User incoming call settlement management method, system and device and storage medium

The user call summary management method addresses the inefficiencies of manual call event summarization by using a pre-trained model to automatically generate accurate call summary labels from user call data, improving accuracy and reducing costs.

CN119090440APending Publication Date: 2024-12-06CTRIP TRAVEL NETWORK TECH SHANGHAI0
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411196976.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-28
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

Existing methods for summarizing user call events require significant manpower and time, struggle with accurate classification and labeling of diverse user intentions, and fail to utilize order history, behavior patterns, and ticket data effectively.

Method used

A user call summary management method involving generating call summary labels from historical data, forming a model training dataset, iteratively training a management model with call content and user feature data, and using a pre-trained Roberta-WWM-ext-large model to automatically generate call summary labels from real-time user calls.

Benefits of technology

Improves the accuracy of user call summaries, reduces manual processing costs, and enhances efficiency by automatically generating and managing call summary labels using call content and user feature data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090440A_ABST
    Figure CN119090440A_ABST
Patent Text Reader

Abstract

The invention provides a user incoming call settlement management method, system and device and a storage medium, and the method comprises the steps: generating a plurality of incoming call settlement labels according to historical user incoming call settlement data; according to the historical user incoming call settlement data, obtaining corresponding incoming call content data and corresponding user feature data; forming a model training data set according to the incoming call content data, the user feature data and the incoming call settlement label; performing multiple iterative training on a preset management model according to the model training data set until the recognition accuracy of the preset management model for an incoming call settlement label corresponding to the incoming call of the user is greater than a preset threshold value; and obtaining a real-time user incoming call event, and obtaining an incoming call settlement label corresponding to the real-time user incoming call event according to the preset management model obtained by training. Through the technical scheme provided by the invention, automatic incoming call settlement label generation and management can be carried out based on the incoming call content data in the incoming call of the user, other associated user feature data and the pre-trained fusion feature model, the accuracy of the incoming call settlement of the user is improved, the manual processing cost is reduced, and the user experience is improved. The call settlement efficiency of the user is improved, and the method has popularization value.
Need to check novelty before this filing date? Find Prior Art

Description

Background Art

[0002] In recent years, with the continuous refinement of product services, the diversity and complexity of user calls for consultation, praise or complaints have continued to grow, which include a variety of different user intentions. Therefore, it is necessary to summarize user call events to analyze the impact of call events corresponding to different user intentions on actual products or services.

[0003] The existing method of summarizing user call events requires a lot of manpower costs and time. At the same time, it is difficult to meet the needs of accurate classification and labeling of different user intentions. It has limitations in identifying diverse user intentions and fails to make full use of the call content and user-related additional information, such as order history, behavior patterns, and ticket data, to effectively summarize the call content. It is time-consuming and labor-intensive, and the summary accuracy is low. Summary of the invention

[0004] In view of the problems in the prior art, the purpose of the present disclosure is to provide a user call summary management method, system, device and storage medium. Specifically, the first aspect of the present disclosure provides a user call summary management method, which may include the following steps:

[0005] Generate several call summary labels based on historical user call summary data;

[0006] According to the historical user call summary data, corresponding call content data and corresponding user feature data are obtained, wherein the call content data includes the conversation content corresponding to the user call, and the user feature data includes one or more combinations of order-related data, product-related data, user-related data, and answer-related data corresponding to the user call;

[0007] A model training data set is formed based on the call content data, user feature data, and call summary labels;

[0008] The preset management model is iteratively trained multiple times according to the model training data set until the recognition accuracy of the call summary label corresponding to the user's incoming call by the preset management model is greater than a preset threshold;

[0009] Acquire real-time user incoming call events, and acquire call summary labels corresponding to the real-time user incoming call events according to the preset management model obtained through training.

[0010] In a possible implementation of the first aspect, the user feature data includes a combination of one or more of order information data, product information data, caller data, connection operation data, and voice interaction operation data associated with the user's incoming call;

[0011] The order information data includes the order status, the departure time and / or departure place corresponding to the order, and the arrival time and / or arrival place corresponding to the order;

[0012] Product information data includes product quantity information and product service types;

[0013] The caller user data includes the user's gender, user age, user type, and the call summary label corresponding to the user's historical calls.

