Instance generation method and device, equipment, storage medium and program product

Abstract

The invention provides a complaint processing instance generation method and device, equipment, a storage medium and a program product, and relates to the technical field of computers. The complaint processing instance generation method comprises the steps of performing key feature extraction on collected historical complaint work order information, and generating a business scene library, an event relation table and a capability mapping table based on an extraction result; based on the business scene library, the event relation table and the capability mapping table, constructing a net-shaped complaint event relation graph, the service information being used for associating the business scene and the complaint processing event; calculating event boundaries of different complaint processing events based on the complaint event relation graph, and constructing a complaint instance framework corresponding to the complaint processing events based on the event boundaries; and filling the complaint instance framework based on the business scene library, the event relation table and the capability mapping table, and generating a complaint processing instance. Through the technical scheme disclosed by the invention, the standardization and automation level of the complaint processing flow can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method for generating a complaint handling instance, a device for generating a complaint handling instance, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] As users' requirements for service quality and efficiency become higher and higher, automated handling of complaint tickets has become an inevitable trend in the development of call centers. Complaint handling instances can be understood as a collection of specific handling methods, processes, and related reference cases set for various types of complaints. When a complaint ticket enters the automated processing process, it is necessary to perform corresponding operations in accordance with the steps and methods specified in the complaint handling instance. However, the complaint handling instance information currently used relies on manual configuration and requires a large amount of operation and maintenance resources. This is especially true for complaint tickets with complex business structures and frequent business changes, which will lead to reduced processing efficiency and increased operating costs of call centers.

[0003] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0004] The purpose of the present disclosure is to provide a method for generating a complaint handling instance, a device for generating a complaint handling instance, an electronic device, a computer-readable storage medium and a computer program product, so as to at least overcome the problem of low efficiency in processing complaint work orders based on complaint handling instances in related technologies to a certain extent.

[0005] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0006] According to one aspect of the present disclosure, a method for generating a complaint handling instance is provided, comprising: extracting key features of collected historical complaint work order information to generate a business scenario library, an event relationship table and a capability mapping table based on the extraction results, wherein the business scenario library is used to record the business scenarios that generate complaints, the event relationship table is used to record the relationship between the business scenarios and complaint handling events, and the capability mapping table is used to record the relationship between the business scenarios and the service information involved in the complaints; constructing a meshed complaint event relationship graph based on the business scenario library, the event relationship table and the capability mapping table, wherein the service information is used to associate the business scenarios with the complaint handling events; calculating the event boundaries of different complaint handling events based on the complaint event relationship graph to construct a complaint instance framework corresponding to the complaint handling events based on the event boundaries; filling the complaint instance framework based on the business scenario library, the event relationship table and the capability mapping table to generate the complaint handling instance.

[0007] In one embodiment of the present disclosure, key features are extracted from the collected historical complaint ticket information to generate a business scenario library, an event relationship table and a capability mapping table based on the extraction results, including: inputting the historical complaint ticket information into a convolutional neural network to classify and summarize the complaint handling events based on the convolutional neural network to obtain summarized event information; calling a large language model to extract the key features from the summarized event information to obtain scenario key information of the business scenario, event key information of the complaint handling event and business key information; clustering the scenario key information to generate the business scenario library based on the clustering results; calling a domain model to extract the relationship between different complaint handling events in the summarized event information based on the scenario key information and the event key information, as well as the relationship between the corresponding business scenario and the complaint handling event to obtain the event relationship table; calling a data association calculation model to detect the association between the scenario key information and the business key information to obtain the capability mapping table.

[0008] In one embodiment of the present disclosure, a meshed complaint event relationship graph is constructed based on the business scenario library, the event relationship table and the capability mapping table, including: determining the complaint handling logical relationship of the complaint handling event based on the business scenario, the complaint handling event and the service information; calling the event relationship model to perform a topology depiction operation based on the complaint handling logical relationship to generate the meshed complaint event relationship graph.

[0009] In one embodiment of the present disclosure, the complaint handling logical relationship of the complaint handling event is determined based on the business scenario, the complaint handling event and the service information, including: integrating the business scenario and the complaint handling event based on the service information to obtain event nodes under different business scenarios; calling a natural language processing model to perform semantic equivalence judgment, semantic similarity judgment and semantic opposite judgment operations on the event nodes to generate event pairs of the complaint handling event; establishing an event framework based on the business scenario library, the event relationship table and the capability mapping table; extracting event motivations from the event framework based on motivation vector parameters; and determining the complaint handling logical relationship based on the event nodes, the event pairs and the event motivations.

[0010] In one embodiment of the present disclosure, the event relationship model is called to perform a topological depiction operation based on the complaint processing logical relationship to generate the meshed complaint event relationship map, including: arranging the different event nodes based on the temporal relationship of the event motivation; sorting the event nodes with the same timing based on the complaint processing logical relationship; merging the event nodes with overlapping relationships, connecting the sorted and processed event nodes, and generating the meshed complaint event relationship map.

[0011] In one embodiment of the present disclosure, the event boundaries of different complaint handling events are calculated based on the complaint event relationship graph to construct a complaint instance framework corresponding to the complaint handling event based on the event boundaries, including: extracting event boundary features of a single complaint handling event based on the complaint event relationship graph; determining event boundaries of different complaint handling processes based on the event boundary features; configuring complaint instance scenarios based on the complaint handling processes within the event boundaries; and calling master data information matching the complaint instance scenarios from the complaint combing library to construct the complaint instance framework based on the complaint instance scenarios and the master data information.

[0012] In one embodiment of the present disclosure, the event boundary features of the single complaint handling event are extracted based on the complaint event relationship graph, including: splitting the complaint event relationship graph based on time-series action monitoring to obtain multiple event fragments based on time-series frames; identifying the event fragments based on event formation to obtain event motivations, event actions and event results as event features of the single complaint handling event; making an approximate judgment between the event features and preset starting features to add a starting boundary annotation to the initially recognized head event based on the judgment result; making an approximate judgment between the event features and preset ending features to add an ending boundary annotation to the initially recognized tail event based on the judgment result to obtain annotated event information; inputting the annotated event information into a DDM-Net model to mine multi-level spatiotemporal features; calculating the confidence of the starting boundary and the ending boundary based on the multi-level spatiotemporal features, and determining the event boundary features based on the calculation results.

[0013] In one embodiment of the present disclosure, event boundaries of different complaint handling processes are determined based on the event boundary features, including: converting the complaint event relationship map into a scatter plot based on an image conversion operation on the event boundary features; calculating the similarity between the event boundary features and the scene key information in the business scenario library; and stroking the scatter points corresponding to the event boundary features whose similarity is greater than a similarity threshold, so as to obtain the event boundaries of the complaint handling process based on the stroking results.

[0014] In one embodiment of the present disclosure, the complaint instance framework is filled in based on the business scenario library, the event relationship table and the capability mapping table to generate the complaint handling instance, including: filling the key features matching the master data information into the byte nodes corresponding to the complaint instance framework based on the business scenario library and the event relationship table, the byte nodes representing the processing features of complaint handling; generating a processing action topology for describing different complaint handling processes based on the event relationship table and the capability mapping table; filling the processing action topology into the field nodes corresponding to the complaint instance framework to generate an instance to be completed; completing the instance to be completed based on a large language model to obtain the complaint handling instance.

