Communication service processing method, electronic device, storage medium and computer product

Abstract

The application discloses a communication service processing method, an electronic device, a storage medium and a computer product. The method comprises the following steps: acquiring first information related to a communication service, wherein the first information comprises multi-dimensional information obtained by performing intent recognition on a conversation related to the communication service; acquiring second information related to the communication service, wherein the second information comprises an execution action of each executable task unit obtained by dividing the first information, and an access condition of each executable task unit; acquiring third information, wherein the third information comprises analysis data generated by analyzing a result obtained by completing the execution action of each executable task unit under the access condition; and inputting the analysis data into a large language model to obtain a processing strategy related to the communication service output by the large language model.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a communication service processing method, electronic device, storage medium and computer product. Background Technology

[0002] The existing support system has established instant messaging and work order scheduling modules. When on-site personnel encounter specific problems during production, they can submit work orders through the support system's mobile application (app) to ask questions and seek technical support from back-end personnel. After receiving the work order on the support system's personal computer (PC), the back-end personnel will communicate with the on-site personnel, verify the live network data based on the specific problem, and provide technical solutions to the on-site personnel based on their personal experience.

[0003] It is evident that the current approach relies on backend processing to address the problems encountered in communication services. Summary of the Invention

[0004] This application provides a communication service processing method, an electronic device, a storage medium, and a computer product, which improve the ability of electronic devices to handle problems encountered in communication services on-site.

[0005] The technical solution of this application embodiment is implemented as follows:

[0006] This application provides a communication service processing method, the method comprising:

[0007] Acquire first information related to communication services, the first information including multi-dimensional information obtained by performing intent recognition on the dialogue related to the communication services;

[0008] Obtain second information related to the communication service, the second information including the execution action of each executable task unit obtained by dividing the first information, and the admission conditions of each executable task unit;

[0009] Obtain third information, which includes analysis data generated by analyzing the results of the execution actions of each executable task unit under the admission conditions;

[0010] The analyzed data is input into a large language model to obtain the processing strategies related to the communication service output by the large language model.

[0011] In the above scheme, obtaining the first information related to communication services includes:

[0012] Obtain the target text by combining the text of the (n+1)th round of dialogue related to the communication service with the text of the previous n rounds of dialogue, where n is a positive integer;

[0013] The target text is input into the large language model to obtain the first information output by the large language model.

[0014] In the above scheme, obtaining the second information related to the communication service includes:

[0015] The first information is input into the large language model to obtain the return results of the execution actions of each executable task unit and related analysis information output by the large language model.

[0016] The interactive prompt information generated by the large language model based on the returned results and the relevant analysis information is obtained. The interactive prompt information is used to guide the target object to perform the target operation and return the result of the target operation. The target operation includes on-site investigation and / or operation of the target device.

[0017] The business data extracted by the large language model based on the result of the target operation is used as the admission condition for each executable task unit.

[0018] In the above scheme, the step of inputting the analyzed data into a large language model to obtain the communication service-related processing strategy output by the large language model includes:

[0019] The analysis data is input into a large language model to obtain a first strategy corresponding to each executable task unit output by the large language model. The first strategy includes first text information that conforms to the communication service logic. The processing strategy includes the first strategy.

[0020] In the above scheme, after inputting the analysis data into the large language model to obtain the first strategy corresponding to each executable task unit output by the large language model, the method further includes:

[0021] A second strategy is generated based on the first strategy. The second strategy includes one or more of the following: second text information, image information, and document information that conform to the communication business logic. The second text information is obtained by expanding the text of the first text information.

[0022] In the above scheme, before inputting the analysis data into the large language model to obtain the first strategy corresponding to each executable task unit output by the large language model, the method further includes:

[0023] The autoregressive language model is trained and then fine-tuned based on the content objective composed of each sample's executable task unit and sample policy to obtain the large language model.

[0024] This application provides a communication service processing apparatus, including:

[0025] An acquisition unit is used to acquire first information related to communication services, the first information including multi-dimensional information obtained by performing intent recognition on the dialogue related to the communication services.

[0026] The acquisition unit is used to acquire second information related to the communication service. The second information includes the execution action of each executable task unit obtained by dividing the first information, and the admission conditions of each executable task unit.

[0027] An acquisition unit is used to acquire third information, which includes analysis data generated by analyzing the results of the execution actions of each executable task unit under the admission conditions.

[0028] The processing unit is used to input the analysis data into the large language model to obtain the processing strategy related to the communication service output by the large language model.

[0029] This application also provides an electronic device, including: a processor and a memory for storing a computer program capable of running on the processor; wherein,

[0030] The processor is configured to execute the steps of any of the above methods when running the computer program.

[0031] This application also provides a storage medium storing a computer program thereon, characterized in that the computer program, when executed by a processor, implements the steps of any of the above methods.

[0032] This application also provides a computer product, including a computer program, which, when executed by a processor, implements the steps of any of the above methods.