[0014] In a possible implementation of the first aspect, forming a model training data set according to the incoming call content data, the user feature data, and the incoming call summary label includes:

[0015] Converting incoming call content data and user feature data into floating-point digital features to serve as input vectors of a preset management model;

[0016] Convert the call summary labels into category vectors to serve as label data for the preset management model.

[0017] In a possible implementation of the first aspect, generating an input vector according to incoming call content data and user feature data includes the following steps:

[0018] According to the call content data, based on data cleaning, the call content text is obtained, and the call content text is formatted to form a standard call text format;

[0019] In the user feature data, the continuous variables in the user feature data are binned and / or discretized, and the non-numerical features in the user feature data are numerated, and the processing results are standardized and normalized to generate a feature matrix corresponding to the user feature data;

[0020] Vectorize the standard incoming call text format to generate embedded vectors in floating point format;

[0021] Vectorize the feature matrix to generate feature vectors in floating point format;

[0022] The embedded vector and the feature vector are fused and concatenated to obtain the input vector.

[0023] In a possible implementation of the first aspect, the preset management model is a Roberta-WWM-ext-large model;

[0024] Acquiring a real-time user incoming call event, and acquiring a call summary label corresponding to the real-time user incoming call event according to a preset management model obtained through training, including the following steps:

[0025] Acquire corresponding incoming call content text and user feature text based on real-time user incoming call events;

[0026] The call content text is sequentially passed through the pre-training layer and attention pooling layer of the Roberta-WWM-ext-large model to form the first semantic vector;

[0027] The user feature text passes through the embedded layer and fully connected layer of the Roberta-WWM-ext-large model in sequence to form the second semantic vector;

[0028] According to the first semantic vector and the second semantic vector, a linear output associated with the call summary label is obtained through a linear layer of the Roberta-WWM-ext-large model;

[0029] The linear output is converted into a probability distribution associated with the call summary label, and the call summary label with the highest probability is selected as the call summary label corresponding to the real-time user call event.

[0030] In a possible implementation of the first aspect above, the loss function corresponding to the preset management model is configured as a cross entropy loss function.

[0031] In a possible implementation of the first aspect, the preset management model is updated based on a preset period; or

[0032] The preset management model self-updates the preset management model based on the label management evaluation result of the real-time user call event.

[0033] A second aspect of the present disclosure provides a user call summary management system, which is used to implement the user call summary management method provided in the first aspect. The user call summary management system may specifically include:

[0034] A label acquisition unit, used to generate a number of call summary labels according to historical user call summary data;

[0035] A data preprocessing unit, configured to obtain corresponding incoming call content data and corresponding user feature data according to historical user incoming call summary data, wherein the incoming call content data includes the conversation content corresponding to the user's incoming call, and the user feature data includes one or more combinations of order-related data, product-related data, user-related data, and answer-related data corresponding to the user's incoming call;

[0036] A data set construction unit, used to form a model training data set based on the incoming call content data, user feature data and the incoming call summary label;

[0037] A model training unit, used to perform multiple iterations of training on the preset management model according to the model training data set, until the recognition accuracy of the call summary label corresponding to the user's incoming call by the preset management model is greater than a preset threshold;

[0038] The incoming call event management unit is used to obtain real-time user incoming call events and obtain the incoming call summary tags corresponding to the real-time user incoming call events according to the preset management model obtained through training.

[0039] A third aspect of the present disclosure provides a user call summary management device, including:

[0040] processor;

[0041] a memory in which executable instructions for the processor are stored;

[0042] The processor is configured to execute the steps of the user call summary management method provided in the first aspect by executing executable instructions.

