Business processing method and device, electronic equipment, storage medium and program product

Abstract

The invention provides a business processing method and device, electronic equipment, a storage medium and a program product, and the method comprises the steps: carrying out the emotion recognition according to the call voice data of a first object in a first business process, and obtaining an emotion recognition result of the first object; performing intention recognition according to the call voice data to obtain an intention recognition result of the first object; performing similarity matching in a preset call library according to the call voice data to obtain a call matching result of the first object; performing behavior prediction according to the service execution state data of the first object corresponding to the first service to obtain a behavior prediction result of the first object; and determining a processing strategy for the first object on the first service according to the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result. According to the embodiment of the invention, the processing strategy better matched with the first object can be determined, so that the processing effect on the first service can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a business processing method and device, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] During business processing, if a user is expected to perform a specific business action, relevant staff need to communicate with the user to determine their intent and take appropriate measures to encourage them to perform the specific business action. Because the specific measures are determined solely by identifying the user's intent, if the user's intent is not accurately identified, it may not be possible to ensure that the user will perform the specific business action, resulting in poor business processing results. Summary of the Invention

[0003] The present disclosure provides a business processing method and device, an electronic device, a computer-readable storage medium, and a computer program product.

[0004] In a first aspect, the present disclosure provides a business processing method, which includes: performing emotion recognition based on call voice data of a first object in a first business process to obtain an emotion recognition result of the first object; performing intent recognition based on the call voice data to obtain an intention recognition result of the first object; performing similarity matching in a preset call library based on the call voice data to obtain a call matching result of the first object; performing behavior prediction based on the business execution status data of the first object corresponding to the first business to obtain a behavior prediction result of the first object; determining a processing strategy for the first object on the first business based on the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result.

[0005] In the second aspect, the present disclosure provides a business processing device, which includes: a first recognition module, used to perform emotion recognition based on call voice data of a first object in a first business process, and obtain an emotion recognition result of the first object; a second recognition module, used to perform intention recognition based on the call voice data, and obtain an intention recognition result of the first object; a matching module, used to perform call similarity matching in a preset call library based on the call voice data, and obtain a call matching result of the first object; a prediction module, used to perform behavior prediction based on the business execution status data of the first object corresponding to the first business, and obtain a behavior prediction result of the first object; a determination module, used to determine the processing strategy for the first object on the first business based on the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result.

[0006] In a third aspect, the present disclosure provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores one or more computer programs executable by the at least one processor, and one or more of the computer programs are executed by the at least one processor to enable the at least one processor to execute the above-mentioned business processing method.

[0007] In a fourth aspect, the present disclosure provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the above-mentioned business processing method when executed by a processor.

[0008] In a fifth aspect, the present disclosure provides a computer program product comprising a computer-readable code, or a non-volatile computer-readable storage medium carrying the computer-readable code. When the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the above-mentioned business processing method.

[0009] The business processing method provided by the embodiment of the present disclosure performs emotion recognition based on the call voice data of the first object in the first business process to obtain the emotion recognition result of the first object; performs intent recognition based on the call voice data to obtain the intention recognition result of the first object; performs call similarity matching in a preset call library based on the call voice data to obtain the call matching result of the first object; performs behavior prediction based on the business execution status data of the first object corresponding to the first business to obtain the behavior prediction result of the first object; determines the processing strategy for the first object in the first business based on the emotion recognition result, intention recognition result, call matching result and behavior prediction result. Since multi-dimensional processing is performed from the aspects of emotion recognition, intention recognition, call matching and behavior prediction, emotion recognition results, intention recognition results, call matching results and behavior prediction results can be obtained, and the processing strategy can be determined based on the above multiple results. Therefore, the determined processing strategy is more matched with the first object, thereby improving the processing effect of the first business.

[0010] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. The above and other features and advantages will become more apparent to those skilled in the art by describing detailed example embodiments with reference to the accompanying drawings. In the accompanying drawings:

[0012] Figure 1 An application scenario diagram of the business processing method and device provided in the embodiments of the present disclosure;

[0013] Figure 2 A flowchart of a business processing method provided in an embodiment of the present disclosure;

[0014] Figure 3 A flowchart of a business processing method provided by an embodiment of the present disclosure;

[0015] Figure 4 A schematic diagram of a processing process of a business processing method provided in an embodiment of the present disclosure;

[0016] Figure 5 A block diagram of a service processing device provided in an embodiment of the present disclosure;

[0017] Figure 6 A block diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] To enable those skilled in the art to better understand the technical solutions of the present disclosure, exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0019] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0020] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0021] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "comprising" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded. Similar words such as "connected" or "connected" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0022] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted as having an idealized or overly formal meaning unless expressly defined as such herein.

[0023] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals. The use of user data in this technical solution complies with relevant national laws and regulations (for example, the "Information Security Technology Personal Information Security Specification", etc.). For example: corresponding prescribed measures are taken to control access to personal information; the display of personal information is subject to prescribed restrictions; the purpose of using personal information does not exceed the scope of direct or reasonable connection; when using personal information, clear identity reference is eliminated to avoid precise positioning of specific individuals.

[0024] When handling transactions, customer service personnel need to communicate with users, identify their intent, and then take appropriate measures to encourage them to perform specific transactions. Failure to accurately identify user intent can result in unexpected transaction results, thereby reducing transaction effectiveness.

[0025] For example, when handling debt collection, agents establish a call connection with customers through intelligent outbound calling and process the conversation content based on an intent recognition model to determine whether the customer has expressed a commitment to repay the debt, and then adopt appropriate collection methods. However, in actual business processing, there are problems such as inaccurate intent recognition or false customer commitments, which can lead to poor collection and affect collection effectiveness.

