Question processing method and device and storage medium

Abstract

The invention discloses a problem processing method and device and a storage medium, relates to the technical field of artificial intelligence, and aims to solve the problem of how to improve the problem processing efficiency. The method comprises the steps of obtaining an input question of a user; based on the input question, obtaining a target knowledge list from a knowledge base; the target knowledge list comprises multiple pieces of knowledge content related to the input question and the similarity between each piece of knowledge content and the input question; determining the confidence degree of the target knowledge list based on the similarity between each knowledge content and the input question; determining a solution of the input problem based on the confidence coefficient; the solution is determined based on a target knowledge list or based on a manual manner.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a problem-solving method, device, and storage medium. Background Art

[0002] In the daily operation and maintenance of an enterprise, a large number of employee issues need to be dealt with, and how to improve the efficiency of handling problems has become a technical problem to be solved.

[0003] At present, although simple automation systems can achieve partial process automation, they lack flexibility in intelligent judgment and problem handling, and their processing capabilities are limited, resulting in low efficiency in problem handling. Summary of the Invention

[0004] The purpose of this application is to provide a problem handling method, device and storage medium, aiming to solve the problem of how to improve the efficiency of problem handling.

[0005] To achieve the above objectives, this application adopts the following technical solutions:

[0006] In a first aspect, the present application provides a problem handling method, which includes: obtaining a user's input question; based on the input question, obtaining a target knowledge list from a knowledge base; the target knowledge list includes multiple knowledge contents related to the input question and the similarity between each knowledge content and the input question; based on the similarity between each knowledge content and the input question, determining the confidence of the target knowledge list; based on the confidence, determining a solution to the input problem; the solution is determined based on the target knowledge list or manually.

[0007] The problem-solving method provided in the embodiments of the present application is divided into two ways to determine the solution to the input problem. One is to automatically obtain a target knowledge list from the knowledge base and determine the solution to the input problem based on the target knowledge list. The other is to manually determine the solution to the input problem. If the knowledge base contains knowledge content that matches the input problem, there is no need to manually search the knowledge base, thereby improving the efficiency of problem solving.

[0008] In one possible implementation, a target knowledge list is obtained from a knowledge base based on an input problem, including: optimizing the input problem using a large language model to obtain a problem description; describing the problem as the optimized input problem; and traversing the knowledge base based on a vector representation of the problem description, filtering out multiple pieces of knowledge related to the problem description from the knowledge base to obtain a target knowledge list.

[0009] In one possible implementation, the confidence of the target knowledge list is determined based on the similarity between each knowledge content and the input question, including: determining the confidence of the target knowledge list based on the weight of each knowledge content, the similarity between each knowledge content and the input question, and the similarity weight index of each knowledge content.

[0010] In a possible implementation, the confidence level of the target knowledge list satisfies a preset formula, which is:

[0011]

[0012] Among them, C represents the confidence of the target knowledge list, w i represents the weight of the i-th knowledge content, s i represents the similarity between the i-th knowledge content and the input question, α i represents the similarity weight index of the i-th knowledge content, n represents the number of knowledge contents in the target knowledge list, and n is a positive integer.

[0013] In a possible implementation, determining a solution to the input problem based on the confidence level includes: determining a solution to the input problem based on a target knowledge list when the confidence level is greater than or equal to a preset threshold.

[0014] In one possible implementation, a solution to the input problem is determined based on a confidence level, including: creating a work order when the confidence level is less than a preset threshold; the work order includes a user identifier and an input problem, and the status of the work order is to be assigned; and determining a solution to the input problem based on the work order.

[0015] In a possible implementation, determining a solution to an input problem based on a work order includes: detecting a status of the work order at a preset time interval; and obtaining a solution to the input problem when the status of the work order is resolved.

[0016] In a second aspect, the present application provides a problem processing device, which includes: a communication unit for obtaining a user's input question; a processing unit for obtaining a target knowledge list from a knowledge base based on the input question; the target knowledge list includes multiple knowledge contents related to the input question and the similarity between each knowledge content and the input question; the processing unit is also used to determine the confidence of the target knowledge list based on the similarity between each knowledge content and the input question; the processing unit is also used to determine a solution to the input problem based on the confidence; the solution is determined based on the target knowledge list or manually.

[0017] In one possible implementation, the processing unit is further used to optimize the input problem using a large language model to obtain a problem description; the problem description is the optimized input problem; the processing unit is further used to traverse the knowledge base based on the vector representation of the problem description, filter out multiple knowledge contents related to the problem description from the knowledge base, and obtain a target knowledge list.

[0018] In a possible implementation, the processing unit is further configured to determine the confidence of the target knowledge list based on the weight of each knowledge content, the similarity between each knowledge content and the input question, and the similarity weight index of each knowledge content.

[0019] In a possible implementation, the confidence level of the target knowledge list satisfies a preset formula, which is:

[0020]

[0021] Among them, C represents the confidence of the target knowledge list, w i represents the weight of the i-th knowledge content, s i represents the similarity between the i-th knowledge content and the input question, α i represents the similarity weight index of the i-th knowledge content, n represents the number of knowledge contents in the target knowledge list, and n is a positive integer.

[0022] In a possible implementation, the processing unit is further configured to determine a solution to the input problem based on the target knowledge list when the confidence level is greater than or equal to a preset threshold.

[0023] In one possible implementation, the processing unit is further used to create a work order when the confidence level is less than a preset threshold; the work order includes a user identifier and an input problem, and the status of the work order is to be assigned; the processing unit is further used to determine a solution to the input problem based on the work order.

[0024] In a possible implementation, the processing unit is further configured to detect the status of the work order at a preset time interval; and the processing unit is further configured to obtain a solution to the input problem when the status of the work order is resolved.

[0025] In a third aspect, the present application provides a problem handling device, which includes: a processor and a communication interface; the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement the problem handling method described in the first aspect and any possible implementation method of the first aspect.

