A work order processing method based on large language model and related device

Abstract

The present application discloses a work order processing method based on a large language model and related devices, which relate to the technical field of large language models, including: after obtaining the work order to be processed, first based on the work order to be processed, calling the user intent generation model to generate the user intent corresponding to the work order to be processed, the user intent generation model is a large language model obtained through training and capable of generating user intent based on the work order; then based on the user intent corresponding to the work order, searching the pre-built work order knowledge base to obtain the work order knowledge retrieval result; finally, according to the work order knowledge retrieval result, obtaining the corresponding reply to the work order. This solution utilizes a large language model, can realize intelligent processing of work orders, and thus can realize more efficient and higher quality work order responses.

Description

Technical Field

[0001] The present application relates to the technical field of large language models, and in particular to a work order processing method and related devices based on a large language model. Background Art

[0002] A work order, also known as a work receipt, is a detailed record document that records the details of one or more work tasks. For example, "User 0102065XXXX, Guo Moumou, number 021 8867XXXX, Family Benefit Package - Home WiFi Service 5 Platinum WiFi 6. This promotion has expired at the end of this month, but the system has automatically extended the bonus. Please deposit 0 yuan; the total bonus is 360 yuan; the 360 ​​yuan will be refunded in 12 installments; this bonus is no longer valid."

[0003] The traditional work order processing method mainly relies on manual processing to generate work order response information. As the number of work orders increases and the categories increase, the traditional manual processing method can no longer achieve more efficient and higher-quality work order responses. Therefore, it is particularly necessary to realize intelligent processing of work orders and thus achieve more efficient and higher-quality work order responses. Summary of the Invention

[0004] In view of the above problems, this application provides a work order processing method and related devices based on a large language model to achieve intelligent processing of work orders, thereby achieving the goal of more efficient and higher-quality work order responses. The specific solution is as follows:

[0005] In a first aspect, the present application provides a work order processing method based on a large language model, comprising:

[0006] Get pending work orders;

[0007] Based on the pending work order, a user intent generation model is called to generate a user intent corresponding to the pending work order, wherein the user intent generation model is a large language model obtained through training and capable of generating user intent based on the work order;

[0008] Based on the user intent corresponding to the work order, a pre-built work order knowledge base is retrieved to obtain a work order knowledge retrieval result;

[0009] According to the work order knowledge retrieval result, a reply corresponding to the work order is obtained.

[0010] In a possible implementation, obtaining a pending work order includes:

[0011] Get the original work order;

[0012] Preprocessing the original work order to obtain a preprocessed work order;

[0013] The pre-processed work order is used as the work order to be processed.

[0014] In a possible implementation, calling a user intent generation model based on the pending work order to generate a user intent corresponding to the pending work order includes:

[0015] Obtain a preconfigured user intent generation prompt format template, wherein the user intent generation prompt format template includes user intent generation requirement information and a work order filling slot;

[0016] Filling the to-be-processed work order into the work order filling slot in the user intention generation prompt format template to obtain the user intention generation prompt;

[0017] The user intention generation prompt is input into the user intention generation model to obtain the user intention corresponding to the to-be-processed work order generated by the user intention generation model.

[0018] In a possible implementation, obtaining a reply corresponding to the work order based on the work order knowledge retrieval result includes:

[0019] Deduplication and secondary sorting are performed on the work order knowledge retrieval results to determine whether there is target work order knowledge, where the target work order knowledge is work order knowledge in the work order knowledge retrieval results that matches the work order;

[0020] If the target work order knowledge exists, a reply corresponding to the work order is obtained based on the target work order knowledge.

[0021] In a possible implementation, if the target work order knowledge does not exist, obtaining a reply corresponding to the work order based on the work order knowledge retrieval result further includes:

[0022] Based on the pending work order, calling a work order structured information extraction model to generate structured information corresponding to the pending work order, wherein the work order structured information extraction model is a large language model obtained through training and capable of extracting work order structured information;

[0023] Based on the pending work order, calling a business category generation model to generate a business category corresponding to the pending work order, wherein the business category generation model is a large language model obtained through training and capable of generating business categories based on work orders;

[0024] The structured information corresponding to the work order to be processed is provided to the business personnel corresponding to the business category corresponding to the work order to be processed, so as to obtain a reply corresponding to the work order.

[0025] In a possible implementation, calling a business category generation model based on the pending work order to generate a business category corresponding to the pending work order includes:

[0026] Obtaining a preconfigured business category generation prompt format template, wherein the business category generation prompt format template includes business category generation requirement information, a work order filling slot, and a business category filling slot;

[0027] Filling the to-be-processed work order into the work order filling slot in the business category generation prompt format template, and filling the preset business category into the business category filling slot to obtain a business category generation prompt;

[0028] The business category generation prompt is input into the business category generation model to obtain the business category corresponding to the to-be-processed work order generated by the business category generation model.

[0029] In a possible implementation, calling a work order structured information extraction model based on the work order to be processed to generate structured information corresponding to the work order to be processed includes:

[0030] Obtaining a preconfigured work order structured information extraction prompt format template, wherein the work order structured information extraction prompt format template includes work order structured information extraction requirement information and work order filling slots;

[0031] Filling the to-be-processed work order into the work order filling slot in the work order structured information extraction prompt format template to obtain a work order structured information extraction prompt;

[0032] The work order structured information extraction prompt is input into the work order structured information extraction model to obtain structured information corresponding to the work order to be processed generated by the work order structured information extraction model.

[0033] A second aspect of the present application provides a work order processing device based on a large language model, comprising:

[0034] A work order acquisition unit is used to obtain work orders to be processed;

[0035] A user intent determination unit is configured to call a user intent generation model based on the pending work order to generate a user intent corresponding to the pending work order, wherein the user intent generation model is a large language model obtained through training and capable of generating user intent based on the work order;

[0036] A retrieval unit, configured to retrieve a pre-built work order knowledge base based on the user intent corresponding to the work order, and obtain a work order knowledge retrieval result;

[0037] The processing unit is used to obtain a reply corresponding to the work order based on the work order knowledge retrieval result.

