Outbound call processing methods, devices, and computer program products

Abstract

This application discloses a method, apparatus, and computer program product for processing outbound call services. Relating to the field of artificial intelligence, the method includes: executing an outbound call task to a user; processing the target service according to a preset service processing procedure; collecting the user's voice signal during the target service processing; extracting voice features and voice content from the voice signal; determining the current service processing node corresponding to the voice content in the preset service processing procedure; determining the service path map corresponding to the preset service processing procedure; determining a path deviation index value based on the current service processing node and the service path map; inputting the voice features into a target model to obtain the user's behavior recognition result; determining a risk index value for the user interrupting the target service based on the behavior recognition result and the path deviation index value; and adjusting the preset service processing procedure according to the risk index value. This application solves the problem of poor accuracy in recognizing user intent in outbound call services in related technologies.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence, and more specifically, to a method, apparatus, and computer program product for processing outbound call services. Background Technology

[0002] Existing intelligent outbound calling systems automate the process through technologies such as data import, automatic dialing, voice recognition interaction, pre-set script responses, and data recording and analysis. However, these systems often only engage in mechanical dialogue, lacking the ability to communicate emotionally and accurately understand and respond to customers' emotional needs, resulting in a poor customer experience. Due to the lack of emotional recognition capabilities, existing systems struggle to provide personalized service plans based on customers' emotional states, limiting further improvements in service quality. Because of these issues, some customers are skeptical of intelligent outbound calling systems and even resist them, impacting their market acceptance.

[0003] There is currently no effective solution to the problem of poor accuracy in recognizing user intent in outbound calling services in related technologies. Summary of the Invention

[0004] The main objective of this application is to provide a method, apparatus, and computer program product for processing outbound call services, so as to solve the problem of poor accuracy in recognizing user intent in outbound call services in related technologies.

[0005] To achieve the above objectives, according to one aspect of this application, a method for processing outbound call services is provided. The method includes: executing an outbound call task to a user; upon detecting user authorization for a target service, processing the target service according to a preset service processing procedure; collecting the user's voice signal during the target service processing; extracting voice features and voice content from the voice signal, wherein the target service is a voice-based service; determining the current service processing node corresponding to the voice content in the preset service processing procedure; determining a service path graph corresponding to the preset service processing procedure; determining a path deviation index value based on the current service processing node and the service path graph, wherein the service path graph contains multiple service processing nodes and edges between the multiple service processing nodes, the edges representing the service processing order; inputting the voice features into a target model to obtain the user's behavior recognition result, wherein the behavior recognition result represents the probability that the user will continue voice interaction; determining a risk index value for the user interrupting the target service based on the behavior recognition result and the path deviation index value; and adjusting the preset service processing procedure according to the risk index value.

[0006] Optionally, extracting speech features from a speech signal includes: extracting fundamental frequency jitter features from the speech signal using Mel-frequency cepstral coefficients; determining the user's speech rate and pause duration based on the fundamental frequency jitter features; determining the spectrum of the speech signal; determining the noise intensity of the speech signal based on the spectrum; and determining at least one of the speech rate, pause duration, and noise intensity as a speech feature.

[0007] Optionally, determining the path deviation index value based on the current business processing node and the business path graph includes: inputting the current business processing node and the business path graph into a graph convolutional network to obtain the path deviation index value. The graph convolutional network preprocesses the business path graph and the current business processing node to obtain an adjacency matrix and a node feature matrix. The activation function of the graph convolutional network takes the adjacency matrix and the node feature matrix as input and outputs the path deviation index value.

[0008] Optionally, determining the risk indicator value for user interruption of target service based on behavior recognition results and path deviation index values ​​includes: inputting behavior recognition results and path deviation index values ​​into a gating weight matrix to obtain an output value, wherein the gating weight matrix is ​​used to determine the contribution ratio of behavior recognition results and path deviation index values ​​in the risk indicator value; calculating the sum of the output value and a preset offset, and inputting the sum into the target activation function to obtain the risk indicator value.

[0009] Optionally, adjusting the preset business processing procedure based on the risk indicator value includes: when the risk indicator value is greater than or equal to the indicator value threshold, determining the initial dialogue script content and initial dialogue process corresponding to the preset business processing procedure; extracting a summary of the initial dialogue script content using a text processing tool, and generating updated dialogue script content based on the summary; removing processes from the initial dialogue process whose correlation index value with the target business is lower than the correlation threshold, and obtaining the updated dialogue process; and processing the target business based on the updated dialogue script content and the updated dialogue process.

[0010] Optionally, adjusting the preset business processing procedure according to the preset strategy includes: when the risk indicator value is greater than or equal to the indicator value threshold and the noise intensity in the voice feature is greater than or equal to the intensity threshold, processing the voice signal through a preset noise reduction algorithm to obtain an updated voice signal; and performing the steps of extracting voice features and voice content from the voice signal based on the updated voice signal.

[0011] Optionally, adjusting the preset business processing procedure according to the preset strategy includes: processing the target business through preset guidance content when the risk indicator value is greater than or equal to the indicator value threshold and the pause duration in the voice feature is greater than or equal to the duration threshold, or when the risk indicator value is greater than or equal to the indicator value threshold and the path deviation indicator value is greater than or equal to the path deviation threshold.

[0012] Optionally, the target model includes an input layer, a convolutional layer, a fully connected layer, and an output layer. The target model outputs the behavior recognition result in the following way: it receives speech features through the input layer, extracts temporal features from the speech features of the input layer through the convolutional layer, maps the temporal features to the probability of the user continuing to engage in voice interaction through the fully connected layer, determines the probability as the behavior recognition result, and outputs the behavior recognition result through the output layer.