[0026] In a fourth aspect, the present application provides a computer-readable storage medium, which stores instructions. When the instructions are executed on a terminal, the terminal executes the problem handling method described in the first aspect and any possible implementation of the first aspect.

[0027] In a fifth aspect, the present application provides a computer program product comprising instructions, which, when executed on a problem handling device, enables the problem handling device to execute the problem handling method as described in the first aspect and any possible implementation of the first aspect.

[0028] In a sixth aspect, the present application provides a chip, which includes a processor and a communication interface, the communication interface and the processor are coupled, and the processor is used to run a computer program or instructions to implement the problem handling method described in the first aspect and any possible implementation method of the first aspect.

[0029] Specifically, the chip provided in this application also includes a memory for storing computer programs or instructions. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0031] Figure 1 A schematic diagram of the composition of a problem handling device provided in an embodiment of the present application;

[0032] Figure 2 A schematic diagram of a module of a problem handling device provided in an embodiment of the present application;

[0033] Figure 3 A flowchart of a problem-solving method provided in an embodiment of the present application;

[0034] Figure 4 A flowchart of another problem-solving method provided in an embodiment of the present application;

[0035] Figure 5 A flowchart of another problem-solving method provided in an embodiment of the present application;

[0036] Figure 6 A flowchart of another problem-solving method provided in an embodiment of the present application;

[0037] Figure 7 A flowchart of another problem-solving method provided in an embodiment of the present application;

[0038] Figure 8 A schematic diagram of the structure of a problem handling device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0039] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0040] In the description of this application, it should be understood that the terms "upper," "lower," "left," "right," "front," "back," "inner," "outer," and the like, indicating directions or positional relationships, are based on the directions or relative positional relationships shown in the accompanying drawings and are intended solely to facilitate the description of this application and simplify the description. They do not indicate or imply that the devices or components referred to must have a specific direction, be constructed, or operate in a specific direction. Therefore, they should not be construed as limitations on this application. Unless otherwise specified, the above-mentioned directionality descriptions may be flexibly set in actual application, provided that the relative positional relationships shown in the accompanying drawings are met.

[0041] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. Throughout this application, unless otherwise specified, "plurality" means two or more.

[0042] In the embodiments of the present application, the terms "comprises," "comprising," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, article, or device comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not preclude the presence of other identical elements in the process, article, or device comprising the element.

[0043] In the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0044] In the description of this specification, specific features, structures, materials or characteristics may be combined in an appropriate manner in any one or more embodiments or examples.

[0045] In the daily operation and maintenance of an enterprise, a large number of employee issues need to be dealt with, and how to improve the efficiency of handling problems has become a technical problem to be solved.

[0046] Currently, the traditional manual processing method requires workers to answer the phone or review online documents and then manually search for solutions from the knowledge base. This method consumes a lot of manpower, is inefficient, and is prone to human error.

[0047] Furthermore, while some simple automation systems can partially automate processes, they lack flexibility in intelligent judgment and problem-solving. For example, they can only match knowledge base content based on preset rules, have limited ability to handle complex problems, and lack a high level of collaborative automation between different links.

[0048] In view of this, an embodiment of the present application provides a problem handling method, which includes: obtaining an input question from a user, and obtaining a target knowledge list from a knowledge base based on the input question. Determine the confidence of the target knowledge list, and determine a solution to the input problem based on the confidence. The solution is determined based on the target knowledge list or manually. That is to say, the problem handling method provided by an embodiment of the present application is divided into two ways to determine the solution to the input problem, one is to automatically obtain the target knowledge list from the knowledge base, and determine the solution to the input problem based on the target knowledge list. The other is to determine the solution to the input problem manually. In the case where the knowledge base has knowledge content that matches the input problem, there is no need to manually search the knowledge base, which can improve the efficiency of problem handling.

[0049] For example, Figure 1 This is a schematic diagram of the composition of a problem handling device 10 provided in an embodiment of the present application. Figure 1 As shown, the problem handling device 10 may include a processor 101 and a bus 102 .

[0050] Furthermore, the problem handling device 10 may further include a communication interface 103 and a memory 104 . The processor 101 , the memory 104 and the communication interface 103 may be connected via a bus 102 .

[0051] The processor 101 is a central processing unit (CPU), a general-purpose processor, a network processor (NP), a digital signal processor (DSP), a microprocessor, a microcontroller, a programmable logic device (PLD), or any combination thereof. The processor 101 may also be other devices with processing functions, such as circuits, devices, or software modules, without limitation.

[0052] The bus 102 is used to transmit information between the components included in the problem handling device 10 .

[0053] Communication interface 103 is used to communicate with other devices or other communication networks. Such other communication networks may be Ethernet, radio access networks (RAN), wireless local area networks (WLAN), etc. Communication interface 103 may be a module, circuit, communication interface, or any other device capable of implementing communication.

[0054] The memory 104 is used to store instructions, where the instructions may be computer programs.

[0055] The memory 104 may be a read-only memory (ROM) or other type of static storage device that can store static information and / or instructions, or a random access memory (RAM) or other type of dynamic storage device that can store information and / or instructions, or an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), magnetic disk storage media or other magnetic storage devices, etc., without limitation.

[0056] It should be noted that memory 104 can exist independently of processor 101 or be integrated with processor 101. Memory 104 can be used to store instructions, program code, or data. Memory 104 can be located within or outside of problem handling device 10, without limitation. Processor 101 is configured to execute instructions stored in memory 104 to implement the problem handling methods provided in the following embodiments of this application.

[0057] It should be noted that the problem handling device 10 can be a desktop computer, a portable computer, a network server, a mobile phone, a tablet computer, a wireless terminal, an embedded device, a chip system or a computer with Figure 1 In addition, Figure 1 The composition shown in the Figure 1 The limitations of each device in Figure 1 In addition to the parts shown, Figure 1 The various devices in the figures may include more or fewer components than shown, or combine certain components, or arrange the components differently.