[0038] In a possible implementation, the work order obtaining unit is specifically configured to:

[0039] Get the original work order;

[0040] Preprocessing the original work order to obtain a preprocessed work order;

[0041] The pre-processed work order is used as the work order to be processed.

[0042] In a possible implementation, the user intention determination unit is specifically configured to:

[0043] Obtain a preconfigured user intent generation prompt format template, wherein the user intent generation prompt format template includes user intent generation requirement information and a work order filling slot;

[0044] Filling the to-be-processed work order into the work order filling slot in the user intention generation prompt format template to obtain the user intention generation prompt;

[0045] The user intention generation prompt is input into the user intention generation model to obtain the user intention corresponding to the to-be-processed work order generated by the user intention generation model.

[0046] In a possible implementation, the processing unit includes:

[0047] a target work order knowledge determination unit, configured to perform deduplication and secondary sorting on the work order knowledge retrieval results to determine whether there is target work order knowledge, wherein the target work order knowledge is work order knowledge in the work order knowledge retrieval results that matches the work order;

[0048] The first processing unit is configured to obtain a reply corresponding to the work order based on the target work order knowledge if the target work order knowledge exists.

[0049] In a possible implementation, if the target work order knowledge does not exist, the processing unit further includes a second processing unit;

[0050] The second processing unit includes:

[0051] A work order structured information extraction unit is configured to call a work order structured information extraction model based on the work order to be processed to generate structured information corresponding to the work order to be processed, wherein the work order structured information extraction model is a large language model obtained through training and capable of extracting work order structured information;

[0052] A business category determination unit is configured to call a business category generation model based on the pending work order to generate a business category corresponding to the pending work order, wherein the business category generation model is a large language model obtained through training and capable of generating business categories based on work orders;

[0053] The manual processing unit is used to provide the structured information corresponding to the work order to be processed to the business personnel corresponding to the business category corresponding to the work order to be processed, so as to obtain a reply corresponding to the work order.

[0054] In a possible implementation, the service category determination unit is specifically configured to:

[0055] Obtaining a preconfigured business category generation prompt format template, wherein the business category generation prompt format template includes business category generation requirement information, a work order filling slot, and a business category filling slot;

[0056] Filling the to-be-processed work order into the work order filling slot in the business category generation prompt format template, and filling the preset business category into the business category filling slot to obtain a business category generation prompt;

[0057] The business category generation prompt is input into the business category generation model to obtain the business category corresponding to the to-be-processed work order generated by the business category generation model.

[0058] In a possible implementation, the work order structured information extraction unit includes:

[0059] Obtaining a preconfigured work order structured information extraction prompt format template, wherein the work order structured information extraction prompt format template includes work order structured information extraction requirement information and work order filling slots;

[0060] Filling the to-be-processed work order into the work order filling slot in the work order structured information extraction prompt format template to obtain a work order structured information extraction prompt;

[0061] The work order structured information extraction prompt is input into the work order structured information extraction model to obtain structured information corresponding to the work order to be processed generated by the work order structured information extraction model.

[0062] The third aspect of the present application provides a computer program product, including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements the work order processing method based on a large language model according to the first aspect or any implementation of the first aspect.

[0063] A fourth aspect of the present application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0064] The memory is used to store computer programs;

[0065] The processor is used to execute the computer program so that the electronic device can implement the work order processing method of the above-mentioned first aspect or any implementation method of the first aspect.

[0066] In a fifth aspect, the present application provides a computer storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the work order processing method based on a large language model according to the above-mentioned first aspect or any implementation of the first aspect.

[0067] By leveraging the aforementioned technical solution, the large language model-based work order processing method and related apparatus provided in this application, after obtaining a pending work order, first calls a user intent generation model based on the pending work order to generate the user intent corresponding to the pending work order. The user intent generation model is a large language model trained to generate user intent based on work orders. Then, based on the user intent corresponding to the work order, a pre-built work order knowledge base is retrieved to obtain work order knowledge retrieval results. Finally, based on the work order knowledge retrieval results, a reply corresponding to the work order is obtained. This solution utilizes a large language model to achieve intelligent processing of work orders, thereby enabling more efficient and higher-quality work order responses. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] The above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that the originals and elements are not necessarily drawn to scale.

[0069] Figure 1 A flowchart of a work order processing method based on a large language model provided in this application;

[0070] Figure 2 A schematic diagram of the structure of a work order processing device based on a large language model provided by this application;

[0071] Figure 3 This is a schematic diagram of the structure of an electronic device provided in this application. DETAILED DESCRIPTION

[0072] The following describes the embodiments of the present application in conjunction with the accompanying drawings. The terms used in the implementation methods of the present application are only used to explain the specific embodiments of the present application and are not intended to limit the present application.

[0073] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0074] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, and this is merely a way of distinguishing the objects of the same attributes when describing them in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0075] A work order, also known as a work receipt, is a detailed record document that records the details of one or more work tasks. For example, "User 0102065XXXX, Guo Moumou, number 021 8867XXXX, Family Benefit Package - Home WiFi Service 5 Platinum WiFi 6. This promotion has expired at the end of this month, but the system has automatically extended the bonus. Please deposit 0 yuan; the total bonus is 360 yuan; the 360 ​​yuan will be refunded in 12 installments; this bonus is no longer valid."

[0076] The traditional work order processing method mainly relies on manual processing to generate work order response information. As the number of work orders increases and the categories increase, the traditional manual processing method can no longer achieve more efficient and higher-quality work order responses. Therefore, it is particularly necessary to realize intelligent processing of work orders and thus achieve more efficient and higher-quality work order responses.