[0043] A fourth aspect of the present disclosure provides a computer-readable storage medium for storing a program, which, when executed, implements the steps of the user call summary management method provided in the first aspect.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] Through the technical solution provided by the present invention, it is possible to automatically generate and manage call summary labels based on the call content data and other related user feature data in the user's calls and based on a pre-trained fusion feature model, thereby improving the accuracy of user call summaries, reducing manual processing costs, and improving the efficiency of user call summaries, which has replicable value. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Other features, objects and advantages of the present disclosure will become more apparent from a reading of the detailed description of non-limiting embodiments made with reference to the following accompanying drawings.

[0047] Figure 1 According to an embodiment of the present disclosure, a flowchart of a method for managing a summary of user incoming calls is provided.

[0048] Figure 2 According to an embodiment of the present disclosure, a flow chart of generating an input vector according to incoming call content data and user feature data is provided.

[0049] Figure 3a According to an embodiment of the present disclosure, a schematic diagram is provided showing how the accuracy of call summary labels of a preset management model changes with training rounds.

[0050] Figure 3b According to an embodiment of the present disclosure, a schematic diagram showing how the model loss of a preset management model changes with training rounds is provided.

[0051] Figure 4According to an embodiment of the present disclosure, a flow chart of obtaining a call summary tag corresponding to a real-time user call event according to a preset management model obtained through training is provided.

[0052] Figure 5 According to an embodiment of the present disclosure, a structural schematic diagram of a user call summary management system is provided.

[0053] Figure 6 According to an embodiment of the present disclosure, a structural schematic diagram of a user call summary management device is provided.

[0054] Figure 7 According to an embodiment of the present disclosure, a schematic diagram of the structure of a computer-readable storage medium is provided. DETAILED DESCRIPTION

[0055] The following describes the implementation methods of the present application through specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in the present application. The present application can also be implemented or applied through other different specific implementation methods, and the details in the present application can also be modified or changed in various ways according to different viewpoints and application systems without departing from the spirit of the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0056] The following is a detailed description of the embodiments of the present application with reference to the accompanying drawings so that those skilled in the art can easily implement the present application. The present application can be embodied in a variety of different forms and is not limited to the embodiments described herein.

[0057] In the representations of the present application, the representations with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials or characteristics represented in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. Moreover, the represented specific features, structures, materials or characteristics may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples represented in the present application and the features of the different embodiments or examples, unless they contradict each other.

[0058] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of the features. In the representation of this application, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0059] In order to clearly describe the present application, components irrelevant to the description are omitted, and the same reference symbols are given to the same or similar components throughout the specification.

[0060] Throughout the specification, when a device is said to be "connected" to another device, this includes not only the case of "direct connection" but also the case of "indirect connection" by placing other elements therebetween. In addition, when a device is said to "include" a certain component, unless otherwise stated, it does not exclude other components, but means that other components may be included.

[0061] When a device is said to be "on" another device, it may be directly on the other device, but there may also be other devices between it. In contrast, when a device is said to be "directly" on another device, there are no other devices between it.

[0062] Although the terms first, second, etc. are used to represent various elements in some examples, these elements should not be limited by these terms. These terms are only used to distinguish one element from another element. For example, the first interface and the second interface, etc. are represented. Moreover, as used in this article, the singular forms "one", "one" and "the" are intended to also include plural forms, unless there is an opposite indication in the context. It should be further understood that the terms "comprising" and "including" indicate the existence of features, steps, operations, elements, components, projects, kinds, and / or groups, but do not exclude the existence, occurrence or addition of one or more other features, steps, operations, elements, components, projects, kinds, and / or groups. The terms "or" and "and / or" used herein are interpreted as inclusive, or mean any one or any combination. Therefore, "A, B or C" or "A, B and / or C" means "any of the following: A; B; C; A and B; A and C; B and C; A, B and C". Only when the combination of elements, functions, steps or operations is inherently mutually exclusive in some way, will there be an exception to this definition.

[0063] The technical terms used herein are only used to refer to specific embodiments and are not intended to limit the present application. The singular form used herein also includes the plural form unless the sentence clearly indicates the contrary meaning. The meaning of "including" used in the specification is to specify specific characteristics, regions, integers, steps, operations, elements and / or components, and does not exclude the existence or addition of other characteristics, regions, integers, steps, operations, elements and / or components.