[0058] In the embodiment of the present application, the chip system can be composed of chips, or can include chips and other discrete devices.

[0059] In addition, the actions and terms involved in the various embodiments of this application can refer to each other without limitation. The message names or parameter names in the messages exchanged between the various devices in the embodiments of this application are only examples, and other names can also be used in specific implementations without limitation.

[0060] For example, Figure 2 This is a module diagram of a problem handling device provided in an embodiment of the present application. Figure 2 As shown, the problem handling device includes: an intelligent agent module, a knowledge base module, and a work order management system module.

[0061] The intelligent agent module is used to receive a user's input question and, based on the input question, retrieve a target knowledge list from the knowledge base. The intelligent agent module is also used to determine the confidence level of the target knowledge list based on the similarity between each piece of knowledge content and the input question. Furthermore, based on the confidence level of the target knowledge list, the intelligent agent module determines a solution to the input question.

[0062] The knowledge base module is used to store various types of knowledge data, including structured data and text data.

[0063] Optionally, the knowledge base module can be a repository for retrieval-augmented generation (RAG) vectors.

[0064] The work order management system module is used to create and manage work orders.

[0065] The problem-solving method provided by the embodiment of the present application is described below with reference to the accompanying drawings. Among them, the actions, terms, etc. involved in the various embodiments of the present application can refer to each other without limitation. The message name or parameter name in the message exchanged between the various devices in the embodiment of the present application is only an example, and other names can also be used in the specific implementation without limitation. The actions involved in the various embodiments of the present application are only an example, and other names can also be used in the specific implementation, such as: "included in" in the embodiment of the present application can also be replaced by "carried on" or "carried in", etc.

[0066] like Figure 3 As shown, the embodiment of the present application proposes a problem solving method, which includes:

[0067] S301: Obtain the user's input question.

[0068] In one possible implementation, the user's input question is obtained through a web form or a telephone voice.

[0069] For example, the above input problems may be: "Cannot log in to the mailbox", "Network connection failed", "Webpage cannot be opened".

[0070] Optionally, after obtaining the user's input question, the user's user ID may be recorded, and the user ID and the user's input question may be stored synchronously.

[0071] S302: Based on the input question, obtain a target knowledge list from the knowledge base.

[0072] The target knowledge list includes multiple knowledge contents related to the input question and the similarity between each knowledge content and the input question.

[0073] In one possible implementation, a large language model is used to identify the input question and obtain the question type corresponding to the input question. The input question is optimized according to the question type to obtain the optimized input question. The optimized input question is converted into a vector representation, and the vector representation is compared with the vector representation corresponding to each knowledge content in the knowledge base to determine the target knowledge. Based on the target knowledge, a target knowledge list is obtained. For details, please refer to the following Figure 4 The embodiments described are not described in detail here.

[0074] Exemplarily, the above-mentioned problem types may include: account issues, network issues, password issues, software issues, hardware issues, and security issues. Account issues may include: registration failure, inability to log in, and account lockout. Network issues may include: connection failure, high network latency, and inability to access a website. Password issues may include: password reset, password expiration, and password authentication failure. Software issues may include: application crashes, system update failures, and software installation failures. Hardware issues may include: screen damage, device overheating, and mouse malfunction. Security issues may include: data loss, virus infection, and firewall failure.

[0075] S303: Determine the confidence level of the target knowledge list based on the similarity between each knowledge content and the input question.

[0076] In one possible implementation, each piece of knowledge in the target knowledge list is analyzed, a weight is assigned to each piece of knowledge based on its importance, and a similarity weight index is determined for each piece of knowledge. The confidence level of the target knowledge list is determined based on the weight of each piece of knowledge, the similarity between each piece of knowledge and the input question, and the similarity weight index of each piece of knowledge.

[0077] Optionally, the confidence of the target knowledge list can be calculated according to the following formula.

[0078]

[0079] Where C represents the confidence of the target knowledge list. i Represents the weight of the i-th knowledge content. i Represents the similarity between the i-th knowledge content and the input question. i Represents the similarity weight index of the i-th knowledge content. n represents the number of knowledge contents in the target knowledge list, and n is a positive integer.

[0080] S304: Determine a solution to the input problem based on the confidence level.

[0081] In one possible implementation, the relationship between the confidence and the preset threshold is determined, and the solution to the input problem is determined based on the judgment result. If the confidence is greater than or equal to the preset threshold (for example, 0.8), the solution to the input problem is determined based on multiple knowledge contents in the target knowledge list. If the confidence is less than the preset threshold, a new work order is created, and after the work order is resolved, the solution to the input problem is obtained. For details, please refer to the following Figure 5 and Figure 6 The embodiments described are not described in detail here.

[0082] In the problem-solving method provided by the present application, the user's input problem is obtained, and a target knowledge list is obtained from the knowledge base based on the input problem. The confidence of the target knowledge list is determined, and the solution to the input problem is determined based on the confidence. Among them, the solution is determined based on the target knowledge list or based on manual methods. That is to say, the problem-solving method provided by the embodiment of the present application is divided into two ways to determine the solution to the input problem, one is to automatically obtain the target knowledge list from the knowledge base, and determine the solution to the input problem based on the target knowledge list. The other is to determine the solution to the input problem manually. In the case where the knowledge base has knowledge content that matches the input problem, there is no need to manually search the knowledge base, which can improve the efficiency of problem handling.

[0083] In one embodiment, Figure 4 As shown, the above S302 can be specifically determined through the following S401 to S402.

[0084] S401. Use a large language model to optimize the input question and obtain a problem description.

[0085] Here, the problem is described as an optimized input problem.

[0086] In one possible implementation, the original input question is preprocessed, including removing special characters, unifying the description, and performing word segmentation. The preprocessed question is then fed into a large language model, which analyzes the question's semantic structure, corrects grammatical errors, removes redundant information, and rephrases the question using standard terminology to produce a problem description.

