Natural language processing and intelligent customer service dialogue management method

Abstract

The invention provides a natural language processing and intelligent customer service dialogue management method, which comprises the following steps of S1, defining business knowledge and business rules of an intelligent customer service module, and establishing a corresponding semantic model; s2, a user request receiving unit in the module is constructed to obtain request content input by a user, the request content is matched according to a predefined semantic model through an analysis unit, and whether the request content belongs to the service category of a professional module or a common service module or not is judged; s3, inputting the request content into an expert module, judging whether the request content is a question in an expert knowledge base or not according to a learning rule of the expert module, and if so, giving an answer by the expert module according to a matching rule; if not, entering the intelligent learning unit, and giving an answer according to a learning rule of the intelligent learning unit and a matching rule; s4, the request content is input into the common service module, and the common service module gives an answer according to the question and answer rule.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent dialogue management, and in particular relates to natural language processing and intelligent customer service dialogue management methods. Background Art

[0002] At present, natural language processing (NLP) is an important branch in the field of artificial intelligence, which involves the study of computer understanding, processing and generation of natural languages ​​(such as English, Chinese, etc.). Intelligent customer service is a customer service system based on artificial intelligence and natural language processing technology, which can understand users' questions and provide accurate answers and solutions. However, although natural language processing and intelligent customer service technology have made significant progress, the existing technology still has some obvious shortcomings. The ambiguity and ambiguity of natural language bring challenges to processing tasks. The same sentence may have different interpretations in different contexts, which requires computers to perform complex contextual reasoning when understanding and processing natural language. In addition, the expression of natural language is often subjective, and different people may have different descriptions and evaluations of the same thing, which increases the difficulty of natural language processing tasks.

[0003] Although the intelligent customer service system has the ability to learn automatically, its learning ability is still limited. The system needs to deal with many different customers and fields, and the situations it faces are complex and changeable. At present, its learning ability cannot fully meet all needs. As a result, in some cases, the intelligent customer service may not accurately understand the user's problem, or cannot provide satisfactory answers and solutions.

[0004] Intelligent customer service also has shortcomings in terms of the accuracy of problem identification. Currently, it is mainly based on the enterprise knowledge base and uses keyword matching to answer user questions. When it cannot fully correspond to customer questions, it is impossible to do a good job of context association and context analysis, and when the user's question is not fully stated, the accuracy of the answer will be affected, reducing the user experience. Intelligent customer service also has shortcomings in information processing, especially in voice and video processing. Although some intelligent customer service can provide voice, video and other responses, the processing ability is weak when analyzing pictures, videos, tables and other content provided by users, and it is difficult to analyze the key points.

[0005] Intelligent customer service is also relatively weak in handling complex issues. It can handle relatively simple and routine issues, but when faced with more complex issues or special situations, it may not be able to give satisfactory answers or solutions, and human customer service intervention is needed to better solve the problem.

[0006] Although natural language processing and intelligent customer service technologies have made significant progress, they still have shortcomings in processing the ambiguity and ambiguity of natural language, learning ability, problem identification accuracy, information processing capabilities, and the ability to handle complex problems. Summary of the invention

[0007] The present invention proposes a natural language processing and intelligent customer service dialogue management method. This technical solution solves the problem of accurate matching and intelligent response of intelligent customer service in natural language processing and dialogue management, especially in distinguishing the service scope of professional modules and general service modules, and using expert modules and intelligent learning units to match and update knowledge, thereby achieving efficient and accurate user request processing.

[0008] The technical solution of the present invention is implemented as follows: a natural language processing and intelligent customer service dialogue management method, the method comprising the following steps:

[0009] S1: Define the business knowledge and business rules of the intelligent customer service module and establish the corresponding semantic model;

[0010] S2: constructing a user request receiving unit in the module to obtain the request content input by the user, matching the request content with a predefined semantic model through an analysis unit, and determining whether the request content belongs to the service scope of the professional module or the general service module;

[0011] S3: Input the request content into the expert module, and determine whether it is a question in the expert knowledge base according to the learning rules of the expert module. If so, the expert module gives an answer according to the matching rules; if not, enter the intelligent learning unit, and give an answer according to the learning rules of the intelligent learning unit and the matching rules;

[0012] S4: Input the request content into the general service module, and the general service module gives an answer according to the question-answering rules; if the general service module or the expert module gives preset answers, it enters the intelligent learning unit to update the knowledge base. If non-preset answers are given, it enters the dynamic learning unit to update the knowledge base through manual intervention and update the rules simultaneously.