[0064] Although not defined differently, all terms, including technical and scientific terms used herein, have the same meaning as those generally understood by those skilled in the art to which this application belongs. Terms defined in commonly used dictionaries are additionally interpreted as having meanings consistent with the relevant technical literature and the contents of the current disclosure, and unless defined, shall not be overly interpreted as ideal or very formal meanings.

[0065] In some embodiments of the present disclosure, Figure 1 According to a flowchart of a method for managing a user's incoming call summary, a method is provided. Figure 1 As shown, the user call summary management method includes:

[0066] Step 101: Generate a number of call summary labels based on historical user call summary data. It is understandable that in the process of summarizing user calls, event labels can be used for classification and summarization. In the embodiments provided in the present disclosure, the labels can be summarized in combination with the actual scenarios of user calls, which will not be described in detail here. In a specific implementation of the present disclosure, the call summary corresponding labels can be presented in a vectorized form, and the label vector Y label Ability to map optional 134 call summary categories to 134-dimensional one-hot encoding

[0067] Step 102: Obtain corresponding call content data and corresponding user feature data based on historical user call summary data. The call content data includes the call content corresponding to the user call, and the user feature data includes one or more combinations of order-related data, product-related data, user-related data, and answer-related data corresponding to the user call. It can be understood that in the embodiments provided by the present disclosure, the summary of the user's incoming call event does not entirely depend on the specific call content of the user's call, but also depends on the multi-dimensional feature data such as order information, product information, user information, etc. associated with the user's call. By integrating these feature data with the call content, the accuracy of the call summary can be significantly improved in the subsequent model training and automated management process.

[0068] Step 103: Generate a model training data set based on the incoming call content data, user feature data, and incoming call summary tags. The specific generation process of the model training data set will be described in detail later and will not be repeated here.

[0069] Step 104: Perform multiple iterations of training on the preset management model according to the model training data set until the recognition accuracy of the preset management model for the call summary label corresponding to the user's incoming call is greater than a preset threshold. The training of the user's call summary management model can use the recognition accuracy as an evaluation indicator. The specific iterative training process will be described in detail later and will not be repeated here.

[0070] Step 105: Acquire a real-time user incoming call event, and acquire a call summary tag corresponding to the real-time user incoming call event according to the preset management model obtained through training.

[0071] It is understandable that, through the above steps 101 to 105, it is possible to accurately and automatically identify summary tags for user incoming call events. The specific implementation of the above steps 101 to 105 will be further described below.

[0072] In the above embodiment, further, in the specific implementation of the aforementioned step 102, the user feature data can be configured to include one or more combinations of order information data, product information data, caller user data, connection operation data, and voice interaction operation data associated with the user's call. Among them, the order information data refers to the order information associated with the user's call event. Taking the travel order as an example, the order information data may include the order status, the departure time and / or departure place corresponding to the order, and the arrival time and / or arrival place corresponding to the order; the product information data is the product information associated with the user's call event, which may specifically include product quantity information and product service type and other contents; the caller user data corresponds to the associated user of the user's call event; the connection operation data includes the answer of the answering customer service to the user's call event and the historical call summary operation data of the answering customer service, which can provide a reference for the label identification of the current call event according to the historical call summary operation data of the answering customer service; the voice interaction operation data includes the interactive data of the user's operation according to the prompt voice in the user's call event, and can also provide a reference for the label identification of the current call event according to the voice operation path in the user's call event, which is not limited here. It can be understood that in the embodiments provided in the present disclosure, user characteristic data can be specifically configured according to the related matters of the actual user call event. For example, in the event of a user calling to inquire about travel order information, the user characteristic data can also include ticket number data associated with the trip, etc., which is not limited here.

