다중 전환 대화를 위한 상황 인지 질의응답 시스템

Abstract

Recently many companies have introducedartificial intelligence chatbots. Furthermore, manyrecent researches have been doing to pursue naturalconversation similar to human-interactiveconversation and to develop sequenced conversationsystem which make possible not only single-turnconversation but also multi-turn conversations.However, in case of changing the sequentialconversation on other subjects while having aconversation on one subject, many chatbot systemsmay not provide proper answers because of, lack ofawareness on the context of the changedconversation, In this study, we propose a newquestion answering system that exploits a subjectclassification model, answer prediction model, andhistorage storage for context-aware multi-turnconversations.