[0013] To achieve the above objectives, according to another aspect of this application, an outbound call service processing apparatus is provided. The apparatus includes: an execution unit, configured to execute an outbound call task to a user, and upon detecting user authorization for a target service, process the target service according to a preset service processing procedure, collect the user's voice signal during the target service processing, and extract voice features and voice content from the voice signal, wherein the target service is a voice-based service; a first determining unit, configured to determine the current service processing node corresponding to the voice content in the preset service processing procedure, determine a service path graph corresponding to the preset service processing procedure, and determine a path deviation index value based on the current service processing node and the service path graph, wherein the service path graph includes multiple service processing nodes and edges between the multiple service processing nodes, the edges representing the service processing order; an input unit, configured to input the voice features into a target model to obtain a user behavior recognition result, wherein the behavior recognition result represents the probability that the user will continue voice interaction; and an adjustment unit, configured to determine a risk index value for the user interrupting the target service based on the behavior recognition result and the path deviation index value, and adjust the preset service processing procedure according to the risk index value.

[0014] In this embodiment, an outbound call task to a user is performed. Upon detecting user authorization for a target service, the target service is processed according to a preset service processing procedure. The user's voice signal is collected during the service processing, and voice features and content are extracted from the voice signal. The target service is a voice-based service. The current service processing node corresponding to the voice content in the preset service processing procedure is determined, and a service path graph corresponding to the preset service processing procedure is determined. A path deviation index value is determined based on the current service processing node and the service path graph. The service path graph contains multiple service processing nodes and edges between them, with edges representing the service processing order. Voice features are input into the target model to obtain user behavior recognition results, which represent the probability that the user will continue voice interaction. Based on the behavior recognition results and path deviation index values, a risk index value for user interruption of target services is determined. The preset service handling process is adjusted according to the risk index value. By introducing real-time behavior recognition and path deviation analysis, combined with the dynamic calculation of risk index values, intelligent optimization of outbound call services is achieved. This improves the responsiveness to users' personalized needs, reduces the risk of service interruption, and thus improves the accuracy of outbound call services in recognizing user intent, thereby solving the technical problem of poor accuracy in recognizing user intent in outbound call services. Attached Figure Description

[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0016] Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing a processing method for outbound call services is shown.

[0017] Figure 2 This is a flowchart of a method for processing outbound call services according to an embodiment of this application;

[0018] Figure 3 This is a schematic diagram of an outbound call service processing device provided according to an embodiment of this application;

[0019] Figure 4 This is a structural block diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0021] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] It should be noted that the information collected in this application (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of this data all comply with relevant laws, regulations, and standards, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse. For example, interfaces are set up between this system and relevant users or organizations, providing users with corresponding access points to choose to agree to or refuse automated decision-making results; if the user chooses to refuse, the process proceeds to the expert decision-making stage.

[0023] Example 1

[0024] According to an embodiment of this application, a method embodiment for processing outbound call services is also provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0025] The method embodiment provided in Embodiment 1 of this application can be executed on a mobile terminal, computer terminal, or similar computing device. Figure 1A hardware block diagram of a computer terminal (or mobile device) for implementing outbound call processing is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, processing devices such as MCU (Microcontroller Unit) or FPGA (Field-Programmable Gate Array), memory 104 for storing data, and transmission device 106 for communication functions. In addition, it may also include: a display, input / output interfaces (I / O interfaces), a USB (Universal Serial Bus) port (which may be included as one of the ports of a BUS (Business Bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0026] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0027] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the outbound call service processing method in this embodiment of the application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the above-mentioned outbound call service processing method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of the above-mentioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0028] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0029] The display may be, for example, a touchscreen LCD display that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0030] Under the above operating environment, this application provides a method for processing outbound call services. Figure 2 This is a flowchart of a method for processing outbound call services according to an embodiment of this application, such as... Figure 2 As shown, the method includes:

[0031] Step S201: Execute an outbound call to the user. If the user is authorized to the target service, the target service is processed according to the preset service processing procedure. The user's voice signal is collected during the processing of the target service, and voice features and voice content are extracted from the voice signal. The target service is a voice-based service.

[0032] In step S201, the financial institution's intelligent outbound calling system executes outbound calls to users, efficiently and securely processing target transactions through voice interaction. Target transactions can be financial institution services, such as account opening or fund trading. At the initial stage of the call, the system uses voice recognition technology to inquire about and confirm the user's authorization for the target transaction, ensuring the legality of the transaction and protecting the customer's privacy and rights. Once user authorization is confirmed, the system begins processing the target transaction according to the preset transaction processing procedure.

[0033] During business processing, the intelligent outbound calling system continuously collects users' voice signals. These signals contain rich information, not only the content of the user's speech but also voice features such as tone, speed, and pauses. These features are crucial for understanding the user's true intentions. Using speech recognition and analysis technology, the system accurately extracts the user's voice content and features from the voice signals. Voice content includes, for example, the user's actual questions or needs, which is the primary basis for business processing. Voice features include, for example, rising intonation may indicate curiosity or excitement; increased speed may show a sense of urgency or anxiety; pauses and weakened tone may indicate uncertainty or hesitation about certain topics. These features help the system gain a deeper understanding of the user's emotional state, providing a basis for subsequent interaction strategies.

[0034] Step S202: Determine the current business processing node corresponding to the voice content in the preset business processing process, determine the business path diagram corresponding to the preset business processing process, and determine the path deviation index value based on the current business processing node and the business path diagram. The business path diagram contains multiple business processing nodes and edges between multiple business processing nodes, and the edges are used to represent the business processing order.

[0035] In step S202, the purpose of voice content analysis is to understand the user's specific needs and what they say. The analyzed voice content is compared with the preset business processing flow to determine which specific business processing node the user is currently at. For example, assuming the target business is a credit card application, the preset business processing flow includes steps such as understanding the user's basic information, verifying the user's identity, explaining the business terms, and collecting necessary application materials. Then, the system will determine whether the user is providing personal information, confirming their identity, or expressing doubts about a certain term based on their current speech, thereby locating the specific business node such as "basic information entry," "identity verification," or "term explanation."