[0015] In one embodiment of the present disclosure, a processing action topology for describing different complaint handling processes is generated based on the event relationship table and the capability mapping table, including: determining the event nodes and event relationship types in the complaint handling process based on the event relationship table, so as to use the event nodes as topological nodes of the processing action topology, and drawing directed edges based on the event relationship types; searching the capability mapping table for service information corresponding to the event nodes, so as to mark the service information on the topological nodes, and generate the processing action topology.

[0016] According to another aspect of the present disclosure, a device for generating a complaint handling instance is provided, comprising: an extraction module, for extracting key features of collected historical complaint work order information, so as to generate a business scenario library, an event relationship table and a capability mapping table based on the extraction results, wherein the business scenario library is used to record the business scenarios that generate complaints, the event relationship table is used to record the relationship between the business scenarios and complaint handling events, and the capability mapping table is used to record the relationship between the business scenarios and the service information involved in the complaints; a construction module, for constructing a meshed complaint event relationship graph based on the business scenario library, the event relationship table and the capability mapping table, wherein the service information is used to associate the business scenarios with the complaint handling events; a calculation module, for calculating the event boundaries of different complaint handling events based on the complaint event relationship graph, so as to construct a complaint instance framework corresponding to the complaint handling events based on the event boundaries; a generation module, for filling the complaint instance framework based on the business scenario library, the event relationship table and the capability mapping table, and generating the complaint handling instance.

[0017] According to another aspect of the present disclosure, there is provided an electronic device, comprising: a processor; and a memory for storing executable instructions of the processor; the processor is configured to execute the above-mentioned method for generating a complaint handling instance by executing the executable instructions.

[0018] According to another aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the method for generating a complaint handling instance is implemented.

[0019] According to another aspect of the present disclosure, a computer program product is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned method for generating a complaint handling instance is implemented.

[0020] The complaint handling instance generation scheme provided by the embodiments of the present disclosure mines key elements from historical complaint ticket information, constructs a comprehensive and structured business scenario library, event relationship table and capability mapping table, and then constructs a complaint event relationship map, thereby realizing the relationship combing of the complaint handling system. By calculating event boundaries and generating complaint handling instances, it is beneficial to improve the standardization and automation level of the complaint handling process, reduce the complexity and uncertainty of manual complaint handling, and thus help improve the efficiency and quality of complaint handling.

[0021] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0023] Figure 1 A schematic diagram showing a system for generating a complaint handling example in an embodiment of the present disclosure;

[0024] Figure 2 A schematic flow chart showing a method for generating a complaint handling instance in an embodiment of the present disclosure;

[0025] Figure 3 A schematic flow chart showing another method for generating a complaint handling instance in an embodiment of the present disclosure;

[0026] Figure 4 A schematic flow chart showing a method for generating a complaint event relationship map in an embodiment of the present disclosure;

[0027] Figure 5 A schematic flow chart showing a method for generating another complaint handling instance in an embodiment of the present disclosure;

[0028] Figure 6 A schematic flow chart showing a method for generating another complaint handling instance in an embodiment of the present disclosure;

[0029] Figure 7 A schematic diagram showing a scatter plot in a generation scheme of a complaint handling example in an embodiment of the present disclosure;

[0030] Figure 8 A schematic flow chart showing a method for generating another complaint handling instance in an embodiment of the present disclosure;

[0031] Fig. 9 A schematic diagram showing a system for generating another example of complaint handling in an embodiment of the present disclosure;

[0032] Fig.10 A schematic flow chart showing a method for generating another complaint handling instance in an embodiment of the present disclosure;

[0033] Fig.11 A schematic flow chart showing a method for generating another complaint handling instance in an embodiment of the present disclosure;

[0034] Fig.12 A schematic diagram of a generating device for a complaint handling example in an embodiment of the present disclosure is shown;

[0035] Fig.13A structural block diagram of an electronic device in an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0036] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0037] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0038] like Figure 1 As shown, complaint instance information is generated by a combination of manual configuration and machine learning. In the manual configuration method, business information is annotated by manually writing scripts or visual interface topology description, and the complaint handling process is described based on business process analysis to obtain an operation trajectory topology description. In the machine learning method, natural language understanding is used to perform deep learning on the annotated business information, and keyword updates are performed in combination with machine learning capabilities to obtain a knowledge graph. Information is integrated based on the knowledge graph to obtain a knowledge graph keyword library. Specifically, the machine learning module extracts existing keywords and performs keyword association expansion based on a specific algorithm (generally using synonyms and near-synonyms algorithms) to enrich the feature information and keyword library in the complaint instance information. Furthermore, the topology of similar events is replicated by manually describing the topology in combination with the association expansion of keywords. Finally, the system integrates the knowledge graph keyword library output by machine learning with the manually drawn operation trajectory topology description to generate complaint instance information.

[0039] The generation and operation of complaint handling instances in the above scheme depend on the operation team, which analyzes the processing of historical complaint tickets, manually annotates the processing results, and orchestrates the process. If the complaint instance library is to be able to sufficiently support the automated processing of complaint tickets, a large amount of operation and maintenance resources will need to be invested. This has limited impact on call centers with relatively simple business structures, but for call centers that need to handle complex business structures and frequent business changes, the investment in operation and maintenance resources will increase exponentially, which not only has high operating costs and low efficiency, but is also prone to errors.

[0040] In order to solve the problem that a large amount of human resources are needed to operate and maintain the complaint handling instance library, a solution for generating complaint handling instances based on business graphs is proposed. In the process of automated processing of complaint tickets, large language models and event relationship topology description models are introduced to generate complaint handling instances. First, the large language model is used to identify historical complaint data, split information, and extract event relationships, and the event information in the complaint ticket is realistically presented and described. Subsequently, the large language model is used to transform the complex and changeable event information into structured and visualized data based on logical relationships. Then, complaint handling instances are generated based on the business graph, covering aspects such as customer information query, complaint content understanding, and processing flow description. The application of the large language model can generate complaint handling instances more objectively and meticulously, increasing the coverage rate of the original complaint automation processing from 50%-60% to 80%, effectively enhancing the ability to generate personalized case complaint handling instances, broadening the coverage of business scenarios for complaint processing automation, improving the efficiency of complaint processing automation, and reducing the operating costs of complaint processing automation. At the same time, it also opens up new directions and ideas for the application of artificial intelligence technology in the field of customer service, and strongly promotes the overall improvement and innovative development of the level of complaint processing automation.

[0041] To facilitate understanding, several terms involved in this application are first explained below.

[0042] Large Language Model (LLM): A large language model refers to a neural network model that contains ultra-large-scale parameters, usually with more than one billion parameters.

[0043] Convolutional Neural Networks (CNN) is a deep learning model that is particularly suitable for processing data with spatial relationships, such as images, audio, and video.

[0044] RPA (Robotic Process Automation): By configuring software robots to simulate repetitive tasks performed by humans on computers, these tasks include but are not limited to data entry, form processing, and file organization. The purpose of RPA is to improve efficiency and reduce human errors. It can automatically execute regular workflows without the need for complex decision-making logic.