[0026] For example, when recommending a communication plan, an agent establishes a call with the customer and processes the conversation based on an intent recognition model to determine whether the customer has expressed a willingness to purchase the plan and then take appropriate follow-up measures. However, in actual business processing, there are issues such as inaccurate intent recognition or false customer commitments, which can affect the effectiveness of the recommendation.

[0027] In view of this, the embodiments of the present disclosure provide a business processing method and device, an electronic device, a computer-readable storage medium, and a computer program product, which can perform multi-dimensional processing from the aspects of emotion recognition, intention recognition, call matching, and behavior prediction, so that emotion recognition results, intention recognition results, call matching results, and behavior prediction results can be obtained, and the processing strategy can be jointly determined based on the above multiple results. Compared with the processing strategy determined only by intention recognition, the processing strategy determined by the embodiments of the present disclosure is more matched with the first object, and therefore can effectively improve the processing effect of the first business.

[0028] Figure 1 This is an application scenario diagram of the business processing method and device provided in the embodiments of the present disclosure.

[0029] like Figure 1 As shown, an application scenario of an embodiment of the present disclosure may include a terminal device 101, a network 103, and a server 102. The network 103 is used as a medium for providing a communication link between the terminal device 101 and the server 102. The network 103 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0030] The user can use the terminal device 101 to interact with the server 102 via the network 103 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0031] The terminal device 101 may be any electronic device having a display screen and supporting web browsing, including but not limited to a smart phone, a tablet computer, a laptop computer, a desktop computer, and the like.

[0032] The server 102 may be a server that provides various services, such as a background management server (for example only) that supports websites browsed by users using the terminal device 101. The background management server may analyze and process received user requests and other data, and feed back the processing results (e.g., web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0033] It should be noted that the business processing method and apparatus provided in the embodiments of the present disclosure can be executed by the server 102. Accordingly, the business processing method and apparatus provided in the embodiments of the present disclosure can be set in the server 102. The business processing method and apparatus provided in the embodiments of the present disclosure can also be executed by a server or server cluster that is different from the server 102 and can communicate with the terminal device 101 and / or the server 102. Accordingly, the business processing method and apparatus provided in the embodiments of the present disclosure can also be set in a server or server cluster that is different from the server 102 and can communicate with the terminal device 101 and / or the server 102.

[0034] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0035] Figure 2 This is a flowchart of a business processing method provided by an embodiment of the present disclosure. Figure 2 , the method comprising:

[0036] Step S21 : performing emotion recognition based on the call voice data of the first subject during the first service process to obtain an emotion recognition result of the first subject.

[0037] The first object is one of the participants in the first business, and the first business is the business to be processed. For example, if the first business is a debt collection business, the first object is the user in arrears or the person being collected. For example, if the first business is a communication package recommendation business, the first object is a potential user or the person being recommended. The above description of the first object and the first business is merely an example and is not limited in the present embodiment.

[0038] In some optional embodiments, an agent establishes a call connection with the first party via an intelligent outbound calling system and engages in a conversation regarding the first service. Based on this, voice data from the first party's call during the first service can be acquired using voice acquisition technology, thereby providing a data foundation for subsequent processing.

[0039] Emotion recognition refers to identifying the emotions of the call recipient (i.e., the first party). During a call, the first party's emotions are typically reflected in their speech. For example, when the first party is excited, their speech rate may be fast and their volume may be loud; for example, when the first party is calm, their speech rate and volume may be moderate. Based on this, by performing appropriate analysis and processing on the call voice data, the first party's emotions can be identified. Moreover, clarifying the first party's emotions helps guide subsequent business processing.

[0040] In some optional embodiments, the emotions recognized based on the call voice data include positive emotions, neutral emotions, and negative emotions, that is, the emotion recognition result includes one of the above three emotions.

[0041] In some optional embodiments, the emotion recognition result of the first subject can be obtained by using speech features and target word detection. The speech features include features that can reflect the emotion of the first subject at the speech level; for example, the speech features may include intonation sub-features, speech rate sub-features, volume sub-features, etc.

[0042] Furthermore, the preset target words are words that can reflect specific emotional characteristics, and there can be one or more preset target words. Target word detection checks whether the call voice data matches the preset target words. For example, "complaint," "report," and other uncivilized words can be set as preset target words. When the first party says these preset target words, it is likely that the first party is in a negative mood.

[0043] In some optional embodiments, a target vocabulary library may be pre-established, and the determined preset target words may be added to the target vocabulary library for centralized management, and the target vocabulary library may be updated according to business needs.

[0044] In some optional embodiments, whether the call voice data hits the preset target word can be determined based on a preset hit rule. The hit rule is used to characterize the number of preset target words contained in the call voice data and the number of occurrences of the preset target words contained in the call voice data when it hits. For example, the hit rule is set as: the number of preset target words = 1 and the number of occurrences > 1, or the number of preset target words > 1. That is, if the first subject only said one preset target word during the call and said the preset target word more than once, the corresponding target word detection result indicates that the call voice data meets the preset hit rule; if the first subject only said one preset target word during the call and said the preset target word once, the corresponding target word detection result indicates that the call voice data does not meet the preset hit rule. Alternatively, if the first subject said multiple preset target words during the call, the corresponding target word detection result indicates that the call voice data meets the preset hit rule.

[0045] It should be noted that when performing target word detection, it is possible to directly determine whether the target word is hit based on the call voice data through voice detection technology, or to first convert the call voice data into corresponding first call text data, and then use the first call text data to perform target word detection through text detection technology. The embodiments of the present disclosure do not limit this.