[0013] In terms of the definition of business knowledge and business rules, this technical solution not only establishes a semantic model, but also clearly distinguishes the service scope of professional modules and general service modules. This distinction enables the system to more accurately determine the type of user request and call the corresponding processing module. However, existing technologies often lack such a clear distinction, resulting in low efficiency and even misjudgment when dealing with complex or professional problems.

[0014] In terms of matching and judging the request content, the technical solution adopts a more refined analysis method. By constructing a user request receiving unit and an analysis unit, the system can match the request content according to a predefined semantic model to determine whether it belongs to the service scope of a professional module or a general service module. This refined matching method greatly improves the processing efficiency and accuracy of the system.

[0015] The technical solution has also made innovations in the application and updating of the knowledge base. The system not only sets up expert modules to handle professional problems, but also introduces intelligent learning units and dynamic learning units to update and improve the knowledge base. When the expert module cannot give an answer, the intelligent learning unit will match and give an answer according to the learning rules; if the answer is a preset answer, the knowledge base will be updated; if it is not a preset answer, it will enter the dynamic learning unit and be updated through manual intervention. This knowledge base application and update mechanism enables the system to continuously learn and improve, and improve its ability to handle complex problems.

[0016] However, existing technologies often rely on fixed knowledge bases and rule bases and lack the ability to self-learn and update. When encountering new knowledge or new problems, existing technologies often cannot give satisfactory answers and may even make mistakes. In addition, existing technologies also have great limitations in updating knowledge bases, which usually require a lot of manual work, are inefficient and prone to errors.

[0017] As a preferred implementation, the business knowledge includes user question and answer knowledge, product knowledge, product feature knowledge, business process knowledge, decision tree knowledge and question and answer rules. The business rules include process rules, logic rules and matching rules. The process rules, logic rules and matching rules are uniformly managed through a rule engine.

[0018] As a preferred implementation, the semantic model is established through the following steps: first, a semantic dictionary is constructed, then part of speech is selected, semantic templates are expanded and modified, and finally, the semantic templates are applied.

[0019] As a preferred implementation, the semantic dictionary is constructed by extracting product characteristics, question-answering rules, question-answering knowledge, and combining user operation experience; after the construction is completed, the mapping relationship between parts of speech, intent, and slots is established through part-of-speech selection training, and the corresponding relationship between parts of speech, intent, and slots is determined; the semantic template is expanded in combination with question-answering rules, question-answering knowledge, and a semantic dictionary; through the correction step of the semantic template, erroneous or inaccurate semantic templates are screened out, and through the application step of the semantic template, the semantic template, question-answering rules, question-answering knowledge, semantic dictionary, and extended question-answer responses are combined to finally complete the establishment of the semantic model.

[0020] As a preferred implementation, the workflow of the expert module is to first construct an algorithm framework for intelligent learning, then define intelligent learning rules within the algorithm framework, and establish an automatic learning system. The automatic learning system automatically completes knowledge extraction by introducing learning rules and stores the knowledge in an expert knowledge base.

[0021] As a preferred implementation, the automatic learning system pre-processes information through knowledge rules, extracts valid keywords in user requests, determines whether the keywords in user requests belong to product features, obtains product data corresponding to product features from question-and-answer rules, constructs FAQ knowledge points, adds the extracted keywords to FAQ, matches the rule base of the rule engine, generates corresponding slot labels, matches the extracted intent keywords with the FAQ knowledge points through the semantic model, completes the match and generates corresponding slot labels, matches the slot labels with the rule engine through the slot labels, generates replies through the semantic model, matches the generated replies with the rule engine, and completes knowledge recommendation.

[0022] As a preferred implementation, the dynamic learning unit connects unanswerable questions to dynamic learning, and obtains keyword information of the request content in the user request; if it is a rule engine, the keyword information that does not appear in the knowledge base of the rule engine is stored in the dynamic knowledge base; the unanswered questions are classified into different dynamic learning pools according to categories and stored in the dynamic knowledge base. The manual customer service determines the keyword information in the dynamic knowledge base, determines the product characteristics to which the keyword information in the dynamic knowledge base belongs, and then obtains the product characteristics, adds them to the dynamic knowledge base, and completes the learning process.