[0045] Below, each step of the method for generating a complaint handling instance in this example implementation will be described in more detail with reference to the accompanying drawings and embodiments.

[0046] Figure 2 A flow chart of a method for generating a complaint handling instance in an embodiment of the present disclosure is shown.

[0047] like Figure 2 As shown, a method for generating a complaint handling instance according to an embodiment of the present disclosure includes:

[0048] Step S202, extract key features of the collected historical complaint work order information to generate a business scenario library, an event relationship table and a capability mapping table based on the extraction results. The business scenario library is used to record the business scenarios that generate complaints, the event relationship table is used to record the relationship between business scenarios and complaint handling events, and the capability mapping table is used to record the relationship between business scenarios and service information involved in the complaint.

[0049] In some embodiments, historical complaint ticket information is preprocessed to obtain complaint ticket data, and the preprocessed complaint ticket data is input into a convolutional neural network. The convolution layer slides on the data through convolution kernels of different sizes to extract local features, and the pooling layer reduces the dimension of the features. After alternating processing of multiple convolution and pooling layers, the network gradually extracts deep key features in the complaint ticket, including key words in the complaint, etc. Based on the extracted features, the complaints are divided into different business scenario categories according to the similarity of business logic and features, and a business scenario library is constructed. The complaint handling events that will be triggered in each business scenario are analyzed, and the corresponding relationships are determined to generate an event relationship table. The service information involved in each business scenario when handling complaints, including business data, business systems, and capability interfaces, is sorted out to form a capability mapping table.

[0050] Step S204, constructing a networked complaint event relationship graph based on the business scenario library, the event relationship table and the capability mapping table, wherein the service information is used to associate business scenarios with complaint handling events.

[0051] In some embodiments, business scenarios and / or complaint handling events are used as nodes. According to the event relationship table, business scenario nodes and corresponding complaint handling event nodes are connected with directed edges to represent the triggering and triggered relationship between them. Combined with the capability mapping table, business scenario nodes and complaint handling event nodes are connected with relevant service information nodes to indicate the dependence and calling relationship on service information when handling complaint handling events under the business scenario, thereby constructing a complex mesh complaint event relationship map, which intuitively presents the various association structures in the entire complaint handling system.

[0052] Step S206, calculating the event boundaries of different complaint handling events based on the complaint event relationship graph, so as to construct a complaint instance framework corresponding to the complaint handling event based on the event boundaries.

[0053] In some embodiments, the complaint event relationship graph is analyzed, and the start and end marks of each complaint handling event are determined as event boundaries in combination with the connection relationship between the nodes in the graph, the data flow, and the business logic. For example, a complaint handling event starts with the relevant business data entering a specific business system and ends with the business system outputting the processing results and triggering the next related event. A complaint instance framework is constructed based on the determined event boundaries. The framework contains the business scenario information area where the complaint handling event is located, the framework structure of the event processing process, placeholders for the service information that needs to be called, and other content, so as to facilitate the subsequent filling of specific information.

[0054] Step S208, fill in the complaint instance framework based on the business scenario library, event relationship table and capability mapping table, and generate a complaint handling instance.

[0055] In some embodiments, corresponding business scenario detailed information is obtained from the business scenario library, such as the specific process of the business scenario, the roles involved, etc., and filled into the business scenario information area of ​​the complaint instance framework, and the detailed processing steps, previous and subsequent connection events, etc. of the complaint handling event are extracted from the event relationship table to improve the event processing process part of the framework. Based on the capability mapping table, the specific details of the service information involved, such as the specific value range of the business data, the operating instructions of the business system, the calling parameters of the capability interface, etc., are filled into the service information part of the framework, thereby generating a complete complaint handling instance with clear processing steps and resource calling information, which can be directly applied to the processing of actual complaint tickets.

[0056] In this embodiment, by mining key elements from historical complaint ticket information, a comprehensive and structured business scenario library, event relationship table and capability mapping table are constructed, and then a complaint event relationship map is constructed, which realizes the relationship combing of the complaint handling system. By calculating event boundaries and generating complaint handling instances, it is beneficial to improve the standardization and automation level of the complaint handling process, reduce the complexity and uncertainty of manual complaint handling, and thus help improve the efficiency and quality of complaint handling.

[0057] In one embodiment of the present disclosure, historical complaint ticket information is collected periodically or in real time through a data collection module, and a large language model is applied through an RPA module to perform event induction on the complaint ticket data to extract event information. The large language model is then applied again to cluster the event information, and key business scenario information is extracted according to categories such as business, service, and product to form a business scenario vocabulary. The domain model is applied to calculate the association between complaint events and business keywords, and the event graph description model is applied to describe the event relationship topology in the data cache module.

[0058] In one embodiment of the present disclosure, an RPA module is used to complete the construction of a complaint handling instance. First, the data collection module is responsible for collecting historical complaint work order information in a regular or real-time manner. This information will serve as the basic data source for subsequent processing.

[0059] Based on the operation of the RPA module, the large language model is used to perform event induction operations on the complaint work order data, and key event information is extracted from a large amount of work order data. The large language model is used again to perform cluster analysis based on the extracted event information. In the clustering process, multiple factors such as business, service, and product are comprehensively considered to extract key business scenario information, and then a business scenario vocabulary is constructed. After that, the domain model is used to deeply calculate the correlation between complaint events and business keywords, and the event graph description model is used to describe the event relationship topology based on these correlations, and finally it is stored in the data cache module for subsequent call and reference at any time.

[0060] In addition, the RPA module can also apply the previously generated business scenario vocabulary to perform clustering calculations. After completing the clustering calculations, the relevant data is pushed to the large language model, which will perform the labeling of business scenario word vectors and vector labeling of related service information. After the labeling is completed, the labeling results are sorted to form a business scenario keyword library. At the same time, the service information is sorted to generate an atomic capability association list, and the sorted and labeled information is cached in the data cache module. The data preprocessing is completed and the instance description process is started.

[0061] In the instance description process, the RPA module inputs the clustered historical complaint tickets into the convolutional neural network, which performs feature extraction operations on the samples in parallel. Then, the clustering algorithm and event boundary algorithm are applied, with the feature similarity within 90% as the boundary condition, to build the complaint event instance framework. After the construction is completed, the RPA module will match and search in the business graph based on the actual instance framework, and extract the supporting system information corresponding to the complaint handling instance execution process. Finally, the RPA module uses the generated business scenario keyword library and atomic capability association list, combined with the previously extracted relevant information, to perform the final generation of the complaint instance, and obtain a complete and practical complaint handling instance, providing standardized and normalized process guidance and data support for complaint handling work.

[0062] like Figure 3 As shown, in one embodiment of the present disclosure, key features are extracted from the collected historical complaint work order information to generate a business scenario library, an event relationship table, and a capability mapping table based on the extraction results, including:

[0063] Step S302: input the historical complaint work order information into the convolutional neural network to classify and summarize the complaint handling events based on the convolutional neural network to obtain summarized event information.