[0046] Exemplarily, performing emotion recognition based on call voice data of a first subject during a first service process to obtain an emotion recognition result for the first subject includes: obtaining speech features and target word detection results from the call voice data, where the speech features are used to characterize speech state information of the first subject during the call, and the target word detection results are used to characterize hit information of preset target words in the call voice data; and performing emotion recognition based on the speech features and the target word detection results to obtain the emotion recognition result. The speech state information may include information such as intonation, speaking rate, and volume.

[0047] In other words, emotion recognition can be performed from both the perspectives of speech features and target words, thereby obtaining more accurate and reasonable emotion recognition results.

[0048] In some optional embodiments, the speech features include at least intonation sub-features, speaking speed sub-features, and volume sub-features; accordingly, emotion recognition is performed based on the speech features and the target word detection results to obtain an emotion recognition result, including: determining the intonation recognition sub-result corresponding to the intonation sub-feature; determining the speaking speed recognition sub-result corresponding to the speaking speed sub-feature; determining the volume recognition sub-result corresponding to the volume sub-feature; fusing the intonation recognition sub-result, speaking speed recognition sub-result, and volume recognition sub-result to obtain a feature recognition result; and determining the emotion recognition result based on the feature recognition result and the target word detection result.

[0049] From this, we can see that for speech features, corresponding analysis and processing can be carried out from at least three perspectives: intonation, speaking rate, and volume. The obtained intonation recognition sub-results, speaking rate recognition sub-results, and volume recognition sub-results can be integrated to obtain feature recognition results at the speech feature level; on this basis, the emotion recognition results are jointly determined according to the feature recognition results and the target word detection results to improve the accuracy of the emotion recognition results.

[0050] In some optional embodiments, the intonation recognition sub-result includes an intonation score, the speaking rate recognition sub-result includes a speaking rate score, and the volume recognition sub-result includes a volume score. The feature recognition result is a feature weighted value determined based on a preset first weight coefficient, the intonation score, the speaking rate score, and the volume score. The target word detection result is determined based on the call voice data and a preset hit rule. Accordingly, the emotion recognition result is determined based on the feature recognition result and the target word detection result, including:

[0051] When the target word detection result represents that the call voice data meets the preset hit rules, the emotion recognition result is determined to be a negative emotion; when the target word detection result represents that the call voice data does not meet the preset hit rules, and the feature weighted value is within the first preset interval, the emotion recognition result is determined to be a positive emotion; when the target word detection result represents that the call voice data does not meet the preset hit rules, and the feature weighted value is within the second preset interval, the emotion recognition result is determined to be a neutral emotion; when the target word detection result represents that the call voice data does not meet the preset hit rules, and the feature weighted value is within the third preset interval, the emotion recognition result is determined to be a negative emotion; wherein, the feature weighted value corresponding to the first preset interval is higher than the feature weighted value corresponding to the second preset interval, and the feature weighted value corresponding to the second preset interval is higher than the feature weighted value corresponding to the third preset interval. Moreover, the feature weighted value corresponding to the first preset interval is higher than the feature weighted value corresponding to the second preset interval, which means that the minimum value of the feature weighted value of the first preset interval is still greater than the maximum value of the feature weighted value of the second preset interval, and the feature weighted value corresponding to the second preset interval is higher than the feature weighted value corresponding to the third preset interval, which means that the minimum value of the feature weighted value of the second preset interval is still greater than the maximum value of the feature weighted value of the third preset interval.

[0052] From this, it can be seen that in some embodiments, if it is determined that the call voice data meets the preset hit rules, the emotion recognition result can be directly determined as a negative emotion without considering the recognition result of the voice feature dimension; if it is determined that the call voice data does not meet the preset hit rules, the emotion recognition result needs to be determined based on the interval in which the feature weight value is located.

[0053] In some optional embodiments, positive or excited emotions are often accompanied by a higher tone, while under negative emotions, the tone becomes lower. Therefore, the fundamental frequency of the speech (in Hertz (Hz)) can be used to characterize the tone sub-feature, and the fundamental frequency is the basic frequency of the vocal cord vibration. In addition, under positive or excited emotions, the speaker's speaking speed is usually relatively stable or moderately faster, while under negative emotions, the speaker may slow down the speaking speed or have an intermittent speaking pattern. Based on this, the speaking speed sub-feature can be characterized by the number of words spoken per minute (Words Per Minute, WPM). Similarly, under positive or excited emotions, the volume is usually moderate, while under negative emotions, the volume may be too high or too low. Therefore, the volume sub-feature can be characterized by the decibel value (dB).

[0054] For the intonation sub-feature, the intonation sub-feature with a fundamental frequency at (70Hz, 150Hz) is determined as a low intonation, the intonation sub-feature with a fundamental frequency at [150Hz, 250Hz] is determined as a medium intonation, and the intonation sub-feature with a fundamental frequency greater than 250Hz is determined as a high intonation. In addition, for low intonation, a intonation score of 1 point is assigned, for medium intonation, a intonation score of 2 points is assigned, and for high intonation, a intonation score of 1 point is assigned. During the processing, the fundamental frequency average value of the first object during the call period can be calculated based on the acquired call voice data, and the fundamental frequency average value can be used as the intonation sub-feature of the first object. It is also determined in which of the above-mentioned fundamental frequency value ranges the fundamental frequency average value is located, thereby determining the intonation score s1 corresponding to the intonation sub-feature.