[0087] Optionally, during the analysis of the preprocessed question, the large language model can also identify the question type corresponding to the question, so as to optimize the input question according to the question type, thereby improving the optimization effect.

[0088] S402. According to the vector representation of the problem description, the knowledge base is traversed, and multiple pieces of knowledge content related to the problem description are screened out from the knowledge base to obtain a target knowledge list.

[0089] In one possible implementation, the problem description is converted into a high-dimensional vector. Each piece of knowledge in the knowledge base is traversed and converted into a vector representation. The similarity between the problem description vector and each piece of knowledge vector is calculated to obtain a similarity value. The knowledge content is sorted based on the similarity score, and the top N pieces of knowledge are selected as target knowledge. These N target pieces of knowledge and the similarity between each piece of knowledge and the input problem are combined to obtain a target knowledge list. N is a positive integer.

[0090] Exemplarily, the similarity between the problem description vector and the knowledge content vector may be calculated using cosine similarity, Euclidean distance, or Pearson correlation coefficient.

[0091] Optionally, after obtaining the target knowledge list, the knowledge content in the target knowledge list may be further optimized, for example, by removing duplicate content, merging similar content, etc.

[0092] In one embodiment, Figure 5 As shown, the above S304 can be specifically determined through the following S501.

[0093] S501 : When the confidence level is greater than or equal to a preset threshold, determine a solution to the input problem based on the target knowledge list.

[0094] In one possible implementation, each piece of knowledge in the target knowledge list is enhanced using a large language model. The enhanced knowledge is then integrated using the large language model again. The model analyzes the connections and complementarities between these pieces of knowledge to generate a comprehensive, holistic solution.

[0095] Furthermore, after obtaining a solution to the input problem, the solution is sent to the user. In addition, user feedback on the solution can also be collected, such as satisfaction, accuracy, or practicality.

[0096] In one embodiment, Figure 6 As shown, the above S304 can be specifically determined through the following S601 to S602.

[0097] S601: When the confidence level is less than a preset threshold, create a work order.

[0098] The work order includes a user ID and an input question, and the status of the work order is to be assigned.

[0099] In one possible implementation, if the confidence level is less than a preset threshold, a work order is created that includes the user ID, the input question, and other relevant information, a unique identifier is assigned to the work order, and the state of the work order is set to pending assignment.

[0100] Optionally, when creating a ticket, you can set a priority for the ticket based on the urgency of the issue to be resolved, for example, low, medium, high, urgent, etc.

[0101] S602: Determine a solution to the input problem based on the work order.

[0102] In one possible implementation, the work order is assigned to a second-line employee. After the work order is assigned, the work order status is checked at preset intervals. If the work order status is resolved, the work order content is extracted to obtain a solution to the input problem.

[0103] Furthermore, after obtaining a solution to the input problem based on the work order, the solution can be optimized to obtain corresponding knowledge content, and the knowledge content can be stored in a knowledge base.

[0104] Figure 7 This is a flowchart of a problem solving method provided in an embodiment of the present application. Figure 7 As shown, the problem handling method provided in the embodiment of the present application may include the following steps.

[0105] S701: The user sends an input question to the agent, and the agent receives the input question from the user.

[0106] S702: The intelligent agent optimizes the input problem and obtains a problem description.

[0107] S703: The agent sends a search request to the knowledge base. Correspondingly, the knowledge base receives the search request from the agent.

[0108] The search request is used to instruct to search for relevant solutions based on the problem description.

[0109] S704: The knowledge base sends the target knowledge list to the agent. Correspondingly, the agent receives the target knowledge list from the knowledge base.

[0110] The target knowledge list includes multiple knowledge contents related to the input question and the similarity between each knowledge content and the input question.

[0111] S705. The intelligent agent determines the confidence of the target knowledge list based on the similarity between each knowledge content and the input question.

[0112] S706: The agent determines whether the confidence rule is satisfied.

[0113] The confidence rule is that the confidence of the target knowledge list is greater than or equal to a preset threshold.

[0114] If the confidence rule is met, the agent executes step S707.

[0115] If the confidence rule is not satisfied, the agent executes step S709.

[0116] S707. The intelligent agent determines a solution to the input problem based on the target knowledge list.

[0117] S708: The agent sends the solution to the user device. Correspondingly, the user device receives the solution from the agent.

[0118] S709: The agent sends a request to create a work order to the work order management system. Correspondingly, the work order management system receives the request from the agent to create a work order.

[0119] The create work order request is used to instruct the creation of a work order.

[0120] S710. The work order management system creates a work order.

[0121] Among them, the work order includes user device identification and input issues.

[0122] S711. The agent detects the status of the work order at a preset time interval.

[0123] S712: When the status of the work order is resolved, the agent obtains a solution to the input problem.

[0124] S713: The agent sends a solution to the user device. Correspondingly, the user device receives the solution from the agent.

[0125] In summary, the problem-solving method provided in the embodiment of the present application can give full play to the advantages of intelligent agents. The intelligent agent has the characteristics of autonomous decision-making and flexible response. After automatically answering employee calls or receiving online input questions, it quickly and accurately queries the knowledge base based on preset rules and real-time conditions. Compared with traditional systems, the query efficiency and accuracy are greatly improved. In addition, the efficiency of distributed collaboration between intelligent agents is reflected. The intelligent agents have clear division of labor and cooperate with each other. In the process of generating answers with the help of large models, they can better integrate information, optimize the content of answers, and ensure that the solutions provided to employees are more targeted and professional.

[0126] Furthermore, the agent's high adaptability and scalability offer the potential for continued growth. When the knowledge base requires updates or business scenarios change, the agent can quickly adapt and easily expand new functionality. For unresolved issues, the agent automatically transfers them to the work order management system. This process is not only efficient, but also leverages the agent's precise intelligent optimization capabilities to rationally arrange the flow of issues based on their type and urgency, improving the rationality of work order processing priorities.