[0073] In the above embodiment, further, in the specific implementation of the aforementioned step 103, in the process of forming a model training data set based on the incoming call content data, user feature data and incoming call summary labels, the multimodal feature data contents of the incoming call content data and user feature data can be converted into floating-point digital features to serve as input vectors of the preset management model during the training process; at the same time, the incoming call summary labels are converted into category vectors to serve as label data that the preset management model needs to apply during the training process. It can be understood that since the incoming call content data and user feature data contain a variety of different data types such as text and categories, in order to unify these multidimensional features and perform model training, these multidimensional features can be converted into floating-point digital features through data preprocessing. The above data preprocessing process will be further described and explained below. Specifically, Figure 2 FIG. 1 shows a flow chart of generating an input vector based on incoming call content data and user feature data. Figure 2 As shown, the following steps may be specifically included:

[0074] Step 201: Based on the call content data, the call content text is obtained based on data cleaning, and the call content text is formatted to form a standard call text format, and then turn to step 203. It can be understood that in the specific execution process of the above step 201, the text can be obtained by data cleaning based on the user's conversation recording, and the stop words, special symbols and other contents in the text that are not related to the specific content of the call are removed, and formatted to make it conform to the input format of subsequent model training; those skilled in the art can select a suitable preprocessing method according to actual needs, and it is not limited here.

[0075] Step 202: In the user feature data, the continuous variables in the user feature data are binned and / or discretized, and the non-numerical features in the user feature data are numerically processed, and the processing results are standardized and normalized to generate a feature matrix corresponding to the user feature data, and then turn to step 204. It can be understood that in the specific execution process of the above step 202, the user features can be classified first, the continuous variables therein are binned and / or discretized, and the data of non-numerical features such as categories are numerically processed, and then unified standardization and normalization are performed to obtain a feature matrix that meets the subsequent model training; those skilled in the art can also select a suitable preprocessing method according to actual needs, which is not limited here.

[0076] It is understandable that the above steps 201 and 202 may be performed synchronously or asynchronously during the specific execution process, which is not limited here.

[0077] Step 203: vectorize the standard incoming call text format to generate an embedded vector in a floating point format.

[0078] Step 204: vectorize the feature matrix to generate a feature vector in floating point format.

[0079] Step 205: The embedded vector and the feature vector are fused and concatenated to obtain an input vector.

[0080] It is understandable that the above steps 203 and 204 can also be performed synchronously or asynchronously during the specific execution process. e And the eigenvector X info In the case of , the input vector X required for subsequent training models can be constructed by fusion and splicing. in , the corresponding specific mathematical expression can be shown as follows:

[0081] X in =concat(X e ,X info ).

[0082] In the above embodiment, further, in the specific implementation of the aforementioned step 104, the preset management model can be configured as a Roberta-WWM-ext-large model. It can be understood that Roberta-WWM-ext-large is an enhanced version of the RoBERTa model (Robustly Optimized BERT Pretraining Approach), which adds the feature of Whole Word Masking (WWM) on the basis of the RoBERTa model, specifically referring to masking a word as a whole, rather than masking each letter in the word. This method can better maintain the semantic structure of the sentence and improve the model's ability to understand the overall concept, and is particularly suitable for the application field of the call summary label summary provided in the present disclosure.

[0083] In the above embodiment, specifically, during the training process of the preset management model, the management model parameters and hyperparameters can be initially preset, wherein the parameters of the RoBERTa model remain frozen, and the remaining newly added model layer parameters are randomly assigned; the relevant training rounds, learning rate and other hyperparameters can be initially configured and optimized by technical personnel in this field based on experience, and are not limited here; and then the training data is provided based on the scheme provided in the above embodiment, which mainly includes operations such as splitting the training set, the validation set and the test set; during the model training process, the script can be called to perform multiple rounds of model training based on the preset accuracy evaluation index; for the model that has completed several rounds of training, the model can be verified on the test set: if the label recognition accuracy corresponding to the test set meets the requirements, the model training is completed; otherwise, it is necessary to readjust the hyperparameters and repeat the above steps until the label recognition effect of the model meets the preset requirements. Exemplary, Figure 3a A schematic diagram showing the change of call summary label accuracy of the preset management model with the training rounds, wherein the horizontal axis represents the training rounds and the vertical axis represents the corresponding label recognition accuracy; Figure 3b A schematic diagram showing the change of model loss of the preset management model with the training rounds is shown, wherein the horizontal axis represents the training rounds and the vertical axis represents the corresponding model loss. Figure 3a and Figure 3b As shown, after 25 to 30 rounds of training, the Roberta-WWM-ext-large model provided by the present disclosure can obtain ideal label recognition accuracy and model loss.

