Knowledge Management Approach for Memory Components Based on User-friendly Conversational System

Abstract

Due to the recent technological development and the growth of computing resources, there are various studies that apply large-scale language models in the field of natural language processing such as conversational systems. Also, there are researches that attempt to maintain a conversation flow and naturally lead a dialogue by treating the contextual information exchanged with users from the perspective of knowledge. In this paper, we propose a method for managing various contextual information such as chat history, situations, and preferred topics based on a knowledge base and generating a conversation customized for the specific user. We design a schema of memory components to deal with the user's contextual information so that implement a Web-based conversational system, which is friendly come to those users. It is expected that these systems using the user-specific memory components will be helpful in such domains like education or customer consultation.