[0033] This application provides a communication service processing method, electronic device, storage medium, and computer product. The method includes: acquiring first information related to communication services, the first information including multi-dimensional information obtained by intention recognition of communication service-related dialogues; acquiring second information related to communication services, the second information including the execution actions of each executable task unit obtained by dividing the first information, and the admission conditions of each executable task unit; acquiring third information, the third information including analysis data generated by analyzing the results obtained by completing the execution actions of each executable task unit under the admission conditions; and inputting the analysis data into a large language model to obtain the communication service-related processing strategy output by the large language model. Attached Figure Description

[0034] Figure 1This is a flowchart illustrating a communication service processing method according to an embodiment of this application;

[0035] Figure 2 This is a schematic diagram illustrating the use of a sliding window mechanism to handle dialogue in an embodiment of this application;

[0036] Figure 3 This is a schematic diagram of the communication service processing flow according to an embodiment of this application;

[0037] Figure 4 This is a schematic diagram illustrating the intent of an embodiment of this application;

[0038] Figure 5 This is a flowchart illustrating the transformation from a simplified to a detailed embodiment of this application.

[0039] Figure 6 This is a schematic diagram of the interface for a business processing scenario in an embodiment of this application;

[0040] Figure 7 This is a schematic diagram illustrating the distributed deployment of the business support system according to an embodiment of this application;

[0041] Figure 8 This is a schematic diagram of the communication service processing device according to an embodiment of this application;

[0042] Figure 9 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0043] The present application will now be described in further detail with reference to the accompanying drawings and embodiments.

[0044] The existing support system has established instant messaging and work order scheduling modules. When field personnel encounter specific problems during production, they can submit work orders through the support system's mobile app and ask questions to the back-end personnel for technical support. After receiving the work order on the support system's PC, the back-end personnel communicate with the field personnel, verify the live network data based on the specific problem, and provide technical solutions to the field personnel based on their personal experience. It is evident that the current system relies on the back-end to handle problems encountered in communication services.

[0045] This application introduces a Large Language Model (LLM) construction support system on the electronic device side, providing an intelligent assistant and building capabilities for multi-turn dialogue, task orchestration, automatic execution, and solution generation. This enhances the ability of electronic devices to handle problems encountered in communication services on-site, thereby increasing the proportion of self-service problem handling by on-site personnel and reducing the workload of back-end personnel.

[0046] The electronic device described in this application can be implemented as a laptop computer, tablet computer, desktop computer, mobile device (e.g., mobile phone, portable music player, personal digital assistant, dedicated messaging device, portable gaming device), intelligent robot, or any terminal with screen display function, or it can be implemented as a server.

[0047] This application provides a communication service processing method, applied to electronic devices, such as... Figure 1 As shown, the method includes:

[0048] Step 101: Obtain first information related to communication services. The first information includes multi-dimensional information obtained by performing intent recognition on dialogues related to communication services.

[0049] In practical applications, electronic devices obtain multi-dimensional information by performing intent recognition on dialogues (which can be multi-turn dialogues) related to communication services. This multi-dimensional information includes multiple attributes of the communication equipment providing the communication services, such as the equipment's serial number (SN) and broadband account.

[0050] Step 102: Obtain second information related to communication services. The second information includes the execution actions of each executable task unit obtained by dividing the first information, and the admission conditions of each executable task unit.

[0051] In practical applications, given the initial information, it can be divided into smaller executable task units T1, T2, ..., TP, with each executable task unit being as small as possible. Each executable task unit has a corresponding admission condition; execution of the corresponding executable task unit will only be triggered when a certain admission condition is met. Here, P is the total number of executable task units.

[0052] In practical applications, the execution of a task is divided into multiple steps. The key business data extracted from the data analysis in step m serves as the admission criteria to determine the executable task unit in step m+1, where m is a positive integer.

[0053] Step 103: Obtain third information, which includes analysis data generated by analyzing the results of the execution actions of each executable task unit under the admission conditions.

[0054] In practical applications, the third information includes integrated information obtained after multiple rounds of dialogue, execution of all executable task units, and data analysis (also known as live network verification data).

[0055] Step 104: Input the analysis data into the large language model to obtain the communication service-related processing strategies output by the large language model.

[0056] In practical applications, LLM refers to a deep learning model trained on large amounts of text data that can generate natural language text or understand the meaning of language text. Large language models can handle a variety of natural language tasks, such as text classification, question answering, and dialogue, and are an important pathway to artificial intelligence.

[0057] In practical applications, the analysis data is input into the large language model, which can call the fault case library for comparison and analysis to obtain the communication service-related processing strategies output by the large language model.

[0058] This application provides a communication service processing method, which includes: acquiring first information related to the communication service, the first information including multi-dimensional information obtained by intention recognition of the communication service related dialogue; acquiring second information related to the communication service, the second information including the execution actions of each executable task unit obtained by dividing the first information, and the admission conditions of each executable task unit; acquiring third information, the third information including analysis data generated by analyzing the results obtained by completing the execution actions of each executable task unit under the admission conditions; and inputting the analysis data into a large language model to obtain the communication service related processing strategy output by the large language model.