[0084] In the above embodiment, specifically, Figure 4 A flowchart of obtaining a call summary label corresponding to a real-time user call event based on a preset management model obtained through training is provided. Figure 4 As shown, the following steps may be specifically included:

[0085] Step 401: based on the real-time user incoming call event, obtain the corresponding incoming call content text and user feature text, and then turn to step 402 and step 403 respectively.

[0086] Step 402 : The call content text passes through the pre-training layer and the attention pooling layer of the Roberta-WWM-ext-large model in sequence to form a first semantic vector, and then turns to step 404 .

[0087] Step 403: The user feature text passes through the embedded layer and the fully connected layer of the Roberta-WWM-ext-large model in sequence to form a second semantic vector, and then turns to step 404. The above steps 402 to 403 can be executed synchronously or asynchronously, which is not limited here.

[0088] Step 404: According to the first semantic vector and the second semantic vector, a linear output associated with the call summary label is obtained through the linear layer of the Roberta-WWM-ext-large model.

[0089] Step 405: convert the linear output into a probability distribution associated with the incoming call summary tag, and select the incoming call summary tag with the highest probability as the incoming call summary tag corresponding to the real-time user incoming call event.

[0090] In the above embodiment, since the classification task of the call summary is performed, the loss function corresponding to the preset management model Roberta-WWM-ext-large model can be configured as a cross entropy loss function.

[0091] In the above embodiment, further, considering that the content of the user's incoming call will change with the update of the order line or product line, the preset management model can be configured to be updated based on a preset period; or based on the label management evaluation result of the real-time user incoming call event, the preset management model is self-updated. Those skilled in the art can select a suitable model update configuration according to actual needs, which is not limited here.

[0092] In some embodiments of the present disclosure, Figure 5 FIG. 1 shows a schematic diagram of the structure of a user call summary management system, which is used to implement the user call summary management method provided in the above embodiment. Figure 5 As shown, the user call summary management system may specifically include:

[0093] The label acquisition unit 501 is used to generate a number of call summary labels according to historical user call summary data.

[0094] The data preprocessing unit 502 is used to obtain corresponding incoming call content data and corresponding user feature data according to the historical user incoming call summary data. The incoming call content data includes the call content corresponding to the user's incoming call, and the user feature data includes one or more combinations of order-related data, product-related data, user-related data, and answer-related data corresponding to the user's incoming call.

[0095] The data set construction unit 503 is used to form a model training data set according to the incoming call content data, user feature data and the incoming call summary label.

[0096] The model training unit 504 is used to perform multiple iterations of training on the preset management model according to the model training data set until the recognition accuracy of the preset management model for the call summary label corresponding to the user's incoming call is greater than a preset threshold.

[0097] The incoming call event management unit 505 is used to obtain real-time user incoming call events and obtain the incoming call summary tags corresponding to the real-time user incoming call events according to the preset management model obtained through training.

[0098] It can be understood that the functions performed by the above-mentioned tag acquisition unit 501 to the incoming call event management unit 505 are the same as the actions performed by the aforementioned steps 101 to 105, and are not repeated here.

[0099] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods or program products. Therefore, various aspects of the present disclosure may be specifically implemented in the following forms, namely: complete hardware implementation, complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, which may be collectively referred to herein as "circuits", "modules" or "platforms".

[0100] Specifically, Figure 6 According to an embodiment of the present disclosure, a schematic diagram of the structure of a user call summary management device is shown. Figure 6 The electronic device 600 according to this embodiment of the present disclosure is described. Figure 6 The electronic device 600 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0101] like Figure 6 As shown, the electronic device 600 is in the form of a general computing device. The components of the electronic device 600 may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including the storage unit 620 and the processing unit 610), a display unit 640, etc.