[0077] In order to achieve intelligent processing of work orders, the inventors of this case initially thought of the following implementation methods:

[0078] Method 1: Work order processing driven by rule engine

[0079] This involves defining matching rules and matching responses to work orders. This solution is simple to use and easy to implement, but it relies heavily on manually formulated matching rules. Maintaining these rules is labor-intensive and costly, and the level of automation is limited, making it difficult to adapt to complex and changing scenarios.

[0080] Method 2: Work order processing based on NLP (Natural Language Processing)

[0081] This involves using machine learning methods to train models using large amounts of historical work order processing data, enabling them to understand work orders and infer solutions. However, this approach is highly data-dependent, has limited knowledge representation and reasoning capabilities, and is limited in its adaptability to diverse work orders.

[0082] Method 3: LLM (Large Language Model)-based work order processing

[0083] 1. Ticket processing, which integrates information retrieval and text generation technologies, dynamically retrieves relevant information from a large number of external knowledge bases and generates accurate and immediate answers based on context. This cutting-edge technology demonstrates remarkable potential in text generation, revolutionizing ticket processing scenarios. However, it also exposes issues such as incomplete coverage of specialized domain knowledge, misunderstanding of intent, and potentially misleading answers. Furthermore, it is difficult to accurately match solutions between highly similar ticket descriptions, resulting in inefficient and ineffective ticket processing.

[0084] In order to solve the above problems, an embodiment of the present application provides a work order processing method. This method can use a large language model to deeply understand the content of the work order, accurately grasp user needs, improve the accuracy of intent recognition, and automatically classify the work order in detail, speed up the problem diversion, and provide an efficient path for subsequent processing. It can also combine RAG (Retrieval-augmented Generation) capabilities, not only to provide answers based on the existing knowledge base, but also to creatively generate responses to ensure the accuracy and innovation of the answers. In addition, it can also extract key information from unstructured work order descriptions, promote the automated creation and processing of work orders, reduce manual intervention, and accelerate process operations. In short, this method can perform intelligent processing of work orders, thereby achieving more efficient and higher-quality work order responses.

[0085] Before introducing this application plan, let me first explain the English terms involved in this article:

[0086] Prompt: Instructions. When communicating with AI (such as a large language model), you need to send instructions to the AI. This can be a text description, such as "Please recommend me a pop song" when you are talking to the AI, or it can be a parameter description in a specific format, such as describing the relevant drawing parameters to ask the AI ​​to draw a picture in a certain format.

[0087] Large language model (LLM) refers to a generative deep neural network model based on the Transformer structure. It has superb language comprehension, generation, logical reasoning, and multi-round dialogue capabilities. Common large language models include the ChatGPT model, the PaLM (Pathways Language Model) model, the Pangu model, the Wenxin Yiyan model, and the Xinghuo Cognitive model.

[0088] LoRA: (Low-Rank Adaptation), a lightweight fine-tuning technology for large language models. LoRA can fine-tune a model with billions of parameters with only thousands of samples, providing an efficient and parameter-economical solution for fine-tuning large language models. It is one of the effective means to deal with the current problems of high cost and resource constraints of fine-tuning large models. LoRA proposes to add two low-rank matrices AB to the weight matrix of each layer without modifying the original model parameters. For the pre-trained weight matrix , we can use a low-rank decomposition to express the parameter update ,Right now:

[0089]

[0090] The advantages of LoRA are:

[0091] Resource-friendly: The GPU memory and computation time required for fine-tuning are significantly reduced.

[0092] Flexibility: Different rank sizes can be selected at different layers to balance performance and resource consumption.

[0093] Simplicity: It is relatively easy to integrate into existing training processes without requiring fundamental changes to the model architecture.

[0094] The following is a detailed introduction to the work order processing method of the embodiment of the present application with reference to the accompanying drawings.

[0095] Reference Figure 1 , Figure 1 A flowchart of a work order processing method provided in an embodiment of the present application is shown as follows: Figure 1 As shown, a work order processing method based on a large language model provided in an embodiment of the present application may include steps S101 to S104, and these steps are described in detail below.

[0096] S101: Get the work order to be processed;

[0097] In this application, the work order to be processed can be a work order in text format. The work order to be processed can be an original work order, or a preprocessed work order obtained after preprocessing the original work order. The original work order and the preprocessed work order can also be work orders in text format.

[0098] In one possible implementation, pre-processing the original work order refers to filtering PII (Personally Identifiable Information) information (such as user name, email address, phone number, number, and other personalized information) in the original work order.

[0099] As an implementation option, pre-set filtering algorithms can be used to pre-process the original work order. For example, the Presidio + SpaCy library can be used to implement rule-based and named entity recognition (NER) PII detection and masking. SpaCy is a natural language processing library that provides a text named entity recognition (NER) model that supports various named entities, such as PERSON, LOCATION, and PHONE_NUMBER. Presidio also supports custom rule-based PII anonymizers for filtering business-specific PII information.

[0100] For ease of understanding, assume that the original work order example is as follows:

[0101] "User 0102065XXXX, Guo Moumou, number 021 8867XXXX, Family Benefit Package - Home WiFi Service 5 Platinum WiFi 6. This promotion has expired at the end of this month, but the system has automatically extended the bonus. Please deposit 0 yuan; the total bonus is 360 yuan; 360 yuan will be returned in 12 installments; this bonus is no longer valid."

[0102] The pre-processed work order example 1 is as follows:

[0103] “ <number>user <person>,Number<PHONE_NUMBER> The Family Benefit Package, Home WiFi Service 5 Platinum WiFi 6, has expired, but the system automatically extends the bonus. Please deposit 0 yuan; the total bonus is 360 yuan; 360 yuan will be returned in 12 months; this bonus is invalid.

[0104] It should be noted that in Example 1, the PII information in the original work order is obtained by anonymizing it into named entities.

[0105] Example 2 of the pre-processed work order is as follows:

[0106] "User, the Family WiFi Service 5 Platinum WiFi 6 benefit package has expired, but the system automatically extends the bonus. Please deposit 0 yuan; the total bonus is 360 yuan; 360 yuan will be returned in 12 installments; this bonus is no longer valid."