[0064] In some embodiments, historical complaint ticket information is preprocessed and converted into a structured data format suitable for convolutional neural network input, and the processed historical complaint ticket information is input into the convolutional neural network. The network extracts local features by sliding the convolution kernel of the convolution layer on the data, performs feature dimensionality reduction through the pooling layer, and integrates the feature information through the fully connected layer. Finally, the complaint handling events are classified and summarized based on the differences in feature performance of different complaint handling events. For example, complaint handling events with similar nature and similar processing processes are grouped together to obtain summarized event information.

[0065] Step S304, calling the large language model to extract key features of the summarized event information, and obtaining scenario key information of the business scenario, event key information of the complaint handling event, and business key information.

[0066] In some embodiments, the summarized event information is input into the large language model, and the large language model extracts key features from this information. For business scenarios, the key information of the scenarios such as the business segments and business process links involved will be extracted. For complaint handling events, the key information of events such as event triggering conditions and core processing actions will be extracted. For business-related parts, key business information such as business core resources and key business rules will be extracted.

[0067] Step S306: cluster the scenario key information to generate a business scenario library based on the clustering processing results.

[0068] In some embodiments, clustering algorithms (such as K-Means clustering, hierarchical clustering, etc.) are used to cluster the scenario key information, and the scenario key information is divided into different clusters according to their similarities in business meaning, feature expression, etc. Each cluster represents a typical business scenario category. Based on the clustering results, the business scenario-related information corresponding to each cluster is sorted and summarized to generate a business scenario library, which is convenient for the subsequent location and analysis of the business scenarios for complaint handling.

[0069] Step S308, calling the domain model to extract the relationship between different complaint handling events in the event information summarized based on the scenario key information and event key information, as well as the relationship between the corresponding business scenarios and complaint handling events, to obtain an event relationship table.

[0070] In some embodiments, the domain model is constructed based on business domain knowledge, and can have a deep understanding of the logical relationship between business scenarios and complaint handling events. It takes scenario key information and event key information as input data, and through its built-in rules, algorithms and predefined business logic, it deeply analyzes the summarized event information, extracts the sequence, causal relationship and other relationships between different complaint handling events, and at the same time determines the attribution, triggering and other relationships between each complaint handling event and the corresponding business scenario, sorts out and records the extracted relationships, and finally forms an event relationship table.

[0071] Step S310: calling a data association calculation model to detect the association relationship between scenario key information and business key information to obtain a capability mapping table.

[0072] In some embodiments, the data association calculation model can be a rule-based association model, a statistical analysis association model, or a machine learning association model, etc., which takes scenario key information and business key information as analysis objects and detects association relationships. For example, by analyzing the business resources frequently used in business scenarios, the system functions called, etc., the corresponding relationships between them in the business operation process are determined, and these association relationships are organized and recorded to obtain a capability mapping table.

[0073] In this embodiment, convolutional neural networks, large language models, domain models and data association computing models are combined to conduct multi-dimensional mining and analysis of historical complaint work order information, generate business scenario libraries, event relationship tables and capability mapping tables, and build a relatively complete complaint handling-related knowledge system, which provides data support and decision-making basis for further optimizing business operations. By introducing convolutional neural networks (CNN) and large language models, a large amount of historical complaint data can be processed in parallel to significantly improve the efficiency of operational analysis. The information induction ability of the large language model can be used to extract events more objectively and accurately, improve the availability of complaint handling instances, and thus improve the execution efficiency of automated complaint handling.

[0074] In one embodiment of the present disclosure, a meshed complaint event relationship map is constructed based on a business scenario library, an event relationship table, and a capability mapping table, including: determining the complaint handling logical relationship of the complaint handling event based on the business scenario, complaint handling event, and service information; calling the event relationship model to perform a topology depiction operation based on the complaint handling logical relationship to generate a meshed complaint event relationship map.

[0075] In this embodiment, common complaint handling events in different business scenarios and their logical connections such as the order of occurrence and causal relationships are sorted out. According to the use of service information in complaint handling events in different business scenarios, the conditional relationship, constraint relationship, etc. between each link are determined. According to the complexity of the business and the characteristics of the data, an appropriate event relationship model is selected. Common models include graph theory-based models (such as directed graph models) and network analysis-based models. Business scenarios and complaint handling events are used as nodes. According to the previously determined complaint handling logical relationship, related nodes are connected with directed edges to construct a topological structure. By constructing a meshed complaint event relationship map, the business scenarios, complaint handling events and the logical relationships between them that were originally scattered, complex and hidden in a large number of business operations and complaint records are presented in an intuitive visual form, so as to facilitate the intuitive generation of complaint handling instances.

[0076] In one embodiment of the present disclosure, determining the complaint handling logic relationship of the complaint handling event based on the business scenario, the complaint handling event and the service information includes:

[0077] Based on the service information, business scenarios and complaint handling events are integrated to obtain event nodes in different business scenarios. Among them, service information includes business data (such as order details, customer information, etc.), business systems (such as e-commerce platform systems, logistics management systems, etc.) and capability interfaces (such as data query interfaces, business operation interfaces), etc. Event nodes are the result of abstracting units with relatively independent functions or actions in the complaint handling process in business scenarios.

[0078] The natural language processing model is called to perform semantic equivalence judgment, semantic similarity judgment and semantic opposite judgment operations on the event nodes to generate event pairs of complaint handling events. Among them, the event pair is a pair of events combined according to semantic equivalence, semantic similarity or semantic opposite relationship after the natural language processing model performs semantic analysis on the event nodes.

[0079] An event framework is established based on the business scenario library, event relationship table and capability mapping table. The event framework is based on the business scenario library, event relationship table and capability mapping table to describe the overall picture of a complaint handling event.

[0080] The event motivation is extracted from the event frame based on the motivation vector parameter, where the event motivation refers to the internal factors that drive the occurrence and development of the event.

[0081] Determine the logical relationship of complaint handling based on event nodes, event pairs and event motivations.

[0082] Among them, based on the basic units of complaint handling events represented by event nodes, the semantic associations embodied by event pairs, and the event driving factors revealed by event motivations, for example, in the "complaint handling of product delivery errors" scenario, the event nodes "customer discovery of product delivery error complaint handling event" and "customer service verification of delivery information event" constitute an event pair, and the event motivation is "correcting delivery errors to protect customer rights and interests and maintain normal business operations." The logical relationship of complaint handling is determined as follows: first, the customer initiates a complaint to trigger the "customer service verification of delivery information event." During the verification process, it is necessary to obtain accurate information through the order query interface (capability interface) based on the order data (business data). If the delivery error is determined, it will further trigger subsequent processing processes such as "arranging for replacement or reissue events." Based on this method, the complaint handling logical relationship chain between each event node is sorted out.

[0083] In this embodiment, event nodes are constructed by integrating service information, event pairs are mined with the help of natural language processing models, and an event framework is established and event motivations are extracted by combining the business scenario library, event relationship table and capability mapping table. Finally, the logical relationship of complaint handling is determined, ensuring the reliability of generating a complaint event relationship map based on the logical relationship of complaint handling.

[0084] In one embodiment of the present disclosure, the event relationship model is called to perform a topology drawing operation based on the complaint processing logical relationship to generate a networked complaint event relationship graph, including:

[0085] Different event nodes are arranged based on the temporal relationship of event motivations; event nodes with the same temporal sequence are sorted based on the logical relationship of complaint processing; event nodes with overlapping relationships are merged, and the sorted and processed event nodes are connected to generate a networked complaint event relationship map.