[0036] A business path diagram is a visual flowchart that depicts the logical sequence of the entire business process from start to finish and the connections between each node. The path diagram includes multiple business processing nodes as well as the "edges" between these nodes; these "edges" represent the order of business processing and transition conditions. Using a business path diagram to guide the intelligent outbound calling process ensures that business is processed in the correct order and allows for flexible navigation to the next node based on user behavior and choices. For example, in a credit card application business path diagram, there is a direct connection between the "Basic Information Entry" node and the "Identity Verification" node, indicating that identity verification should proceed immediately after basic information entry is completed. However, if a user raises additional requests or questions at a certain node, the system needs to be able to identify this deviation and make appropriate process adjustments. The path deviation metric measures the degree of difference between the user's actual behavior and the preset business process. It is calculated by comparing the user's current business node with the node they should reach according to the normal process. A higher deviation metric value means a more serious deviation from the preset process, potentially leading to business processing disruptions and a poor customer experience.

[0037] Step S203: Input the voice features into the target model to obtain the user's behavior recognition result, wherein the behavior recognition result is used to characterize the probability that the user will continue to engage in voice interaction.

[0038] In step S203, the target model can be a deep learning model, such as a convolutional neural network, a long short-term memory network, or a temporal convolutional network. During the interaction between the intelligent outbound calling system and the user, the system continuously captures and analyzes the user's voice features. These voice features are input into the target model, which uses its internal algorithms and parameters to perform deep learning, analyzing the behavioral patterns behind these features. Subsequently, a numerical value or probability, i.e., the user behavior recognition result, is output to characterize the user's willingness or probability to continue voice interaction.

[0039] For example, if the model detects that a user's speech speed increases and their tone becomes more hurried, it may indicate that the user is eager for an answer or to complete an action. In this case, the behavior recognition result tends to give a higher probability of continuing the interaction. Conversely, if the user's speech speed slows down, pauses more frequently, and their voice contains a tone of doubt or dissatisfaction, the behavior recognition result will predict a lower probability of continuing the interaction, prompting the system to take appropriate measures, such as providing more explanation, offering reassurance, or directly transferring the user to human customer service.

[0040] Step S204: Determine the risk index value of user interruption of target service based on behavior recognition results and path deviation index value, and adjust the preset service handling process according to the risk index value.

[0041] In step S204, the risk indicator value is calculated by combining the user's behavior recognition results and path deviation indicator value, aiming to assess the likelihood of the user interrupting the target service. The behavior recognition results reflect the user's willingness and probability of continuing voice interaction, while the path deviation indicator value measures the gap between the user's actual actions and the preset process. For example, if the behavior recognition results show a low probability of the user continuing interaction, while the path deviation indicator value is high, it indicates that the user may be confused or dissatisfied with the current process, posing a high risk of interruption. The preset service processing flow is adjusted based on the risk indicator value. This adjustment aims to reduce the risk of interruption, ensure smooth service completion, and improve user satisfaction. Adjustment strategies may include: simplifying the process: If the user shows high levels of confusion or dissatisfaction, the system can skip some unnecessary steps or stages and directly proceed to key service nodes to reduce user waiting time and avoid further frustration or misunderstanding. Providing additional information: When the user lacks sufficient understanding or has doubts about a certain service node, the system can automatically provide supplementary information or guidance to help the user better understand the current situation and reduce the possibility of interruption. When the risk indicator value reaches a certain critical point, the system will automatically transfer the call to a human customer service representative for intervention. This is especially suitable for complex problem-solving scenarios, providing a higher level of personalized service and effectively mitigating the risk of interruption.

[0042] The outbound call processing method provided in this application embodiment executes an outbound call task to a user. Upon detecting user authorization for a target service, it processes the target service according to a preset service processing flow. The method collects the user's voice signal during the target service processing, extracts voice features and voice content from the voice signal, where the target service is a voice-based service. It determines the current service processing node corresponding to the voice content in the preset service processing flow, determines the service path graph corresponding to the preset service processing flow, and determines a path deviation index value based on the current service processing node and the service path graph. The service path graph contains multiple service processing nodes and edges between these nodes, with the edges representing the service path. The system optimizes outbound calling services by introducing real-time behavior recognition and path deviation analysis, combined with dynamic calculation of risk indicators. This improves the responsiveness to personalized user needs, reduces the risk of service interruption, and enhances the accuracy of outbound calling services in recognizing user intent. This addresses the technical problem of poor accuracy in recognizing user intent in outbound calling services.

[0043] Speech features may include speech rate, pause duration, and noise intensity. Optionally, in the outbound call service processing method provided in this application embodiment, extracting speech features from the speech signal includes: extracting fundamental frequency jitter features from the speech signal using Mel-frequency cepstral coefficients; determining the user's speech rate and pause duration based on the fundamental frequency jitter features; determining the spectrum of the speech signal; determining the noise intensity of the speech signal based on the spectrum; and determining at least one of the speech rate, pause duration, and noise intensity as a speech feature.

[0044] In some embodiments, the fundamental frequency jitter of each 40ms frame of speech signal is captured in real time using Mel-frequency cepstral coefficients. This data is analyzed and tracked to track the rhythm of the customer's conversation and ambient sound characteristics. Specifically, when the conversation rhythm shows increased speech rate, ambient sound features include traffic noise, and semantic keywords such as "simply put," it can be determined that a more concise explanation and other corresponding business processing, such as sending a text message, are needed. By analyzing the frequency change rate in the fundamental frequency jitter characteristics, the system can estimate the user's speech rate. Changes in speech rate can reflect the user's state; for example, increased speech rate may indicate tension or anxiety, while decreased speech rate may suggest impatience or a state of deep thought. Long periods of silence in the speech signal are identified as pause durations. Measuring pause duration helps understand the user's reaction time; longer pauses may indicate difficulty in understanding the question or hesitation about a topic.

[0045] In addition to the features mentioned above, the system also assesses environmental noise intensity by determining the spectrum of the speech signal. The level of noise intensity indirectly reflects the noise level of the call environment, thus affecting the user's call experience and interaction efficiency. For example, frequent background noise may indicate that the user is in a noisy environment, which can interfere with the user's ability to hear information and reduce the clarity of the conversation. Integrating these three features—speech rate, pause duration, and noise intensity—constitutes a comprehensive speech feature set that can comprehensively reflect the user's current communication state and the influence of the external environment. The system inputs these features into a target model (such as a deep neural network), and the model, based on trained weights and thresholds, outputs the user's willingness and probability to continue voice interaction, i.e., the behavior recognition result.

