■ The picture shows a safe robot patrol in an office building in Tokyo, Japan。 Picture Source: CFP

Human -machine dialogue system is one of the areas where artificial intelligence technology is most widely used in Japan。As early as the late 1980s，Japan applies voice recognition and synthetic technology to the development of human -machine dialogue system。In recent years，Japan encourages discipline cross -integration，Research on the human -machine dialogue system with linguistics theory，promoted the deep development of related research in this field。

　　Natural language processing builds human -machine communication bridge  

​​Artificial intelligence is a computer or a machine controlled by a computer，Simulation、Extension and expansion of human intelligence，Perception environment、Get knowledge and use knowledge to obtain the best results、Method、Technology and Application System，Core technology is mainly machine learning。Machine learning is a summary rule and judgment standard from the data，Then predict in accordance with these rules and standards、Artificial intelligence technology that judges unknown data，Contains technology such as deep learning。Deep learning refers to the multi -layer artificial neural network and its training method，The essence is to learn more useful features by constructing many hidden -layer machine learning models and massive training data，The accuracy of the final improvement or prediction。The motivation for research deep learning is to establish a neural network for the establishment of analog human brain for analysis and learning，bet365 Play online games The operating mechanism of imitating the human brain to explain the data。Current，Application deep learning technology has made major progress with image recognition、voice recognition、Forecast and natural language processing, etc.。where，Research in the field of natural language treatment involves natural language，That is the language used by people，closely related to linguistics。

Natural language processing is based on language as the object，Use computer technology to analyze、Technology of understanding and processing natural language。It is a communication bridge connecting machines and humans，It aims to realize good communication of human -machine。Natural language processing is divided into natural language understanding and natural language generation。The former means that the machine should have the ability to understand the meaning of natural language text，But due to the diversity of natural language、The characteristics of ambiguity and dependence context，Therefore, natural language understanding technology cannot reach the level of human understanding language; the latter requires that the machine can be expressed in natural language.、Thought, etc.，Including content OK、Sentence aggregation、Sentence aggregation、grammar、Reference expression generation and language implementation。Typical applications of natural language processing technology have emotional analysis、voice recognition、Machine translation and chat robot，where，Chat robot is a typical human -machine dialogue system。

　　Linguistics to help human -machine dialogue system research and development  

Starting from the beginning of artificial intelligence research，People are committed to developing a highly intelligent human -machine dialogue system。Human -machine dialogue system Bet365 lotto review is an intelligent system that allows machines to understand the natural language of human beings and can interact with humans。Its technical structure involves voice recognition、Semantic understanding、Dialogue management、Natural language generation and voice synthesis。According to the degree of openness，Can be divided into an open domain man -machine dialogue system and vertical domain man -machine dialogue system;，Can be divided into target -oriented human -machine dialogue systems and non -target -oriented human -machine dialogue systems; according to function，Can be divided into task -oriented man -type dialogue system and non -mission -oriented human -machine dialogue system。

The Japanese Renpu Dialogue System Research has shown vigorous vitality since the early 1990s，Get a series of results。The current focus of research and development is the non -task -to -dial -type chat -type man -machine dialogue system，Related research involves linguistics、Psychology、Sociology and other disciplines。In recent years，Institute of National Mandarin in Japan、Researchers from scientific research institutions and groups from the Society of Artificial Intelligence Society, from the perspective of different academic sciences，Actively research and development with natural language processing technology as the base，Human -machine dialogue system that adapts to different needs of users。For example，In the study related to linguistics，Researchers use the characteristics of phonology in voice science (such as promotion in Japanese、Dialing sound and other changes deformation laws)，Gives a variety of character personality and emotional characteristics of intelligence，Make them choose the appropriate language according to the user's personality。They also use vocabulary -related texts (Text Representation)bet365 best casino games 、Improve the topic expansion procedure and topic conversion program，Based on the principle of noun replacement, give the system to answer language agility。The predicate structure in the analysis sentence law、grid assistant matching, etc.，Researchers have designed it that can be extracted by extraction、Recommended information to actively turn on the dialogue system; polite models based on liking，The system that develops adjustable man -machine distance or human -machine relationship。In multi -mode field，Researchers comprehensively analyze the user's secretive speech and natural language.。