[0127] Throughout the entire problem-handling process, agents also play a crucial role in improving reliability and fault tolerance. When regularly tracking the resolution status of second-line work orders, even if some agents fail, others can promptly take over, ensuring continuity of tracking and ensuring the automated and stable operation of the entire process, from issue acceptance to feedback, thereby comprehensively improving problem-handling efficiency and service quality.

[0128] The problem-solving method provided in the embodiments of the present application is applicable to the operation and maintenance departments of various enterprises and institutions, especially organizations with a large number of employees and frequent consulting issues, such as the information technology (IT) operation and maintenance departments and human resources service departments of large enterprises. In addition, with the acceleration of the digital transformation of enterprises, the demand for efficient operation and maintenance services is growing. This application can effectively improve service quality and efficiency, reduce labor costs, and has broad market application prospects. At the same time, with the continuous development of artificial intelligence technology, it can be further optimized and upgraded to adapt to more complex scenarios and business needs.

[0129] Optionally, the above problem-solving method can be implemented through code. Specifically, this may include code for importing information, configuring information, managing the work order system, building a knowledge base, data models, tools, agents, scheduled tasks, and main execution logic. Each of these code components is described below.

[0130] 1-1. Import information code

[0131] For example, the import information code may be as follows:

[0132]

[0133]

[0134] The first line of code above imports Python's regular expression module re, which is used for string matching, searching, and replacing operations.

[0135] The second line of code above imports the pymysql library, which is used to connect to and operate the MySQL database.

[0136] The third line of code above imports the redis library, which is used to connect to and operate the Redis database.

[0137] The fourth line of code above indicates importing Type from the typing module, which is used to specify the type of variables or function parameters.

[0138] The fifth line of code imports BaseModel and Field from the pydantic library. BaseModel defines the data model and provides data validation and serialization capabilities. Field defines the model fields and their attributes.

[0139] The sixth line of code above indicates importing BackgroundScheduler from the apscheduler library. BackgroundScheduler is a background scheduler used to execute tasks in the background at regular intervals.

[0140] The seventh line of code imports AgentExecutor and Tool from the langchain library. AgentExecutor is the agent executor, used to run the agent and process user input. Tool is used to define the tools that the agent can call.

[0141] The eighth line of code imports StructuredChatAgent from the langchain library. StructuredChatAgent is a structured chat agent that handles user requests according to a predefined process.

[0142] The ninth line of code imports prompt-related classes from the langchain library, including ChatPromptTemplate, SystemMessage, and HumanMessagePromptTemplate. ChatPromptTemplate is a chat prompt template that defines the prompt format for the agent. SystemMessage is a system message template that defines the agent's behavior rules. HumanMessagePromptTemplate is a human message template that defines the prompt format for user input.

[0143] The tenth line of code above imports FAISS from the langchain_community library. FAISS is a vector retrieval library for efficiently storing and retrieving vector representations.

[0144] The eleventh line of code above imports RecursiveCharacterTextSplitter from the langchain_text_splitters library. RecursiveCharacterTextSplitter is a text splitter used to split long documents into smaller chunks.

[0145] The twelfth line of code above imports HuggingFaceEmbeddings from the langchain_community library. HuggingFaceEmbeddings is an embedding model loader used to convert text into vector representations.

[0146] The thirteenth line of code above indicates importing the DEEPSEEK_API_KEY variable (i.e., the DeepSeek API key) from the config module for calling the large language model.

[0147] 1-2. Configuration information code

[0148] For example, the configuration information code may be as follows:

[0149]

[0150]

[0151] The first line of code above indicates the definition of configuration information.

[0152] The second line of code above specifies that the Redis database is running on the local machine.

[0153] The third line of code indicates that the MySQL database is running on the local machine.

[0154] The fourth line of code above indicates that the user name used to connect to the MySQL database is "ops".

[0155] The fifth line of code above indicates that the password used to connect to the MySQL database is "pass".

[0156] The sixth line of code indicates that the name of the MySQL database to be connected is specified as "itsm".

[0157] The seventh line of code above specifies a confidence threshold of 0.8.

[0158] The eighth line of code above specifies that the model used to generate text embeddings is the "BAAI / bge-base-zh" model.

[0159] The ninth line of code specifies that the large language model is the "deepseek-chat" model.

[0160] The tenth line of code above defines the DEEPSEEK_API_KEY variable.

[0161] 1-3. Work Order Management System Code

[0162] For example, the work order management system code may be as follows:

[0163]

[0164]

[0165] The first line of code above defines a class named ITSMClient, which is used to interact with the work order management system.

[0166] The second through seventh lines of code above define the initialization method for ITSMClient. Specifically, the third line creates a database connection object and uses the pymysql library to connect to the MySQL database. The fourth line retrieves the MySQL database host address from the CONFIG configuration information. The fifth line retrieves the MySQL database username from the CONFIG configuration information. The sixth line retrieves the MySQL database password from the CONFIG configuration information. The seventh line retrieves the name of the MySQL database to connect to from the CONFIG configuration information.

[0167] The above lines 9 to 16 of code define a method for creating a new work order. This method accepts a user ID and a problem description and returns the ID of the newly created work order. Specifically, the tenth line of code creates a database cursor using a context manager. The eleventh and twelfth lines of code define SQL insert statements for inserting work order information into the tickets table. The initial state of the work order is pending. The thirteenth line of code executes the SQL statement to insert the work order information. The fourteenth line of code gets the last ID of the inserted record. The fifteenth line of code commits the transaction to ensure that the data is written to the database. The sixteenth line of code returns the formatted work order ID in the format of INC-<ticket_id> .

