Work order processing method and system based on AI large model

Abstract

The invention discloses a work order processing method and device based on an AI large model, and relates to the technical field of data processing. The method comprises the following steps: extracting user appeal information from a target complaint work order to obtain a target appeal type; extracting complaint commodity information from the target complaint work order, and generating a work order component set corresponding to the complaint commodity information; matching the work order component set with a plurality of component relation trees to determine a target component relation tree; obtaining an emotion category of the user from the target complaint work order, generating an appeal scene according to the emotion category and the target component relation tree, inputting the user appeal information and the complaint commodity information into the appeal scene to execute logic judgment, and obtaining an appeal result; determining at least one processing scheme according to the appeal result and the current actual state of the product corresponding to the target complaint work order; and sending the appeal result and the processing scheme to a customer service staff processing the target complaint work order. By implementing the technical scheme provided by the invention, the complaint work order processing efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, specifically to a work order processing method and system based on an AI large model. Background Technology

[0002] In the context of the rapid development of the internet and e-commerce, the strategic importance of customer service in enterprise operations is becoming increasingly prominent. When a buyer raises an inquiry or complaint, the request is converted into a work order and enters the processing flow. In handling work orders, customer service personnel need to integrate multi-source heterogeneous data, including basic order information, logistics tracking information, customer call records, real-time weather data, branch registration information, and claims-related information. Based on established business rules and accumulated historical experience, they then develop targeted solutions to address the problems.

[0003] In existing technologies, customer service work order processing mainly relies on standardized workflows. However, in actual applications, each work order has different requirements. When faced with special or complex work orders, standardized workflows often fail to provide effective processing guidance, causing customer service personnel to spend more time on problem analysis and solution development, which affects the efficiency of work order processing. Summary of the Invention

[0004] This application provides a work order processing method and system based on an AI large model. This method effectively solves the problem that standardized processes in customer service work order processing are difficult to handle when they become complex, and improves the efficiency of complaint work order processing.

[0005] Firstly, this application provides a work order processing method based on an AI large model. The method includes: extracting user request information from a target complaint work order; matching the user request information with various request types in a preset request table to obtain a target request type; extracting complaint product information from the target complaint work order to generate a work order component set corresponding to the complaint product information, wherein the work order component set includes multiple product attribute components in the complaint product information; obtaining multiple component relationship trees corresponding to the target request type, each component relationship tree including a standard component set composed of multiple standard components and the arrangement rules between the standard components; matching the work order component set with multiple component relationship trees, and determining the component relationship tree with the highest component matching degree as the target component relationship tree; obtaining the user's emotion category from the target complaint work order; generating a request scenario based on the emotion category and the target component relationship tree; inputting the user request information and complaint product information into the request scenario to perform logical judgment and obtain a request result; determining at least one processing solution based on the request result and the current actual status of the product corresponding to the target complaint work order; and sending the request result and processing solution to the customer service personnel handling the target complaint work order.

[0006] By employing the above technical solutions, user request information is extracted from target complaint work orders. This information is then matched with request types in a pre-defined request table to quickly identify the user's target request type. Utilizing the natural language processing capabilities of AI large-scale models, key information is accurately extracted from complex language expressions and semantically matched with the pre-defined request table to identify the user's target request type. This dynamic matching method based on actual content avoids the limitations of fixed classification of request types in traditional standard processes. Multiple product attribute components are extracted from the complained product information to construct a work order component set, enabling fine-grained analysis of specific product issues. Based on the user request type, multiple component relationship trees are obtained, and a matching algorithm selects the target component relationship tree with the highest component matching degree, dynamically selecting the most suitable logical structure for the current work order from the standard component set. User emotion categories are obtained from the complaint work orders and combined with the target component relationship tree to generate corresponding request scenarios. Logical judgments are executed within these scenarios, and a processing solution is dynamically generated based on the request result and the actual product status. Finally, the request result and processing solution are sent to customer service personnel, forming a complete closed loop from request classification, problem analysis, solution generation to execution feedback. This closed-loop processing approach ensures that customer service personnel can quickly understand and implement the optimal solution, thereby improving the efficiency and quality of complaint handling.

[0007] Optionally, the complaint product information is extracted from the target complaint work order to generate a work order component set corresponding to the complaint product information. The work order component set includes multiple product attribute components from the complaint product information. Specifically, this includes: extracting multiple complaint product information related to the complaint product from the target complaint work order, where the complaint product is the product targeted by the target complaint work order; classifying each complaint product information to obtain multiple product attributes; filtering out product type attributes, product claim attributes, and product problem attributes from the multiple product attributes; identifying the product type attributes, product claim attributes, and product problem attributes to obtain multiple text recognition results; determining the product attribute components based on the text recognition results, and summarizing the multiple product attribute components into the work order component set.

[0008] By adopting the above technical solution, multiple product information related to the complained product is extracted from the target complaint work order, ensuring the integrity and accuracy of the product data involved in the processing. The product information is classified into multiple product attributes, realizing multi-dimensional structured processing of product information. Key attributes related to the complaint are screened from multiple product attributes, and irrelevant information is eliminated, ensuring the focus of data analysis. Text recognition is performed on product type attributes, product claim attributes, and product problem attributes in sequence, resulting in multiple text recognition results. The identified product attributes are integrated into product attribute components and summarized into the work order component set, realizing modular management of product information. The construction of the work order component set allows complex complaint issues to be decomposed into multiple independent components, facilitating the automation, standardization, and efficiency of the processing flow.

[0009] Optionally, a request scenario is generated based on the emotion category and the target component relationship tree. Specifically, this includes: if the emotion category is an abnormal emotion category, obtaining the original component set from the target component relationship tree, retrieving the usage frequency of each original component in the original component set (the usage frequency being the number of times each original component appears in the standard component set); obtaining the historical reference component set from the historical processing schemes, and determining a set of original components, where the set of original components is the intersection of the original component set and the historical reference component set; when the usage frequency of the target component in the historical processing scheme is greater than the preset usage frequency, and the contribution level in the historical processing scheme is greater than the preset contribution level, the target component is confirmed as a core component, and the target component is any component in the set of original components; when the usage frequency is less than or equal to the preset usage frequency, and the contribution level is less than or equal to the preset contribution level, the target component is confirmed as an auxiliary component; and adjusting the connection relationship of each original component in the target component relationship tree based on the core component and the auxiliary component to obtain the request scenario.

[0010] By adopting the above technical solution, and targeting users' abnormal emotions, the original component set of the target component relationship tree is dynamically obtained and analyzed in conjunction with historical processing solutions. This helps to accurately identify the core issues of user demands. A reference component set is obtained from historical processing solutions and its intersection with the original component set is calculated. This allows full utilization of successful processing experience from historical data. Based on the usage frequency and contribution of the target components, the components are divided into core components and auxiliary components. The connection relationships of the original components in the target component relationship tree are adjusted based on the core components and auxiliary components, making the relationship tree more closely fit the current demand scenario. This dynamic demand scenario construction optimizes the user experience, especially in complex or abnormal emotional scenarios, and can significantly improve user satisfaction with the service.

[0011] Optionally, user request information and complaint product information are input into the request scenario to perform logical judgment. Specifically, this includes: parsing the target complaint order based on the user request information to obtain the complaint address, which is the trajectory address of the target order when the user filed the complaint; obtaining the product claim component, product type component, and product issue component from the complaint product information; inputting the product claim component, product type component, and product issue component into the request scenario; determining whether a single component conforms to the corresponding component logic rules in the request scenario, where a single component can be any one of the product claim component, product type component, and product issue component; if a single component conforms to the corresponding component logic rules in the request scenario, then the single component result is obtained; and combining the complaint address and multiple component results to obtain the request result.