[0107] It should be noted that in Example 2, the PII information in the original work order is anonymized to a null value.

[0108] It should be further explained that, in this application, filtering the PII information in the original work order can not only effectively protect the personal identity information in the work order data, but also improve the accuracy of the next step of user intention identification.

[0109] S102: Based on the pending work order, calling a user intent generation model to generate a user intent corresponding to the pending work order, wherein the user intent generation model is a large language model obtained through training and capable of generating user intent based on the work order;

[0110] In this application, the large language model can be fine-tuned and trained in advance using work order samples and the user intent labels corresponding to the work order samples to obtain a large language model that has the ability to generate user intent based on the work order as a user intent generation model.

[0111] In one feasible approach, LoRA (Low-Rank Adaptation, a lightweight fine-tuning technology for large language models) can be used to fine-tune the large language model.

[0112] S103: Based on the user intent corresponding to the work order, a pre-built work order knowledge base is searched to obtain a work order knowledge search result;

[0113] In this application, in one possible implementation, the work order knowledge base includes multiple original FAQ (Frequently Asked Questions) knowledge, each of which includes questions and answers. For ease of understanding, examples of original FAQ knowledge are as follows:

[0114] question,answer

[0115] "Business processing reminder: User information does not exist or has expired", "This situation is caused by the user information not being synchronized with billing. A full CTOB data synchronization will be performed on the member reporting the error."

[0116] In order to ensure that the correct FAQ knowledge can be matched as much as possible when different expressions are used in the work order to represent the same user intention, in the present application, in another possible implementation, for each original FAQ knowledge, the question in the original FAQ knowledge can be summarized and extracted to obtain a summary of the question; the summary of the question and the answer in the original FAQ knowledge are combined to generate processed FAQ knowledge, and each processed FAQ knowledge is combined into a work order knowledge base.

[0117] It should be noted that extracting summaries from questions in the original FAQ knowledge and obtaining summaries of the questions can be achieved using a summary extraction model. The summary extraction model is a large language model with summary extraction capabilities obtained through training. Considering the working principle of the large language model, when using the summary extraction model to extract summaries from questions in the original FAQ knowledge and obtaining summaries of the questions, it is necessary to generate a summary extraction prompt based on the questions in the original FAQ knowledge. The summary extraction prompt is input into the summary extraction model, and the summary extraction model can then output a summary of the question. For ease of understanding, an example of a summary extraction prompt is as follows:

[0118] You are an intelligent information extraction assistant that follows the following principles:

[0119] 1. Extract specific information strictly based on "context" and do not allow adding fabricated elements to the answer

[0120] context:

[0121] {input}.

[0122] Where "{input}" is the slot to be filled with the question.

[0123] In one possible implementation, a combined retrieval strategy of sparse retrieval (BM25 algorithm) and vector retrieval (CosSimilarity algorithm) can be used to recall FAQ knowledge (original FAQ knowledge or processed FAQ knowledge) as the work order knowledge retrieval result.

[0124] S104: Obtain a reply corresponding to the work order based on the work order knowledge retrieval result.

[0125] In this application, in one achievable method, the work order knowledge retrieval results can be deduplicated and re-ranked to determine whether target work order knowledge exists, where the target work order knowledge is work order knowledge in the work order knowledge retrieval results that matches the work order. The redundant model and the rerank model can be used to dedupe and re-rank the work order knowledge retrieval results.

[0126] If the target work order knowledge exists, a response corresponding to the work order is obtained based on the target work order knowledge. Specifically, the answer in the target work order knowledge can be returned as the response corresponding to the work order. Alternatively, the target work order knowledge is input into a large language model, which then generates a response corresponding to the work order.

[0127] If the target work order knowledge does not exist, manual processing can be prompted.

[0128] The work order processing method provided in this embodiment, after obtaining a pending work order, first calls a user intent generation model based on the pending work order to generate the user intent corresponding to the pending work order. The user intent generation model is a large language model trained to generate user intent based on the work order. Then, based on the user intent corresponding to the work order, the pre-built work order knowledge base is searched to obtain work order knowledge retrieval results. Finally, based on the work order knowledge retrieval results, a response corresponding to the work order is obtained. This solution utilizes a large language model to achieve intelligent processing of work orders, thereby achieving more efficient and higher-quality work order responses.

[0129] Taking into account the working principle of the large language model, when calling the user intent generation model based on the pending work order to generate the user intent corresponding to the pending work order, it is necessary to build a user intent generation prompt based on the pending work order, input the user intent generation prompt into the user intent generation model, and the user intent generation model can generate and output the user intent corresponding to the pending work order. As an implementable method, in this application, a user intent generation prompt format template can be preconfigured, and the user intent generation prompt format template includes user intent generation requirement information and work order filling slots. Among them, the user intent generation requirement information is used to describe the task and the expected output format. For ease of understanding, an example of a user intent generation prompt format template is as follows:

[0130] You are a virtual ticket assistant. Your task is to help ticket handlers extract user intent from ticket text descriptions.

[0131] # Please follow the following rules when extracting

[0132] 1. Try to keep the original text and express it in one sentence;

[0133] 2. Do not include personal information such as address, phone number, ID number, account number, order number, or serial number in your answer;

[0134] 3. Do not include modal particles or emotional words in your answers;

[0135] 4. Do not include how to solve the current problem in your answer;

[0136] 5. Do not include the company name in your answer;

[0137] 6. Do not include the name of your mobile service plan in your answer;

[0138] 7. Do not include the payment amount or broadband bandwidth in your answer.

[0139] # Work Order

[0140] {input}

[0141] Extract user intent from the above content, focusing on user requests and system prompts. Be sure to strictly adhere to the above extraction rules and limit the output to no more than 40 characters.

[0142] Among them, {input} corresponds to the work order filling slot, and the rest is the user intent generation requirement information.