[0046] This embodiment utilizes voice feature extraction and analysis to enable the intelligent outbound calling system to provide more intelligent and user-friendly services, effectively improving business completion rates and customer satisfaction.

[0047] The path deviation index value can be generated based on a graph convolutional network. Optionally, in the outbound call service processing method provided in this application embodiment, determining the path deviation index value based on the current service processing node and the service path graph includes: inputting the current service processing node and the service path graph into a graph convolutional network to obtain the path deviation index value. The graph convolutional network preprocesses the service path graph and the current service processing node to obtain an adjacency matrix and a node feature matrix. The activation function of the graph convolutional network takes the adjacency matrix and the node feature matrix as input and outputs the path deviation index value.

[0048] In some embodiments, the adjacency matrix describes the connections between nodes in the business path graph. Each row and column corresponds to a business node, and the elements in the matrix represent the strength or presence of connections between nodes. The node feature matrix contains specific attributes of each business node, representing the dynamic state information of each node, composed of real-time customer behavior and historical statistics. The graph convolutional network first performs feature propagation based on the adjacency matrix and the node feature matrix. Through layer-by-layer iteration, the information of each node can flow along the edges of the path graph to adjacent nodes, a process similar to message passing on a graph. After feature propagation, the graph convolutional network aggregates the features of each node to form a higher-level node representation. By aggregating the features of different nodes, the graph convolutional network can capture global business process information and local path deviations.

[0049] The output layer of the graph convolutional network calculates a path deviation metric based on the aggregated node representations. Specifically, the model compares the expected and actual positions of the current business processing node with the changes in the relationship between that node and other nodes to quantify the degree of deviation of user behavior from the preset process. This metric reflects the consistency between user behavior and the business process and is used for subsequent risk assessment and process optimization.

[0050] For example, the path deviation index value is calculated using the following formula:

[0051] ;

[0052] in, It is the path deviation index value. : Adjacency matrix with self-loops, where A is a business path graph, representing the compliant jump relationship between business nodes (such as "identity verification" and "product introduction"), which is predefined by business rules. yes diagonal matrix, The diagram shows the node feature matrix of the l-th layer of the convolutional network. The dynamic state information of each node consists of real-time customer behavior and historical statistics. : Trainable weight matrix. σ: ReLU activation function.

[0053] The graph convolutional network in this embodiment exhibits unique advantages in calculating path deviation index values. It can evaluate the degree of deviation between user behavior and preset processes in real time through in-depth analysis of the business path graph, thereby assisting the intelligent outbound calling system to make more accurate decisions, optimize business processes, and improve customer satisfaction.

[0054] Risk indicator values ​​can be determined based on behavior recognition results and path deviation indicator values. Optionally, in the outbound call service processing method provided in this application embodiment, determining the risk indicator value of user interruption of target service based on behavior recognition results and path deviation indicator values ​​includes: inputting behavior recognition results and path deviation indicator values ​​into a gating weight matrix to obtain an output value, wherein the gating weight matrix is ​​used to determine the contribution ratio of behavior recognition results and path deviation indicator values ​​in the risk indicator value; calculating the sum of the output value and a preset offset, and inputting the sum into a target activation function to obtain the risk indicator value.

[0055] In some embodiments, the gating weight matrix acts as a regulator, determining the contribution ratio of behavior recognition results and path deviation indicators when calculating the risk index value for user interruption of target services. This allows the system to dynamically adjust the weights of these two indicators based on actual conditions, thus more accurately reflecting the degree of interruption risk. The system first inputs the behavior recognition results and path deviation indicators into the gating weight matrix. After processing by the gating weight matrix, the two input features are weighted according to their respective importance, generating an output value that integrates user interaction intention and process deviation risk. The preset offset is a static value that reflects the system's basic assessment of interruption risk without any input features. Adding the preset offset to the output value aims to adjust the baseline of the calculation results, ensuring that the risk index value fluctuates within a reasonable range, taking into account the inherent risks of the business.

[0056] The target activation function is used here to map the sum of the output value and a preset offset to a specific output range, such as between 0 and 1. The activation function can be Sigmoid, Tanh, or ReLU. After passing the sum of the output value and the preset offset through the target activation function, the system obtains a risk indicator value between 0 and 1. This value intuitively represents the probability that the user will interrupt the target service; the closer it is to 1, the higher the risk of interruption.

[0057] For example, the formula for calculating the risk indicator value is as follows:

[0058] ;

[0059] Where σ is the sigmoid function, which compresses the output to a probability range. It is a gating weight matrix. High gating controls the behavior recognition result and are suitable for detecting local temporal anomalies. Low gating values ​​control the path deviation index value and are suitable for detecting path violations. It is a preset offset that controls the initial offset of the risk indicator value. If it is negative, the model tends to output low risk initially.

[0060] This embodiment enhances the system's flexibility and responsiveness by calculating and evaluating risk indicators of user interruption of target services in real time, thereby improving service quality and efficiency, reducing unnecessary business interruptions, and enhancing the overall user experience and satisfaction.

[0061] Adjusting the preset business processing flow may include adjusting the dialogue script content and the dialogue flow. Optionally, in the outbound call processing method provided in this application embodiment, adjusting the preset business processing flow according to the risk indicator value includes: when the risk indicator value is greater than or equal to the indicator value threshold, determining the initial dialogue script content and the initial dialogue flow corresponding to the preset business processing flow; extracting a summary of the initial dialogue script content using a text processing tool, and generating updated dialogue script content based on the summary; removing processes from the initial dialogue flow whose correlation index value with the target business is lower than the correlation threshold, to obtain the updated dialogue flow; and processing the target business based on the updated dialogue script content and the updated dialogue flow.

[0062] In some embodiments, the initial dialogue script includes standard greetings, explanatory text, questions, and preset response options. The initial dialogue flow describes the sequence of steps required to complete the target business. The initial dialogue script content is summarized using a text processing tool to condense the lengthy script into more concise and easily understood core information. During the summary extraction process, the text processing tool identifies key sentences and removes repetitive or irrelevant information fragments, retaining the core logic and business points of the dialogue. The updated dialogue script content generated based on the summary significantly reduces script complexity while maintaining information integrity, helping to improve user comprehension and acceptance, and reducing user patience loss due to excessive script length.