[0023] After adopting the above technical solution, the beneficial effect of the present invention is that in the definition of business knowledge and business rules, the technical solution not only establishes a semantic model, but also clearly distinguishes the service scope of professional modules and general service modules. This distinction enables the system to more accurately determine the type of user request, thereby calling the corresponding processing module. However, the prior art often lacks such a clear distinction, resulting in low efficiency when dealing with complex or professional problems, and may even cause misjudgment. In matching and judging the request content, the technical solution adopts a more refined analysis method. By constructing a user request receiving unit and an analysis unit, the system can match the request content according to a pre-defined semantic model, thereby judging whether it belongs to the service scope of a professional module or a general service module. This refined matching method greatly improves the processing efficiency and accuracy of the system. In the application and updating of the knowledge base, the technical solution has also been innovative. The system not only sets up an expert module to handle professional problems, but also introduces an intelligent learning unit and a dynamic learning unit to update and improve the knowledge base. When the expert module cannot give an answer, the intelligent learning unit will match and give an answer according to the learning rules; if the answer is a preset answer, the knowledge base will be updated; if it is not a preset answer, it will enter the dynamic learning unit and be updated through manual intervention. This knowledge base application and update mechanism enables the system to continuously learn and improve, and improve its ability to handle complex problems. However, existing technologies often rely on fixed knowledge bases and rule bases, and lack the ability to self-learn and update. When encountering new knowledge or new problems, existing technologies often cannot give satisfactory answers, and may even make mistakes. In addition, existing technologies also have great limitations in updating knowledge bases, which usually require a lot of manual work, are inefficient and prone to errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0025] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0026] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0027] Example:

[0028] like Figure 1 As shown in the figure, the natural language processing and intelligent customer service dialogue management method is a customer service system management method that integrates advanced artificial intelligence technology and natural language processing technology. Its core is to achieve accurate identification and efficient response to user requests by building an intelligent customer service module. The method mainly includes multiple steps such as defining business knowledge and rules, building a user request receiving and analysis unit, interaction between expert modules and intelligent learning units, and collaborative work between ordinary service modules and dynamic learning units. Below, we will explain its working principle in detail and demonstrate its workflow through a specific work scenario.

[0029] Here’s how it works:

[0030] Define business knowledge and rules: The intelligent customer service module first needs to define its business knowledge and business rules. This includes clarifying the business areas, types of questions, and corresponding answering rules that the customer service system needs to handle. On this basis, a corresponding semantic model is established to convert user requests into a format that the system can understand. This step is the foundation of the intelligent customer service system and determines the range of questions that the system can handle and the quality of its answers.

[0031] Build a user request receiving and analysis unit: The user request receiving unit is responsible for capturing the request content entered by the user, which can be achieved through text input boxes, voice input, etc. The analysis unit is responsible for matching and parsing the captured request content according to the pre-defined semantic model. By analyzing the keywords, phrases, and contextual information in the user request, the system can determine whether the request content belongs to the service scope of the professional module or the general service module. This step ensures that the system can accurately identify the user's intention and provide a basis for subsequent processing.

[0032] Interaction between the expert module and the intelligent learning unit: For requests that fall within the scope of professional module services, the system inputs them into the expert module for processing. The expert module has a built-in rich expert knowledge base to store common questions and their answers in professional fields. The system matches the request content according to the learning rules of the expert module. If a matching question is found, the corresponding answer is given directly. If no matching question is found, it enters the intelligent learning unit for processing. The intelligent learning unit analyzes and learns the request content through machine learning algorithms and tries to give a reasonable answer. At the same time, the intelligent learning unit will continuously optimize its learning rules and knowledge base based on user feedback and the quality of the answer.

[0033] Collaborative work between the general service module and the dynamic learning unit: For requests that fall within the service scope of the general service module, the system inputs them into the general service module for processing. The general service module has built-in general question-and-answer rules and answer libraries to handle common customer service issues. The system matches the request content according to the question-and-answer rules and gives corresponding answers. If the answer is a preset answer, the system will decide whether to enter the intelligent learning unit to update the knowledge base based on the quality of the answer and user feedback. If the answer is not a preset answer, or the user feedback is unsatisfactory, the system enters the dynamic learning unit for processing. The dynamic learning unit updates and improves the knowledge base and rules through manual intervention to ensure that the system can continue to provide high-quality services.