[0055] For the speech speed sub-feature, the speech speed sub-feature with a speech speed less than 120WPM is determined as a low speech speed, the speech speed sub-feature with a speech speed between [120WPM, 160WPM] is determined as a medium speech speed, and the speech speed sub-feature with a speech speed greater than 160WPM is determined as a high speech speed. In addition, for the low speech speed, a speech speed score of 1 point is assigned, for the medium speech speed, a speech speed score of 3 points is assigned, and for the high speech speed, a speech speed score of 2 points is assigned. During the processing, the average speech speed of the first subject during the call can be calculated based on the acquired call voice data, and the average speech speed value can be used as the speech speed sub-feature of the first subject. It is also determined in which of the above-mentioned speech speed value ranges the average speech speed value falls, thereby determining the speech speed score s2 corresponding to the speech speed sub-feature.

[0056] For the volume sub-feature, a volume sub-feature with a volume less than 50dB is determined to be low volume, a volume sub-feature with a volume between [50dB, 70dB] is determined to be medium volume, and a volume sub-feature with a volume greater than 70dB is determined to be high volume. Furthermore, a volume score of 1 is assigned to low volume, 2 points to medium volume, and 2 points to high volume. During processing, the average volume of the first subject during the call can be calculated based on the acquired call voice data. This average volume can be used as the volume sub-feature of the first subject. It is then determined within which of the aforementioned volume value ranges the average volume falls, thereby determining the volume score s3 corresponding to the volume sub-feature.

[0057] Furthermore, if the preset first weight coefficient includes w11 corresponding to the intonation sub-feature (or intonation score), w12 corresponding to the speaking speed sub-feature (or speaking speed score), and w13 corresponding to the volume sub-feature (or volume score), then the above-mentioned first weight coefficient is used to perform a weighted operation with the intonation score, speaking speed score, and volume score: S = w11×s1+w12×s2+w13×s3, and the weighted result S can be obtained, which is the feature weighted value.

[0058] In addition, it is necessary to determine whether the call voice data meets the preset hit rules. If it is determined to meet the preset hit rules, the emotion recognition result can be directly determined as negative emotion. If it is determined not to meet the preset hit rules, it is necessary to determine in which preset range the feature weight value S falls, so as to obtain the emotion recognition result of the first object.

[0059] Assuming that the first preset interval is [2, 2.3], the second preset interval is [1.3, 2), and the third preset interval is [1, 1.3), if the feature weighted value S∈[2, 2.3], the emotion recognition result is determined to be a positive emotion; if the feature weighted value S∈[1.3, 2), the emotion recognition result is determined to be a neutral emotion; if the feature weighted value S∈[1, 1.3), the emotion recognition result is determined to be a negative emotion.

[0060] In some optional embodiments, the operations of extracting the intonation sub-feature, generating the intonation score, extracting the speech rate sub-feature, generating the speech rate score, extracting the volume sub-feature, generating the volume score, etc., can all be performed by a preset speech recognition model. The speech recognition model is a pre-trained model that has feature extraction and certain computing capabilities.

[0061] Step S22: perform intention recognition based on the call voice data to obtain an intention recognition result of the first object.

[0062] Among them, intent recognition refers to parsing the action or request that the user wants to perform from the information corresponding to the user (ie, the first object) (such as natural language text data).

[0063] In some optional embodiments, the intention recognition result includes the first object's willingness level to perform the first business, wherein the willingness level can reflect the degree to which the first object is willing to perform the first business, that is, the degree to which the first object is willing to perform the first business (such as the degree to which the first object is willing to repay) can be determined through intention recognition.

[0064] In some optional embodiments, intention recognition is performed based on call voice data to obtain an intention recognition result of the first object, including: obtaining first call text data corresponding to the call voice data; identifying the execution intention of the first object for the first business based on the first call text data; and classifying the execution intention to obtain a willingness level corresponding to the execution intention.

[0065] For example, when handling debt collection business, intent recognition is mainly to identify the degree to which the first party is willing to repay the loan; when handling recommendation business, intent recognition is mainly to identify the degree to which the first party is willing to handle the business recommended by the customer service.

[0066] In some optional embodiments, the call voice data can be converted into corresponding first call text data using methods such as automatic speech recognition (ASR) technology. Then, a knowledge document library for intent recognition is constructed, so that the initial intent recognition model can learn semantic recognition capabilities, intent recognition capabilities, etc. based on the knowledge document library, thereby completing the intent recognition task based on the prompt word (Prompt). The intent recognition model includes a deep learning model such as a large language model (LLM), which is not limited in the present embodiment.

[0067] Exemplarily, an intent recognition model is constructed based on LLM, and the intent recognition model is trained so that it can understand the text, clarify the meaning of the text, and learn the classification criteria for different intent types, so that the execution intention that best matches the first call text data can be determined based on the classification criteria, and then the willingness level corresponding to the execution intention can be determined.

[0068] For example, in a debt collection scenario, the first party's predetermined execution intentions for the first business include: willingness to repay (unconditionally), willingness to repay (conditionally), inability to repay, refusal to repay, and other. Willingness to repay (unconditionally) typically corresponds to the first party directly expressing a willingness to repay (e.g., "Okay, I'll repay right away"); willingness to repay (conditionally) typically corresponds to the first party's willingness to repay but attaching conditions such as a repayment time or installment plan (e.g., "I'll repay tomorrow. Can we discuss paying the interest first?"); inability to repay typically corresponds to the first party being unable to repay due to objective reasons (e.g., "I'm short of money lately and can't repay"); refusal to repay typically corresponds to the first party's subjective unwillingness to repay (e.g., "I just won't repay"); and other refers to execution intentions that fall outside the aforementioned categories. Furthermore, the willingness level corresponding to willingness to repay (unconditionally) is determined to be 1, willingness to repay (conditionally) is determined to be 2, inability to repay is determined to be 3, refusal to repay is determined to be 4, and other is determined to be 3.