[0012] By employing the aforementioned technical solution, user complaint information is analyzed, and the complaint address of the target order is extracted. This allows for rapid location of the order at the time of the complaint. The product claim component, product type component, and product issue component are extracted from the complained product information, breaking down complex product information into independently analyzable modules. Inputting each component into the complaint scenario and determining whether it conforms to the corresponding logical rules avoids errors caused by abnormal component data or non-compliance with rules. If a single component conforms to the logical rules, a single component result is obtained. Multiple component results are combined with the complaint address to form the final complaint result. Combining the complaint address with component results creates a complete complaint result, integrating geographical, product, and issue information involved in the user complaint. Optionally, after sending the request result and handling solution to the customer service personnel handling the target complaint ticket, the method further includes: obtaining the initial number of historical request types within a preset time period; determining the number of historical request types corresponding to the same historical period based on the preset time period, where the historical request type and the target request type are the same type; dividing the difference between the initial number of requests and the number of historical requests by the preset number of requests to obtain the request growth rate; obtaining the number of complaints sent by users for historical request types within the preset time period, and retrieving the average number of complaints for historical request types within the preset time period; dividing the difference between the number of complaints and the average number of complaints by the average number of complaints to obtain the complaint growth rate; if the request growth rate is greater than the preset request growth rate, and the complaint growth rate is greater than the preset complaint growth rate, then obtaining the target handling solution corresponding to the historical request type; retrieving user evaluation information for the target handling solution, modifying the target handling solution based on the evaluation information to obtain the final handling solution, and replacing the target handling solution corresponding to the historical request type with the final handling solution.

[0013] By adopting the above technical solution, the complaint growth rate is calculated, and the changing trend of target complaint types within a preset time period is monitored in real time. This allows for the rapid detection of abnormal growth in a certain type of complaint. The calculation of the complaint growth rate combines the number of user complaints for a specific complaint type with the historical average, enabling accurate identification of abnormal growth in user complaints. When both the complaint growth rate and the demand growth rate exceed preset thresholds, the target handling solutions corresponding to historical complaint types are retrieved and corrected based on user evaluation information. This allows for the acquisition of user evaluation information on the target handling solutions, correction of aspects where users are dissatisfied, and prevention of the recurrence of the same problems. This significantly improves user satisfaction and reduces the risk of escalating complaints and further user churn.

[0014] Optionally, the target handling plan is revised based on the evaluation information to obtain the final handling plan. Specifically, this includes: obtaining the user's first evaluation information on the target complaint ticket and the user's second evaluation information on historical complaint tickets. The historical ticket database for each user includes the target complaint ticket and historical complaint tickets, with the historical complaint ticket being the previous complaint ticket of the target complaint ticket; extracting satisfaction score, timeliness score, and resolution effectiveness score from the first and second evaluation information respectively; performing a weighted summation of the satisfaction score, timeliness score, and resolution effectiveness score to obtain the first score and the second score. The first score is the score corresponding to the target complaint ticket, and the second score is the score corresponding to the historical complaint ticket; if the first score is lower than the second score, calculating the difference between the first score and the average score, with the average score being the comprehensive score of all users for the target handling plan. When the difference is greater than the preset difference, the indicators corresponding to the first score and the average score are compared to obtain the abnormal indicators. Improvement steps are generated based on the abnormal indicators, and the target steps in the target processing plan are replaced with improvement steps to obtain the final processing plan.

[0015] By employing the aforementioned technical solution, comparing the first evaluation information of the target complaint work order with the second evaluation information of historical complaint work orders allows for dynamic monitoring of changes in user satisfaction. When the first score is lower than the second score, user dissatisfaction or regression with the current work order handling solution can be quickly identified, providing a timely feedback mechanism for problem improvement. By calculating the difference between the first score and the average score, and combining it with a preset difference threshold, irrelevant score fluctuations can be effectively filtered out, ensuring that only handling solutions significantly below the average level are marked as abnormal, thereby improving the accuracy of anomaly detection and avoiding misjudgments. By comparing each indicator corresponding to the first score and the average score one by one, the specific indicators causing the score decline can be accurately located. Improvement steps are generated based on the abnormal indicators, and by replacing the target steps in the target handling solution with the improvement steps, targeted optimization is achieved, ensuring that the handling solution can quickly adapt to changes in user needs, improving the effectiveness of the final solution and user acceptance.

[0016] Optionally, at least one processing plan is determined based on the complaint result and the current actual status of the product corresponding to the target complaint ticket. Specifically, this includes: determining at least one initial processing plan based on the complaint result and the current actual status of the product corresponding to the target complaint ticket; retrieving the user profile corresponding to the user, obtaining the consumption level and historical complaint list from the user profile, where the consumption level corresponds to the user's annual consumption amount; classifying the historical complaint list to obtain complaint transportation category and complaint product quality category; obtaining the first and second percentages of each of the complaint transportation category and complaint product quality category in the historical complaint list; if the first percentage is greater than the second percentage, and the consumption level is high consumption level, then triggering a supplementary logistics service plan for the high consumption level complaint transportation category, merging the supplementary logistics service plan with the initial processing plan to obtain the processing plan; if the first percentage is greater than the second percentage, and the consumption level is low consumption level, then triggering a points compensation plan for the low consumption level complaint transportation category, merging the points compensation plan with the initial processing plan to obtain the processing plan.

[0017] By adopting the above technical solutions, based on user profiles and historical complaint data, it is possible to dynamically identify users' consumption preferences and main complaint categories, providing differentiated compensation plans for users with different consumption levels. This achieves precise matching of personalized services and enhances user satisfaction. For high-consumption users, a supplementary logistics service plan is provided, demonstrating preferential treatment and attention to high-value users; for low-consumption users, a points compensation plan is offered, which reduces compensation costs while encouraging continued consumption and increasing user stickiness. This differentiated strategy effectively improves the experience for users with different consumption levels, enhancing user recognition and trust in the platform. By integrating the initial processing plan and supplementary plans, user requests can be comprehensively resolved in one go, significantly improving processing efficiency.

[0018] In a second aspect, this application provides a work order processing device based on an AI large model. The device includes an acquisition unit, a processing unit, and a sending unit. The acquisition unit extracts user request information from a target complaint work order, matches the user request information with various request types in a preset request table to obtain a target request type; extracts complaint product information from the target complaint work order, generates a work order component set corresponding to the complaint product information, wherein the work order component set includes multiple product attribute components in the complaint product information; acquires multiple component relationship trees corresponding to the target request type, each component relationship tree including a standard component set composed of multiple standard components, and the arrangement rules between the standard components; the processing unit matches the work order component set with multiple component relationship trees, and determines the component relationship tree with the highest component matching degree as the target component relationship tree; obtains the user's emotion category from the target complaint work order, generates a request scenario based on the emotion category and the target component relationship tree, inputs the user request information and product attribute information into the request scenario to perform logical judgment, and obtains the request result; determines at least one processing solution based on the request result and the current actual status of the product corresponding to the target complaint work order; and the sending unit sends the request result and processing solution to the customer service personnel handling the target complaint work order.

[0019] In a third aspect, this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory, causing the electronic device to perform any of the methods described above in this application.

[0020] In a fourth aspect, this application provides a computer-readable storage medium storing instructions that, when executed, perform any of the methods described above in this application.

[0021] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: 1. Extract user request information from target complaint work orders and match it with request types in a pre-set request table. This allows for rapid identification of the user's target request type. Utilizing the natural language processing capabilities of AI large-scale models, key information is accurately extracted from complex language expressions and semantically matched with the pre-set request table to identify the user's target request type. This dynamic matching method based on actual content avoids the limitations of fixed classification of request types in traditional standard processes. 2. Extract multiple product attribute components from the complained product information to construct a work order component set, enabling fine-grained analysis of specific product issues. 3. Obtain multiple component relationship trees based on the user request type and select the target component relationship tree with the highest component matching degree using a matching algorithm. This achieves dynamic selection of the most suitable logical structure for the current work order from the standard component set. 4. Obtain the user's emotional category from the complaint work order and generate a corresponding request scenario based on the target component relationship tree. Execute logical judgments within the request scenario and dynamically generate a solution based on the request result and the actual product status. Finally, send the request result and solution to customer service personnel, forming a complete closed loop from request classification, problem analysis, solution generation to execution feedback. This closed-loop processing approach ensures that customer service personnel can quickly understand and implement the optimal solution, thereby improving the efficiency and quality of complaint handling.

