several techniques such as database redesign and load balancing for improving the server performance was examined. For the determination of the most effective technique for enhancing the chatbot performance, the total request per second and average latency were measured. By evaluating and comparing the performance of the first chatbot and the newly developed chatbot, it was verified that the chatbot system met the functional and performance requirements. Thus, this study shows what requirements should be set to operate the chatbot system and which method is effective to improve the chatbot's performance.