[0069] Based on this, the first call text data corresponding to the call voice data is input into the intention recognition model. After text understanding, intention recognition and other processing, the execution intention of the first object can be determined, and the willingness level corresponding to the execution intention can be determined, thereby obtaining the intention recognition result.

[0070] Step S23: performing call similarity matching in a preset call database based on the call voice data to obtain a call matching result of the first object.

[0071] In some optional embodiments, the preset call library is a database constructed using the second call text data within a preset time period (for example, historical call text data within a certain time period in the past) and the corresponding business execution results. The purpose of constructing the preset call library is to draw on or refer to historical information to evaluate the current business handling results of the first object. Among them, the business execution results correspond to the first business. For example, if the first business is a collection business, the business execution results may include the historical collection rate; if the first business is a package recommendation business, the business execution results may include the recommendation success rate.

[0072] In some optional embodiments, the business processing method may also include: obtaining second call text data and business execution results of multiple second objects in the process of processing the first business; clustering the second call text data based on text similarity to obtain multiple text clusters; for each text cluster, determining the target behavior execution rate of the text cluster according to the business execution results corresponding to the second call text data in the text cluster; and obtaining a preset call library based on the text clusters and the target behavior execution rates of the text clusters.

[0073] Exemplarily, for debt collection services, the business execution results may include historical collection rates, and the target behavior execution rate corresponds to the average collection rate. First, calls with collections can be screened out from the debt collection business scenario, and multiple second call text data and historical collection rates corresponding to these calls can be obtained. The second call text data are clustered based on text similarity to obtain multiple text clusters, and the second call text data contained in each text cluster have a high similarity. Furthermore, for each text cluster, the average is calculated based on the historical collection rates corresponding to the multiple second call text data in the text cluster to obtain the average collection rate corresponding to the text cluster, and the average collection rate is used as the target behavior execution rate for the text cluster. In the above manner, a preset call library can be obtained, which includes multiple text clusters and the target behavior execution rate corresponding to each text cluster.

[0074] It should be noted that the preset call library can be updated regularly or irregularly as needed to ensure the accuracy of call matching results. For example, the second call text data and service execution results in the preset call library can be updated, and / or the second call text data therein can be re-clustered as needed.

[0075] In some optional embodiments, call similarity matching is performed in a preset call library based on call voice data to obtain a call matching result of the first object, including: obtaining first call text data corresponding to the call voice data; determining a text cluster in the preset call library that matches the first call text data based on semantic similarity; and determining the call matching result based on the target behavior execution rate corresponding to the matched text cluster.

[0076] That is to say, after obtaining the first call text data of the first object, the text cluster that matches the first call text data (for example, the text cluster with the highest semantic similarity to the first call text data in the preset call library) can be determined by calculating the semantic similarity, and the target behavior execution rate corresponding to the text cluster can be determined as the target behavior execution rate of the first object, thereby obtaining a call matching result.

[0077] For example, for the debt collection scenario, the preset call library is determined based on multiple second call text data and the corresponding historical collection rates, including multiple text clusters, each text cluster corresponds to a target behavior execution rate (i.e., average collection rate), and the target behavior execution rate is essentially the average of multiple historical collection rates corresponding to the text cluster. After obtaining the first call text data of the first object, the semantic similarity between the first call text data and each text cluster can be calculated to determine the text cluster with the highest semantic similarity and the target behavior execution rate of the text cluster. The target behavior execution rate is used as the call matching result, indicating that the collection rate of the first object is the same as or relatively close to the value of the target behavior execution rate. In the above processing, the correlation between semantic similarity and collection rate is utilized, and the average collection rate of the text cluster with the highest semantic similarity is used to estimate the current collection rate of the first object.

[0078] For example, for business recommendation scenarios (such as recommending communication business packages to users), the preset call library is determined based on multiple second call text data and corresponding recommendation success rates, including multiple text clusters, each text cluster corresponds to a target behavior execution rate, and the target behavior execution rate is essentially the average of multiple recommendation success rates corresponding to the text cluster. After obtaining the first call text data of the first object, the semantic similarity between the first call text data and each text cluster can be calculated to determine the text cluster with the highest semantic similarity and the target behavior execution rate of the text cluster. The target behavior execution rate is used as the call matching result, indicating that the probability of the first object receiving the recommended service is the same as or relatively close to the value of the target behavior execution rate. In the above processing, the correlation between semantic similarity and recommendation success rate is utilized, and the recommendation success rate of the current first object is estimated by the average recommendation success rate of the text cluster with the highest semantic similarity.

[0079] It should be noted that in some optional embodiments, a similarity threshold may be pre-set. If the similarity between the first call text data of the first object and each text cluster is less than the similarity threshold, the first call text data is deemed to not match any of the text clusters, thereby obtaining a call matching result indicating a call mismatch.

[0080] Step S24 : performing behavior prediction based on the business execution status data of the first object corresponding to the first business to obtain a behavior prediction result of the first object.

[0081] The service execution status data may reflect the execution status of the first service by the first subject. For example, in a debt collection scenario, the service execution status data may include the first subject's overdue amount and number of days overdue; for example, in a communication package service recommendation scenario, the service execution status data may include the first subject's remaining call minutes, the number of call minutes within the package, and the average actual call minutes.