[0022] 2. Extract multiple product information related to the complained product from the target complaint work order to ensure the completeness and accuracy of product data involved in the processing; classify product information into multiple product attributes to achieve multi-dimensional structured processing of product information; filter out key attributes related to the complaint from multiple product attributes, eliminate irrelevant information, and ensure the focus of data analysis; perform text recognition on product type attributes, product claim attributes, and product problem attributes in sequence to obtain multiple text recognition results; integrate the identified product attributes into product attribute components and summarize them into the work order component set to achieve modular management of product information; the construction of the work order component set allows complex complaint issues to be decomposed into multiple independent components, which facilitates the automation, standardization, and efficiency of the processing flow. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the first process of a work order processing method based on an AI large model provided in an embodiment of this application; Figure 2 This is a schematic diagram of a work order processing method based on an AI large model provided in an embodiment of this application; Figure 3 This is a schematic diagram of the second process of a work order processing method based on an AI large model provided in an embodiment of this application; Figure 4This is a schematic diagram of the third process of a work order processing method based on an AI large model provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a work order processing device based on an AI large model provided in an embodiment of this application; Figure 6 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application.

[0024] Explanation of reference numerals in the attached drawings: 501, acquisition unit; 502, processing unit; 503, transmission unit; 600, electronic device; 601, processor; 602, memory; 603, user interface; 604, network interface; 605, communication bus. Detailed Implementation

[0025] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0026] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0027] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0028] Therefore, how to solve the problem that standardized processes in customer service tickets are unable to cope with increasing complexity is a challenge. This application provides a ticket processing method based on an AI large-scale model, applied to a server. The server in this application is a platform providing ticket processing services for customer service personnel. Figure 1 This is a first flowchart illustrating a work order processing method based on an AI large model provided in an embodiment of this application. (Refer to...) Figure 1 The method includes the following steps S101-S107.

[0029] S101: Extract user request information from the target complaint work order, match the user request information with each request type in the preset request table to obtain the target request type.

[0030] In step S101 above, after a consumer selects the goods they want to buy in an online store and completes payment, a goods order is generated. Subsequently, based on actual needs, an appropriate logistics method is selected to transport the goods in the goods order. From the shipment of the goods order until the goods are signed for by the consumer, the entire process continues. Consumers can initiate corresponding complaints based on their own needs; "user" refers to the consumer. Users can send complaints via phone or text message. The system determines the corresponding order information based on the user's complaint and then creates a target complaint ticket based on the complaint and order information. The target complaint ticket is created based on the user's complaint. The system extracts the user's request information from the target complaint ticket. Natural language processing (NLP) technology is used to segment and semantically analyze the text content of the target complaint ticket, identifying keywords and semantic features. For example, when a user submits a complaint like "The package hasn't been delivered; it's been delayed for three days. This logistics is too slow," key information such as "package," "not delivered," "delayed," and "slow logistics" will be extracted and integrated to form a structured user request message. Figure 2 The complaint ticket shown can be used to determine the complainant's information based on the user's ID information, then to query the recipient and address filled in the order based on the user's ID, and then to determine the trajectory address of the package during the time when the user sent the complaint content, and finally to determine the complainant based on the trajectory address.

[0031] Furthermore, after obtaining user request information, the system intelligently matches this information with a pre-set request form. This form stores various standardized request types, including but not limited to requests to change address, cancel shipment, verify receipt, expedite delivery, request door-to-door delivery, request delivery to a designated location, request pre-delivery phone call, complain about courier attitude, complain about customer service attitude, request refund of shipping fees, complain about information leakage, stop harassment, request invoices, claim compensation, inquire about branch information, inquire about shipping services, unjustified return and redelivery, request to send a pickup code, request return, request to change contact number, request to pick up a package not belonging to the recipient, verify damage, customer request for self-pickup, verify missing / inconsistent contents, check weight, complain about courier point, verify lost items, verify delays, and request to contact the courier. A vector similarity calculation method is used to convert the extracted user request information into feature vectors, which are then compared with the standard vectors corresponding to each request type in the pre-set request form. By setting a similarity threshold, the request type with the highest similarity exceeding the threshold is selected as the target request type.

[0032] In the example above, the work order can be matched to the target request type "expedited delivery". When a user's request information corresponds to multiple request types, priority can be assigned to each request type based on historical processing experience, and the request type with the highest priority can be directly output as the target request type. For example, if a user simultaneously requests "expedited delivery" and "request door-to-door delivery", since "expedited delivery" has a lower priority than "request door-to-door delivery", "request door-to-door delivery" will be output as the target request type first. Through this standardized request extraction and matching process, the core user request can be quickly and accurately understood, providing a clear direction for subsequent work order processing. At the same time, this standardized processing method also facilitates the system's accumulation and analysis of processing experience for various requests, continuously optimizing processing strategies.

[0033] S102: Extract the complaint product information from the target complaint work order and generate a work order component set corresponding to the complaint product information. The work order component set includes multiple product attribute components from the complaint product information.

[0034] In step S102 above, relevant information about the complained product is first located from the target complaint work order. Regular expressions and semantic recognition technology can be used to accurately identify basic information such as product number, product name, and specifications within the work order. Then, a set of work order components corresponding to the complained product information is generated. This set includes multiple product attribute components from the complained product information. Specifically, this includes: extracting multiple complained product information related to the complained product from the target complaint work order; classifying each complained product information to obtain multiple product attributes; filtering out product type attributes, product claim attributes, and product problem attributes from these attributes; recognizing the product type attributes, product claim attributes, and product problem attributes to obtain multiple text recognition results; determining the product attribute components based on the text recognition results; and summarizing these multiple product attribute components into the work order component set.

[0035] Specifically, natural language processing (NLP) technology is used to extract multiple product information related to the complained-about product from the target complaint order. This information may be scattered across different parts of the order. For example, in a complaint order like "My 128GB XX phone, purchased the day before yesterday, has a black screen. It's still under warranty; please handle it as soon as possible," it's necessary to identify and extract all information fragments related to the complained-about product. To achieve structured information processing, the extracted product information is intelligently categorized. Using a pre-set classification model, this information is classified into different product attributes, such as type attributes (fresh produce, furniture, clothing), basic attributes (brand, model, specifications), status attributes (usage time, warranty status), problem attributes (fault symptoms, impact level), and claim attributes (no-reason return, warranty period). Deep learning models can be used for attribute recognition during this process to ensure classification accuracy. From the various product attributes obtained, we focused on identifying product type attributes, product claim attributes, and product issue attributes. Product type attributes determine the specific area of ​​expertise required (e.g., electronic products require technical support, while clothing requires quality inspection and certification); product claim attributes influence compensation strategies (e.g., differences in handling methods before and after the warranty period); and product issue attributes directly relate to the specific content of the solution. We then conducted in-depth analysis of each of these three types of attributes.

[0036] For product type attributes, matching is performed using a product knowledge graph to accurately identify the specific category and hierarchical relationship of the product. For product claim attributes, a comprehensive judgment is made based on order information and warranty policy. For product problem attributes, matching analysis is performed using a problem feature library. This series of recognition processes generates multiple text recognition results, each containing the specific category of the attribute. Based on the text recognition results, product attribute components are further determined. For example, when "xx mobile phone 128G version" is recognized, a product type attribute component containing information such as brand, model, and capacity is generated; when "within warranty period" is recognized, a product claim attribute component containing information such as warranty status and purchase time is generated. "Claim received" can indicate warranty status, while "no claim received" can indicate no warranty coverage; when "black screen" is recognized, a product problem attribute component containing information such as fault type and severity is generated. "Damaged part" can also be used to represent the fault type. These generated product attribute components are uniformly summarized into a work order component set, forming a complete componentized data structure. This information extraction and component-based processing method transforms fragmented product information in work orders into structured data components, significantly improving the efficiency and accuracy of subsequent processing. This approach not only accurately understands the specific demands of complaints but also facilitates the accumulation of standardized processing experience.