[0059] In some embodiments, step 101 obtains first information related to communication services, including:

[0060] Get the target text by combining the text of the (n+1)th round of dialogue related to communication services with the text of the previous n rounds of dialogue, where n is a positive integer;

[0061] Input the target text into the large language model to obtain the first information output by the large language model.

[0062] In practical applications, the processing of multi-turn dialogues adopts a sliding window mechanism. For the (n+1)th turn of dialogue, when the large language model answers the question in the current turn, it combines the text of the question in the current turn with the text of the previous n turns of historical dialogue to obtain a new question text, i.e., the target text, which is then input into the large language model.

[0063] See Figure 2 As shown, taking the sliding window mechanism for processing two adjacent rounds of dialogue as an example, during human-computer interaction, the user engages in multiple rounds of dialogue with the electronic device, with new dialogues constantly being generated. The sliding window 201 slides to the side indicated by arrow 202 (e.g., the right side) to obtain new question text as the target text. This application uses the sliding window mechanism to predict the current answer by splicing the previous dialogue and the current dialogue (i.e., the current question) when processing multi-round dialogues. When the dialogue content exceeds the window length, the earliest sentence is discarded to ensure the continuity and relevance of the dialogue and improve the accuracy of the first piece of information obtained.

[0064] See Figure 3 As shown, in an implementable communication service processing flow, taking a three-stage example, the communication service processing method provided in this application embodiment will be described: the three stages include the "identifying the problem" stage 301, the "planning and execution" stage 302, and the "output result" stage 303;

[0065] In the "Identify the Problem" step 301, LLM completes information collection through multiple rounds of dialogue with users, identifies the user's true intentions, and clarifies the problem that the user really wants to solve.

[0066] In the "Clarify the Problem" stage 301, the user asks a question on the spot, the electronic device recognizes the intent and judges whether the information dimensions are complete. If not, the question is completed through multiple rounds of interaction between the user and the electronic device; if yes, the process proceeds to the "Planning and Execution" stage 302.

[0067] Dialogue Example 1 is shown in Table 1 below. In this example, the LLM guides the user to input device information based on the scene recognition results.

[0068]

[0069] Table 1

[0070] Dialogue Example 2 is shown in Table 2 below. In this example, LLM guides the user to input the problem scenario based on the recognition results.

[0071]

[0072] Table 2

[0073] LLM (Local Modem) gathers information through multi-turn dialogue, identifies the user's true intent, and clarifies the problem the user truly wants to solve. In other words, user intent is obtained through multi-turn dialogue to complete the question and then analyze it. Through these multi-turn dialogues, LLM stores multiple dimensions of the current device's attributes in short-term memory, such as the device's serial number (SN) and broadband account information.

[0074] See Figure 4As shown, a user initiates a conversation (i.e., asks a question on the spot), and the question can be in the form of text, voice, or images. The LLM, in conjunction with its image character recognition model and speech-to-text capabilities, perceives the user input. The LLM recognizes and integrates information from text, image, and voice inputs, obtaining the following: the question (including phenomenon, category, misalignment code, etc.), the device (including type, identifier, etc.), and other information. It can also combine expert experience from long-term memory (also known as domain knowledge, such as user questioning habits, case studies, interface documentation, device manuals, etc.) to determine whether the information provided by the user is complete. If the information is incomplete, the LLM generates guiding words to guide the user to supplement the missing information in a new round of dialogue. This continues until the LLM infers and recognizes complete question information from the user's conversation context.

[0075] The "Planning and Execution" stage 302 includes task orchestration, task execution, and data analysis. The actual production scenarios for the planning and execution stage are quite complex and may involve various situations. In response to these situations, this stage may also include on-site user operations and multiple rounds of interaction.

[0076] Task orchestration refers to domain experts using multiple sets of natural language to describe the problem-solving process under different user intentions, including the application programming interfaces (APIs) involved in each executable task unit, the system's interaction with the data returned by the system, and the judgment rules. Domain knowledge is then imported into the LLM application through pre-training, fine-tuning, or prompt word engineering.

[0077] Task execution refers to the interaction between the LLM and dozens of supporting systems to access live data from multiple data sources, including data querying, modification, and deletion operations. Furthermore, access control is introduced during task execution, firstly by controlling access to functional modules, and secondly by implementing domain-based control over data.

[0078] Data analysis refers to analyzing the results returned by the LLM from dozens of supporting systems. This involves parsing the data according to the interface definition specifications and extracting key business data, which refers to the entry conditions for executable task units in the process. After each subtask is completed, the data generated needs to be analyzed. Combining this with the expert experience learned by the LLM, at certain points in the process, it's necessary to guide installation and maintenance personnel to conduct on-site inspections. The LLM outputs guiding words to engage in multiple rounds of dialogue with the personnel to gather more information to assist in troubleshooting or resolving issues. The LLM then places this integrated information into the context for reasoning about the next executable task, until the task is completed.

[0079] In practical applications, during the entire diagnostic and treatment process, if intermediate steps require manual on-site verification or operation, and a self-closing loop cannot be achieved through API scheduling, LLM precisely guides the human to conduct on-site investigation or operation of a specific device. It then guides the person to return the verification or operation results via dialogue. The person can return this information through natural language text or on-site photos. Through multi-turn dialogue and interaction, LLM obtains the "admission conditions" for a specific "executable action A." LLM extracts this information from the dialogue, places it as environmental parameters within the context, and continues to consider the next step according to the thought process.