[0168] Lines 17 to 22 of the code above define a method for checking the status of a work order. This method accepts a work order ID and returns a dictionary containing the work order's status and solution. If the work order does not exist, None is returned. Specifically, line 18 creates a database cursor using a context manager. Line 19 defines a SQL query statement for obtaining the work order's status and solution. Line 20 executes the SQL statement to query the work order information. Line 21 retrieves the first row of the query result. Line 22 returns a dictionary containing the work order's status and solution if the result exists, otherwise None is returned.

[0169] Lines 23 through 27 define a method for retrieving pending tickets. This method returns a list containing the ID of each pending ticket and the corresponding user ID. Line 24 creates a database cursor using a context manager. Line 25 defines a SQL query to retrieve all unresolved tickets. Line 26 executes the SQL query to retrieve ticket information. Line 27 iterates over the query results, formats each row of data into a dictionary, and returns a list containing these dictionaries.

[0170] 1-4. Knowledge base system code

[0171] For example, the knowledge base system code may be as follows:

[0172]

[0173]

[0174] The first line of code above defines a class named KnowledgeBase, which is used to implement a semantically enhanced knowledge base system.

[0175] The second to fifth lines of code above define the initialization method for KnowledgeBase, which is automatically called when the KnowledgeBase object is created. Specifically, the third line of code initializes a HuggingFaceEmbeddings object, which is used to convert text into vector representations. The model_name parameter is obtained from the CONFIG configuration information. The fourth line of code calls the private method _load_documents to load knowledge items and stores the results in self.documents. The fifth line of code calls the private method _init_vector_db to initialize the vector database and stores the results in self.vector_db.

[0176] The sixth to eleventh lines of code above define a private method for loading knowledge entries, which returns a list containing sample knowledge entries.

[0177] The above lines 12 to 15 define a private method for initializing the vector database. Specifically, the 13th line creates a RecursiveCharacterTextSplitter object to split long text into smaller blocks, specifying a maximum of 200 characters per block. The 14th line uses text_splitter to split each knowledge entry in self.documents into smaller blocks and stores the results in splits. The 15th line uses the FAISS.from_documents method to convert the split text blocks into vectors and store them in the vector database.

[0178] The above lines 16 to 28 of code represent the definition of a public method search, which is used to retrieve relevant solutions based on the user query. This method returns a dictionary containing the retrieved solutions and their related information. Specifically, the 17th line of code represents the use of the vector_db.similarity_search_with_score method to find the knowledge items most similar to the query in the vector database. query is the user's query string, and k=3 specifies that the 3 most similar knowledge items are returned. docs is a list, each element is a tuple, including the content and similarity score of the similar knowledge item. Lines 18 to 22 of code represent the creation of a list solutions containing solutions. For each retrieved knowledge item, its content, similarity score, and a serial number are extracted. The 23rd line of code represents the calculation of the average similarity score of all knowledge items and stores it in avg_score. Lines 24 to 28 of code represent the return of a dictionary, including a list of solutions, confidence, and confidence flag.

[0179] 1-5. Data model code

[0180] For example, the above data model code may be as follows:

[0181]

[0182]

[0183] The first line of code defines a class named ParseInputModel, which inherits from BaseModel and is used to represent the data structure of the user's original input.

[0184] The second line of code above defines a class called KnowledgeQueryModel, which inherits from BaseModel and is used to represent the data structure of the query description.

[0185] The third line of code above defines a class named TicketCheckModel, which inherits from BaseModel and is used to represent the data structure of a ticket check request.

[0186] 1-6. Toolset Code

[0187] For example, the above toolset code can be as follows:

[0188]

[0189]

[0190]

[0191] The first four lines of code define a tool called input_parser, which is used to parse the user-entered job number and problem type. The input data model accepted by the tool is ParseInputModel.

[0192] The above fifth to fifteenth lines of code represent the definition of a private method _run, which is used to execute the core logic of the input_parser tool. This method accepts a string parameter raw_input and returns a dictionary. Specifically, the sixth to ninth lines of code represent the definition of a dictionary patterns, which contains patterns for regular expression matching. Among them, the user_id pattern is used to match the work number or employee number, and the problem_type pattern is used to match the problem type. The tenth and eleventh lines of code represent the use of the re.search method to find the matching work number and problem type in raw_input. The twelfth to fifteenth lines of code represent the return of a dictionary containing the parsed work number and problem description. If the work number or problem type is not found, the default values ​​of unknown work number and others are used.

[0193] The above lines 16 to 19 of code define a tool called knowledge_query, which is used to retrieve solutions from the knowledge base. The input data model accepted by the tool is KnowledgeQueryModel.

[0194] The above lines 20 to 22 indicate calling the initialization method of the parent class and instantiating a KnowledgeBase object.

[0195] Lines 23 and 24 define a private method, _run, which executes the core logic of the knowledge_query tool. This method accepts a string argument, problem_desc, and calls the search method on KnowledgeBase to return the result.

[0196] The 25th to 28th lines of code above define a tool named ticket_creator, which is used to create second-line work tickets. The input data model accepted by the tool is KnowledgeQueryModel.

[0197] The above lines 29 to 31 of code indicate using the parent class initialization method and instantiating an ITSMClient object.

[0198] Lines 32 and 33 define a private method, _run, which executes the core logic of the ticket_creator tool. This method accepts a string parameter, problem_desc, and calls the ITSMClient's create_ticket method, returning the result.

[0199] The 34th to 37th lines of code above define a tool called ticket_checker, which is used to check the status of ticket processing. The input data model accepted by the tool is TicketCheckModel.

[0200] The 38th to 40th lines above indicate calling the initialization method of the parent class and instantiating an ITSMClient object.

[0201] Lines 41 and 42 define a private method, _run, which executes the core logic of the ticket_checker tool. This method accepts a string parameter, ticket_id, and calls the ITSMClient's check_ticket_status method to return the result.

