Continuation of Chat Dialogue Using a Question Corpus: Chain of Question Generation based on Interrogative Estimation

Abstract

In recent years, with the advent of natural language processing technology, there has been a growing interest in dialogue systems. However, in chat dialogue between a system and a user, the system rarely chains questions, making it difficult for the user to maintain the conversation. In human dialogue, the introduction of a topic typically involves a sequence of questions and responses. In this study, we focus on this process and aim to facilitate natural chat dialogue by generating a chain of questions that promote communication. Here, we construct a corpus of questions from existing dialogue corpora. The question corpus consists of the dialogue history, the questions themselves, the nouns within the questions, and the corresponding responses. Based on this corpus, we estimate the appropriate interrogative using a related word as input. We generate a question using the obtained related word and the estimated interrogative, then further utilize the user's responses to create a chain of questions. The experimental results demonstrate the advantages of our proposed method compared to a baseline derived from previous studies.