[0080] As can be seen, the multi-round interaction in this application is reflected in the "identify the problem" stage 301 and the "planning and execution" stage 302. In the entire process of problem identification, diagnosis and handling, LLM introduces multi-round dialogue. In the two key stages of "identify the problem" and "planning and execution", the user is guided to give feedback through text output, and the user's feedback is used as the input for subsequent process handling.

[0081] This application employs a series of processes to automate problem handling: refining the problem through multi-round dialogue, understanding the problem through intent recognition, automatically solving the problem through planning and execution, and providing recommended solutions. Furthermore, this application can incorporate preset solutions into a prompt model, improving the accuracy and usability of the solution recommendation function in the self-service problem handling device.

[0082] In some embodiments, step 102 obtains second information related to communication services, including:

[0083] The first information is input into the large language model to obtain the return results of the execution actions of each executable task unit output by the large language model and related analysis information;

[0084] The system obtains interactive prompts generated by the large language model based on the returned results and related analysis information. These prompts guide the target object to perform the target operation and return the result of the target operation. The target operation includes on-site investigation and / or operation of the target device.

[0085] Obtain business data extracted from the results of the target operation based on the large language model, and use it as the admission condition for each executable task unit.

[0086] In practical applications, upon receiving initial information (e.g., the problem Q that the user wants to solve), LLM plans Q into smaller executable task units T1, T2, ..., TP, with each executable task unit being as small as possible. From the expert experience in long-term memory, it infers the execution action A of each executable task unit and the admission conditions of each executable task unit under problem Q.

[0087] For example, Table 3 provides the executable task units, admission conditions, and execution actions planned for the user question Q (i.e., user intent Q):

[0088]

[0089] Table 3

[0090] In some scenarios, during the "Planning and Execution" stage (302), the actual production environment is complex and may present multiple situations. Possibility 1: Some devices are not connected to the network and cannot be queried from the information management system via automated means. Understanding the status of these devices requires manual on-site inspection. This includes: 1) Faults in some non-intelligent facilities, such as bent network lines, loose interfaces, power failures, etc. 2) Device malfunctions, resulting in a disconnection from the uplink device, making it impossible to obtain device status and perform deeper diagnostics. For example, a frozen or malfunctioning optical modem causing a disconnection from the uplink OLT. Possibility 2: Some operations require on-site user intervention and cannot be handled remotely. After guiding the user to complete the operation, the diagnostic or processing flow still needs to continue. For example, the user needs to change their password on-site before the user is logged out. Here, Table 4 is used as an example to illustrate the dialogue in the "Planning and Execution" stage:

[0091]

[0092]

[0093] Table 4

[0094] It should be noted that when calling the optical modem's self-discovery tool, the result is an array, and for example, the corresponding values ​​in R / F / S / P are different.

[0095] As can be seen from the above, this application introduces multi-round dialogue into the self-service problem handling process. The support staff can supplement the on-site situation through text dialogue, photos, screenshots, etc., and take into account the status of devices not connected to the network and operations that require manual intervention during the process of troubleshooting problems on the live network.

[0096] In some embodiments, step 104 inputs the analyzed data into the large language model to obtain the communication service-related processing strategy output by the large language model, including:

[0097] The analysis data is input into the large language model to obtain the first strategy corresponding to each executable task unit output by the large language model. The first strategy includes the first text information that conforms to the communication business logic; the processing strategy includes the first strategy.

[0098] In practical applications, the first strategy is called abbreviation scheme. For example, a short text is analyzed and identified by combining the keyword analysis and expert experience from the existing case library to select unique corpus that conforms to the business logic. Based on this corpus, LLM is trained, and the model parameters are tuned and optimized by referring to business cases. The number of model layers and parameter settings are modified. Under the premise that the previous process is accurate, the accuracy of LLM inferring abbreviation scheme by combining live network verification data is as high as about 98%.

[0099] This application, based on data verified on the current network and combined with fault cases from a large model trained / fine-tuned by domain experts, provides specific recommended solutions according to different equipment states. To avoid providing only a correct but meaningless reference based on probability calculations by the large model, and to improve the accuracy of the recommended solutions, enabling the large model to provide more constructive and practical suggestions for specific production scenarios, the solution is divided into two parts: a simplified solution and solution details.

[0100] In some embodiments, before inputting the analyzed data into the large language model to obtain the first strategy corresponding to each executable task unit output by the large language model, the communication service processing method further includes:

[0101] The autoregressive language model is trained and then fine-tuned based on the content objective composed of each sample's executable task unit and sample policy to obtain a large language model.

[0102] In practical applications, the content is the result of each executable task unit, the target pair is a shortened scheme, and the content and target pair form a content-target pair.