[0143] In another embodiment of the present application, a specific implementation method of calling a user intent generation model based on the pending work order to generate the user intent corresponding to the pending work order is described. The method may include the following steps:

[0144] S201: Obtain a pre-configured user intent generation prompt format template, wherein the user intent generation prompt format template includes user intent generation requirement information and a work order filling slot;

[0145] S202: Filling the to-be-processed work order into the work order filling slot in the user intention generation prompt format template to obtain a user intention generation prompt;

[0146] For ease of understanding, let's assume the work order to be processed is "User 130XXXX3074 applied for China Unicom's mobile phone portability service at Wuyi Wuyitang Guanpeng Mobile Store Store No. 7 on August 18, 2023, and participated in the 2023 New Network 5G Gold Coin Contract_128 Yuan activity. This terminal activity was processed normally, but there is no gold coin exchange record (screenshot attached). Please help verify the reason. Thank you! Contact: Chen Moumou 138XXXX0402", then the user intent to generate a prompt can be as follows:

[0147] You are a virtual ticket assistant. Your task is to help ticket handlers extract user intent from ticket text descriptions.

[0148] # Please follow the following rules when extracting

[0149] 1. Try to keep the original text and express it in one sentence;

[0150] 2. Do not include personal information such as address, phone number, ID number, account number, order number, or serial number in your answer;

[0151] 3. Do not include modal particles or emotional words in your answers;

[0152] 4. Do not include how to solve the current problem in your answer;

[0153] 5. Do not include the company name in your answer;

[0154] 6. Do not include the name of your mobile service plan in your answer;

[0155] 7. Do not include the payment amount or broadband bandwidth in your answer.

[0156] # Work Order

[0157] User 130XXXX3074 applied for a China Unicom mobile phone transfer service at the Wuyitang Guanpeng Mobile Store (store number 7) on August 18, 2023. This service was also offered as part of the 2023 5G Gold Coin Contract for Newcomers (¥128). This terminal activity was handled normally, but there is no record of the gold coin exchange (screenshot attached). Please help investigate the cause. Thank you! Contact: Chen 138XXXX0402

[0158] Extract user intent from the above content, focusing on user requests and system prompts. Be sure to strictly adhere to the above extraction rules and limit the output to no more than 40 characters.

[0159] S203: Input the user intention generation prompt into the user intention generation model to obtain the user intention corresponding to the to-be-processed work order generated by the user intention generation model.

[0160] For example, the user intent corresponding to the above-mentioned work order to be processed generated by the user intent generation model may be "no gold coin redemption record after the China Unicom transfer service is processed."

[0161] In the above embodiment, it is pointed out that obtaining a reply corresponding to the work order based on the work order knowledge retrieval results includes: deduplicating and secondary sorting the work order knowledge retrieval results to determine whether there is target work order knowledge, where the target work order knowledge is the work order knowledge that matches the work order in the work order knowledge retrieval results; if the target work order knowledge exists, obtaining a reply corresponding to the work order based on the target work order knowledge. If the target work order knowledge does not exist, obtaining a reply corresponding to the work order based on the work order knowledge retrieval results also includes:

[0162] S301: Based on the pending work order, call a work order structured information extraction model to generate structured information corresponding to the pending work order, wherein the work order structured information extraction model is a large language model obtained through training and capable of extracting work order structured information;

[0163] In this application, the large language model can be fine-tuned and trained in advance using work order samples and the structured information labels corresponding to the work order samples to obtain a large language model with the ability to extract work order structured information as a business category generation model.

[0164] In one feasible approach, LoRA (Low-Rank Adaptation, a lightweight fine-tuning technology for large language models) can be used to fine-tune the large language model.

[0165] S302: Based on the pending work order, calling a business category generation model to generate a business category corresponding to the pending work order, wherein the business category generation model is a large language model obtained through training and capable of generating business categories based on work orders;

[0166] In this application, the large language model can be fine-tuned and trained in advance using work order samples and the business category labels corresponding to the work order samples to obtain a large language model that has the ability to generate business categories based on work orders as a business category generation model.

[0167] In one feasible approach, LoRA (Low-Rank Adaptation, a lightweight fine-tuning technology for large language models) can be used to fine-tune the large language model.

[0168] S303: providing the structured information corresponding to the pending work order to a business person corresponding to the business category corresponding to the pending work order for processing, so as to obtain a reply corresponding to the work order.

[0169] Taking into account the working principle of the large language model, when calling the business category generation model based on the pending work order to generate the business category corresponding to the pending work order, it is necessary to build a business category generation prompt based on the pending work order, input the business category generation prompt into the business category generation model, and the business category generation model can generate and output the business category corresponding to the pending work order. As an implementable method, in the present application, a business category generation prompt format template can be pre-configured, and the business category generation prompt format template includes business category generation requirement information, work order filling slot and business category filling slot. Among them, the business category generation requirement information is used to describe the task and the expected output format. For ease of understanding, an example of a business category generation prompt format template is as follows:

[0170] You are an intelligent assistant tasked with accurately classifying a work order into a specific business category based on the work order text description provided by the user. Please select the best matching business category from the available options as your answer.

[0171] #Work Order:

[0172] {input}

[0173] #Alternative business categories:

[0174] {classifications}.

[0175] Among them, {input} corresponds to the work order filling slot, {classifications} corresponds to the business category filling slot, and the rest is the business category generation requirement information. In another embodiment of the present application, the specific implementation method of calling the business category generation model based on the pending work order to generate the business category corresponding to the pending work order is described. This method may include the following steps:

[0176] S401: Obtain a pre-configured service category generation prompt format template, wherein the service category generation prompt format template includes service category generation requirement information, a work order filling slot, and a service category filling slot;

[0177] S402: Filling the to-be-processed work order into the work order filling slot in the business category generation prompt format template, and filling the preset business category into the business category filling slot to obtain a business category generation prompt;

[0178] For ease of understanding, assume the pending work order is "User 130XXXX3074 applied for China Unicom number portability at Wuyi Wuyitang Guanpeng Mobile Store Store No. 7 on August 18, 2023, and participated in the 2023 New Network 5G Gold Coin Contract_128 Yuan activity. This terminal activity was processed normally, but there is no gold coin exchange record (screenshot attached). Please help verify the reason, thank you! Contact: Chen Moumou 138XXXX0402", and the preset service categories are "Number Portability, Points Exchange, Business Inquiry, Arrears Payment, Data Package". The service category generation prompt can be as follows:

[0179] You are an intelligent assistant tasked with accurately classifying a work order into a specific business category based on the work order text description provided by the user. Please select the best matching business category from the available options as your answer.