[0082] It should be noted that the above service execution status data is only an example, and the embodiments of the present disclosure do not limit this.

[0083] In some optional embodiments, performing behavior prediction based on business execution status data of a first object corresponding to a first business to obtain a behavior prediction result for the first object includes: inputting the business execution status data of the first business into a preset behavior prediction model to obtain the behavior prediction result; wherein the behavior prediction model is trained based on multiple business execution status data within a preset time period and the business execution results corresponding to each business execution status data. The preset time period can be a certain time period in the past, and accordingly, the multiple business execution status data within the preset time period are essentially multiple historical business execution status data, which is not limited in the present embodiment.

[0084] In some optional embodiments, in a collection scenario, a trained behavior prediction model can be obtained by training a preset behavior prediction model using the business execution status data of multiple second objects within a preset time period and the business execution results corresponding to each business execution status data.

[0085] For example, the behavior prediction model can use the formula P=1 / [1+e (a-b×A-c×D) ] is represented by, P represents the collection rate, and the collection rate P is the behavior prediction result, a represents the basic collection probability, b represents the influence factor of each unit of overdue amount on the collection probability, c represents the influence factor of each unit of overdue days on the collection probability, A represents the overdue amount, and D represents the number of overdue days. Among them, the basic collection probability a, the influence factor b of each unit of overdue amount on the collection probability, and the influence factor c of each unit of overdue days on the collection probability can be determined by model training using a large amount of sample data (such as historical data) (for example, a=5, b=0.02, c=0.2), and can also be adjusted accordingly according to actual business needs.

[0086] For example, when the overdue amount of the first object is 100 and the number of overdue days is 3, input it into the above-mentioned behavior prediction model (taking a=5, b=0.02, c=0.2 as an example), the collection rate P=0.917 can be output, indicating that the collection rate of the first object is 91.7%.

[0087] Step S25 , determining a processing strategy for the first object in the first service based on the emotion recognition result, the intention recognition result, the call matching result, and the behavior prediction result.

[0088] The processing strategy is a solution adopted when executing the first service on the first object. An accurate and reasonable processing strategy helps improve the processing efficiency and effect of the first service.

[0089] In some optional embodiments, a processing strategy for the first object on the first business is determined based on the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result, including: quantifying the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result respectively to obtain a first score corresponding to the emotion recognition result, a second score corresponding to the intention recognition result, a third score corresponding to the call matching result and a fourth score corresponding to the behavior prediction result; obtaining a fusion score based on the first score, the second score, the third score, the fourth score and a preset second weight coefficient; determining the processing strategy for the first object on the first business based on the correspondence between the fusion score and the preset score strategy, the score strategy correspondence being used to characterize the correspondence between the fusion score and the processing strategy.

[0090] It can be seen that when determining the processing strategy, the results of dimensions such as emotion recognition, intention recognition, call matching and behavior prediction are comprehensively considered, and the importance of different dimensions is weighed. A corresponding second weight coefficient is set for each dimension, and the fusion score is obtained by weighted summation, making the fusion score more accurate and reasonable, so that a processing strategy that is more compatible with the first object can be determined.

[0091] It should be noted that, for the call matching result, if it indicates a mismatch (ie, there is no matching text cluster in the preset call library), when quantizing it, its value can be quantized to zero or other preset values.

[0092] In some optional embodiments, if the emotion recognition results include positive emotions, neutral emotions and negative emotions, the first score corresponding to positive emotions is 1 point, neutral emotions are 0.5 points, and negative emotions are 0 points; the second score corresponding to the willingness level 1 in the intention recognition result is 1 point, the second score corresponding to the willingness level 2 is 0.8 points, the second score corresponding to the willingness level 3 is 0.2 points, the second score corresponding to the willingness level 4 is 0 points, and the second score corresponding to the willingness level 5 is 0.2 points; the value of the target behavior execution probability corresponding to the call matching result is taken as the third score; the value of the collection rate in the behavior prediction result is taken as the fourth score.

[0093] Assuming the preset second weight coefficients include w21 corresponding to the emotion recognition result, w22 corresponding to the intent recognition result, w23 corresponding to the call matching result, and w24 corresponding to the behavior prediction result, a fused score is obtained by weighting the first, second, third, and fourth scores using the second weight coefficients. In a debt collection scenario, this fused score can represent the repayment probability of the first subject.

[0094] Furthermore, the corresponding relationship between the scoring strategies can be expressed in Table 1.

[0095] Table 1 Correspondence between score strategies

[0096] Fusion score Probability level Treatment strategy [0.8，1) High probability level Natural payment [0.6，0.8) Medium probability level Robot follow-up (0，0.6) Low probability level Manual follow-up by agents

[0097] It can be seen from this that after determining the fusion score, the corresponding processing strategy can be determined in combination with Table 1, so as to perform the corresponding collection operation based on the processing strategy.

[0098] Similar to the debt collection scenario, in the business recommendation scenario, a fusion score can be obtained through corresponding weighted operations. This fusion score can represent the probability of the first object receiving the recommended business. Further combined with the score strategy correspondence, the corresponding processing strategy can be determined, and then the corresponding recommendation operation can be executed according to the processing strategy.

[0099] In some optional embodiments, to improve business processing efficiency and effectiveness, after determining the processing strategy, a call summary may be performed based on the aforementioned emotion recognition results, intent recognition results, call matching results, and behavior prediction results to generate a portrait of the first subject. This portrait information is then provided to the relevant terminals or personnel executing the processing strategy, enabling them to quickly and comprehensively grasp the relevant information about the first subject and facilitate the execution of the corresponding processing strategy. The portrait information of the first subject may reflect information such as the status of the first subject during the execution of the first business.