[0063] Simultaneously, the initial dialogue flow is reviewed, removing nodes or steps whose relevance to the target business is below a certain threshold. The relevance metric measures the contribution of each process node to advancing core business objectives. If some nodes have excessively low metric values, it means they are not essential for completing the target business, or are not the optimal path for the user in the current context. After this removal, the updated dialogue flow is more focused on the key steps of the target business, avoiding unnecessary detours or information overload, further improving process efficiency and user satisfaction.

[0064] For example, if the original dialogue script contains a lot of product introductions and promotional content, but the risk indicator value shows that the user may be tired or impatient, the system will automatically remove this part of the content and only retain information related to specific business operations, such as account verification and transaction confirmation steps. At the same time, the system will simplify the expression and use more straightforward and friendly language to communicate with the user.

[0065] This embodiment effectively reduces interruption rates and improves business completion and customer experience by rapidly adjusting its dialogue strategy and business processes when high-risk situations are detected to better adapt to the user's current needs and status.

[0066] Adjusting the preset business processing procedure may also include noise reduction processing of the voice signal. Optionally, in the outbound call processing method provided in the embodiments of this application, adjusting the preset business processing procedure according to the preset strategy includes: when the risk index value is greater than or equal to the index value threshold and the noise intensity in the voice feature is greater than or equal to the intensity threshold, processing the voice signal through a preset noise reduction algorithm to obtain an updated voice signal; and performing the step of extracting voice features and voice content from the voice signal based on the updated voice signal.

[0067] In some embodiments, the purpose of the denoising algorithm is to filter out or reduce background noise in the speech signal, enhance speech clarity, and thus improve the accuracy of speech recognition and sentiment analysis. Denoising algorithms include, but are not limited to: spectral subtraction: recovering a clean speech signal by estimating the noise spectrum and subtracting it from the spectrum of the original speech signal; and wavelet transform: utilizing the multi-resolution characteristics of wavelet analysis to identify and suppress high-frequency noise while preserving low-frequency speech components. The choice of denoising algorithm can be determined based on the type of noise (e.g., steady-state noise, transient noise) and scenario requirements (real-time performance, resource consumption). Once the denoising algorithm has finished processing, the original speech signal will be updated to obtain a clearer version of the speech signal. Subsequently, the system needs to re-execute the speech feature and speech content extraction steps based on this updated speech signal.

[0068] This embodiment, by performing noise reduction processing on voice signals, can flexibly optimize business processes and interaction strategies when facing complex call conditions and high-risk scenarios, thereby improving service quality and user satisfaction.

[0069] Adjusting the preset business processing procedure may also include modifying the business processing path. Optionally, in the outbound call business processing method provided in the embodiments of this application, adjusting the preset business processing procedure according to the preset strategy includes: when the risk indicator value is greater than or equal to the indicator value threshold and the pause duration in the voice feature is greater than or equal to the duration threshold, or when the risk indicator value is greater than or equal to the indicator value threshold and the path deviation indicator value is greater than or equal to the path deviation threshold, the target business is processed through preset guidance content.

[0070] In some embodiments, when a risk indicator value reaches or exceeds a preset threshold, and voice feature analysis indicates that the user's pause duration also exceeds a threshold, it suggests that the user may be confused or impatient with the current conversation, or exhibiting hesitation or a negative emotional response. The conversation is guided according to a preset script, which typically includes direct answers to user concerns, concise guidance on business processes, or provision of additional service options. For example, if a user lingers too long at a particular business step, the system may automatically provide a simplified explanation of the process or ask the user if they wish to skip the current step and proceed directly to the part they are more interested in.

[0071] On the other hand, when the risk indicator value also reaches or exceeds the threshold, and the path deviation indicator value also exceeds the path deviation threshold, it indicates that the user has deviated from the normal business path during the business process, possibly due to information misunderstanding, operational errors, or dissatisfaction with the existing process. The process correction mechanism should be activated immediately to analyze the user's current position and the reasons for deviation, dynamically adjust the business process diagram, and replan a path that better suits the user's current state and needs. For example, if a user gets lost in a complex menu hierarchy, the system will simplify the menu structure and guide the user back to the key service node. Next, the system will use a preset re-guidance strategy, providing clear instructions or prompts through the updated business path diagram to help the user return to the correct business process. The re-guidance strategy may include providing a visual operation guide, sending relevant links to the user's mobile phone, or clearly explaining the next steps and expected results via voice.

[0072] This embodiment improves user satisfaction, reduces the risk of business interruption, and ensures business continuity and efficiency through a dynamic adjustment mechanism based on user behavior and feedback.

[0073] Optionally, in the outbound call service processing method provided in the embodiments of this application, the target model includes an input layer, a convolutional layer, a fully connected layer, and an output layer. The target model outputs the behavior recognition result in the following manner: receiving speech features through the input layer, extracting temporal features from the speech features of the input layer through the convolutional layer; mapping the temporal features to the probability of the user continuing to engage in voice interaction through the fully connected layer, determining the probability as the behavior recognition result, and outputting the behavior recognition result through the output layer.

[0074] In some embodiments, the target model can be a temporal convolutional network model, and the process for generating behavior recognition results can be represented by the following formula:

[0075] ;

[0076] in, : Voice characteristics of the customer at time step t. : Convolution kernel weight matrix (K kernels, input dimension n). : Inflation factor (l is the number of layers). σ: ReLU activation function. b is the bias parameter.

[0077] The target model can be adaptively optimized based on deep learning and reinforcement learning algorithms. The decision network and target network are initialized, with the target network structure cloned from the decision network. Parameters are periodically synchronized, ReLU activation is used, and the final layer outputs linearly. A circular storage experience pool mechanism is constructed to continuously store key data units from the state transition process. Each data unit completely records a quadruple of initial state, action performed, reward obtained, and adjusted state. When the accumulated experience reaches a threshold capacity, a batch processing strategy combining random sampling and priority selection is adopted, prioritizing samples with larger prediction errors for training, significantly improving data utilization efficiency and breaking temporal correlations.