[0180] #Work Order:

[0181] User 130XXXX3074 applied for a China Unicom mobile phone transfer service at the Wuyitang Guanpeng Mobile Store (store number 7) on August 18, 2023. This service was also offered as part of the 2023 5G Gold Coin Contract for Newcomers (¥128). This terminal activity was handled normally, but there is no record of the gold coin exchange (screenshot attached). Please help investigate the cause. Thank you! Contact: Chen 138XXXX0402

[0182] #Alternative business categories:

[0183] {Number portability, points redemption, business enquiries, overdue payments, data packages}

[0184] S403: Input the business category generation prompt into the business category generation model to obtain the business category corresponding to the to-be-processed work order generated by the business category generation model.

[0185] Exemplarily, the business category corresponding to the above-mentioned work order to be processed generated by the business category generation model can be "number portability service".

[0186] Taking into account the working principle of the large language model, when calling the work order structured information extraction model based on the work order to be processed and generating the structured information corresponding to the work order to be processed, it is necessary to build a work order structured information extraction prompt based on the work order to be processed, and input the work order structured information extraction prompt into the work order structured information extraction model. The work order structured information extraction model can then generate and output the structured information corresponding to the work order to be processed. As an implementable method, in this application, a work order structured information extraction prompt format template can be pre-configured, and the work order structured information extraction prompt format template includes work order structured information extraction requirement information and work order filling slots. Among them, the work order structured information extraction requirement information is used to describe the task and the desired output format. For ease of understanding, an example of a work order structured information extraction prompt format template is as follows:

[0187] You are a virtual ticket assistant. Your task is to help ticket processing specialists extract ticket structured information from the ticket text description.

[0188] # Please follow the following rules when extracting

[0189] 1. Strictly extract specific information based on "known text" and do not allow adding fabricated elements to the answers

[0190] 2. Fill in blanks if there are no relevant fields

[0191] 3. Output in json format

[0192] # Work Order

[0193] {input}.

[0194] Among them, {input} corresponds to the work order filling slot, and the rest is the work order structured information extraction requirement information.

[0195] In another embodiment of the present application, a specific implementation method of calling a work order structured information extraction model based on the work order to be processed to generate structured information corresponding to the work order to be processed is described. The method may include the following steps:

[0196] S501: Obtain a pre-configured work order structured information extraction prompt format template, wherein the work order structured information extraction prompt format template includes work order structured information extraction requirement information and work order filling slots;

[0197] S502: Fill the to-be-processed work order into the work order filling slot in the work order structured information extraction prompt format template to obtain a work order structured information extraction prompt;

[0198] For ease of understanding, let's assume the work order to be processed is "User 130XXXX3074 applied for China Unicom's mobile phone portability service at Wuyi Wuyitang Guanpeng Mobile Store Store No. 7 on August 18, 2023, and participated in the 2023 New Network 5G Gold Coin Contract_128 Yuan activity. This terminal activity was processed normally, but there is no gold coin exchange record (screenshot attached). Please help verify the reason. Thank you! Contact: Chen Moumou 138XXXX0402", then the user intent to generate a prompt can be as follows:

[0199] You are a virtual ticket assistant. Your task is to help ticket processing specialists extract ticket structured information from the ticket text description.

[0200] # Please follow the following rules when extracting

[0201] 1. Strictly extract specific information based on "known text" and do not allow adding fabricated elements to the answers

[0202] 2. Fill in blanks if there are no relevant fields

[0203] 3. Output in json format

[0204] # Work Order

[0205] User 130XXXX3074 applied for a China Unicom mobile phone data portability service at the Guanpeng Mobile Store, Branch 7, Wuyi, on August 18, 2023. They also participated in the 2023 New 5G Gold Coin Contract (128 yuan) promotion. While the terminal activity was processed normally, there is no record of the gold coin exchange (screenshot attached). Please help investigate the cause. Thank you! Contact: Chen Moumou (138XXXX0402).

[0206] S503: Input the work order structured information extraction prompt into the work order structured information extraction model to obtain structured information corresponding to the work order to be processed generated by the work order structured information extraction model.

[0207] Exemplarily, the structured information corresponding to the above-mentioned work order to be processed generated by the work order structured information extraction model may be:

[0208] { "time": "August 18, 2023",

[0209] "User": "130XXXX3074",

[0210] "Phone": "138XXXX0402",

[0211] "Ticket Text": "User 130XXXX3074 applied for China Unicom's mobile phone data portability service at Wuyi Wuyitang Guanpeng Mobile Store Store No. 7 and participated in the 2023 New 5G Gold Coin Contract - 128 yuan promotion. The terminal activity was processed normally during the process, but there was no record of the gold coin exchange. A screenshot is attached. Please investigate the cause. Contact: Chen Moumou, Phone: 138XXXX0402."}

[0212] The above describes a work order processing method provided in an embodiment of the present application. The following describes an apparatus for executing the above work order processing method.

[0213] See also Figure 2 , Figure 2 This is a structural diagram of a work order processing device based on a large language model provided in an embodiment of the present application. Figure 2 As shown, the work order processing device includes:

[0214] A work order acquisition unit 11 is used to acquire work orders to be processed;

[0215] A user intent determination unit 12 is configured to call a user intent generation model based on the work order to be processed to generate a user intent corresponding to the work order to be processed, wherein the user intent generation model is a large language model obtained through training and capable of generating user intent based on the work order;

[0216] A retrieval unit 13 is configured to retrieve a pre-built work order knowledge base based on the user intent corresponding to the work order to obtain a work order knowledge retrieval result;

[0217] The processing unit 14 is configured to obtain a reply corresponding to the work order based on the work order knowledge retrieval result.