[0037] S103: Obtain multiple component relationship trees corresponding to the target demand type. Each component relationship tree includes a set of standard components composed of multiple standard components, as well as the orchestration rules between the standard components.

[0038] In step S103 above, after determining the target complaint type corresponding to the target complaint work order, multiple component relationship trees corresponding to the target complaint type are retrieved from a preset component relationship database. The preset component relationship database stores multiple component relationship trees corresponding to different complaint types. For example, when the target complaint type is "expedited processing", multiple component relationship trees containing the standard processing flow for expedited processing will be retrieved. These component relationship trees are obtained based on the summary and optimization of historical processing cases.

[0039] Each component relationship tree consists of two core elements: a set of standard components and orchestration rules. The set of standard components contains multiple standard components required to process a specific type of work order; these standard components are the smallest processing units that have been systematically organized. Taking expedited orders as an example, its set of standard components might include a "claims assessment component," a "problem assessment component," and a "type assessment component," etc. These standard components have clearly defined input parameter requirements and output result formats to ensure the standardization of the processing. The orchestration rules define the logical relationships and execution order between the standard components. A directed acyclic graph data structure is used to describe these rules, clearly specifying the processing logic such as dependencies, conditional jumps, and parallel execution between components. For example, the "claims assessment component" will be executed first to assess whether compensation should be granted; based on the assessment result, it will determine whether to initiate the preparatory work of the "problem assessment component" and the "type assessment component."

[0040] When constructing the component relationship tree, the specificities of different processing scenarios were considered. For the same target request type, multiple component relationship trees are prepared to handle different processing situations. For example, for expedited delivery orders, separate component relationship trees are built for different scenarios such as regular express delivery, fresh food delivery, and large item logistics. Each component relationship tree contains the optimal standard component combination and processing flow for that scenario. Through this component relationship tree-based processing mechanism, the standardization and intelligence of work order processing can be achieved. The modular design of standard components ensures the reusability and maintainability of the processing process, while flexible orchestration rules guarantee the adaptability and optimization space of the processing flow.

[0041] S104: Match the work order component set with multiple component relationship trees, and determine the component relationship tree with the highest component matching degree as the target component relationship tree.

[0042] In step S104 above, after determining the work order component set based on the complaint product information, each product attribute component in the work order component set is directly compared with standard components in multiple component relationship trees. During the comparison process, multiple dimensions such as component attribute characteristics and logical relationships are comprehensively considered. For example, when the work order component set includes components such as "high-end mobile phone," "under warranty," and "black screen failure," the matching degree between these components and the corresponding standard components in different component relationship trees is calculated. A weighted calculation method can be used to evaluate the component matching degree. First, different weight values ​​are set for different types of components; for example, the weight of the core issue component is higher than that of the auxiliary information component. Then, the similarity score between each component and the standard components is calculated, including text similarity and attribute matching degree. For a scenario like "high-end mobile phone black screen failure," key indicators such as product category matching degree and failure type matching degree are emphasized.

[0043] By employing vectorization calculation methods, the features of each component in the work order component set are transformed into multi-dimensional vectors, and the standard components in the component relationship tree are also transformed into corresponding feature vectors. Accurate matching scores are obtained by calculating the cosine similarity or Euclidean distance between vectors. For example, when a set of standard components in a component relationship tree is found to be highly applicable in handling functional faults in high-end electronic products, that component relationship tree will receive a high matching score. After calculating the matching scores of all component relationship trees, the component relationship tree with the highest matching score is selected as the target component relationship tree. This target component relationship tree serves as the primary basis for subsequent work order processing, ensuring the scientific and effective nature of the processing flow. For example, once the most suitable component relationship tree for handling "high-end mobile phone black screen faults" is determined, subsequent processing can be carried out according to the defined standard components and orchestration rules. This multi-dimensional matching-based selection method significantly improves the accuracy of work order processing. Through precise component matching, the most suitable processing solution can be selected for different types of work orders, increasing the accuracy of work order processing to over 95%. Simultaneously, the adaptive nature of this matching mechanism ensures that the system can continuously optimize the matching strategy and continuously improve processing effectiveness.

[0044] S105: Obtain the user's emotion category from the target complaint ticket, generate a request scenario based on the emotion category and the target component relationship tree, input the user's request information and the complained product information into the request scenario to execute logical judgment and obtain the request result.

[0045] In S105 above, the user's emotional state has a significant impact on the selection and execution of the handling plan when processing the target complaint ticket. First, sentiment analysis technology is used to identify the user's emotional category from the target complaint ticket, including multiple dimensions such as text tone analysis, keyword recognition, and semantic understanding. When abnormal emotional categories such as anger or anxiety are detected, more targeted handling strategies need to be adopted.

[0046] Furthermore, a request scenario is generated based on the emotion category and the target component relationship tree. Specifically, this includes: if the emotion category is an abnormal emotion category, obtaining the original component set from the target component relationship tree, retrieving the usage frequency of each original component in the original component set (the usage frequency being the number of times each original component appears in the standard component set); obtaining the historical reference component set from the historical processing schemes, and determining a set of original components, where the set of original components is the intersection of the original component set and the historical reference component set; when the usage frequency of the target component in the historical processing scheme is greater than the preset usage frequency, and the contribution level in the historical processing scheme is greater than the preset contribution level, the target component is confirmed as a core component, and the target component is any component in the set of original components; when the usage frequency is less than or equal to the preset usage frequency, and the contribution level is less than or equal to the preset contribution level, the target component is confirmed as an auxiliary component; and adjusting the connection relationship of each original component in the target component relationship tree based on the core component and the auxiliary component to obtain the request scenario.

[0047] Specifically, for complaint tickets categorized as "abnormal emotions," the initial step involves retrieving a set of original components from the established target component relationship tree. These original components form the basic building blocks of the processing plan, each with a specific processing function. The usage frequency data of these original components in historical ticket processing is then retrieved, i.e., the number of times each component is used in the standard component set. For example, for "verify damage" tickets, the historical usage frequency of components such as "refund processing," "exchange processing," and "apology explanation" is tallied. To optimize the processing plan, a set of historical reference components is retrieved from the historical processing plan database. These historical plans are summaries of successful handling cases in similar situations. By calculating the intersection of the original component set and the historical reference component set, a set of original components with proven practical application is obtained. This intersection operation ensures that the selected components not only meet the processing requirements of the current ticket but also have good historical processing results. For each target component in this set of original components, its importance is evaluated. The evaluation criteria include two key indicators: usage frequency and contribution level, with the contribution level quantified numerically. When a target component's usage frequency in historical processing exceeds a preset usage frequency (e.g., usage rate exceeding 60%), and its contribution exceeds a preset contribution level (e.g., customer satisfaction improvement exceeding 80%), the target component is identified as a core component. For example, when processing "query weight" work orders, if the "professional quality inspection" component has both high usage frequency and contribution level, it is identified as a core component. Conversely, when neither the usage frequency nor the contribution level of a target component meets the preset standards, it is identified as an auxiliary component. Although these auxiliary components are not critical links in the processing, they can play a supplementary and optimization role in specific situations. For example, the "soothing wording" component may be used as an auxiliary component in conjunction with the core component. If only one of the usage frequency or contribution level meets the preset standards, no adjustment is needed for the target component. For example, if the "soothing wording" component initially corresponds to an auxiliary component in the target relationship tree, and if the "soothing wording" component only meets the preset standards in terms of usage frequency but not in terms of contribution level, then the "soothing wording" component remains an auxiliary component. Based on the identification results of core and auxiliary components, the connection relationships of each original component in the target component relationship tree are dynamically adjusted. Adjustment principles include prioritizing core components, optimizing the connection order between core components, and rationally scheduling the use of auxiliary components. This adjustment creates a processing flow that better suits the current needs and scenarios. For example, when handling quality complaints from angry users, the component order might be adjusted, bringing the "compensation plan" component forward while strengthening the connections between core components such as the "quality inspection report."