[0086] like Figure 4 As shown, a map of the relationship map of complaint events is drawn, including:

[0087] The event features of multiple events in the data cache (including event 1, event 2, ..., event n) are identified and labeled, and the convolutional neural network model (CNN) extraction method is applied to locate the key information of the event through vector labeling, and feature extraction is performed based on the unit vector.

[0088] The natural language understanding (NLP) model is applied to judge whether events form an "event pair" relationship based on semantic equivalence, similarity, and opposition, so as to obtain event pairs based on the judgment results, as well as event relationship classification and event association identification of event pairs. Among them, event relationship classification includes co-reference relationship, temporal relationship, causal relationship and subordinate relationship, and temporal relationship includes precedence, succession and overlap.

[0089] Furthermore, based on the reading comprehension paradigm, the event framework is established to determine the invalid elements in the event features and isolate them. Based on the existing motivation vector parameters and the semantic equivalence method, the event motivation is extracted. Finally, the event logic calculation is performed based on the event motivation, valid event features and related keywords, including:

[0090] First, they are arranged in chronological order, and then in logical order if they are in sequential position. If events overlap, the information is sorted according to the event merging processing method. According to the event pairs related by motivation, the established event relationship model is retrieved for pairing and depiction. The event pairs and their elements are connected to finally form a complaint event relationship map.

[0091] like Figure 5 As shown, in one embodiment of the present disclosure, the event boundaries of different complaint handling events are calculated based on the complaint event relationship graph, so as to construct a complaint instance framework corresponding to the complaint handling event based on the event boundary, including:

[0092] Step S502: extracting event boundary features of a single complaint handling event based on the complaint event relationship graph.

[0093] Among them, the event boundary characteristics of a single complaint handling event refer to a series of feature sets that can define the start, key nodes in the development process, and the end state of a single complaint handling event, including time-related factors, business data-related factors, participating role-related factors, and business system-related factors.

[0094] Step S504: determining event boundaries of different complaint handling processes based on event boundary characteristics.

[0095] Among them, the event boundary refers to the boundary that determines the transition of the entire complaint handling process from one stage to another based on the boundary characteristics of a single complaint handling event.

[0096] Step S506, configuring the complaint instance scenario based on the complaint handling process within the event boundary.

[0097] Among them, the complaint instance scenario is a specific scenario constructed according to the complaint handling process within the event boundary, which is used to simulate and display various situations in the actual complaint handling process.

[0098] Step S508, calling the master data information matching the complaint instance scenario from the complaint combing library to build a complaint instance framework based on the complaint instance scenario and the master data information.

[0099] Among them, the complaint sorting library refers to a knowledge base that stores a large amount of information related to complaint handling, and the master data information refers to the key data and information related to the complaint instance scenario extracted from the complaint sorting library, which can be matched according to factors such as the business type of the complaint instance scenario, the processing process stage, and the products or services involved.

[0100] In this embodiment, by generating a complaint event relationship graph, event boundary features are extracted, and based on the event boundary features, the event boundaries of the complaint handling process are determined, the complaint instance scenario is configured and the master data information is called to build a complaint instance framework, thereby realizing a clear complaint handling guidance framework from a complex complaint graph, which helps to concretize the abstract complaint handling business process and improve the degree of automation of the complaint handling process.

[0101] like Figure 6 As shown, in one embodiment of the present disclosure, the event boundary features of a single complaint handling event are extracted based on the complaint event relationship graph, including:

[0102] Step S602, splitting the complaint event relationship graph based on time sequence action monitoring to obtain multiple event fragments based on time sequence frames.

[0103] Among them, the Temporal Action Detection (TAD) algorithm is used to detect and locate the time interval of event actions in the sequence and identify the event action category. The temporal frame refers to the time interval used by the temporal action detection algorithm to divide the event, and the event fragment is the unit obtained by splitting the complaint event relationship map through the temporal action detection algorithm.

[0104] Step S604, identifying the event fragment based on the event formation, obtaining the event motivation, event action and event result as the event features of a single complaint handling event.

[0105] Step S606: perform an approximate judgment between the event feature and the preset starting feature, so as to add a starting boundary mark to the initially recognized header event based on the judgment result.

[0106] Among them, approximate judgment refers to comparing and judging the event features identified from the event fragment with the preset starting features and ending features to detect whether they are similar or close.

[0107] Step S608: Approximately judge the event feature and the preset end feature, and add an end boundary mark to the initially identified tail event based on the judgment result to obtain marked event information.

[0108] Among them, the marked event information refers to the information set formed by detailed marking of complaint handling events.

[0109] Step S610: input the labeled event information into the DDM-Net model to mine multi-level spatiotemporal features.

[0110] Among them, the DDM-Net model is a deep learning model used to process data with dual modality. It extracts the temporal and spatial characteristics of events from different levels and angles by performing in-depth analysis on the input annotated event information. The multi-level spatiotemporal features can more comprehensively describe the characteristics of event boundaries.

[0111] In some embodiments, a temporal action monitoring (TAD) algorithm is used to split the generated action graph and output a number of time frames / event segments. Within the split time segments, information is formed into pairs for identification, i.e., motivation->action->result is segment information.

[0112] Furthermore, based on [time slice 1, ..., time slice n], the boundary calculation is performed on each slice, the time slice close to the first step is marked with START, and the time slice close to the last step is marked with END. The DDM-Net model is applied to construct a multi-level spatiotemporal feature library, and the calculated START and END feature information is verified to be correct, and irrelevant and redundant feature parameters are deleted.

[0113] Step S612, calculating the confidence of the start boundary and the end boundary based on the multi-level spatiotemporal features, and determining the event boundary features based on the calculation results.

[0114] The confidence level is an indicator used to measure the accuracy or reliability of the determination of the start boundary and the end boundary.

[0115] In one embodiment of the present disclosure, determining event boundaries of different complaint handling processes based on event boundary features includes:

[0116] Based on the image conversion operation of event boundary features, the complaint event relationship map is converted into a scatter plot; the similarity between the event boundary features and the key scene information in the business scenario library is calculated; the scattered points corresponding to the event boundary features whose similarity is greater than the similarity threshold are stroked to obtain the event boundary of the complaint handling process based on the stroke results.

[0117] In some embodiments, based on the complaint event relationship graph or the samples generated based on the complaint work order information, all extracted event boundary features are converted into a graph to form a dense scatter plot, such as Figure 7 As shown, the horizontal axis and vertical axis of the scatter plot are the first quantitative feature and the second quantitative feature respectively. The first quantitative feature can be the number of associations between the event boundary feature and other events, and the second quantitative feature can be an event diversity index. The first quantitative feature can be a quantitative attribute of the event boundary feature, and the second quantitative feature can be another quantitative feature associated with the horizontal axis attribute. The business scenario keyword library extracted based on the business graph is used as the event boundary classification, and the event boundary is outlined from the scatter plot. The similarity between the feature information and the keyword>90% is used as the judgment standard for the outline. Events in multiple event boxes are marked as compound events, and events that cannot be marked in the event box are marked as special events.

