A Dialogic Syntax Approach To The Coherence Achievement In Human-to-AI-Software Verbal Interactions

With the development of artificial intelligence(AI),the intelligent chat software such as Microsoft XiaoIce(微软小冰)has come into our daily life,and human-to-AI-software verbal interaction has become a research hotspot nowadays.Scholars have studied the existing robots,revealed the shortcomings of human-to-AI-software verbal interaction,and also put forward valuable suggestions.However,few scholars have studied human-to-AI-software verbal interaction from the perspective of linguistics.In particular,few scholars analyze human-to-AI-software verbal interaction from the perspective of discourse coherence.Therefore,this thesis attempts to study human-to-AI-software verbal interaction from the viewpoint of coherence achievement.The problem of discourse coherence is the focus of linguistic research.The research on the coherence achievement in human-to-AI-software verbal interactions is helpful to promote the development of human-machine dialogue system.Since the publication of Cohesion in English by Halliday and Hasan,coherence research has attracted much attention.However,for nearly half a century,since most scholars have been devoted to the study of monologic discourse coherence,now the coherence of discourse interaction needs more research.In recent years,cognitive functional linguistics has paid more and more attention to discourse interaction,focusing on how people understand each other in interaction.Du Bois’ dialogic syntax theory is the latest theoretical achievement in the field of cognitive functional linguistics.We try to use Du Bois’ dialogic syntax theory to analyze the coherence achievement in human-to-AI-software verbal interactions,which is a new theoretical perspective for coherence research.Paul Thagard has made great achievements in coherence analysis.He has systematically explored cognitive coherence in his book,Coherence in Thought and Action.Thagard has studied cognitive coherence mainly from the perspective of the relationship between thought and action,but paid little attention to the coherence inlanguage form.We can still use Thagard’s coherence theory to explain the coherence in human-to-AI-software verbal interactions.Du Shihong has made a study on the coherence of dialogues.He believes that the ideal object of coherence research is discourse in interaction.He proposed a new approach,that is,Line Identification for Coherence Analysis(LICA),which divides coherence into dictional,emotional,intentional,and rational coherence.It is because of these coherence in the dialogues that people can communicate normally.Discourse interaction is a coherent interaction between the minds of both sides of the dialogue.The two sides of the dialogue must construct each other’s minds in order to form a unity of discourse interaction.Therefore,the problem of other minds plays an influential role in dialogue analysis.Based on Du Bois’ dialogic syntax theory,Thagard’s theory of coherence and Du’s theory of coherence,this research attempts to address three issues.First,is there any coherence achievement in human-to-AI-software verbal interactions? Second,what kind of coherence can be found in human-to-AI-software verbal interactions?Third,How to explain the lack of coherence in human-to-AI-software verbal interactions? In order to solve the above questions,we take Microsoft XiaoIce as the research object,collect the dialogues between users and Microsoft XiaoIce on the WeChat and QQ platforms,to analyze the coherence in human-to-AI-software verbal interactions.This thesis mainly consists of three parts.The first part is the literature review part,which recapitulates some previous research accomplishments.For instance,the study of artificial intelligence,especially the study of Microsoft XiaoIce and human-to-AI-software verbal interaction;the study of the problem of other minds includes the theories of mind and solutions to the problem of other minds;the study of coherence mainly combs various coherence theories.The second part is the theoretical foundations part.This part introduces three theories,Du Bois’ dialogic syntax theory,Thagard’s theory of coherence and Du’s theory of coherence.The third part is the case study part,which analyzes the coherence of the dialogues between human and Microsoft XiaoIce under the guidance of the theories introduced in the second part.According to Du Bois’ dialogic syntax theory,it is concluded that there is a certain coherence achievement in the dialogues between human and Microsoft XiaoIce,andthis coherence is closely related to the explicit resonance resources.Based on Du’s theory of coherence,we can draw a conclusion that there are dictional,intentional,rational,and emotional coherence in the dialogues between human and Microsoft XiaoIce.According to Thagard’s theory of coherence,to a great extent,the lack of the mind of Microsoft XiaoIce leads to the lack of the coherence in the dialogues between human and Microsoft XiaoIce.This research finds that there is a certain coherence achievement in human-to-AI-software verbal interactions,and we can find dictional,intentional,rational,and emotional coherence in human-to-AI-software verbal interactions;nevertheless,because AI softwares have no minds,they cannot think like human beings,and cannot really understand human’s words,so the human-to-AI-software verbal interaction often lacks coherence,which affects the fluency of dialogues.This thesis provides a linguistic basis for the future study of human-to-AI-software verbal interaction.In the future,researchers can pay more attention to linguistic phenomena to accelerate the upgrade and development of human-machine dialogue system.