[0048] Furthermore, the user's request information and the complained product information are input into the request scenario to execute logical judgments. Specifically, this includes: parsing the target complaint order based on the user's request information to obtain the complaint address, which is the location of the target order when the user filed the complaint; obtaining the product claim component, product type component, and product issue component from the complained product information; inputting the product claim component, product type component, and product issue component into the request scenario to determine whether a single component conforms to the corresponding component logic rules in the request scenario, where a single component can be any one of the product claim component, product type component, and product issue component; if a single component conforms to the corresponding component logic rules in the request scenario, the single component result is obtained; and combining the complaint address and multiple component results to obtain the request result.

[0049] Specifically, the user request information in the target complaint order is analyzed. Key information can be extracted using natural language processing technology, including determining the complaint address, which reflects the specific location of the target order in the logistics trajectory. Different handling measures are required when the complaint occurs at different trajectory addresses. Then, product claim components, product type components, and product issue components are obtained from the previously extracted complaint product information. These three types of components contain key information such as compensation conditions, product attributes, and problem characteristics. The product claim component includes information such as claim status and order timeliness; the product type component includes information such as whether it is a fresh product and temperature control requirements; and the product issue component includes information on whether there are any abnormal conditions such as damage or loss. The product claim components, product type components, and product issue components are input into the constructed complaint scenario to begin matching and verifying the component logic rules. The complaint scenario pre-sets logic rules for different components, which define the applicable conditions and processing logic of the components in a specific scenario. Rule verification is performed on individual components. When an individual component meets the corresponding component logic rules in the complaint scenario, the processing result for that component is generated. For example, the product claim component includes two judgment conditions: "Claim exists" and "Claim does not exist." If an order is identified as having claim attributes, such as mandatory compensation for delays, then the product claim component meets the rule condition for "Claim exists," and the result "Claim exists" is output. The product type component includes two judgment conditions: "Fresh produce" and "Non-fresh produce." If an order is identified as cold chain food, then the product type component meets the rule condition for "Fresh produce," and the result "Fresh produce" is output. The product problem type includes two judgment conditions: "Problem exists" and "Problem does not exist." If a package is identified as intact and undamaged, then the product problem type component meets the rule condition for "Problem does not exist," and the result "Problem does not exist" is output. The "Problem exists" product problem type component includes "Damaged" and "Missing" components; relevant content is extracted from the order, and the corresponding component result is output. For example... Figure 3As shown, since the target request type is "urging delivery", the trajectory address corresponds to a certain transit station. That is, the complaint order is stopped at a certain central station by default. Then, the set of work order components corresponding to "urging delivery" is obtained. First, the target relationship tree that matches the set of work order components is determined. Then, the request scenario is generated according to the emotion category and the target relationship tree. Then, the product compensation component, product type component and product problem component are obtained from the complaint product information. Then, it is determined whether each component is a core component or an auxiliary component to determine the position of each component in the request scenario. Then, the positions are connected to obtain a request scenario.

[0050] Finally, the complaint address is combined with the results of multiple components to form a complete request result. This combination process considers the correlation and priority between different pieces of information. In the example above, when the order is identified as "stuck in the center" from the complaint address, it is positioned as a logistics center stagnation scenario. Then, multiple component results are obtained, and the center stagnation is combined with the results of multiple components to obtain the request result "a fresh goods shipment with no issues and eligible for compensation under the center stagnation scenario." This scenario- and rule-based processing mechanism can accurately identify the processing needs of different types of orders and provide precise solutions. Practice shows that this processing method can improve the processing time of fresh goods expedited orders by 40%, increase user satisfaction by 35%, and effectively reduce losses caused by logistics delays.

[0051] S106: Determine at least one solution based on the results of the complaint and the current actual status of the product corresponding to the target complaint work order.

[0052] In S106 above, the complaint result includes the complaint address and processing suggestions for each component. It retrieves the current actual status of the product corresponding to the target complaint order in real time, including key data such as logistics trajectory and delivery time. For example, for a complaint result stating "a fresh produce item with no issues and claims pending due to center stagnation," it checks the actual storage time, storage temperature, and packaging integrity of the fresh produce. Based on the matching analysis of the complaint result and the current actual status, a processing plan is generated. This plan can also be adjusted by incorporating the user's historical complaint records, thereby improving the user experience.

[0053] In addition, at least one processing plan is determined based on the results of the complaint and the current actual status of the product corresponding to the target complaint ticket. Specifically, this includes: determining at least one initial processing plan based on the results of the complaint and the current actual status of the product corresponding to the target complaint ticket; retrieving the user profile corresponding to the user, obtaining the consumption level and historical complaint list from the user profile, where the consumption level is the level corresponding to the user's annual consumption amount; classifying the historical complaint list to obtain complaint transportation category and complaint product quality category; obtaining the first and second percentages of each of the complaint transportation category and complaint product quality category in the historical complaint list; if the first percentage is greater than the second percentage, and the consumption level is high consumption level, then triggering the logistics service supplementary plan for the complaint transportation category for high consumption level, merging the logistics service supplementary plan with the initial processing plan to obtain the processing plan; if the first percentage is greater than the second percentage, and the consumption level is low consumption level, then triggering the points compensation plan for the complaint transportation category for low consumption level, merging the points compensation plan with the initial processing plan to obtain the processing plan.

[0054] Specifically, an initial handling plan is first generated based on the requested outcome and the current actual status of the product. This initial plan includes basic problem-solving measures. For example, for fresh produce orders stuck at the logistics center, the initial handling plan might include expedited delivery and basic compensation. To optimize the handling plan, user profile data is then retrieved. This user profile data includes consumption level and a list of historical complaints. Consumption level is calculated based on the user's annual spending; for example, users with annual spending exceeding 100,000 yuan might be classified as high-consumption users, while those with annual spending below 10,000 yuan might be classified as low-consumption users. The user's historical complaint list is then intelligently categorized into complaints about transportation and complaints about product quality. The proportion of complaints about transportation and complaints about product quality in the historical complaint list is then calculated to obtain the first and second proportions. For example, if 60% of a user's complaints are about transportation (first proportion) and 40% are about quality (second proportion), it indicates that the user encounters more logistics-related problems. If the first proportion of complaints about transportation is greater than the second proportion of complaints about product quality, a differentiated supplementary plan will be implemented based on the user's consumption level. For high-spending users, supplementary logistics service plans will be triggered, such as VIP express delivery, end-to-end temperature monitoring, and dedicated follow-up services. These supplementary services will be intelligently merged with the initial handling plan to form a more comprehensive solution that meets the service quality requirements of high-end users. For low-spending users, points compensation plans will be triggered, such as providing extra points rewards and coupons. These compensation plans will also be merged with the initial handling plan to form a comprehensive solution that meets the needs of this type of user. For example, in addition to addressing current logistics delays, points compensation can also be used to improve user satisfaction.

[0055] Furthermore, when the first proportion is less than the second proportion, for high-spending users, supplementary solutions related to product quality complaints are triggered. These supplementary solutions include arranging for professional quality inspectors to conduct on-site testing, providing original factory parts replacement services, and opening a green channel for expedited replacements—all high-end service options. For example, when a high-spending user complains about electronic product quality issues, value-added services such as on-site service by original factory engineers and quality upgrade compensation will be added to the initial solution. These high-standard supplementary quality solutions are intelligently merged with the initial solution to form a more comprehensive solution that meets the stringent quality requirements of high-end users. When the first proportion is less than the second proportion, for low-spending users, cost-effective supplementary quality solutions are triggered. These supplementary solutions include providing convenient after-sales service outlets, standardized quality issue handling procedures, and basic product quality assurance. When low-spending users report product quality issues, supplementary content such as appointments at nearby after-sales service outlets and standardized repair services will be added to the initial solution. These supplementary quality solutions will also be merged with the initial solution to form a comprehensive solution suitable for the needs of this type of user.