[0100] In some optional embodiments, profile information of the first object may be selectively generated. For example, profile information of the first object is generated only when a specific processing strategy (e.g., customer service manual follow-up) is determined based on the correspondence between the fusion score and the score strategy. This is primarily to assist in the execution of difficult-to-execute processing strategies by providing profile information, thereby reducing the execution difficulty and improving the effectiveness of business processing.

[0101] The LLM model can be used to generate the first subject's portrait information. For example, adaptive prompt words can be pre-written, and through model training, the LLM can learn and summarize the generation capabilities, thereby using the input emotion recognition results, intent recognition results, call matching results, and behavior prediction results to create a portrait of the first subject and output the portrait information of the first subject.

[0102] For example, the portrait information of the first object includes: the first object's identity is ID1, the communication mood is negative, and the intention to refuse repayment is clearly expressed during the call, accompanied by target words, the current overdue amount is 1,088 yuan, it has been overdue for 3 months, and it is a high-age account. Please adopt targeted collection strategies in the communication, it is recommended to soothe the emotions, explore the reasons for refusing repayment, and clearly inform the risks of continuing overdue.

[0103] For example, the portrait information of the first object includes: the first object's identity is ID2, the communication mood is relatively positive, and during the call, he expressed his intention to apply for a package, but he has doubts about the package charges. He is a potential customer. Please adopt targeted recommendation strategies in communication, clarify the content of his doubts, give detailed explanations, and continue to inform him of the benefits of applying for a package.

[0104] According to the business processing method of the embodiment of the present disclosure, emotion recognition is performed based on the call voice data of the first object in the first business process to obtain the emotion recognition result of the first object; intent recognition is performed based on the call voice data to obtain the intention recognition result of the first object; call similarity matching is performed in a preset call library based on the call voice data to obtain the call matching result of the first object; behavior prediction is performed based on the business execution status data of the first object corresponding to the first business to obtain the behavior prediction result of the first object; based on the emotion recognition result, intent recognition result, call matching result and behavior prediction result, the processing strategy for the first object in the first business is determined. Since multi-dimensional processing is performed from the aspects of emotion recognition, intent recognition, call matching and behavior prediction, emotion recognition results, intent recognition results, call matching results and behavior prediction results can be obtained, and the processing strategy can be determined based on the above multiple results. Therefore, the determined processing strategy is more matched with the first object, thereby improving the processing effect of the first business.

[0105] Figure 3 A flowchart of a business processing method provided by an embodiment of the present disclosure. Figure 3 , the method comprising:

[0106] In step S301, the customer service establishes a call connection with the first party through the outbound call system to process a first service, and obtains the call voice data of the first party during the process of processing the first service.

[0107] Step S302: Acquire speech features and target word detection results of call speech data.

[0108] Step S303 : performing emotion recognition based on the speech features and the target word detection result to obtain an emotion recognition result of the first object.

[0109] Step S304: Acquire first call text data corresponding to the call voice data.

[0110] Step S305 : identifying the execution intention of the first object for the first service based on the first call text data.

[0111] Step S306: classify the execution intentions, obtain the willingness levels corresponding to the execution intentions, and determine the intention recognition results.

[0112] Step S307: determining a text cluster matching the first call text data in a preset call library based on semantic similarity.

[0113] Step S308: Determine the call matching result based on the target behavior execution rate corresponding to the matched text cluster.

[0114] Step S309 : performing behavior prediction based on the business execution state data of the first object corresponding to the first business to obtain a behavior prediction result of the first object.

[0115] Step S310, determining a first score corresponding to the emotion recognition result, a second score corresponding to the intention recognition result, a third score corresponding to the call matching result, and a fourth score corresponding to the behavior prediction result.

[0116] Step S311: Obtain a fusion score based on the first score, the second score, the third score, the fourth score, and a preset second weight coefficient.

[0117] Step S312: Determine a processing strategy for the first object in the first service according to the correspondence between the fusion score and the preset score strategy.

[0118] After the processing strategy is determined, relevant personnel may execute the processing strategy to complete the first business.

[0119] Figure 4 A schematic diagram of the processing process of a business processing method provided in an embodiment of the present disclosure.

[0120] Reference Figure 4 The customer service representative establishes a call connection with the first party through the outbound call system to handle the first service. The customer service representative can use voice collection technology to obtain the call voice data of the first party during the process of handling the first service. The obtained call voice data can be analyzed and processed at the voice level and the corresponding text level.

[0121] like Figure 4 As shown, for call speech data, on the one hand, its speech features are extracted and emotion recognition is performed based on the speech features to obtain feature recognition results. On the other hand, target word detection is performed on the call speech data using a pre-set target word library to determine whether the call speech data meets the preset hit rules and obtain target word detection results. Furthermore, after obtaining the feature recognition results and target word detection results, the emotion recognition results can be obtained based on both.

[0122] In addition, the call voice data can be converted into corresponding first call text data, and corresponding data processing can be performed based on the first call text data from the two dimensions of intent recognition and call matching. When performing intent recognition, the intention recognition model can be used to semantically understand the first call text data, thereby determining the execution intention of the first object, and dividing the execution intention into corresponding willingness levels to obtain the intention recognition result. In addition, the first call text data can be matched with the text clusters in the preset call library for semantic similarity to determine which text cluster in the preset call library has the highest semantic similarity with the first call text data, and then the call matching result of the first object can be determined based on the target behavior execution rate corresponding to the text cluster with the highest semantic similarity.