[0034] Workflow in a specific work scenario: Taking the intelligent customer service system of an e-commerce platform as an example, we demonstrate the workflow of this method in a specific work scenario. User makes a request: The user enters a question through the chat window of the e-commerce platform: "I want to return a product, what should I do?" This question is a common customer service question on the e-commerce platform, so the system classifies it as a service of the general service module.

[0035] Analyze user requests: The user request receiving unit captures the user input and passes it to the analysis unit. The analysis unit parses and matches the user request according to the pre-defined semantic model. By analyzing the keyword "return" and context information in the user request, the system determines that this is a question about the return process.

[0036] General service module processing: The system inputs the user request into the general service module for processing. The general service module matches the request content according to the question-answering rules and finds a preset answer related to the return process: "Hello, the return process is as follows: 1. Log in to your account; 2. Enter the order management page; 3. Find the order to be returned; 4. Click the return button and follow the prompts." The system returns this answer to the user.

[0037] User feedback and knowledge base update: After receiving the answer, the user gives satisfactory feedback. Based on the user's feedback and the quality of the answer, the system determines that the answer is valid and decides not to update the knowledge base. However, if the user expresses dissatisfaction with the answer in the feedback, or raises a new question, the system will enter the intelligent learning unit or dynamic learning unit for processing.

[0038] Processing of the intelligent learning unit: Suppose in another scenario, the user asks a more specific question: "What should I pay attention to when returning a product?" This question does not find a completely matching answer in the ordinary service module. At this time, the system inputs it into the intelligent learning unit for processing. The intelligent learning unit analyzes and learns the user's request through machine learning algorithms and tries to give a reasonable answer. At the same time, the intelligent learning unit will continuously optimize its learning rules and knowledge base based on user feedback and the quality of the answer.

[0039] Processing of dynamic learning units: If the answer given by the intelligent learning unit still cannot meet the user's needs, or the user feedback is very unsatisfactory, the system will enter the dynamic learning unit for processing. The dynamic learning unit analyzes and answers the questions through human intervention, and updates the knowledge base and rules. For example, in this scenario, the human customer service may give a more detailed and accurate answer based on the user's feedback and the specific requirements of the return process, and add it to the knowledge base so that it can be directly used when handling similar problems later.

[0040] The natural language processing and intelligent customer service dialogue management method achieves accurate identification and efficient response to user requests through multiple steps such as building an intelligent customer service module, defining business knowledge and rules, building a user request receiving and analysis unit, interaction between expert modules and intelligent learning units, and collaborative work between ordinary service modules and dynamic learning units. In a specific work scenario, this method can flexibly select processing methods and update knowledge bases based on user request content and feedback, ensuring that the system can continue to provide high-quality services.

[0041] The business knowledge includes user question and answer knowledge, product knowledge, product feature knowledge, business process knowledge, decision tree knowledge and question and answer rules. The business rules include process rules, logic rules and matching rules. The process rules, logic rules and matching rules are uniformly managed through the rule engine. In the preferred implementation, the business knowledge not only covers traditional fields such as user question and answer knowledge, product knowledge, product feature knowledge, business process knowledge, but also specifically adds decision tree knowledge and question and answer rules to form a more complete and systematic knowledge system. In terms of business rules, by integrating process rules, logic rules and matching rules, and using the rule engine for unified management, flexible configuration and efficient execution of rules are achieved. Compared with the prior art, this integration and unified management method significantly improves the flexibility and scalability of the system, enables the system to adapt to business changes more quickly, and reduces maintenance costs. The prior art often manages these knowledge and rules in a decentralized manner, resulting in low efficiency in knowledge updating and rule adjustment, and it is difficult to ensure consistency and accuracy.

[0042] The semantic model is established through the following steps: first, a semantic dictionary is constructed, followed by simultaneous selection of parts of speech, expansion of semantic templates and modification of semantic templates, and finally application of semantic templates. By combining user operation experience and business knowledge to construct a semantic dictionary, the accuracy and practicality of the dictionary are ensured. Subsequently, the selection of parts of speech, expansion and modification of semantic templates are performed simultaneously. This series of steps not only enhances the expressive power of the semantic model, but also improves its accuracy and robustness. In particular, the mapping relationship between parts of speech, intent, and slots is established, and the expansion of semantic templates in combination with question-answering rules and knowledge enables the model to better understand and process user requests. Compared with the prior art, this method pays more attention to the internal structure and logical relationship of the model, avoids the limitations brought about by simple template matching, and improves the generalization ability and adaptability of the model.