[0056] S107: Send the results of the request and the solution to the customer service personnel handling the target complaint ticket.

[0057] In S107 above, the request results and handling solutions are integrated to form a structured work order processing guide. The request results can also be linked to the target complaint work order, and the handling solution can be displayed in the corresponding position on the target complaint work order. An intelligent push mechanism can be used to ensure that information is accurately delivered to the customer service personnel handling this type of work order. The pushed information uses a structured display method, which facilitates quick understanding and execution by customer service personnel, significantly improving the efficiency and quality of work order processing.

[0058] In one possible implementation, to continuously optimize the effectiveness of the processing solution, it is necessary to dynamically adjust the processing measures by analyzing historical data and user feedback, promptly identify shortcomings in the processing solution, and provide more accurate solutions. The process involves: obtaining the initial number of historical request types within a preset time period; determining the historical number of historical request types corresponding to the same historical period based on the preset time period, assuming the historical request type and the target request type are the same type; dividing the difference between the initial number of requests and the historical number of requests by the preset number of requests to obtain the request growth rate; obtaining the number of complaints sent by users for the historical request type within the preset time period, and retrieving the average number of complaints for the historical request type within the preset time period; dividing the difference between the number of complaints and the average number of complaints by the average number of complaints to obtain the complaint growth rate; if the request growth rate is greater than the preset request growth rate, and the complaint growth rate is greater than the preset complaint growth rate, then obtaining the target processing solution corresponding to the historical request type; retrieving user evaluation information for the target processing solution, revising the target processing solution based on the evaluation information, obtaining the final processing solution, and replacing the target processing solution corresponding to the historical request type with the final processing solution.

[0059] Specifically, the system obtains the initial number of requests for a specific historical request type within a preset time period (e.g., the last 30 days) and compares it with the historical number of requests of the same type during the same period in the past (e.g., the same period last year). For example, for the historical request type "urging delivery," the system calculates the number of requests in the current 30 days and the number of requests in the same period last year. The request growth rate is calculated by dividing the difference between these two periods by the preset number of requests. For example, if the initial number of requests is 150 and the historical number for the same period is 100, meaning the preset number of requests is 100, then the request growth rate is 50%. Simultaneously, the system obtains the number of complaints initiated by users for historical request types within the preset time period and compares it with the average number of complaints for historical request types within the preset time period. The growth rate per day is calculated by dividing the difference between these two values ​​by the average number of complaints. For example, if a user currently has 30 complaints and the average number of complaints is 20, then the complaint growth rate is 50%. When the detected growth rate of requests exceeds the preset growth rate (e.g., 30%), and the growth rate of complaints exceeds the preset growth rate (e.g., 25%), it indicates that the target handling solutions for historical request types cannot resolve user complaints, and the handling solutions need to be optimized and adjusted. The process involves obtaining the target handling solutions currently in use for historical request types and retrieving user evaluation information for these solutions, including satisfaction ratings, written feedback, and resolution effectiveness. Based on this evaluation information, the target handling solutions are revised to obtain the final handling solution. This final handling solution then replaces the original target handling solution for handling subsequent similar requests. This dynamic optimization mechanism significantly improves the effectiveness of the handling solutions, continuously refining them through data collection and analysis, forming a virtuous cycle of optimization.

[0060] Furthermore, the analysis of user feedback to identify shortcomings in the target handling plan and formulate improvement measures, along with the modification of the target handling plan based on the feedback information to obtain the final handling plan, specifically includes: obtaining the first feedback information of users on the target complaint work order and the second feedback information of users on historical complaint work orders. The historical work order database corresponding to the user includes the target complaint work order and historical complaint work orders, with the historical complaint work order being the previous complaint work order of the target complaint work order; extracting satisfaction score, timeliness score, and resolution effect score from the first and second feedback information respectively; performing a weighted summation of the satisfaction score, timeliness score, and resolution effect score to obtain the first score and the second score, where the first score is the score corresponding to the target complaint work order and the second score is the score corresponding to the historical complaint work order; if the first score is lower than the second score, calculating the difference between the first score and the average score, where the average score is the comprehensive score of all users on the target handling plan; when the difference is greater than a preset difference, comparing the indicators corresponding to the first score and the average score respectively to obtain abnormal indicators; generating improvement steps based on the abnormal indicators, replacing the target steps in the target handling plan with the improvement steps to obtain the final handling plan.

[0061] Specifically, two key evaluation data points are retrieved from the user's historical work order database: the first evaluation information of the target complaint work order and the second evaluation information of the historical complaint work order. The target complaint work order is the most recently processed complaint work order in the historical work order database, and the historical complaint work order is the complaint work order that precedes the target complaint work order in the historical work order database. This continuous comparison of evaluation data can reflect the changing trends of user experience. For example, if the user's most recent complaint was about a logistics delay, the work order for the user's previous complaint about a logistics delay is retrieved, and the first evaluation information for the most recent logistics delay complaint and the second evaluation information for the previous logistics delay complaint are obtained. Then, satisfaction score, timeliness score, and resolution effectiveness score are extracted from the first and second evaluation information. These scores are weighted and calculated, for example, assigning a weight of 40% to the satisfaction score, 30% to the timeliness score, and 30% to the resolution effectiveness score. The first score of the target complaint work order and the second score of the historical complaint work order are obtained by weighted summation. For example, if the target complaint ticket has a satisfaction rating of 7, a timeliness rating of 7, and a resolution effectiveness rating of 9, the weighted first score is 7.6. If the historical complaint ticket has a satisfaction rating of 9, a timeliness rating of 9, and a resolution effectiveness rating of 9, the weighted first score is 8.1. When the first score is lower than the second score, it indicates a declining user experience, requiring further analysis of the cause. The difference between the first score and the average score (the comprehensive score of all users for the solution) will be calculated. For example, if the first score is 6 and the average score is 8, the difference is 2. When this difference exceeds a preset difference (e.g., 1.5), a deep analysis process will be initiated. A detailed comparison of the indicators corresponding to the first score and the average score will be performed to identify abnormal indicators. For example, the comparison might reveal that the timeliness score is significantly lower than the average level, which is marked as an abnormal indicator. Based on the identified abnormal indicators, corresponding improvement steps will be generated. For example, for the problem of low timeliness scores, improvement steps such as "increasing processing priority" and "optimizing process nodes" might be generated. The final processing solution is formed by replacing the relevant target steps in the original processing plan with newly generated improved steps. For example, the "standard processing flow" in the original plan is replaced with a "priority processing channel" to improve processing efficiency. This dynamic optimization mechanism significantly improves the effectiveness of the processing solution. Data shows that using the improved processing solution, user satisfaction increased by an average of 42%, and processing timeliness improved by 35%. By analyzing changes in continuous evaluation data, fluctuations in service quality can be detected in a timely manner, allowing for precise improvements. Furthermore, by comparing individual evaluations with the group average level, personalized needs can be identified, providing more targeted solutions.

[0062] In this embodiment, after creating a target complaint ticket based on user request information, user request information can be extracted from the target complaint ticket. Customer service personnel can directly analyze the user request information, or, based on the above analysis process of the target complaint ticket, input the user request information into a preset request table for querying to obtain the target request type. Then, based on the complaint product information, a ticket component set is determined. Based on the ticket component set, the target relationship tree with the highest matching degree is determined. Then, a request scenario is constructed based on the emotion category as the target relationship number. Subsequently, the user request information and the complaint product information are input into the request scenario for logical judgment to obtain the request result. Then, based on the request result, the corresponding handling solution is matched. Finally, the handling solution, the request result, and the target complaint ticket are associated, and the associated target complaint ticket is sent to customer service personnel so that customer service personnel can directly process the target complaint ticket based on the handling solution, saving customer service personnel the process of manual analysis and summarization, thereby improving the processing efficiency of the target complaint ticket. Figure 4 As shown.