[0118] In this embodiment, the complaint event relationship map is transformed into a scatter plot by performing an image conversion operation on the event boundary features, thereby achieving an intuitive visual presentation of complex relationships. The similarity between the event boundary features and the key scene information in the business scenario library is calculated to associate the business scenarios with the complaint handling process, and then the event boundaries of the complaint handling process are determined along the scatter plot edges corresponding to the event boundary features whose similarity is greater than a threshold. This makes the boundary determination more accurate and effectively defines the boundaries of the complaint handling process.

[0119] like Figure 8 As shown, in one embodiment of the present disclosure, a complaint instance framework is filled based on a business scenario library, an event relationship table, and a capability mapping table to generate a complaint handling instance, including:

[0120] Step S802, based on the business scenario library and the event relationship table, the key features matching the master data information are filled into the byte nodes corresponding to the complaint instance framework, and the byte nodes represent the processing features of the complaint processing.

[0121] In some embodiments, taking the telecommunications complaint scenario as an example, the business scenario library contains various scenario information such as package services, network signals, and call charges. The event relationship table clarifies the relationship between events such as customer complaints about abnormal package charges and customer service inquiries about package details and verification of billing records. First, find the key features that match the master data information (such as specific package types, customer levels, etc.) from the telecommunications business scenario library and the event relationship table. For example, in the case where a customer of a high-end package complains about slow network speeds, the key features such as the network priority protection policy corresponding to the package are matched. Then these features are filled into the byte nodes corresponding to the complaint instance framework. The byte nodes may represent processing features such as complaint processing priority and special processing requirements, so that the framework initially has customized processing feature information for the complaint.

[0122] Step S804: Generate a processing action topology for describing different complaint handling processes based on the event relationship table and the capability mapping table.

[0123] Among them, the complaint event relationship map is a complex network structure, including multiple business scenarios, numerous complaint handling events and rich service information. The processing action topology is used to describe a specific complaint handling process, with the processing actions in the process as topological nodes. These nodes are connected by directed edges, indicating the sequence and causal relationship of the processing actions.

[0124] In some embodiments, the sequence of events in the telecommunications complaint handling process is sorted out based on the event relationship table, and the service information required for each process is determined in combination with the capability mapping table. For example, the technical department needs to call the network monitoring system to view signal strength, base station load and other data (business data) when handling network problems, and operate the network optimization system (business system) through a specific capability interface, thereby generating a processing action topology, such as taking customer service acceptance as the starting node, connecting to the technical department's network detection node, and then connecting to the feedback processing result node, etc., to describe the processing actions of different complaint handling processes and their logical relationships.

[0125] Step S806, fill the processing action topology into the field node corresponding to the complaint instance framework to generate an instance to be improved.

[0126] In some embodiments, the constructed processing action topology is filled into the field node corresponding to the complaint instance framework. The field node can be a part that describes the processing steps and other contents. For example, the network detection action and the subsequent action sequence are filled into the processing step field node, and the information of the technical department, customer service department, etc. involved is filled into the relevant department field node, thereby generating an instance to be improved that covers the processing action flow framework but may not have perfect details.

[0127] Step S808: Complete the instance to be completed based on the large language model to obtain a complaint handling instance.

[0128] In some embodiments, the large language model supplements the missing information of each link in the processing action topology based on its learning of telecommunications business knowledge and complaint handling text, so as to make the entire complaint handling instance complete and clear, and generate a complaint handling instance that can be directly used in telecommunications complaint handling work.

[0129] In this embodiment, by filling the key features matching the master data information into the byte nodes of the complaint instance framework based on the business scenario library and the event relationship table, the main features of the complaint processing can be quickly determined, and the processing action topology is generated based on the event relationship table and the capability mapping table and filled into the corresponding field nodes to construct the instance to be improved, so that the framework of the complaint handling process can be quickly established, and the instance to be improved can be further completed with the help of the large language model, which can efficiently supplement the detailed information and improve the entire complaint handling instance, thereby improving the efficiency and accuracy of the complaint handling instance construction.

[0130] In one embodiment of the present disclosure, a processing action topology for describing different complaint handling processes is generated based on an event relationship table and a capability mapping table, including:

[0131] Based on the event relationship table, the event nodes and event relationship types in the complaint handling process are determined, and the event nodes are used as the topological nodes of the processing action topology. Directed edges are drawn based on the event relationship types. The service information corresponding to the event nodes is searched from the capability mapping table to mark the service information on the topological nodes and generate a closed-loop processing action topology.

[0132] In some embodiments, information related to the recorded complaint handling process is extracted from the event relationship table to determine the event nodes. The event nodes cover the key links of complaint handling from the beginning to the end, and the relationship types between event nodes are clarified, including causal relationships, sequential relationships, collaborative relationships, etc. The determined event nodes are set as topological nodes of the processing action topology, and then directed edges are drawn between the corresponding topological nodes based on the event relationship types. The information of the event nodes is searched from the capability mapping table to find the service information corresponding to each event node, and the service information is marked in detail on the corresponding topological nodes to form a closed-loop processing action topology.

[0133] In this embodiment, by constructing a processing action topology based on the event relationship table, a graph of the complaint handling process is constructed. By annotating service information on topological nodes with the help of a capability mapping table, accurate positioning and efficient allocation of resources are achieved, thereby improving processing efficiency and optimizing resource utilization based on complaint handling instances.

[0134] like Fig. 9As shown, a device for generating complaint handling instances based on a business graph according to an embodiment of the present disclosure includes: a data collection module 902, an RPA module 904, a convolutional neural network 906, a large language model 908, an algorithm module 910, a computing module 912, a data cache module 914 and an interface module 916.

[0135] The data collection module 902 is used to extract the processed complaint ticket information from the complaint ticket processing system on a regular or real-time basis.

[0136] RPA module 904 is used for event relationship topology description, complaint handling instance description process triggering and execution supervision.

[0137] The convolutional neural network 906 is used for concurrent processing of massive amounts of data to improve the efficiency of event extraction and event relationship description.

[0138] The large language model 908 implements event extraction, event relationship extraction and event relationship calculation based on its information induction ability and event calculation ability.

[0139] The algorithm module 910 is used to store algorithm and model modules, mainly including clustering algorithms, event relationship algorithms, event graph description models, event boundary algorithms, etc.

[0140] The computing module 912 is generally a server GPU as a computing unit, which provides computing power support for the execution of convolutional neural network and large language model tasks.

[0141] The data cache module 914 is used for temporary caching of historical complaint ticket data, generated business graphs, and complaint instances.

[0142] The interface module 916 is used as a channel for data interaction between the present patented device and other modules and systems.

[0143] like Fig.10 As shown, according to the present disclosure, a private event relationship topology depiction method includes:

[0144] Step S1002: The data collection module periodically or in real time extracts processed historical complaint work order information as a sample.

[0145] In step S1004, the RPA module inputs the sample into the convolutional neural network for concurrent processing, and extracts features of the input complaint work order information through the convolutional layer of the convolutional neural network.

[0146] Step S1006, calling the large language model to merge based on similar features, and performing event information summarization, including the time when the event occurred, the business / service involved, the service information involved, the customer's demands and the processing plan.