[0218] In a possible implementation, the work order obtaining unit is specifically configured to:

[0219] Get the original work order;

[0220] Preprocessing the original work order to obtain a preprocessed work order;

[0221] The pre-processed work order is used as the work order to be processed.

[0222] In a possible implementation, the user intention determination unit is specifically configured to:

[0223] Obtain a preconfigured user intent generation prompt format template, wherein the user intent generation prompt format template includes user intent generation requirement information and a work order filling slot;

[0224] Filling the to-be-processed work order into the work order filling slot in the user intention generation prompt format template to obtain the user intention generation prompt;

[0225] The user intention generation prompt is input into the user intention generation model to obtain the user intention corresponding to the to-be-processed work order generated by the user intention generation model.

[0226] In a possible implementation, the processing unit includes:

[0227] a target work order knowledge determination unit, configured to perform deduplication and secondary sorting on the work order knowledge retrieval results to determine whether there is target work order knowledge, wherein the target work order knowledge is work order knowledge in the work order knowledge retrieval results that matches the work order;

[0228] The first processing unit is configured to obtain a reply corresponding to the work order based on the target work order knowledge if the target work order knowledge exists.

[0229] In a possible implementation, if the target work order knowledge does not exist, the processing unit further includes a second processing unit;

[0230] The second processing unit includes:

[0231] A work order structured information extraction unit is configured to call a work order structured information extraction model based on the work order to be processed to generate structured information corresponding to the work order to be processed, wherein the work order structured information extraction model is a large language model obtained through training and capable of extracting work order structured information;

[0232] A business category determination unit is configured to call a business category generation model based on the pending work order to generate a business category corresponding to the pending work order, wherein the business category generation model is a large language model obtained through training and capable of generating business categories based on work orders;

[0233] The manual processing unit is used to provide the structured information corresponding to the work order to be processed to the business personnel corresponding to the business category corresponding to the work order to be processed, so as to obtain a reply corresponding to the work order.

[0234] In a possible implementation, the service category determination unit is specifically configured to:

[0235] Obtaining a preconfigured business category generation prompt format template, wherein the business category generation prompt format template includes business category generation requirement information, a work order filling slot, and a business category filling slot;

[0236] Filling the to-be-processed work order into the work order filling slot in the business category generation prompt format template, and filling the preset business category into the business category filling slot to obtain a business category generation prompt;

[0237] The business category generation prompt is input into the business category generation model to obtain the business category corresponding to the to-be-processed work order generated by the business category generation model.

[0238] In a possible implementation, the work order structured information extraction unit includes:

[0239] Obtaining a preconfigured work order structured information extraction prompt format template, wherein the work order structured information extraction prompt format template includes work order structured information extraction requirement information and work order filling slots;

[0240] Filling the to-be-processed work order into the work order filling slot in the work order structured information extraction prompt format template to obtain a work order structured information extraction prompt;

[0241] The work order structured information extraction prompt is input into the work order structured information extraction model to obtain structured information corresponding to the work order to be processed generated by the work order structured information extraction model.

[0242] An electronic device is also provided in an embodiment of the present application. Figure 3 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic device in the embodiments of the present application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 3 The electronic device shown is merely an example and should not limit the functions and scope of use of the embodiments of the present application.

[0243] like Figure 3 As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 602 or programs loaded from a storage device 608 into a random access memory (RAM) 603. When the electronic device is powered on, the RAM 603 also stores various programs and data required for the operation of the electronic device. The processing device 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0244] Typically, the following devices may be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a memory card, a hard disk, etc.; and a communication device 609. The communication device 609 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Figure 3 The electronic device is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead.

[0245] An embodiment of the present application also provides a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any one of the work order processing methods based on a large language model provided in the embodiment of the present application.

[0246] A computer-readable storage medium is also provided in an embodiment of the present application. The computer storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any one of the work order processing methods based on a large language model provided in the embodiment of the present application.

[0247] It should also be noted that the device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed across multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided in this application, the connection relationship between the modules indicates that there is a communication connection between them, which can be specifically implemented as one or more communication buses or signal lines.

[0248] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course can also be implemented by special hardware including application-specific integrated circuits, special CPUs, special memories, special components, etc. In general, all functions performed by computer programs can be easily implemented with corresponding hardware, and the specific hardware structures used to implement the same function can also be diverse, such as analog circuits, digital circuits or special circuits, etc. However, for the present application, software program implementation is a better implementation method in most cases. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer's floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk or optical disk, etc., and includes a number of instructions to enable a computer device (which can be a personal computer, training equipment, or network equipment, etc.) to execute the methods described in each embodiment of the present application.

[0249] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware, or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0250] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website, a computer, a training device or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website, a computer, a training device or a data center. The computer-readable storage medium can be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium can be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).< / person> < / number>