[0063] This application also provides a work order processing device based on an AI large model. Figure 5 This is a schematic diagram of the structure of a work order processing device based on an AI large model provided in an embodiment of this application. (Refer to...) Figure 5 The device includes an acquisition unit 501, a processing unit 502, and a sending unit 503.

[0064] Unit 501 extracts user request information from the target complaint work order, matches the user request information with each request type in the preset request table to obtain the target request type; extracts complaint product information from the target complaint work order, generates a work order component set corresponding to the complaint product information, wherein the work order component set includes multiple product attribute components in the complaint product information; and obtains multiple component relationship trees corresponding to the target request type, each component relationship tree including a standard component set composed of multiple standard components, and the arrangement rules between each standard component. Processing unit 502 matches the work order component set with multiple component relationship trees, and determines the component relationship tree with the highest component matching degree as the target component relationship tree; obtains the user's emotion category from the target complaint work order, generates a request scenario based on the emotion category and the target component relationship tree, inputs the user's request information and product attribute information into the request scenario to perform logical judgment and obtain the request result; and determines at least one processing solution based on the request result and the current actual status of the product corresponding to the target complaint work order. Sending unit 503 sends the request result and handling solution to the customer service personnel handling the target complaint ticket.

[0065] In one possible implementation, the acquisition unit 501 is used to extract multiple complaint product information related to the complained product from the target complaint work order, wherein the complained product is the product targeted by the target complaint work order; the processing unit 502 is used to classify each complaint product information to obtain multiple product attributes; filter out product type attributes, product claim attributes, and product problem attributes from the multiple product attributes; identify the product type attributes, product claim attributes, and product problem attributes to obtain multiple text recognition results; determine product attribute components based on the text recognition results, and summarize the multiple product attribute components into the work order component set.

[0066] In one possible implementation, the acquisition unit 501 is used to, if the emotion category is an abnormal emotion category, obtain the original component set from the target component relationship tree, retrieve the usage frequency corresponding to each original component in the original component set, where the usage frequency is the number of times each original component appears in the standard component set; obtain the historical reference component set from the historical processing scheme, and determine a set of original components, wherein the set of original components is the intersection of the original component set and the historical reference component set; the processing unit 502 is used to, when the usage frequency of the target component in the historical processing scheme is greater than the preset usage frequency, and the contribution level in the historical processing scheme is greater than the preset contribution level, confirm the target component as a core component, and the target component is any component in the set of original components; when the usage frequency is less than or equal to the preset usage frequency, and the contribution level is less than or equal to the preset contribution level, confirm the target component as an auxiliary component; adjust the connection relationship of each original component in the target component relationship tree based on the core component and the auxiliary component to obtain the demand scenario.

[0067] In one possible implementation, the processing unit 502 is used to parse the target complaint order according to the user's request information to obtain the complaint address, which is the trajectory address of the target order when the user made the complaint; the acquisition unit 501 is used to obtain the product claim component, product type component, and product problem component from the complaint product information, input the product claim component, product type component, and product problem component into the request scenario, and determine whether a single component conforms to the corresponding component logic rules in the request scenario, wherein a single component is any one of the product claim component, product type component, and product problem component; if a single component conforms to the corresponding component logic rules in the request scenario, the single component result is obtained; the complaint address and multiple component results are combined to obtain the request result.

[0068] In one possible implementation, the acquisition unit 501 is used to acquire the initial number of historical request types within a preset time period, determine the number of historical request types corresponding to the same historical period based on the preset time period, and the historical request type and the target request type are of the same type; the processing unit 502 is used to divide the difference between the initial number of requests and the number of historical requests by the preset number of requests to obtain the request growth rate; the acquisition unit 501 is used to acquire the number of complaints sent by users for historical request types within the preset time period, and retrieve the average number of complaints for historical request types within the preset time period; the processing unit 502 is used to divide the difference between the number of complaints and the average number of complaints by the average number of complaints to obtain the complaint growth rate; if the request growth rate is greater than the preset request growth rate, and the complaint growth rate is greater than the preset complaint growth rate, then the target processing solution corresponding to the historical request type is acquired; the user's evaluation information on the target processing solution is retrieved, the target processing solution is modified based on the evaluation information to obtain the final processing solution, and the target processing solution corresponding to the historical request type is replaced with the final processing solution.

[0069] In one possible implementation, the acquisition unit 501 is used to acquire the user's first evaluation information on the target complaint work order and the user's second evaluation information on historical complaint work orders. The historical work order database corresponding to the user includes the target complaint work order and historical complaint work orders, and the historical complaint work order is the previous complaint work order of the target complaint work order. The processing unit 502 is used to extract the satisfaction score, timeliness score, and resolution effect score from the first evaluation information and the second evaluation information respectively, and perform a weighted summation calculation on the satisfaction score, timeliness score, and resolution effect score to obtain the first score and the second score. The first score is the score corresponding to the target complaint work order, and the second score is the score corresponding to the historical complaint work order. If the first score is lower than the second score, the difference between the first score and the average score is calculated. The average score is the comprehensive score of all users for the target handling solution. When the difference is greater than a preset difference, it is confirmed that the indicators corresponding to the first score and the average score are compared to obtain the abnormal indicators. The sending unit 503 is used to generate improvement steps based on the abnormal indicators, replace the target steps in the target handling solution with the improvement steps, and obtain the final handling solution.

[0070] In one possible implementation, the processing unit 502 is used to determine at least one initial processing plan based on the complaint result and the current actual status of the product corresponding to the target complaint work order; the acquisition unit 501 is used to retrieve the user profile corresponding to the user, and obtain the consumption level and historical complaint list from the user profile, where the consumption level is the level corresponding to the user's annual consumption amount; the processing unit 502 is used to classify the historical complaint list to obtain the complaint transportation category and the complaint product quality category; the acquisition unit 501 is used to obtain the first percentage and the second percentage of the complaint transportation category and the complaint product quality category in the historical complaint list respectively; if the first percentage is greater than the second percentage and the consumption level is high consumption level, the processing unit 502 is used to trigger the logistics service supplementary plan for the complaint transportation category for the high consumption level, and merge the logistics service supplementary plan with the initial processing plan to obtain the processing plan; if the first percentage is greater than the second percentage and the consumption level is low consumption level, the processing unit 502 is used to trigger the points compensation plan for the complaint transportation category for the low consumption level, and merge the points compensation plan with the initial processing plan to obtain the processing plan.

[0071] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0072] This application also discloses an electronic device. (See reference...) Figure 6 , Figure 6 This application provides a schematic diagram of the structure of an electronic device. The electronic device 600 may include: at least one processor 601, at least one network interface 604, a user interface 603, a memory 602, and at least one communication bus 605.

[0073] The communication bus 605 is used to enable communication between these components.

[0074] The user interface 603 may include a display screen and a camera. Optionally, the user interface 603 may also include a standard wired interface and a wireless interface.

[0075] The network interface 604 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0076] The processor 601 may include one or more processing cores. The processor 601 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 602, and by calling data stored in memory 602. Optionally, the processor 601 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 601 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and application requests; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 601 and may be implemented as a separate chip.

[0077] The memory 602 may include random access memory (RAM) or read-only memory. Optionally, the memory 602 may include a non-transitory computer-readable storage medium. The memory 602 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 602 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch functionality, sound playback functionality, image playback functionality, etc.), instructions for implementing the various method embodiments described above, etc.; the data storage area may store data involved in the various method embodiments described above, etc. Optionally, the memory 602 may also be at least one storage device located remotely from the aforementioned processor 601.

[0078] like Figure 6 As shown, the memory 602, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a mobile phone-based application for monitoring abnormal temperature.

[0079] exist Figure 6In the electronic device 600 shown, the user interface 603 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 601 can be used to call the application stored in the memory 602 based on the temperature anomaly monitoring of the mobile phone. When executed by one or more processors, the electronic device performs one or more of the methods described in the above embodiments.

[0080] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0081] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0082] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some service interfaces; indirect couplings or communication connections between devices or units may be electrical or other forms.

[0083] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0084] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0085] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0086] The above description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truths. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure.