[0147] Step S1008: call the large language model to extract key information from the summarized event information.

[0148] Key information extraction includes: Business scenario word extraction: extracting keywords related to business, product, and service names, such as traffic packages.

[0149] Event keyword extraction: Extract the root cause keywords of the complaint, such as: cannot be used, wrong purchase.

[0150] Business data extraction: Extract the data brought in during event processing, such as: balance of 100 yuan.

[0151] Step S1010, using the synonym calculation algorithm and clustering algorithm, systematically cluster and integrate the extracted business scenario information, classify semantically similar and business-related keywords into one category, form a business scenario keyword library after multiple rounds of processing, and store it in the data cache module.

[0152] Step S1012: Apply the domain model to the summarized event content, calculate the relationship information between the event and the business scenario, and form an event relationship table to cache in the data cache module.

[0153] Among them, the domain model is an analytical model that supports relational calculations to describe the objects involved in the business and their relationships through master data identification and association relationships.

[0154] Step S1014: Apply the data association calculation model to calculate the association relationship between the business scenario and the service information, and form a capability mapping table to cache in the data cache module.

[0155] Among them, service information includes business data, business systems, capability interfaces, etc.

[0156] Step S1016, based on the business scenario keyword library, event relationship table and capability mapping table stored in the data cache module, use the event relationship model to calculate and reason the logical relationship between the business and the service, apply the topology description algorithm, execute the event relationship topology description, and cache the description results to the data cache module.

[0157] Step S1018: The data cache module pushes the generated event relationship topology back to the complaint handling system through the interface module.

[0158] like Fig.11 As shown, according to another embodiment of the present disclosure, a method for generating a complaint handling instance includes:

[0159] Step S1102: The RPA module starts the complaint instance generation operation.

[0160] In step S1104, the RPA module obtains historical complaint ticket data from the data cache module as sample information, and inputs it into the convolutional neural network. During the operation of the convolutional neural network, the large language model is called to perform business scenario vector labeling tasks on the sample information.

[0161] Step S1106, based on the sample information of the business scenario vector that has been labeled, perform sample clustering operation according to the characteristics of similar business scenario vectors, and after the sample clustering is completed, cache it in the data cache module.

[0162] In step S1108, the RPA model applies a large language model to merge the content of the clustered samples, performs event content summarization, and applies an event boundary algorithm to the summarized content to calculate the event boundary, with feature similarity within 90% as the event boundary.

[0163] Step S1110, based on the event boundary, divide the complaint instance scenario, call the master data information of the complaint combing library, and build the instance content framework.

[0164] Step S1112: extract keywords based on the business scenario library and event relationship information and enter them into the corresponding bytes of the instance box.

[0165] Step S1114, based on the event relationship table and the capability mapping table, the event graph model and the topology description algorithm are applied to perform the processing action topology description.

[0166] Step S1116, after completing the processing action description, enter it into the corresponding field of the complaint instance framework.

[0167] Step S1118, applying the large language model to complete the information not filled in the instance framework to obtain a complaint handling instance.

[0168] It should be noted that the above figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0169] Refer to the following Fig.12 To describe a generation device 1200 of a complaint handling instance according to an embodiment of the present disclosure. Fig.12 The generation device 1200 of the complaint handling instance shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiment of the present disclosure.

[0170] The generation device 1200 of the complaint handling instance is expressed in the form of a hardware module or a software module. The components of the generation device 1200 of the complaint handling instance may include but are not limited to: an extraction module 1202, which is used to extract key features of the collected historical complaint work order information, so as to generate a business scenario library, an event relationship table and a capability mapping table based on the extraction results, the business scenario library is used to record the business scenario that generates the complaint, the event relationship table is used to record the relationship between the business scenario and the complaint handling event, and the capability mapping table is used to record the relationship between the business scenario and the service information involved in the complaint; a construction module 1204, which is used to construct a networked complaint event relationship map based on the business scenario library, the event relationship table and the capability mapping table, wherein the service information is used to associate the business scenario with the complaint handling event; a calculation module 1206, which is used to calculate the event boundaries of different complaint handling events based on the complaint event relationship map, so as to construct a complaint instance framework corresponding to the complaint handling event based on the event boundary; a generation module 1208, which is used to fill the complaint instance framework based on the business scenario library, the event relationship table and the capability mapping table to generate a complaint handling instance.

[0171] like Fig.13 As shown, the electronic device 1300 is in the form of a general computing device. The components of the electronic device 1300 may include but are not limited to: at least one processing unit 1310, at least one storage unit 1320, and a bus 1330 connecting different system components (including the storage unit 1320 and the processing unit 1310).

[0172] The storage unit stores program codes, which can be executed by the processing unit 1310, so that the processing unit 1310 performs the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification. For example, the processing unit 1310 can perform the following steps: Figure 2 The scheme described in .

[0173] The storage unit 1320 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 13201 and / or a cache storage unit 13202 , and may further include a read-only storage unit (ROM) 13203 .

[0174] The storage unit 1320 may also include a program / utility 13204 having a set (at least one) of program modules 13205, such program modules 13205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0175] Bus 1330 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0176] The electronic device 1300 may also communicate with one or more external devices 1370 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may also communicate with one or more devices that enable a user to interact with the electronic device 1300, and / or communicate with any device that enables the electronic device 1300 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed via an input / output (I / O) interface 1350. Furthermore, the electronic device 1300 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 1360. As shown, the network adapter 1360 communicates with other modules of the electronic device 1300 via a bus 1330. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device 1300, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0177] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or an electronic device, etc.) to execute the method according to the implementation of the present disclosure.

[0178] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present disclosure may also be implemented in the form of a program product, which includes a program code, and when the program product is run on an electronic device, the program code is used to enable the electronic device to perform the steps according to various exemplary implementations of the present disclosure described in the above "Exemplary Method" section of the present specification.

[0179] According to the program product for implementing the above method according to the embodiment of the present disclosure, it can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on an electronic device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0180] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0181] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0182] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0183] The program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the IoT terminal, as an independent software package, partially on the user computing device and partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0184] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of a module or unit described above can be further divided into multiple modules or units to be embodied. In addition, although the various steps of the method in the present disclosure are described in a specific order in the accompanying drawings, this does not require or imply that these steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired result. In addition or alternatively, some steps can be omitted, multiple steps can be combined into one step for execution, and / or one step can be decomposed into multiple steps for execution, etc. Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions so that a computing device (which can be a personal computer, a server, a mobile terminal, or an electronic device, etc.) executes the method according to the embodiment of the present disclosure. Those skilled in the art will easily think of other embodiments of the present disclosure after considering the specification and practicing the disclosure here. This application is intended to cover any variants, uses or adaptive changes of the present disclosure, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the technical field that are not disclosed in the present disclosure. The specification and embodiments are only regarded as exemplary, and the true scope and spirit of the present disclosure are indicated by the attached claims.