[0103] In a feasible scenario, the steps for handling business processes that cannot access the internet are as follows (for ease of reading, some steps have been omitted, but key steps are retained to illustrate the process of obtaining the recommended solution):

[0104] Step 1: Check the account's online status by using the "Check Account Online Duration" tool;

[0105] If the returned "Online Time" is empty, the account is offline, proceed to step 2;

[0106] If the returned "Online Time" is not empty, the account is online, proceed to step 3;

[0107] Step 2: Check account verification status. Use the "Account Online / Offline Records" tool to view the most recent record.

[0108] If the returned "Package Type" is not "Authentication Failed", then proceed to step 12;

[0109] If the returned "Package Type" content is "Authentication Failed" and the authentication result is "Binding Information Error", then proceed to step 4;

[0110] If the returned "Package Type" content is "Authentication Failed" and the authentication result is "User's online authentication has been rejected", then proceed to step 5;

[0111] If the returned "Package Type" content is "Authentication Failed" and the authentication result is "Incorrect Password", then proceed to step 6;

[0112] If the returned "Package Type" content is "Authentication Failed" and the authentication result is "User Locked", then proceed to step 7;

[0113] If the returned "Package Type" content is "Authentication failed" and the authentication result is "User does not exist", then proceed to step 8;

[0114] If the returned "Package Type" content is "Authentication Failed", and the authentication result is not any of the following: "Binding Information Error", "User Online Authentication Rejected", "Password Error", "User Locked", or "User Does Not Exist", then proceed to step 35.

[0115] Step 3: Analyze the account disconnection records. Use the "Account Login / Outbound Records" tool to view the most recent record.

[0116] If the returned "Package Type" does not fall under either "Authentication Failed" or "Billing Ended", then proceed to step 24.

[0117] If the returned "Package Type" is "Authentication Failed" and the authentication result is "User Online Authentication Rejected", then proceed to step 5;

[0118] If the returned "Package Type" is "Billing End" and the authentication result is "User Offline (Session Timeout)", then proceed to step 23.

[0119] Step 24: Determine if there are frequent disconnection conditions by calling the "Account Login / Outbound Record" tool;

[0120] If the "frequent disconnections" field is True in the first dictionary of the returned data list, then execute step 28.

[0121] If the first dictionary in the returned data list is set to False for the "Frequent disconnections" field, proceed to step 30.

[0122] Step 30: Determine if the optical power is abnormal by calling the "Optical Modem and OLT Light Reception" tool;

[0123] If the tool results show "ONU received optical power" as "normal" and "OLT received ONU optical power" as "normal", proceed to step 31.

[0124] If the tool results show "ONU received optical power" as "normal" and "OLT received ONU optical power" as "strong light", proceed to step 36.

[0125] If the tool results show "ONU received optical power" as "normal" and "OLT received ONU optical power" as "weak", proceed to step 37.

[0126] Step 36:

[0127] Diagnostic results: OLT receiving ONU optical power: strong; ONU receiving optical power: normal;

[0128] Recommended solution: Please implement high-power beam correction.

[0129] The process is complete.

[0130] The above is the complete business logic.

[0131] Output example:

[0132] The question is an input question that must be answered.

[0133] Based on the step number given in the question or the step number to be executed mentioned in the question, please extract the original complete fragment of the corresponding step from the business logic. Please ignore other content in the question and do not modify any content of the extracted step fragment.

[0134] Example 1:

[0135] Question:

[0136] Step 1.

[0137] Answer:

[0138] Step 1: Check the account's online status, determine if the online time is empty, and use the "Check Account Online Time" tool;

[0139] If the returned "Online Time" is empty, the account is offline, proceed to step 2;

[0140] If the returned "Online Time" is not empty, the account is online, proceed to step 3;

[0141] Example 2:

[0142] Question:

[0143] Thought process: In the results returned by the tool for querying account online time, the "online time" field is not empty.

[0144] The selected next step number is: step3.

[0145] Answer:

[0146] Step 3: Analyze the account disconnection records. Use the "Account Login / Outbound Records" tool to view the most recent record.

[0147] If the returned "Package Type" is "Authentication Failed" and the authentication result is "User Online Authentication Rejected", then proceed to step 5;

[0148] If the returned "Package Type" is "Billing Ended" and the authentication result is "User Offline (Session Timeout)", then proceed to step 23;

[0149] If the returned "Package Type" is not "Authentication Failed" or "Billing Ended", then proceed to step 24;

[0150] It should be noted that: based on the step number given in the question or the step number to be executed mentioned in the question, please extract the original complete fragment of the corresponding step from the business logic. Please ignore other content in the question and do not modify any content of the extracted step fragment.

[0151] For example, the current problem:

[0152] Question: {question}

[0153] Answer:

[0154] """

[0155] During the above processing, a large language model is trained using business-specific data with an efficient parameter fine-tuning scheme (such as LoRa). During the fine-tuning process, a set of parameters, i.e., the model's total parameters Φ, is found to maximize Equation 1.