Claims

1. A work order processing method based on an AI large model, characterized in that, The method includes: Extract user request information from the target complaint work order, and match the user request information with each request type in the preset request table to obtain the target request type; Extract the complaint product information from the target complaint work order and generate a work order component set corresponding to the complaint product information, wherein the work order component set includes multiple product attribute components in the complaint product information; Obtain multiple component relationship trees corresponding to the target demand type. Each component relationship tree includes a set of standard components composed of multiple standard components, and the orchestration rules between the standard components. The work order component set is matched with multiple component relationship trees, and the component relationship tree with the highest component matching degree is determined as the target component relationship tree; The user's emotion category is obtained from the target complaint ticket. A request scenario is generated based on the emotion category and the target component relationship tree. The user's request information and the complained product information are input into the request scenario to perform logical judgment and obtain the request result. Determine at least one solution based on the results of the complaint and the current status of the product corresponding to the target complaint work order; The target complaint ticket will send the requested outcome and the proposed solution to the customer service personnel handling the target complaint ticket.

2. The method according to claim 1, characterized in that, The step involves extracting complaint product information from the target complaint work order and generating a work order component set corresponding to the complaint product information. The work order component set includes multiple product attribute components from the complaint product information, specifically including: Extract information on multiple complained products associated with the complained product from the target complaint order, wherein the complained product is the product targeted by the target complaint order; The information on the complained products is categorized to obtain various product attributes; Filter out product type attributes, product claim attributes, and product problem attributes from the various product attributes described; The product type attribute, the product claim attribute, and the product problem attribute are identified to obtain multiple text recognition results; Based on the text recognition results, product attribute components are determined, and multiple product attribute components are summarized into the work order component set.

3. The method according to claim 1, characterized in that, The step of generating the appeal scenario based on the emotion category and the target component relationship tree specifically includes: If the emotion category is an abnormal emotion category, then the original component set is obtained from the target component relationship tree, and the usage frequency corresponding to each original component in the original component set is retrieved. The usage frequency is the number of times each original component appears in the standard component set. Obtain a set of historical reference components from historical processing schemes and determine a set of original components, wherein the set of original components is the intersection of the set of original components and the set of historical reference components; When the frequency of use of the target component in the historical processing scheme is greater than the preset frequency of use, and the degree of contribution in the historical processing scheme is greater than the preset degree of contribution, the target component is confirmed as a core component, and the target component is any one of the original components in the group. When the usage frequency is less than or equal to the preset usage frequency, and the contribution level is less than or equal to the preset contribution level, the target component is confirmed as an auxiliary component; The connection relationships of each original component in the target component relationship tree are adjusted based on the core component and the auxiliary component to obtain the requested scenario.

4. The method according to claim 3, characterized in that, The step of inputting the user request information and the complained product information into the request scenario to perform logical judgment specifically includes: The target complaint ticket is parsed based on the user's request information to obtain the complaint address, which is the trajectory address of the target order when the user filed the complaint. Obtain the product claim component, product type component, and product issue component from the complained product information. Input the product claim component, product type component, and product issue component into the complaint scenario and determine whether a single component conforms to the corresponding component logic rules in the complaint scenario. The single component is any one of the product claim component, product type component, and product issue component. If the individual component conforms to the corresponding component logic rules in the requested scenario, then the individual component result is obtained; The complaint address and multiple component results are combined to obtain the request result.

5. The method according to claim 1, characterized in that, After sending the requested outcome and the proposed solution to the customer service representative handling the target complaint ticket, the method further includes: Get the initial number of historical request types within a preset time period, and determine the number of historical requests corresponding to the historical request type in the same historical period based on the preset time period, wherein the historical request type is the same as the target request type. The difference between the initial number of requests and the historical number of requests is divided by the preset number of requests to obtain the request growth rate. Get the number of complaints sent by the user for the historical request type within the preset time period, and retrieve the average number of complaints for the historical request type within the preset time period. The complaint growth rate is obtained by dividing the difference between the number of complaints and the average number of complaints by the average number of complaints. If the growth rate of the demand is greater than the preset growth rate of the demand, and the growth rate of the complaint is greater than the preset growth rate of the complaint, then the target processing solution corresponding to the historical demand type is obtained. The system retrieves the user's evaluation information on the target processing solution, modifies the target processing solution based on the evaluation information, obtains the final processing solution, and replaces the target processing solution corresponding to the historical request type with the final processing solution.

6. The method according to claim 5, characterized in that, The step of revising the target processing scheme based on the evaluation information to obtain the final processing scheme specifically includes: Obtain the user's first evaluation information on the target complaint work order, and obtain the user's second evaluation information on historical complaint work orders. The historical work order database corresponding to the user includes the target complaint work order and the historical complaint work orders. The historical complaint work order is the previous complaint work order of the target complaint work order. Satisfaction score, timeliness score, and resolution effect score are extracted from the first evaluation information and the second evaluation information, respectively. The satisfaction score, timeliness score, and resolution effect score are weighted and summed to obtain a first score and a second score. The first score is the score corresponding to the target complaint work order, and the second score is the score corresponding to the historical complaint work orders. If the first score is lower than the second score, the difference between the first score and the average score is calculated. The average score is the comprehensive score of all users for the target handling solution. When the difference is greater than a preset difference, it is confirmed that the indicators corresponding to the first score and the average score are compared to obtain the abnormal indicators. Based on the abnormal indicators, improvement steps are generated, and the target steps in the target processing scheme are replaced with the improvement steps to obtain the final processing scheme.

7. The method according to claim 1, characterized in that, The step of determining at least one processing solution based on the requested outcome and the current actual status of the product corresponding to the target complaint order specifically includes: Determine at least one initial handling plan based on the results of the complaint and the current actual status of the product corresponding to the target complaint work order; Retrieve the user profile corresponding to the user, and obtain the consumption level and historical complaint list from the user profile. The consumption level is the level corresponding to the user's annual consumption amount. The historical complaint list is categorized to obtain complaint categories for transportation and complaint categories for product quality. Obtain the first and second percentages of the complaint transportation category and the complaint product quality category in the historical complaint list, respectively. If the first proportion is greater than the second proportion, and the consumption level is a high consumption level, then the logistics service supplement plan for the high consumption level for the complaint transportation category is triggered, and the logistics service supplement plan is merged with the initial processing plan to obtain the processing plan; If the first proportion is greater than the second proportion, and the consumption level is a low consumption level, then the points compensation scheme for the low consumption level for the complaint transportation category is triggered, and the points compensation scheme is merged with the initial processing scheme to obtain the processing scheme.

8. A work order processing device based on an AI large model, characterized in that, The device includes an acquisition unit (501), a processing unit (502), and a sending unit (503). The acquisition unit (501) extracts user request information from the target complaint work order, and matches the user request information with each request type in the preset request table to obtain the target request type. Extract complaint product information from the target complaint work order, generate a work order component set corresponding to the complaint product information, wherein the work order component set includes multiple product attribute components in the complaint product information; obtain multiple component relationship trees corresponding to the target appeal type, each component relationship tree includes a standard component set composed of multiple standard components, and the arrangement rules between each standard component; The processing unit (502) matches the work order component set with multiple component relationship trees, and determines the component relationship tree with the highest component matching degree as the target component relationship tree; obtains the user's emotion category from the target complaint work order, generates a request scenario based on the emotion category and the target component relationship tree, inputs the user's request information and the product attribute information into the request scenario to perform logical judgment, and obtains the request result; and determines at least one processing solution based on the request result and the current actual status of the product corresponding to the target complaint work order. The sending unit (503) sends the request result and the processing solution to the customer service personnel who are handling the target complaint work order.

9. An electronic device, characterized in that, The device includes a processor (601), a memory (602), a user interface (603), and a network interface (604). The memory (602) is used to store instructions. The user interface (603) and the network interface (604) are used to communicate with other devices. The processor (601) is used to execute the instructions stored in the memory (602) to cause the electronic device (600) to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-7.

Images