[0078] For example, dynamic gating output: risk index value = 0.68 (full value 1.0, >0.5 requires intervention), generating decision quadruple: state vector = [response delay 2.8s, deviation 0.6, risk value 0.68, business node position 2]; action space candidates: action 0 (maintain path): excluded due to excessive risk; action 1 (compress script): activated (because the risk value is in the range of 0.5-0.7); action 3 (transfer to human): not triggered temporarily if the threshold of 0.9 is not reached.

[0079] After selecting a data augmentation strategy, negative loss is used as the reward signal. After each fixed number of training rounds, a batch of empirical data is randomly sampled from the experience replay pool, and the target Q-value is calculated to enable the system to learn a better strategy. When the training reaches convergence or the set maximum number of rounds, the augmentation strategy with the highest estimated Q-value is extracted as the final strategy. Customer feedback on outbound call services is collected, and the system is continuously optimized based on the feedback results to improve the accuracy of intent recognition and the level of service personalization.

[0080] The target model in this embodiment enables the system to gain a deeper understanding of user intentions, thereby making more accurate and human-centered decisions in areas such as dialogue management and service strategy formulation. For example, the system can adjust the dialogue script in a timely manner based on user behavior recognition results, introducing more emotionally reassuring statements or simplifying business descriptions to improve user satisfaction and reduce service interruptions.

[0081] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0082] Example 2

[0083] This application also provides an outbound call service processing apparatus. It should be noted that the outbound call service processing apparatus of this application can be used to execute the outbound call service processing method provided in this application. The outbound call service processing apparatus provided in this application will be described below.

[0084] According to an embodiment of this application, an apparatus for implementing the above-described outbound call service processing method is also provided. Figure 3 This is a schematic diagram of an outbound call service processing device provided according to an embodiment of this application, such as... Figure 3 As shown, the device includes:

[0085] The execution unit 301 is used to perform outbound call tasks to users. When the user is detected to have authorized the target service, the target service is processed according to the preset service processing procedure. The user's voice signal is collected during the processing of the target service, and voice features and voice content are extracted from the voice signal. The target service is a voice-based service.

[0086] The first determining unit 302 is used to determine the current business processing node corresponding to the voice content in the preset business processing process, determine the business path diagram corresponding to the preset business processing process, and determine the path deviation index value based on the current business processing node and the business path diagram. The business path diagram contains multiple business processing nodes and edges between multiple business processing nodes, and the edges are used to represent the business processing order.

[0087] The input unit 303 is used to input speech features into the target model to obtain the user's behavior recognition result, wherein the behavior recognition result is used to characterize the probability that the user will continue to engage in voice interaction;

[0088] The adjustment unit 304 is used to determine the risk index value of user interruption of target business based on behavior recognition results and path deviation index value, and adjust the preset business processing process according to the risk index value.

[0089] The outbound call processing apparatus provided in this application embodiment executes an outbound call task to a user through an execution unit 301. Upon detecting user authorization for a target service, it processes the target service according to a preset service processing flow, collects the user's voice signal during the target service processing, and extracts voice features and voice content from the voice signal. The target service is a voice-based service. A first determining unit 302 determines the current service processing node corresponding to the voice content in the preset service processing flow, determines the service path graph corresponding to the preset service processing flow, and determines a path deviation index value based on the current service processing node and the service path graph. The service path graph includes multiple service processing nodes and edges between these nodes, with the edges representing... The business processing sequence is as follows: Input unit 303 inputs voice features into the target model to obtain the user's behavior recognition result, where the behavior recognition result is used to characterize the probability that the user will continue voice interaction; Adjustment unit 304 determines the risk index value of the user interrupting the target business based on the behavior recognition result and the path deviation index value, and adjusts the preset business processing process according to the risk index value. By introducing real-time behavior recognition and path deviation analysis, combined with the dynamic calculation of the risk index value, intelligent optimization of outbound call business is realized, which achieves the goal of improving the responsiveness to users' personalized needs and reducing the risk of business interruption. This achieves the technical effect of improving the accuracy of outbound call business in recognizing user intent, and thus solves the technical problem of poor accuracy of outbound call business in recognizing user intent.

[0090] Optionally, in the outbound call processing apparatus provided in this application embodiment, extracting speech features from the speech signal includes: extracting fundamental frequency jitter features from the speech signal using Mel frequency cepstral coefficients, determining the user's speech rate and pause duration based on the fundamental frequency jitter features; determining the spectrum of the speech signal, determining the noise intensity of the speech signal based on the spectrum; and determining at least one of the speech rate, pause duration, and noise intensity as a speech feature.

[0091] Optionally, in the outbound call service processing apparatus provided in this application embodiment, determining the path deviation index value based on the current service processing node and the service path graph includes: inputting the current service processing node and the service path graph into a graph convolutional network to obtain the path deviation index value. The graph convolutional network preprocesses the service path graph and the current service processing node to obtain an adjacency matrix and a node feature matrix. The activation function of the graph convolutional network takes the adjacency matrix and the node feature matrix as input and outputs the path deviation index value.

[0092] Optionally, in the outbound call service processing apparatus provided in this application embodiment, determining the risk index value of user interruption of target service based on behavior recognition results and path deviation index values ​​includes: inputting behavior recognition results and path deviation index values ​​into a gating weight matrix to obtain an output value, wherein the gating weight matrix is ​​used to determine the contribution ratio of behavior recognition results and path deviation index values ​​in the risk index value; calculating the sum of the output value and a preset offset, inputting the sum into a target activation function to obtain the risk index value.

[0093] Optionally, in the outbound call processing apparatus provided in this application embodiment, adjusting the preset business processing flow according to the risk indicator value includes: when the risk indicator value is greater than or equal to the indicator value threshold, determining the initial dialogue script content and initial dialogue flow corresponding to the preset business processing flow; extracting a summary of the initial dialogue script content using a text processing tool, and generating updated dialogue script content based on the summary; removing processes from the initial dialogue flow whose correlation index value with the target business is lower than the correlation threshold, to obtain the updated dialogue flow; and processing the target business based on the updated dialogue script content and the updated dialogue flow.