[0156]

[0157] By adding noise ε to the loss function, we obtain Formula 2 as follows, which enhances the generalization ability of the model's inference:

[0158]

[0159] Adding a multilayer perceptron (MLP) layer to the last layer of the model yields Equation 3, which is used for classification:

[0160]

[0161] Among them, P Φ (y|x,y) represents a pre-trained autoregressive language model, Z={(x i ,y i )}i=1,…,N represents the content-target pair for each downstream task. The content is the result of each executable task unit, and the target pair is a shortened scheme, such as {“Please perform optical power rectification”, “Please call 10086 to change the password”, “Please close the port”,..}.

[0162] In some embodiments, after inputting the analyzed data into the large language model to obtain the first strategy corresponding to each executable task unit output by the large language model, the communication service processing method further includes:

[0163] The second strategy is generated based on the first strategy. The second strategy includes one or more of the following: second text information, image information, and document information that conform to the communication business logic. The second text information is obtained by expanding the first text information.

[0164] In practical applications, the second strategy, known as the detailed solution, builds upon the abbreviated solution by describing the specific operational steps in detail, providing more comprehensive explanations of certain concepts, or offering additional reference information. The detailed solution can be output in formats such as text, images, and tables. A large library of case studies and device documentation is injected into the large language model through pre-training or external knowledge bases. The system then combines the current network verification data with the abbreviated solution to deduce the detailed solution using the large language model.

[0165] See Figure 5 As shown, taking the abbreviated solution "Please rectify the optical power" as an example, the problem is expanded based on the abbreviated solution, the knowledge base text is searched, and finally the image and text are linked to obtain the output detailed solution, that is, the detailed measures on how to rectify the strong light.

[0166] See Figure 6 As shown, in a feasible business processing scenario, with Figure 6Taking the interface shown as an example, the process from initiating a diagnosis to presenting the results is explained: The large language model processes the user's input. The first interface, 601, displays: "Diagnosing the problem of broadband account A being unable to access the internet. Please wait. Intelligent diagnosis in progress..." The second interface, 602, displays: "Diagnosis result: Account binding information error; information as follows: 4. Account online status: Offline; 5. Authentication information: Authentication identifier, authentication time, authentication information, reason for authentication failure (specific details of each item are not shown); 6. Optical modem status: Offline." The third interface, 603, displays: "Recommended solution: Port closed. Please handle according to local customs." The fourth interface, 604, displays: "More (showing solution details)," presented in text, such as describing specific operation steps; in images, such as processing requirements for different cities; and in documents, such as operation specifications. In this embodiment, the large language model returns the results of intelligent problem processing to the user. The problem processing results are divided into several parts: guiding words, data analysis results, data query results, abbreviated solutions, and solution details. It should be noted that the first interface 601, the second interface 602, the third interface 603, and the fourth interface 604 can be defined and combined according to the actual situation. This application does not make any specific limitations on this.

[0167] In practical applications, after the dialogue ends, experts in the field can evaluate the intelligent processing of the large language model, continuously supplement their experience, and use the domain knowledge provided by the experts to perform post-processing corrections on the large language model.

[0168] See Figure 7 As shown, the business support system can be distributed across multiple servers. The business server 701 is responsible for the unified control and scheduling of the search engine. The component server 702 stores embedded information such as prior rules or knowledge, API information, and database tables. This information is used to search for the closest knowledge entries in the vector library as supplementary background knowledge for model input before model calls. The component server 702 stores short-term memory in an in-memory database and long-term memory in a structured database. The business server 701 can communicate with different information management platforms through a unified API. The model server 703 can deploy one or more large language models after resource evaluation as needed. The entire system can be intelligently deployed based on cloud and containerization technologies, offering high portability and facilitating continuous deployment.

[0169] The business support system refines issues through multi-round dialogue, understands issues through intent recognition, automatically resolves issues through planning and execution, provides recommended solutions, and ultimately achieves an automated problem-solving process. In troubleshooting issues on the live network, it can take into account the status of devices not yet connected to the network and situations requiring manual intervention.

[0170] This application also provides a communication service processing device, such as... Figure 8 As shown, the communication service processing device includes:

[0171] The acquisition unit 801 is used to acquire first information related to communication services, the first information including multi-dimensional information obtained by performing intent recognition on dialogues related to communication services.

[0172] The acquisition unit 801 is used to acquire second information related to communication services. The second information includes the execution actions of each executable task unit obtained by dividing the first information, and the admission conditions of each executable task unit.

[0173] The acquisition unit 801 is used to acquire third information, which includes analysis data generated by analyzing the results of the execution actions of each executable task unit under the admission conditions.

[0174] The processing unit 802 is used to input the analysis data into the large language model and obtain the communication service-related processing strategies output by the large language model.

[0175] In some embodiments, the acquisition unit 801 is used to acquire the target text obtained by combining the text of the (n+1)th round of dialogue related to the communication service with the text of the previous n rounds of dialogue, where n is a positive integer;

[0176] The processing unit 802 is used to input the target text into the large language model and obtain the first information output by the large language model.

[0177] In some embodiments, the acquisition unit 801 is used to acquire the target text obtained by combining the text of the (n+1)th round of dialogue related to the communication service with the text of the previous n rounds of dialogue, where n is a positive integer;

[0178] The processing unit 802 is used to input the target text into the large language model and obtain the first information output by the large language model.