[0202] 1-7. Agent system code

[0203] For example, the above agent system code can be as follows:

[0204]

[0205]

[0206] The first six lines of code define a tools list for storing tool class instances. Specifically, the second line creates an InputParserTool instance to parse the user-entered work number and problem type. The third line creates a KnowledgeQueryTool instance to retrieve solutions from the knowledge base. The fourth line creates a TicketCreatorTool instance to create second-tier work tickets. The fifth line creates a TicketCheckerTool instance to check the work ticket processing status.

[0207] Lines 7 through 23 of the code above define a system template, which serves as the agent's system instructions, guiding the agent in processing user requests according to the process. This system template includes the processing flow, processing rules, and tool usage formats. The tool usage format provides an example of the output structure of each tool, helping the agent to correctly interpret and use the tool's output.

[0208] The 24th to 28th lines of code above define the prompt template used when the agent interacts with the user. This prompt template is a ChatPromptTemplate object. Specifically, the 24th line of code defines a class method that accepts a message list to construct the prompt template. The 25th line of code creates a system message object with the content system_template. The 26th line of code creates a human message template object using the template "{input}". {input} will be replaced by the actual user input. The 27th line of code creates a system message object with the content "Current context:\n{agent_scratchpad}". {agent_scratchpad} will be replaced by the actual context content.

[0209] It's important to note that intelligent agents possess autonomy, responsiveness, initiative, and social capabilities. Autonomy refers to independent operation without human intervention, autonomously setting sub-goals and execution paths. Responsiveness refers to real-time perception of environmental changes and dynamic strategy adjustment. Initiative refers to proactively initiating goal-oriented behaviors. For example, intelligent customer service proactively inquires about user needs. Social capabilities refer to multi-agent collaboration and natural human-machine interaction.

[0210] Furthermore, the key technology stack for intelligent agents includes the perception layer, the cognitive layer, the decision layer, and the execution layer. The perception layer includes multimodal input processing and environmental sensor data fusion. The cognitive layer includes knowledge graph construction and retrieval, as well as memory mechanisms. The decision layer includes a dynamic workflow engine (e.g., LangGraph), a reinforcement learning policy network, and a risk-return assessment model. The execution layer includes tool or function calls and multimodal output generation.

[0211] 1-8. Scheduled task code

[0212] For example, the above scheduled task code can be as follows:

[0213]

[0214]

[0215] The first line of code above defines a class called AutoScheduler, which is used to manage automated task scheduling.

[0216] Lines 2 through 5 define the AutoScheduler initialization method, which accepts an executor parameter. Specifically, the third line creates a BackgroundScheduler instance to run the scheduled task in the background. The fourth line assigns the passed executor object to the instance variable self.executor to execute the task. The fifth line calls the _setup_jobs method to set up the scheduled task.

[0217] The sixth and seventh lines of code above define a private method, _setup_jobs, for setting up scheduled tasks. self.check_tickets specifies the method to execute, and 'cron' uses a cron trigger, allowing tasks to be executed at specific intervals. hour = 9 sets the task to execute at 9:00 AM every day.

[0218] Lines 8 to 15 of the code above define a private method, check_tickets, for checking the status of work tickets. Specifically, line 9 creates an ITSMClient instance for interacting with the work ticket management system. Line 10 iterates over all pending work tickets. Lines 11 to 13 call the invoke method of the executor, passing in an input dictionary containing the work ticket IDs to be checked. Line 14 checks whether the status in the execution result is resolved. Line 15 calls the _notify_user method to notify the user if the work ticket has been resolved.

[0219] The sixteenth and seventeenth lines of code above define a private method _notify_user to notify the user that the work order has been resolved.

[0220] Lines 18 and 19 define a private method called start to start the scheduler. Specifically, the scheduler's start method is called to begin executing the configured tasks.

[0221] 1-9. Main execution logic code

[0222] For example, the main execution logic code may be as follows:

[0223]

[0224]

[0225] The first line of code checks whether the current script is the main program entry point. The following code will only be executed if the script is run directly.

[0226] The second through seventh lines of code initialize the large language model. Specifically, the second line imports the ChatDeepSeek class from the langchain_deepseek module. The third line creates a ChatDeepSeek instance for interacting with the large language model. The fourth line retrieves the large language model from the CONFIG configuration file. The fifth line sets the temperature parameter to 0. The sixth line retrieves the DeepSeek API key from the CONFIG configuration file.

[0227] The eighth line of code creates an agent. StructuredChatAgent.from_llm_and_tools creates a structured chat agent using the large language model and tools. llm = llm passes the initialized large language model instance to the agent. tools = tools provides a list of tools to the agent. prompt = prompt provides a prompt template to the agent.

[0228] The ninth line of code creates an executor. AgentExecutor creates an executor instance. agent = agent passes the agent instance to the executor. tools = tools provides a list of tools to the executor. verbose = True enables verbose output.

[0229] The tenth line of code above creates an AutoScheduler instance and passes the executor to it.

[0230] The eleventh line of code above indicates starting the scheduler and starting to execute the set tasks.

[0231] The above lines 12 to 16 of code define a test case list to simulate user requests, where each case is a string.

[0232] Lines 17 through 20 of the code above execute the test case and output the results. Specifically, line 17 iterates through each test case. Line 18 outputs the currently processed user request. Line 19 calls the invoke method of the executor, passing in the current case and executing the request. Line 20 outputs the execution result, showing the agent's response to the user request.

[0233] It is understandable that the above-mentioned problem handling method can be implemented by a problem handling device. In order to realize the above-mentioned functions, the problem handling device includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the modules and algorithm steps of each example described in the embodiments disclosed herein, the embodiments disclosed in this application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the embodiments disclosed in this application.

[0234] The embodiments disclosed in this application can divide the functional modules of the problem processing device generated by the above method example. For example, each functional module can be divided according to each function, or two or more functions can be integrated into one processing module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiments disclosed in this application is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0235] Figure 8 This is a schematic diagram of the structure of a problem handling device provided in an embodiment of the present application. Figure 8 As shown, the problem handling device 80 can be used to perform Figure 3-Figure 7 The problem handling method shown in FIG. 80 includes a communication unit 801 and a processing unit 802 .