[0094] Optionally, in the outbound call service processing apparatus provided in this application embodiment, adjusting the preset service processing procedure according to the preset strategy includes: when the risk indicator value is greater than or equal to the indicator value threshold and the noise intensity in the voice feature is greater than or equal to the intensity threshold, processing the voice signal through a preset noise reduction algorithm to obtain an updated voice signal; and performing the step of extracting voice features and voice content from the voice signal based on the updated voice signal.

[0095] Optionally, in the outbound call service processing device provided in this application embodiment, adjusting the preset service processing procedure according to the preset strategy includes: when the risk indicator value is greater than or equal to the indicator value threshold and the pause duration in the voice feature is greater than or equal to the duration threshold, or when the risk indicator value is greater than or equal to the indicator value threshold and the path deviation indicator value is greater than or equal to the path deviation threshold, the target service is processed through preset guidance content.

[0096] Optionally, in the outbound call service processing apparatus provided in the embodiments of this application, the target model includes an input layer, a convolutional layer, a fully connected layer, and an output layer. The target model outputs the behavior recognition result in the following manner: receiving speech features through the input layer, extracting temporal features from the speech features of the input layer through the convolutional layer; mapping the temporal features to the probability of the user continuing to engage in voice interaction through the fully connected layer, determining the probability as the behavior recognition result, and outputting the behavior recognition result through the output layer.

[0097] It should be noted that the execution unit 301, the first determining unit 302, the input unit 303, and the adjustment unit 304 mentioned above correspond to steps S201 to S204 in Embodiment 1. The four units and the corresponding steps implement the same instances and application scenarios, but are not limited to the content disclosed in Embodiment 1. It should be noted that the above modules or units can be hardware or software components stored in memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above modules or units can also be part of a device and run in the computer terminal 10 provided in Embodiment 1.

[0098] Example 3

[0099] Embodiments of this application may provide an electronic device. Figure 4 This is a structural block diagram of an electronic device according to an embodiment of this application. Figure 4 As shown, the electronic device may include: one or more ( Figure 4 (Only one is shown) Processor 402, memory 404, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0100] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the methods and apparatus in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby implementing the above-described methods. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0101] The processor can access information and applications stored in the memory via a transmission device to perform the following steps: Execute an outbound call to the user; upon detecting user authorization for the target service, process the target service according to a preset service processing procedure; collect the user's voice signal during the target service processing; extract voice features and voice content from the voice signal, where the target service is a voice-based service; determine the current service processing node corresponding to the voice content in the preset service processing procedure; determine the service path graph corresponding to the preset service processing procedure; determine the path deviation index value based on the current service processing node and the service path graph, where the service path graph contains multiple service processing nodes and edges between them, with edges representing the service processing order; input the voice features into the target model to obtain the user's behavior recognition result, where the behavior recognition result represents the probability that the user will continue voice interaction; determine the risk index value for the user interrupting the target service based on the behavior recognition result and the path deviation index value; and adjust the preset service processing procedure according to the risk index value.

[0102] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: extracting the fundamental frequency jitter feature from the speech signal through the Mel frequency cepstral coefficients, determining the user's speech rate and pause duration based on the fundamental frequency jitter feature; determining the spectrum of the speech signal, determining the noise intensity of the speech signal based on the spectrum; and determining at least one of the speech rate, pause duration, and noise intensity as a speech feature.

[0103] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: input the current business processing node and business path graph into the graph convolutional network to obtain the path deviation index value. The graph convolutional network preprocesses the business path graph and the current business processing node to obtain the adjacency matrix and the node feature matrix. The activation function of the graph convolutional network takes the adjacency matrix and the node feature matrix as input and outputs the path deviation index value.

[0104] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: inputting the behavior recognition result and path deviation index value into the gating weight matrix to obtain the output value, wherein the gating weight matrix is ​​used to determine the contribution ratio of the behavior recognition result and path deviation index value in the risk index value; calculating the sum of the output value and the preset offset, and inputting the sum into the target activation function to obtain the risk index value.

[0105] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: when the risk indicator value is greater than or equal to the indicator value threshold, determine the initial dialogue script content and initial dialogue process corresponding to the preset business handling process; extract the summary of the initial dialogue script content through a text processing tool, and generate the updated dialogue script content based on the summary; remove processes from the initial dialogue process whose correlation index value with the target business is lower than the correlation threshold, and obtain the updated dialogue process; handle the target business based on the updated dialogue script content and the updated dialogue process.

[0106] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: when the risk index value is greater than or equal to the index value threshold and the noise intensity in the speech features is greater than or equal to the intensity threshold, process the speech signal through a preset noise reduction algorithm to obtain an updated speech signal; and perform the step of extracting speech features and speech content from the speech signal based on the updated speech signal.

[0107] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: when the risk indicator value is greater than or equal to the indicator value threshold and the pause duration in the voice feature is greater than or equal to the duration threshold, or when the risk indicator value is greater than or equal to the indicator value threshold and the path deviation indicator value is greater than or equal to the path deviation threshold, the target business is handled through preset guidance content.

[0108] The processor can also call the information and application programs stored in the memory through the transmission device to perform the following steps: receiving speech features through the input layer, extracting temporal features from the speech features of the input layer through the convolutional layer; mapping the temporal features to the probability of the user continuing to engage in voice interaction through the fully connected layer, determining the probability as the behavior recognition result, and outputting the behavior recognition result through the output layer.