[0179] In some embodiments, the processing unit 802 is used to input the first information into the large language model to obtain the return results of the execution actions of each executable task unit output by the large language model and related analysis information;

[0180] The acquisition unit 801 is used to acquire interactive prompt information generated by the large language model based on the returned results and related analysis information. The interactive prompt information is used to guide the target object to perform the target operation and return the result of the target operation. The target operation includes on-site investigation and / or operation of the target device.

[0181] The acquisition unit 801 is used to acquire business data extracted from the results of the target operation based on the large language model, as the admission condition for each executable task unit.

[0182] In some embodiments, the processing unit 802 is used to input the analysis data into the large language model to obtain a first strategy corresponding to each executable task unit output by the large language model. The first strategy includes first text information that conforms to the communication service logic. The processing strategy includes the first strategy.

[0183] In some embodiments, the processing unit 802 is used to generate a second strategy based on a first strategy. The second strategy includes one or more of the following: second text information, image information, and document information that conform to communication service logic. The second text information is obtained by text expansion of the first text information.

[0184] In some embodiments, the processing unit 802 is used to train the autoregressive language model and fine-tune the autoregressive language model based on the content objective composed of each sample executable task unit and sample policy to obtain a large language model.

[0185] To implement the communication service processing method of the embodiments of this application, the embodiments of this application also provide an electronic device, such as... Figure 9 As shown, the electronic device 900 includes: a communication interface 901 and a processor 902; wherein,

[0186] Communication interface 901;

[0187] The processor 902, connected to the communication interface 901, is used to execute the above-mentioned information service processing method when running a computer program;

[0188] Memory 903 stores computer programs that can run on processor 902.

[0189] The communication interface is used to: acquire first information related to communication services, the first information including multi-dimensional information obtained by performing intent recognition on dialogues related to communication services;

[0190] Obtain second information related to communication services. The second information includes the execution actions of each executable task unit obtained by dividing the first information, and the admission conditions of each executable task unit.

[0191] Obtain third information, which includes analysis data generated by analyzing the results of the execution actions of each executable task unit under the admission conditions;

[0192] Processor 902 is used to: input analysis data into a large language model and obtain communication service-related processing strategies output by the large language model.

[0193] In some embodiments, the communication interface 901 is used to: obtain the target text obtained by combining the text of the (n+1)th round of dialogue related to the communication service with the text of the previous n rounds of dialogue, where n is a positive integer;

[0194] Processor 902 is used to: input target text into a large language model and obtain the first information output by the large language model.

[0195] In some embodiments, the processor 902 is configured to: input the first information into the large language model, and obtain the return results of the execution actions of each executable task unit output by the large language model and related analysis information;

[0196] Communication interface 901 is used to: obtain interactive prompts generated by the large language model based on the returned results and related analysis information. The interactive prompts are used to guide the target object to perform the target operation and return the result of the target operation. The target operation includes on-site investigation and / or operation of the target device.

[0197] Communication interface 901 is used to: obtain business data extracted from the results of the target operation based on the large language model, as an admission condition for each executable task unit.

[0198] In some embodiments, the processor 902 is configured to: input analysis data into a large language model to obtain a first strategy corresponding to each executable task unit output by the large language model, wherein the first strategy includes first text information conforming to communication service logic; and process the strategy including the first strategy.

[0199] In some embodiments, the processor 902 is configured to: generate a second strategy based on a first strategy, the second strategy including one or more of the following: second text information, image information, document information conforming to communication service logic, wherein the second text information is obtained by text expansion of the first text information.

[0200] In some embodiments, processor 902 is configured to: train an autoregressive language model and fine-tune the autoregressive language model based on a content objective composed of each sample executable task unit and sample policy to obtain a large language model.

[0201] The methods disclosed in the embodiments of this application can be applied to or implemented by the processor 902. The processor 902 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor 902 or by instructions in the form of software. The processor 902 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 902 can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware decoding processor, or being executed by a combination of hardware and software modules in the decoding processor. The software modules may be located in a storage medium, which is located in the memory 903. The processor 902 reads the information in the memory 903 and combines it with its hardware to complete the steps of the aforementioned method.

[0202] In an exemplary embodiment, the electronic device 900 may be implemented by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers (MCUs), microprocessors, or other electronic components to perform the aforementioned method.

[0203] It is understood that the memory in the embodiments of this application can be volatile memory or non-volatile memory, or both. Specifically, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), ferromagnetic random access memory (FRAM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); magnetic surface memory can be disk storage or magnetic tape storage. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as Static Random Access Memory (SRAM), Synchronous Static Random Access Memory (SSRAM), Dynamic Random Access Memory (DRAM), Synchronous Dynamic Random Access Memory (SDRAM), Double Data Rate Synchronous Dynamic Random Access Memory (DDRSDRAM), Enhanced Synchronous Dynamic Random Access Memory (ESDRAM), Sync Link Dynamic Random Access Memory (SLDRAM), and Direct Rambus Random Access Memory (DRRAM).The memories described in the embodiments of this application are intended to include, but are not limited to, these and any other suitable types of memories.