[0043] The construction of the semantic dictionary is to extract product characteristics, question-answering rules, question-answering knowledge, and combine user operation experience to construct the semantic dictionary; after the construction is completed, the mapping relationship between part of speech, intent, and slot is established through part of speech selection training, and the corresponding relationship between part of speech, intent, and slot is determined; the semantic template is expanded in combination with question-answering rules, question-answering knowledge, and semantic dictionaries; through the correction step of the semantic template, the wrong or inaccurate semantic template is screened out, and through the application step of the semantic template, the semantic template, question-answering rules, question-answering knowledge, semantic dictionary, and extended question-answering replies are combined to finally complete the establishment of the semantic model. By combining semantic templates, question-answering rules, question-answering knowledge, semantic dictionaries, and extended question-answering replies, a comprehensive semantic processing system is formed. This comprehensive application method not only improves the accuracy and diversity of replies, but also realizes intelligent question-answering processing. Compared with the prior art, the prior art often relies on fixed template matching or simple keyword search, resulting in the singleness and inaccuracy of replies. The preferred implementation method introduces more knowledge and rules, as well as intelligent processing methods, so that the system can respond to various user requests more flexibly and provide richer and more personalized replies.

[0044] The workflow of the expert module is to first construct an algorithm framework for intelligent learning, then define intelligent learning rules within the algorithm framework, and establish an automatic learning system. The automatic learning system automatically completes knowledge extraction and stores it in the expert knowledge base by introducing learning rules. An algorithm framework for intelligent learning is constructed, and intelligent learning rules are defined within the framework to establish an efficient and automated learning system. This system can automatically complete knowledge extraction and store it in the expert knowledge base, thereby achieving continuous updating and optimization of knowledge. Compared with the prior art, the expert modules in the prior art often rely on manual intervention or a fixed knowledge base, resulting in slow knowledge updating and difficulty in adapting to new business scenarios. The preferred implementation significantly improves the speed and accuracy of knowledge updating by introducing intelligent learning algorithms and automated processing procedures, allowing the system to adapt to business changes more quickly.

[0045] The automatic learning system pre-processes information through knowledge rules, extracts valid keywords in user requests, determines whether keywords in user requests belong to product features, obtains product data corresponding to product features from question-and-answer rules, constructs FAQ knowledge points, adds extracted keywords to FAQ, matches the rule base of the rule engine, generates corresponding slot labels, matches the extracted intent keywords to FAQ knowledge points through semantic models, completes matching and generates corresponding slot labels, matches the slot labels to the rule engine through slot labels, generates replies through semantic models, matches the generated replies to the rule engine, and completes knowledge recommendation. Information is pre-processed through knowledge rules, valid keywords are extracted, and FAQ knowledge points are constructed in combination with product features and question-and-answer rules. Through the collaborative work of semantic models and rule engines, the system can understand and process user requests more deeply, and generate accurate slot labels and replies. Compared with the prior art, the automatic learning systems in the prior art often lack the ability to deeply process and intelligently match information, resulting in inaccuracy and inefficiency in replies. The preferred implementation scheme improves the system's ability to understand and process user requests by introducing more intelligent processing methods and knowledge rules, making responses more accurate and efficient.

[0046] The dynamic learning unit connects the unanswered questions to the dynamic learning, and obtains the keyword information of the request content in the user request; if it is a rule engine, the keyword information that does not appear in the knowledge base of the rule engine is stored in the dynamic knowledge base; the unanswered questions are divided into different dynamic learning pools according to categories and stored in the dynamic knowledge base. The manual customer service determines the keyword information in the dynamic knowledge base, determines the product characteristics to which the keyword information in the dynamic knowledge base belongs, and then obtains the product characteristics, adds them to the dynamic knowledge base, and completes the learning process. In the workflow of the dynamic learning unit, the preferred implementation method obtains keyword information by connecting to the unanswered questions, and divides them into different dynamic learning pools according to categories. The manual customer service determines the product characteristics based on the keyword information and adds them to the dynamic knowledge base, thereby realizing flexible knowledge update and personalized learning. Compared with the prior art, the dynamic learning unit in the prior art often lacks flexibility and personalized processing capabilities, resulting in slow knowledge update and difficulty in meeting the personalized needs of users. The preferred implementation method introduces the collaborative work of the dynamic learning pool and the manual customer service, so that the system can respond to new knowledge needs in user requests more quickly and provide personalized replies and services. This flexibility and personalized processing not only improves user satisfaction, but also enhances the system's competitiveness and market adaptability.