Claims

1. A work order processing method based on a large language model, characterized in that: include: Get pending work orders; Based on the work order to be processed, a user intent generation model is called to generate a user intent corresponding to the work order to be processed, wherein the user intent generation model is a large language model obtained through training and capable of generating user intent based on the work order; wherein the large language model is fine-tuned and trained in advance using work order samples and user intent labels corresponding to the work order samples, and a large language model capable of generating user intent based on the work order is obtained as the user intent generation model; Based on the user intent corresponding to the work order, a pre-built work order knowledge base is retrieved to obtain a work order knowledge retrieval result; Obtaining a reply corresponding to the work order according to the work order knowledge retrieval result; Wherein, if the target work order knowledge does not exist, the target work order knowledge is the work order knowledge matching the work order in the work order knowledge retrieval result, then obtaining a reply corresponding to the work order according to the work order knowledge retrieval result includes: Based on the work order to be processed, a work order structured information extraction model is called to generate structured information corresponding to the work order to be processed, wherein the work order structured information extraction model is a large language model obtained through training and capable of extracting work order structured information; wherein the large language model is fine-tuned and trained in advance using work order samples and structured information labels corresponding to the work order samples, and the large language model capable of extracting work order structured information is obtained as a business category generation model; Based on the work order to be processed, a business category generation model is called to generate a business category corresponding to the work order to be processed, wherein the business category generation model is a large language model obtained through training and has the ability to generate business categories based on work orders; wherein the large language model is fine-tuned and trained in advance using work order samples and business category labels corresponding to the work order samples, so as to obtain a large language model with the ability to generate business categories based on work orders as the business category generation model; the business category generation prompt input to the business category generation model includes: business category generation requirement information for describing the task and the expected output format, the work order to be processed and a preset business category, and the business category corresponding to the work order to be processed is one of the preset business categories; The structured information corresponding to the work order to be processed is provided to the business personnel corresponding to the business category corresponding to the work order to be processed, so as to obtain a reply corresponding to the work order.

2. The method according to claim 1, characterized in that: The obtaining of the work order to be processed includes: Get the original work order; Preprocessing the original work order to obtain a preprocessed work order; The preprocessed work order is used as the work order to be processed.

3. The method according to claim 1, characterized in that The calling of a user intent generation model based on the work order to be processed to generate a user intent corresponding to the work order to be processed includes: Obtain a preconfigured user intent generation prompt format template, wherein the user intent generation prompt format template includes user intent generation requirement information and a work order filling slot; Fill the to-be-processed work order into the work order filling slot in the user intention generation prompt format template to obtain the user intention generation prompt; The user intention generation prompt is input into the user intention generation model to obtain the user intention corresponding to the to-be-processed work order generated by the user intention generation model.

4. The method according to claim 1, characterized in that: Obtaining a reply corresponding to the work order according to the work order knowledge retrieval result includes: De-duplicate and re-sort the work order knowledge retrieval results to determine whether there is target work order knowledge, where the target work order knowledge is the work order knowledge in the work order knowledge retrieval results that matches the work order; If the target work order knowledge exists, a reply corresponding to the work order is obtained based on the target work order knowledge.

5. The method according to claim 1, characterized in that The calling of a business category generation model based on the work order to be processed to generate a business category corresponding to the work order to be processed includes: Obtain a preconfigured service category generation prompt format template, wherein the service category generation prompt format template includes service category generation requirement information, a work order filling slot, and a service category filling slot; Fill the to-be-processed work order into the work order filling slot in the business category generation prompt format template, and fill the preset business category into the business category filling slot to obtain a business category generation prompt; The business category generation prompt is input into the business category generation model to obtain the business category corresponding to the to-be-processed work order generated by the business category generation model.

6. The method according to claim 1, characterized in that Based on the work order to be processed, calling the work order structured information extraction model to generate structured information corresponding to the work order to be processed includes: Obtaining a preconfigured work order structured information extraction prompt format template, wherein the work order structured information extraction prompt format template includes work order structured information extraction requirement information and work order filling slots; Filling the to-be-processed work order into the work order filling slot in the work order structured information extraction prompt format template to obtain a work order structured information extraction prompt; The work order structured information extraction prompt is input into the work order structured information extraction model to obtain structured information corresponding to the work order to be processed generated by the work order structured information extraction model.

7. A work order processing device based on a large language model, characterized in that: include: A work order acquisition unit is used to acquire work orders to be processed; A user intent determination unit is used to call a user intent generation model based on the work order to be processed to generate a user intent corresponding to the work order to be processed, wherein the user intent generation model is a large language model obtained through training and capable of generating user intent based on the work order; wherein the large language model is fine-tuned and trained in advance using work order samples and user intent labels corresponding to the work order samples, and a large language model capable of generating user intent based on the work order is obtained as the user intent generation model; A retrieval unit, configured to retrieve a pre-built work order knowledge base based on the user intent corresponding to the work order, and obtain a work order knowledge retrieval result; A processing unit, configured to obtain a reply corresponding to the work order according to the work order knowledge retrieval result; Wherein, if the target work order knowledge does not exist, the processing unit includes a second processing unit; The second processing unit comprises: A work order structured information extraction unit is used to call a work order structured information extraction model based on the work order to be processed to generate structured information corresponding to the work order to be processed, wherein the work order structured information extraction model is a large language model with the ability to extract work order structured information obtained through training; wherein the large language model is fine-tuned and trained in advance using work order samples and structured information labels corresponding to the work order samples, and the large language model with the ability to extract work order structured information is obtained as a business category generation model; A business category determination unit is used to call a business category generation model based on the work order to be processed, and generate a business category corresponding to the work order to be processed, wherein the business category generation model is a large language model obtained through training and has the ability to generate business categories based on work orders; wherein the large language model is fine-tuned and trained in advance using work order samples and business category labels corresponding to the work order samples, and a large language model with the ability to generate business categories based on work orders is obtained as the business category generation model; the business category generation prompt input to the business category generation model includes: business category generation requirement information for describing a task and an expected output format, the work order to be processed, and a preset business category, and the business category corresponding to the work order to be processed is one of the preset business categories; The manual processing unit is used to provide the structured information corresponding to the work order to be processed to the business personnel corresponding to the business category corresponding to the work order to be processed, so as to obtain a reply corresponding to the work order.

8. A computer program product, characterized in that It includes computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements the work order processing method based on a large language model as described in any one of claims 1 to 6.

9. An electronic device, characterized in that: The method comprises at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program so that the electronic device can implement the work order processing method based on a large language model as described in any one of claims 1 to 6.

10. A computer storage medium, characterized in that: The computer storage medium carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can implement the work order processing method based on a large language model as described in any one of claims 1 to 6.

Images