[0204] In an exemplary embodiment, this application also provides a storage medium, namely a computer storage medium, specifically a computer-readable storage medium, such as a memory 903 storing a computer program, which can be executed by the processor 902 of the electronic device 900 to complete the steps of the aforementioned method. The computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, Flash Memory, magnetic surface memory, optical disc, or CD-ROM.

[0205] In an exemplary embodiment, this application also provides a computer product including a computer program that can be executed by a processor 902 of an electronic device 900 to complete the steps of the aforementioned method.

[0206] It should be noted that terms such as "first" and "second" are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0207] Furthermore, the technical solutions described in the embodiments of this application can be combined arbitrarily without conflict.

[0208] The above description is merely a preferred embodiment of this application and is not intended to limit the scope of protection of this application.

Claims

1. A communication service processing method, characterized in that, The method includes: Acquire first information related to communication services, the first information including multi-dimensional information obtained by performing intent recognition on the dialogue related to the communication services; Obtain second information related to the communication service. The second information includes the execution action of each executable task unit obtained by dividing the first information, and the admission condition of each executable task unit. The execution of the executable task unit corresponding to the admission condition will be triggered only when a certain admission condition is met. Obtain third information, which includes analysis data generated by analyzing the results of the execution actions of each executable task unit under the admission conditions; The analyzed data is input into a large language model to obtain the processing strategy related to the communication service output by the large language model. The acquisition of the first information related to communication services includes: A sliding window mechanism is adopted, in which the sliding window slides to the side where the new dialogue is generated, splices the text of the (n+1)th round of dialogue related to the communication service and the text of the previous n rounds of dialogue, discards the dialogue content that exceeds the length of the window, and obtains the new question text as the target text, where n is a positive integer; The target text is input into the large language model to obtain the first information output by the large language model; The step of obtaining the second information related to the communication service includes: The first information is input into the large language model to obtain the return results of the execution actions of each executable task unit and related analysis information output by the large language model. The interactive prompt information generated by the large language model based on the returned results and the relevant analysis information is obtained. The interactive prompt information is used to guide the target object to perform the target operation and return the result of the target operation. The target operation includes on-site investigation and / or operation of the target device. The business data extracted by the large language model based on the result of the target operation is used as the admission condition for each executable task unit.

2. The method according to claim 1, characterized in that, The step of inputting the analyzed data into a large language model to obtain the communication service-related processing strategy output by the large language model includes: The analysis data is input into a large language model to obtain a first strategy corresponding to each executable task unit output by the large language model. The first strategy includes first text information that conforms to the communication service logic. The processing strategy includes the first strategy.

3. The method according to claim 2, characterized in that, After inputting the analysis data into the large language model to obtain the first strategy corresponding to each executable task unit output by the large language model, the method further includes: A second strategy is generated based on the first strategy. The second strategy includes one or more of the following: second text information, image information, and document information that conform to the communication service logic. The second text information is obtained by expanding the first text information.

4. The method according to claim 2, characterized in that, Before inputting the analysis data into the large language model to obtain the first strategy corresponding to each executable task unit output by the large language model, the method further includes: The autoregressive language model is trained and then fine-tuned based on the content objective composed of each sample's executable task unit and sample policy to obtain the large language model.

5. A communication service processing apparatus, characterized in that, include: An acquisition unit is used to acquire first information related to communication services, the first information including multi-dimensional information obtained by performing intent recognition on the dialogue related to the communication services. The acquisition unit is used to acquire second information related to the communication service. The second information includes the execution action of each executable task unit obtained by dividing the first information, and the admission condition of each executable task unit. The execution of the executable task unit corresponding to the admission condition will be triggered only when a certain admission condition is met. An acquisition unit is used to acquire third information, which includes analysis data generated by analyzing the results of the execution actions of each executable task unit under the admission conditions. The processing unit is used to input the analysis data into the large language model to obtain the communication service-related processing strategy output by the large language model. The acquisition unit is configured to use a sliding window mechanism to slide the window toward the side where the new dialogue is generated, splice the text of the (n+1)th round of dialogue related to the communication service with the text of the previous n rounds of dialogue, discard dialogue content that exceeds the length of the window, and obtain a new question text as the target text, where n is a positive integer. The processing unit is used to input the target text into the large language model and obtain the first information output by the large language model; The processing unit is used to input the first information into the large language model to obtain the return result of the execution action of each executable task unit output by the large language model and related analysis information. The acquisition unit is used to acquire interactive prompt information generated by the large language model based on the returned results and the relevant analysis information. The interactive prompt information is used to guide the target object to perform the target operation and return the result of the target operation. The target operation includes on-site investigation and / or operation of the target device. The acquisition unit is used to acquire business data extracted by the large language model based on the result of the target operation, as the admission condition for each executable task unit.

6. An electronic device, characterized in that, include: A processor and memory for storing computer programs that can run on the processor; wherein, The processor, when running the computer program, performs the steps of the method according to any one of claims 1 to 4.

7. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

8. A computer product comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images