[0047] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A natural language processing and intelligent customer service dialogue management method, characterized in that: The method comprises the following steps: S1: Define the business knowledge and business rules of the intelligent customer service module and establish the corresponding semantic model; S2: constructing a user request receiving unit in the module to obtain the request content input by the user, matching the request content with a predefined semantic model through an analysis unit, and determining whether the request content belongs to the service scope of the professional module or the general service module; S3: Input the request content into the expert module, and determine whether it is a question in the expert knowledge base according to the learning rules of the expert module. If so, the expert module gives an answer according to the matching rules; if not, enter the intelligent learning unit, and give an answer according to the learning rules of the intelligent learning unit and the matching rules; S4: Input the request content into the common service module, and the common service module gives an answer according to the question-answering rules; If the general service module or the expert module gives only preset answers, it will enter the intelligent learning unit to update the knowledge base. If it gives only non-preset answers, it will enter the dynamic learning unit to update the knowledge base through manual intervention and update the rules simultaneously.

2. The method for natural language processing and intelligent customer service dialogue management according to claim 1, characterized in that: The business knowledge includes user question and answer knowledge, product knowledge, product feature knowledge, business process knowledge, decision tree knowledge and question and answer rules. The business rules include process rules, logic rules and matching rules. The process rules, logic rules and matching rules are uniformly managed through a rule engine.

3. The method for natural language processing and intelligent customer service dialogue management according to claim 1, characterized in that: The semantic model is established through the following steps: first, a semantic dictionary is constructed, then part of speech is selected, semantic templates are expanded and modified, and finally, the semantic templates are applied.

4. The method for natural language processing and intelligent customer service dialogue management according to claim 3, characterized in that: The construction of semantic dictionary is to extract product features, question-answering rules, question-answering knowledge and combine with user operation experience to construct semantic dictionary; after the construction is completed, the mapping relationship between part of speech, intention and slot is established through part of speech selection training, and the corresponding relationship between part of speech, intention and slot is determined; the semantic template is expanded by combining question-answering rules, question-answering knowledge and semantic dictionary; Through the correction step of the semantic template, erroneous or inaccurate semantic templates are screened out. Through the application step of the semantic template, the semantic template, question and answer rules, question and answer knowledge, semantic dictionary and extended question and answer replies are combined to finally complete the establishment of the semantic model.

5. The method for natural language processing and intelligent customer service dialogue management according to claim 1, characterized in that: The workflow of the expert module is to first construct an algorithm framework for intelligent learning, then define intelligent learning rules within the algorithm framework, and establish an automatic learning system. The automatic learning system automatically completes knowledge extraction by introducing learning rules and stores it in the expert knowledge base.

6. The method for natural language processing and intelligent customer service dialogue management according to claim 5, characterized in that: The automatic learning system pre-processes information through knowledge rules, extracts valid keywords in user requests, determines whether the keywords in user requests belong to product features, obtains product data corresponding to product features from question-and-answer rules, constructs FAQ knowledge points, adds the extracted keywords to FAQ, matches the rule base of the rule engine, generates corresponding slot labels, matches the extracted intent keywords with the FAQ knowledge points through the semantic model, completes the match and generates corresponding slot labels, matches the slot labels with the rule engine through the slot labels, generates replies through the semantic model, matches the generated replies with the rule engine, and completes knowledge recommendation.

7. The method for natural language processing and intelligent customer service dialogue management according to claim 1, characterized in that: The dynamic learning unit connects the unanswerable questions to dynamic learning to obtain keyword information of the request content in the user request; If it is a rule engine, the keyword information that does not appear in the rule engine's knowledge base is stored in the dynamic knowledge base; the unanswered questions are classified into different dynamic learning pools and stored in the dynamic knowledge base. The manual customer service determines the keyword information in the dynamic knowledge base and the product characteristics to which the keyword information in the dynamic knowledge base belongs, then obtains the product characteristics and adds them to the dynamic knowledge base to complete the learning process.

Images