[0102] The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 performs the steps according to various exemplary embodiments of the present disclosure. For example, the processing unit 610 can perform the following steps: Figure 1 The relevant steps of the user call summary management method shown in .

[0103] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access memory (RAM) 6201 and / or a cache storage unit 6202 , and may further include a read-only storage unit (ROM) 6203 .

[0104] The storage unit 620 may also include a program / utility 6204 having a set (at least one) of program modules 6205, such program modules 6205 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.

[0105] Bus 630 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.

[0106] The electronic device 600 may also communicate with one or more external devices 700 (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 600, and / or communicate with any device that enables the electronic device 600 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 650. Furthermore, the electronic device 600 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via a network adapter 660. The network adapter 660 may communicate with other modules of the electronic device 600 via a bus 630. 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 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0107] The embodiment of the present disclosure also provides a computer-readable storage medium for storing a program, and the steps of the method for managing user call summary are implemented when the program is executed. In some possible implementations, various aspects of the present disclosure can also be implemented in the form of a program product, which includes a program code. When the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary implementations of the present disclosure described in the above-mentioned method for managing user call summary of this specification.

[0108] Specifically, Figure 7 According to an embodiment of the present disclosure, a schematic diagram of the structure of a computer-readable storage medium is shown. Figure 7As shown, a program product 800 for implementing the above-mentioned user call summary method according to an embodiment of the present disclosure is described, which can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited thereto, and 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.

[0109] The program product may employ 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 thereof. More specific examples (a 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 thereof.

[0110] Computer readable storage media may include data signals propagated in baseband or as part of a carrier wave, in which readable program codes are 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. The readable storage medium 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, device, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0111] The program code for executing the specific implementation operations of the user call summary method provided by the aforementioned embodiment of the present disclosure can be written in any combination of one or more programming languages, and the programming language includes an object-oriented programming language, such as Java, C++, etc., and also includes a conventional procedural programming language, such as "C" language or a similar programming language. The program code can be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on the remote computing device, or completely on the remote computing device or server. In the case of a remote computing device, the remote computing device can 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 can be connected to an external computing device (for example, using an Internet service provider to connect through the Internet).

[0112] To sum up, through the technical solution provided by the present invention, it is possible to automatically generate and manage call summary labels based on the call content data and other related user feature data in the user's calls and based on the pre-trained fusion feature model, thereby improving the accuracy of user call summaries, reducing manual processing costs, and improving the efficiency of user call summaries, which has replicable value.

[0113] The above content is a further detailed description of the present disclosure in combination with specific preferred implementation modes, and it cannot be determined that the specific implementation of the present disclosure is limited to these descriptions. For ordinary technicians in the technical field to which the present disclosure belongs, several simple deductions or substitutions can be made without departing from the concept of the present disclosure, which should be regarded as falling within the scope of protection of the present disclosure.

Claims

1. A method for managing user call summary, It is characterized in that The steps include: Generate several call summary labels based on historical user call summary data; Acquire corresponding incoming call content data and corresponding user feature data according to the historical user incoming call summary data, wherein the incoming call content data includes the conversation content corresponding to the user's incoming call, and the user feature data includes one or more combinations of order-related data, product-related data, user-related data, and answer-related data corresponding to the user's incoming call; Forming a model training data set according to the incoming call content data, the user feature data and the incoming call summary label; Performing multiple iterations of training on the preset management model according to the model training data set until the recognition accuracy of the preset management model for the call summary label corresponding to the user's incoming call is greater than a preset threshold; A real-time user incoming call event is obtained, and the incoming call summary tag corresponding to the real-time user incoming call event is obtained according to the preset management model obtained through training.