[0109] This application provides a scheme for executing outbound calls to users. Upon detecting user authorization for a target service, the scheme processes the target service according to a preset service processing flow. It collects the user's voice signal during the service processing, extracts voice features and voice content from the voice signal, where the target service is a voice-based service. The scheme determines the current service processing node corresponding to the voice content in the preset service processing flow, determines the corresponding service path graph, and determines a path deviation index value based on the current service processing node and the service path graph. The service path graph contains multiple service processing nodes and edges between them, representing the service processing order. The voice features are input into a target model to obtain the user's behavior recognition result, which represents the probability that the user will continue voice interaction. Based on the behavior recognition result and the path deviation index value, a risk index value for user interruption of the target service is determined, and the preset service processing flow is adjusted according to the risk index value. By introducing real-time behavior recognition and path deviation analysis, combined with the dynamic calculation of risk index values, intelligent optimization of outbound calling services was achieved. This improved the ability to respond to personalized user needs and reduced the risk of service interruption, thereby enhancing the accuracy of outbound calling services in recognizing user intent and solving the technical problem of poor accuracy in recognizing user intent.

[0110] Those skilled in the art will understand that Figure 4 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 4 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 4 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 4 The different configurations shown.

[0111] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0112] Example 4

[0113] Embodiments of this application also provide a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the outbound call service processing method provided in Embodiment 1.

[0114] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0115] This application also provides a computer program product, which, when executed on a data processing device, is a program suitable for performing processing steps of outbound call services.

[0116] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0117] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0119] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0120] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0121] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0122] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for processing outbound call services, characterized in that, include: The system performs outbound calls to users. When user authorization for a target service is detected, the system processes the target service according to a preset service processing procedure. The system collects the user's voice signal during the processing of the target service and extracts voice features and voice content from the voice signal. The target service is a voice-based service. The current business processing node corresponding to the voice content in the preset business processing flow is determined, and the business path graph corresponding to the preset business processing flow is determined. Based on the current business processing node and the business path graph, the path deviation index value is determined. The business path graph contains multiple business processing nodes and edges between the multiple business processing nodes, and the edges are used to represent the business processing order. The voice features are input into the target model to obtain the user's behavior recognition result, wherein the behavior recognition result is used to characterize the probability that the user will continue to engage in voice interaction; Based on the behavior recognition results and the path deviation index value, a risk index value for the user to interrupt the target service is determined, and the preset service processing procedure is adjusted according to the risk index value.

2. The method according to claim 1, characterized in that, Extracting speech features from the speech signal includes: The fundamental frequency jitter feature is extracted from the speech signal using Mel frequency cepstral coefficients, and the user's speech rate and pause duration are determined based on the fundamental frequency jitter feature; Determine the spectrum of the speech signal, and determine the noise intensity of the speech signal based on the spectrum; The speech feature is determined by at least one of the speech rate, the pause duration, and the noise intensity.

3. The method according to claim 1, characterized in that, Determining the path deviation index value based on the current business processing node and the business path map includes: The current business processing node and the business path graph are input into a graph convolutional network to obtain the path deviation index value. The graph convolutional network preprocesses the business path graph and the current business processing node to obtain an adjacency matrix and a node feature matrix. The activation function of the graph convolutional network takes the adjacency matrix and the node feature matrix as input and outputs the path deviation index value.

4. The method according to claim 1, characterized in that, The risk indicator value for determining the user's interruption of the target service based on the behavior recognition result and the path deviation indicator value includes: The behavior recognition result and the path deviation index value are input into the gating weight matrix to obtain the output value, wherein the gating weight matrix is ​​used to determine the contribution ratio of the behavior recognition result and the path deviation index value in the risk index value; Calculate the sum of the output value and the preset offset, and input the sum into the target activation function to obtain the risk index value.

5. The method according to claim 1, characterized in that, Adjusting the preset business processing procedure based on the risk indicator value includes: If the risk indicator value is greater than or equal to the indicator value threshold, determine the initial dialogue script content and initial dialogue process corresponding to the preset business processing process; Extract a summary of the initial dialogue script content using a text processing tool, and generate an updated dialogue script content based on the summary; The updated dialogue flow is obtained by removing processes from the initial dialogue flow whose relevance index value to the target business is lower than the relevance threshold. The target business is processed based on the updated dialogue script content and the updated dialogue flow.

6. The method according to claim 1, characterized in that, Adjusting the preset business processing procedure based on the risk indicator value includes: When the risk indicator value is greater than or equal to the indicator value threshold and the noise intensity in the speech feature is greater than or equal to the intensity threshold, the speech signal is processed by a preset noise reduction algorithm to obtain an updated speech signal. Based on the updated speech signal, perform the steps of extracting speech features and speech content from the speech signal.

7. The method according to claim 1, characterized in that, Adjusting the preset business processing procedure based on the risk indicator value includes: If the risk indicator value is greater than or equal to the indicator value threshold and the pause duration in the voice feature is greater than or equal to the duration threshold, or if the risk indicator value is greater than or equal to the indicator value threshold and the path deviation indicator value is greater than or equal to the path deviation threshold, the target business is processed through preset guidance content.

8. The method according to claim 1, characterized in that, The target model includes an input layer, a convolutional layer, a fully connected layer, and an output layer. The target model outputs the behavior recognition result in the following manner: The speech features are received through the input layer, and temporal features are extracted from the speech features of the input layer through the convolutional layer. The temporal features are mapped to the probability that the user will continue to interact via voice through the fully connected layer, the probability is determined as the behavior recognition result, and the behavior recognition result is output through the output layer.

9. A processing device for outbound call services, characterized in that, include: An execution unit is used to perform outbound call tasks to users. When the user is detected to have authorized the target service, the execution unit processes the target service according to the preset service processing procedure. The execution unit collects the user's voice signal during the processing of the target service and extracts voice features and voice content from the voice signal. The target service is a voice-based service. The first determining unit is used to determine the current business processing node corresponding to the voice content in the preset business processing process, determine the business path diagram corresponding to the preset business processing process, and determine the path deviation index value based on the current business processing node and the business path diagram. The business path diagram includes multiple business processing nodes and edges between the multiple business processing nodes, and the edges are used to represent the business processing order. An input unit is used to input the speech features into a target model to obtain the user's behavior recognition result, wherein the behavior recognition result is used to characterize the probability that the user will continue to engage in voice interaction; The adjustment unit is used to determine the risk index value of the user interrupting the target service based on the behavior recognition result and the path deviation index value, and adjust the preset service processing procedure according to the risk index value.

10. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the steps of the outbound call service processing method according to any one of claims 1 to 8.

Images