Claims

1. A method for generating a complaint handling instance, characterized in that: include: Extract key features from the collected historical complaint work order information to generate a business scenario library, an event relationship table, and a capability mapping table based on the extraction results. The business scenario library is used to record the business scenarios that generate complaints, the event relationship table is used to record the relationship between the business scenarios and complaint handling events, and the capability mapping table is used to record the relationship between the business scenarios and the service information involved in the complaints; Building a networked complaint event relationship graph based on the business scenario library, the event relationship table and the capability mapping table, wherein the service information is used to associate the business scenario with the complaint handling event; Calculating event boundaries of different complaint handling events based on the complaint event relationship graph, so as to construct a complaint instance framework corresponding to the complaint handling event based on the event boundaries; Based on the business scenario library, the event relationship table and the capability mapping table, the complaint instance framework is filled to generate the complaint handling instance.

2. The method for generating a complaint handling instance according to claim 1, characterized in that: Extract key features from the collected historical complaint ticket information to generate a business scenario library, event relationship table, and capability mapping table based on the extraction results, including: Inputting the historical complaint work order information into a convolutional neural network, so as to classify and summarize the complaint handling events based on the convolutional neural network to obtain summarized event information; Calling a large language model to extract the key features of the summarized event information to obtain the scenario key information of the business scenario, the event key information of the complaint handling event, and the business key information; Performing clustering processing on the scenario key information to generate the business scenario library based on the clustering processing result; Calling the domain model to extract the relationship between different complaint handling events in the summarized event information based on the scenario key information and the event key information, as well as the relationship between the corresponding business scenarios and the complaint handling events, to obtain the event relationship table; The data association calculation model is called to detect the association relationship between the scenario key information and the business key information to obtain the capability mapping table.

3. The method for generating a complaint handling instance according to claim 1, characterized in that: A networked complaint event relationship graph is constructed based on the business scenario library, the event relationship table, and the capability mapping table, including: Determining a complaint handling logic relationship of the complaint handling event based on the business scenario, the complaint handling event and the service information; The event relationship model is called to perform a topology drawing operation based on the complaint processing logical relationship to generate the network complaint event relationship map.

4. The method for generating a complaint handling instance according to claim 3, characterized in that: Determining a complaint handling logic relationship of the complaint handling event based on the business scenario, the complaint handling event, and the service information includes: Integrate the business scenarios and the complaint handling events based on the service information to obtain event nodes under different business scenarios; Calling a natural language processing model to perform semantic equivalence judgment, semantic similarity judgment, and semantic opposite judgment operations on the event node to generate an event pair of the complaint handling event; Establishing an event framework based on the business scenario library, the event relationship table and the capability mapping table; extracting event motivation from the event frame based on motivation vector parameters; Based on the event nodes, the event pairs and the event motivations, the complaint handling logical relationship is determined.

5. The method for generating a complaint handling instance according to claim 4, characterized in that: The event relationship model is called to perform a topology drawing operation based on the complaint processing logical relationship to generate the networked complaint event relationship graph, including: Arranging different event nodes based on the temporal relationship of the event motivation; For the event nodes with the same time sequence, sorting is performed based on the complaint processing logical relationship; The event nodes with overlapping relationships are merged, and the sorted and processed event nodes are connected to generate the networked complaint event relationship map.

6. The method for generating a complaint handling instance according to claim 1, characterized in that: The event boundaries of different complaint handling events are calculated based on the complaint event relationship graph, so as to construct a complaint instance framework corresponding to the complaint handling event based on the event boundaries, including: Extracting event boundary features of a single complaint handling event based on the complaint event relationship graph; Determining event boundaries of different complaint handling processes based on the event boundary characteristics; Configuring a complaint instance scenario based on the complaint handling process within the event boundary; The master data information matching the complaint instance scenario is called from the complaint combing library to construct the complaint instance framework based on the complaint instance scenario and the master data information.

7. The method for generating a complaint handling instance according to claim 6, characterized in that: Extracting the event boundary features of the single complaint handling event based on the complaint event relationship graph includes: The complaint event relationship graph is split based on time sequence action monitoring to obtain multiple event fragments based on time sequence frames; Performing event-based recognition on the event fragments to obtain event motivation, event action and event result as event features of the single complaint handling event; Approximately judging the event feature and the preset starting feature, so as to add a starting boundary mark to the initially recognized header event based on the judgment result; Approximately judging the event feature and the preset end feature, and adding an end boundary mark to the initially identified tail event based on the judgment result to obtain marked event information; Inputting the labeled event information into the DDM-Net model to mine multi-level spatiotemporal features; The confidence of the start boundary and the end boundary is calculated based on the multi-level spatiotemporal features, and the event boundary features are determined based on the calculation results.

8. The method for generating a complaint handling instance according to claim 7, characterized in that: Based on the event boundary characteristics, event boundaries of different complaint handling processes are determined, including: Based on an image conversion operation on the event boundary features, converting the complaint event relationship map into a scatter plot; Calculating the similarity between the event boundary feature and the scene key information in the business scene library; Scattered points corresponding to the event boundary features whose similarity is greater than a similarity threshold are stroked to obtain the event boundary of the complaint handling process based on the stroke result.

9. The method for generating a complaint handling instance according to claim 6, characterized in that: Filling the complaint instance framework based on the business scenario library, the event relationship table and the capability mapping table to generate the complaint handling instance includes: Based on the business scenario library and the event relationship table, the key features matching the master data information are filled into the byte nodes corresponding to the complaint instance framework, and the byte nodes represent the processing features of the complaint processing; Generating a processing action topology for describing different complaint handling processes based on the event relationship table and the capability mapping table; Fill the processing action topology into the field node corresponding to the complaint instance framework to generate an instance to be completed; The instance to be completed is completed based on the large language model to obtain the complaint handling instance.

10. The method for generating a complaint handling instance according to claim 9, characterized in that: Generating a processing action topology for describing different complaint handling processes based on the event relationship table and the capability mapping table includes: Determine the event nodes and event relationship types in the complaint handling process based on the event relationship table, use the event nodes as topological nodes of the processing action topology, and draw directed edges based on the event relationship types; The service information corresponding to the event node is searched in the capability mapping table to mark the service information on the topology node and generate the processing action topology.

11. A device for generating a complaint handling instance, characterized in that: include: An extraction module is used to extract key features from the collected historical complaint work order information, so as to generate a business scenario library, an event relationship table and a capability mapping table based on the extraction results. The business scenario library is used to record the business scenarios that generate complaints, the event relationship table is used to record the relationship between the business scenarios and the complaint handling events, and the capability mapping table is used to record the relationship between the business scenarios and the service information involved in the complaints; A construction module, used to construct a networked complaint event relationship map based on the business scenario library, the event relationship table and the capability mapping table, wherein the service information is used to associate the business scenario with the complaint handling event; A calculation module, used for calculating the event boundaries of different complaint handling events based on the complaint event relationship graph, so as to construct a complaint instance framework corresponding to the complaint handling event based on the event boundaries; A generation module is used to fill in the complaint instance framework based on the business scenario library, the event relationship table and the capability mapping table to generate the complaint handling instance.

12. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; Wherein, the processor is configured to execute the method for generating a complaint handling instance as described in any one of claims 1 to 10 by executing the executable instructions.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for generating a complaint handling instance according to any one of claims 1 to 10 is implemented.

14. A computer program product having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for generating a complaint handling instance according to any one of claims 1 to 10 is implemented.

Images