With the help of linguistics theory，The study of the Japanese machine dialogue system surpassed the traditional research mainly based on voice recognition and semantic judgment，Start focusing on voice、Semantic、Multi -mode development and system language design at three levels of linguistics。Human -machine dialogue system is a two -way communication between humans and machines，One of its important functions is to actively recommend information for users.。So，​​Some researchers start from the perspective of the system，By writing the correlation and discreteness of the judgment of the vocabulary、Program and other procedures from the user's past chat record，Design and development of a system that can actively mention the topic。Some researchers use discourse similarity、Maximum an average information and other methods to verify the intention of the user's topic，Provide help for the correct identification of users to identify the user correctly。Overall，The study of the Japanese human machine dialogue system can be divided into two categories according to the research paradigm: one is to examine the natural language characteristics in interpersonal exchanges first，Design program based on its survey results，bet365 live casino games Then use the dialogue between the user and the system to verify whether the system language is natural，further optimize the program according to the test results; the other is to build a corpus based on a large number of text information on the webpage，This is used as a system language data to design a human -machine dialogue system，Some studies after writing the program，Verify the natural nature of the system language by investigating the dialogue between the user and the system。

　　Discourse polite theory provides problem solutions  

Although the research results of the human -machine dialogue system in Japan involved multiple words in the system language、Different types of speech behavior and diversity topics，But there are almost no artificial intelligence technical experts who can give the wheel with the highest frequency of use、Reasonable solutions for speech behavior sequence and topic expansion mode。Dialogue theory in the field of linguistics provides a basis for solving the above problems。

The theory of polite discourse is proposed by Japanese scholar Uzumi Mami Mayumi，It is one of the most influential polite theories in the Japanese linguistic industry。Basic concept of discourse in this theory，Through the age of limited session scenarios and session participants、Identity、Gender and other factors，Typical language elements in different scenes can be determined、Explanation of specific speech behavior chains and topic expansion mode。Current，Discussion of speech behavior based on the basic state of interpersonal dialogue (such as invitations、Request、Refusal、Apology, etc.)、Language Elements (Knight Word、Attopia、Polite term) and topic introduction of topics。Taking request behavior as an example，The expression of the Chinese mother tongue is chain chain: attract attention → expressing the patriotic → explanation Bet365 app download situation → proposed requests → statement auxiliary behavior → additional explanation → insert other topics → re -request → express for → end the session。Artificial Intelligence Technology Experts can directly apply the above conclusion to the program writing of the human -machine dialogue system in conjunction with the actual situation，Difficult to judge the process of solving typical language elements and specific speech behaviors、Production Language Machine Conditioning and Rich Questions。，You can also expand the topic identified in natural language research (such as the topic expansion mode of the topic in the first meeting of Japanese college students: Knight → Introduce personal information → talk about college life and related content) Import the human -machine dialogue system，​​When solving the topic of the system, ethical and moral issues may violate users' personal privacy。

Cross -domain collaboration between Japanese artificial intelligence technical experts and linguists shows that only humans understand how to run their own language，The simulation of the machine's simulation of human language can only be based on。Although deep learning can provide gratifying prediction results，But it is still difficult to explain the specific semantics of each dimension of the result vector of the result。So，Human -machine dialogue system should not only be used as statistical results under big data led，Linguistics results in the field of interpersonal dialogue should be used，Make it facing the natural language processing in the field of fusion deep learning、Image recognition、voice recognition、Four the development of the four major technical directions，Gradually develop into a multi -mode man -machine dialogue system with high reasoning capabilities。

　　(This article is the Study on the Research on the Phasement of the Japanese Social Science Foundation Western Project "The Research on the Tips for the Personal Dialogue System for the Human Machine Dialogue" (21xyy009)  

　　(Author Unit: School of Culture and Economics, Xi'an University of Foreign Languages ​​University; Graduate School of Xi'an University of Foreign Studies)