2. The method for managing user incoming call summaries according to claim 1, It is characterized in that The user characteristic data includes one or more combinations of order information data, product information data, caller user data, connection operation data, and voice interaction operation data associated with the user's incoming call; The order information data includes the order status, the departure time and / or the departure place corresponding to the order, and the arrival time and / or the arrival place corresponding to the order; The product information data includes product quantity information and product service type; The incoming call user data includes user gender, user age, user type, and the incoming call summary label corresponding to the user's historical incoming calls.

3. The method for managing user incoming call summaries according to claim 1, It is characterized in that The forming of a model training data set according to the incoming call content data, the user feature data and the incoming call summary label comprises: Converting the incoming call content data and the user feature data into floating-point digital features to serve as input vectors of the preset management model; The call summary label is converted into a category vector to serve as label data of the preset management model.

4. The method for managing user incoming call summaries as claimed in claim 3, It is characterized in that Generating the input vector according to the incoming call content data and the user feature data comprises the following steps: According to the incoming call content data, based on data cleaning, to obtain the incoming call content text, and format the incoming call content text to form a standard incoming call text format; In the user feature data, performing binning and / or discretization processing on continuous variables in the user feature data, performing numerical processing on non-numerical features in the user feature data, and performing standardization and normalization processing on the processing results to generate a feature matrix corresponding to the user feature data; Performing vectorization processing on the standard incoming call text format to generate an embedded vector in a floating point format; Performing vectorization processing on the feature matrix to generate a feature vector in a floating point format; The embedded vector and the feature vector are fused and concatenated to obtain the input vector.

5. The method for managing user incoming call summaries according to claim 1, It is characterized in that The preset management model is the Roberta-WWM-ext-large model; The step of acquiring a real-time user incoming call event and acquiring the incoming call summary tag corresponding to the real-time user incoming call event according to the preset management model obtained through training includes the following steps: Acquire corresponding incoming call content text and user feature text based on the real-time user incoming call event; The incoming call content text is sequentially passed through the pre-training layer and the attention pooling layer of the Roberta-WWM-ext-large model to form a first semantic vector; The user feature text passes through the embedded layer and the fully connected layer of the Roberta-WWM-ext-large model in sequence to form a second semantic vector; According to the first semantic vector and the second semantic vector, obtaining a linear output associated with the call summary label through a linear layer of the Roberta-WWM-ext-large model; The linear output is converted into a probability distribution associated with the incoming call summary label, and the incoming call summary label with the highest probability is selected as the incoming call summary label corresponding to the real-time user incoming call event.

6. The method for managing user incoming call summaries as claimed in claim 5, It is characterized in that The loss function corresponding to the preset management model is configured as a cross entropy loss function.

7. The method for managing user incoming call summaries according to claim 1, It is characterized in that The preset management model is updated based on a preset period; or The preset management model self-updates the preset management model based on the label management evaluation result of the real-time user incoming call event.

8. A user call summary management system, used to implement the user call summary management method according to any one of claims 1 to 7, It is characterized in that include: A label acquisition unit, used to generate a number of call summary labels according to historical user call summary data; A data preprocessing unit, configured to obtain corresponding incoming call content data and corresponding user feature data according to the historical user incoming call summary data, wherein the incoming call content data includes the conversation content corresponding to the user's incoming call, and the user feature data includes one or more combinations of order-related data, product-related data, user-related data, and answer-related data corresponding to the user's incoming call; A data set construction unit, configured to form a model training data set according to the incoming call content data, the user feature data and the incoming call summary label; A model training unit, configured to perform multiple iterations of training on a preset management model according to the model training data set, until the recognition accuracy of the preset management model for the call summary label corresponding to the user's incoming call is greater than a preset threshold; The incoming call event management unit is used to obtain real-time user incoming call events and obtain the incoming call summary label corresponding to the real-time user incoming call event according to the preset management model obtained through training.

9. A user call summary management device, It is characterized in that include: processor; a memory storing executable instructions of the processor; Wherein, the processor is configured to execute the steps of the user call summary management method as described in any one of claims 1 to claim 7 by executing the executable instructions.

10. A computer-readable storage medium for storing a program, It is characterized in that When the program is executed by a processor, the steps of the user call summary management method described in any one of claims 1 to 7 are implemented.