[0236] The communication unit 801 is used to obtain the user's input question; the processing unit 802 is used to obtain a target knowledge list from the knowledge base based on the input question; the target knowledge list includes multiple knowledge contents related to the input question and the similarity between each knowledge content and the input question; the processing unit 802 is also used to determine the confidence of the target knowledge list based on the similarity between each knowledge content and the input question; the processing unit 802 is also used to determine the solution to the input problem based on the confidence; the solution is determined based on the target knowledge list or manually.

[0237] In one possible implementation, the processing unit 802 is further used to optimize the input problem using a large language model to obtain a problem description; the problem description is the optimized input problem; the processing unit 802 is further used to traverse the knowledge base based on the vector representation of the problem description, filter out multiple knowledge contents related to the problem description from the knowledge base, and obtain a target knowledge list.

[0238] In a possible implementation, the processing unit 802 is further configured to determine the confidence of the target knowledge list based on the weight of each knowledge content, the similarity between each knowledge content and the input question, and the similarity weight index of each knowledge content.

[0239] In a possible implementation, the confidence level of the target knowledge list satisfies a preset formula, which is:

[0240]

[0241] Among them, C represents the confidence of the target knowledge list, w i represents the weight of the i-th knowledge content, s i represents the similarity between the i-th knowledge content and the input question, α i represents the similarity weight index of the i-th knowledge content, n represents the number of knowledge contents in the target knowledge list, and n is a positive integer.

[0242] In a possible implementation, the processing unit 802 is further configured to determine a solution to the input problem based on the target knowledge list when the confidence level is greater than or equal to a preset threshold.

[0243] In one possible implementation, the processing unit 802 is further used to create a work order when the confidence level is less than a preset threshold; the work order includes a user identifier and an input problem, and the status of the work order is to be assigned; the processing unit 802 is also used to determine a solution to the input problem based on the work order.

[0244] In a possible implementation, the processing unit 802 is further configured to detect the status of the work order at a preset time interval; the processing unit 802 is further configured to obtain a solution to the input problem when the status of the work order is resolved.

[0245] Through the description of the above embodiments, those skilled in the art will clearly understand that for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. The specific working processes of the above-described systems, devices, and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0246] The present disclosure also provides a computer-readable storage medium having instructions stored thereon. When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the problem handling method provided in the above-mentioned embodiment of the present disclosure.

[0247] The embodiments of the present disclosure also provide a computer program product containing instructions, which, when executed on an electronic device, enables the electronic device to execute the problem handling method provided by the embodiments of the present disclosure.

[0248] Among them, the computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a register, a hard disk, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above, or any other form of computer-readable storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be an integral part of the processor. The processor and the storage medium can be located in an application-specific integrated circuit (ASIC). In the embodiments of the present application, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0249] The above are only specific embodiments of the present application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A problem solving method, characterized in that: The method comprises: Get the user's input question; Based on the input question, a target knowledge list is obtained from a knowledge base; the target knowledge list includes a plurality of knowledge contents related to the input question and a similarity between each knowledge content and the input question; Determining the confidence of the target knowledge list based on the similarity between each knowledge content and the input question; Based on the confidence level, a solution to the input problem is determined; the solution is determined based on the target knowledge list or manually.

2. The method according to claim 1, characterized in that The step of obtaining a target knowledge list from a knowledge base based on the input question includes: Optimizing the input problem using a large language model to obtain a problem description; the problem description is the optimized input problem; According to the vector representation of the problem description, the knowledge base is traversed, and a plurality of knowledge contents related to the problem description are screened out from the knowledge base to obtain the target knowledge list.

3. The method according to claim 1, characterized in that Determining the confidence of the target knowledge list based on the similarity between each knowledge content and the input question includes: The confidence of the target knowledge list is determined based on the weight of each knowledge content, the similarity between each knowledge content and the input question, and the similarity weight index of each knowledge content.

4. The method according to claim 3, characterized in that The confidence level of the target knowledge list satisfies a preset formula, which is: Where C represents the confidence of the target knowledge list, w i represents the weight of the i-th knowledge content, s i represents the similarity between the i-th knowledge content and the input question, α i represents the similarity weight index of the i-th knowledge content, n represents the number of knowledge contents in the target knowledge list, and n is a positive integer.

5. The method according to claim 1 or 2, characterized in that Determining a solution to the input problem based on the confidence level includes: In a case where the confidence level is greater than or equal to a preset threshold, a solution to the input problem is determined based on the target knowledge list.

6. The method according to claim 1 or 2, characterized in that Determining a solution to the input problem based on the confidence level includes: When the confidence level is less than a preset threshold, a work order is created; the work order includes a user identifier and the input question, and the status of the work order is pending assignment; Based on the work order, a solution to the input problem is determined.

7. The method according to claim 6, characterized in that Determining a solution to the input problem based on the work order includes: Checking the status of the work order at preset time intervals; When the status of the work order is resolved, a solution to the input problem is obtained.

8. A problem handling device, characterized in that: The device includes: a communication unit and a processing unit; The communication unit is used to obtain the user's input question; The processing unit is configured to obtain a target knowledge list from a knowledge base based on the input question; the target knowledge list includes a plurality of knowledge contents related to the input question and a similarity between each knowledge content and the input question; The processing unit is further configured to determine the confidence level of the target knowledge list based on the similarity between each knowledge content and the input question; The processing unit is further configured to determine a solution to the input problem based on the confidence level; the solution is determined based on the target knowledge list or manually.

9. A problem handling device, characterized in that: include: A processor and a communication interface; the communication interface is coupled to the processor, and the processor is used to run a computer program or instruction to implement the problem handling method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and when the computer instructions are executed on a computer, the computer is caused to execute the problem handling method according to any one of claims 1 to 7.

Images