[0123] Furthermore, the business execution status data of the first object corresponding to the first business can be obtained, and behavior prediction can be performed based on the business execution status data and using a behavior prediction model to obtain a behavior prediction result of the first object.

[0124] After obtaining the emotion recognition results, intent recognition results, call matching results, and behavior prediction results, these four results are quantified to obtain corresponding first, second, third, and fourth scores. These four scores are then weighted using a preset second weighting coefficient to obtain a fused score, which represents the likelihood of repayment by the first subject. After obtaining the fused score, a processing strategy for the first subject in the first service is determined based on the correspondence between the fused score and the preset score strategy.

[0125] like Figure 4 As shown, if the fusion score corresponds to a high probability level (the probability level represents the high or low probability of repayment), the natural repayment processing strategy is adopted; if the fusion score corresponds to a medium probability level, the robot follow-up processing strategy is adopted; if the fusion score corresponds to a low probability level, the agent manual follow-up processing strategy is adopted.

[0126] In addition, in some optional embodiments, for situations where the processing strategy is manual follow-up by the agent, a preset portrait model can also be used. The portrait model provides portrait information of the first object based on the four processing results of emotion recognition results, intention recognition results, call matching results, and behavior prediction results, for reference by the agent when executing the processing strategy.

[0127] According to the business processing method of the embodiment of the present disclosure, the first object can be analyzed and processed from the dimensions of voice, semantics, business execution status, etc. through emotion recognition, intention recognition, call matching, and behavior prediction, thereby obtaining emotion recognition results, intention recognition results, call matching results, and behavior prediction results, and can comprehensively and comprehensively determine the processing strategy for the first object in the first business based on the above four results, so that the determined processing strategy is more matched with the first object, thereby improving the processing effect of the first business.

[0128] It should be noted that in the embodiment of the present disclosure, the personal information of the first object is obtained after obtaining his authorization and consent.

[0129] It is understood that the above-mentioned various method embodiments mentioned in this disclosure can be combined with each other to form combined embodiments without violating the principle logic. Due to space limitations, this disclosure will not go into details. It is understood by those skilled in the art that in the above-mentioned methods of specific implementation, the specific execution order of each step should be determined by its function and possible internal logic.

[0130] In addition, the present disclosure also provides a business processing device, an electronic device, and a computer-readable storage medium, all of which can be used to implement any business processing method provided by the present disclosure. The corresponding technical solutions and descriptions are referred to the corresponding records in the method part and will not be repeated here.

[0131] Figure 5 A block diagram of a service processing device provided in an embodiment of the present disclosure.

[0132] Reference Figure 5 , an embodiment of the present disclosure provides a business processing device, and the business processing device 500 includes the following modules.

[0133] The first recognition module 501 is configured to perform emotion recognition based on the call voice data of the first subject during the first service process to obtain an emotion recognition result of the first subject.

[0134] The second recognition module 502 is configured to perform intention recognition based on the call voice data to obtain an intention recognition result of the first object.

[0135] The matching module 503 is configured to perform call similarity matching in a preset call database according to the call voice data to obtain a call matching result of the first object.

[0136] The prediction module 504 is configured to perform behavior prediction based on the business execution state data of the first object corresponding to the first business, and obtain a behavior prediction result of the first object.

[0137] The determination module 505 is used to determine a processing strategy for the first object in the first service based on the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result.

[0138] In some optional embodiments, the first recognition module 501 performs emotion recognition based on the call voice data of the first object in the first business process to obtain an emotion recognition result of the first object, including: obtaining the voice features and target word detection results of the call voice data, the voice features are used to characterize the voice state information of the first object during the call, and the target word detection results are used to characterize the hit information of the call voice data for the preset target words; performing emotion recognition based on the voice features and the target word detection results to obtain an emotion recognition result.

[0139] In some optional embodiments, the speech features include intonation sub-features, speaking speed sub-features, and volume sub-features; accordingly, emotion recognition is performed based on the speech features and the target word detection results to obtain emotion recognition results, including: determining the intonation recognition sub-result corresponding to the intonation sub-feature; determining the speaking speed recognition sub-result corresponding to the speaking speed sub-feature; determining the volume recognition sub-result corresponding to the volume sub-feature; fusing the intonation recognition sub-result, speaking speed recognition sub-result, and volume recognition sub-result to obtain a feature recognition result; and determining the emotion recognition result based on the feature recognition result and the target word detection result.

[0140] In some optional embodiments, the intonation recognition sub-result includes an intonation score, the speech rate recognition sub-result includes a speech rate score, the volume recognition sub-result includes a volume score, and the feature recognition result is a feature weighted value determined based on a preset first weight coefficient, the intonation score, the speech rate score, and the volume score; accordingly, the target word detection result is determined based on the call voice data and a preset hit rule, and the hit rule is used to characterize the number of preset target words contained in the call voice data in a hit situation and the number of occurrences of the preset target words; determining the emotion recognition result based on the feature recognition result and the target word detection result includes: when the target word detection result indicates that the call voice data meets the preset hit rule, determining the emotion recognition result as a negative emotion; When the target word detection result indicates that the call voice data does not meet the preset hit rules and the feature weighted value is within the first preset interval, the emotion recognition result is determined to be a positive emotion; when the target word detection result indicates that the call voice data does not meet the preset hit rules and the feature weighted value is within the second preset interval, the emotion recognition result is determined to be a neutral emotion; when the target word detection result indicates that the call voice data does not meet the preset hit rules and the feature weighted value is within the third preset interval, the emotion recognition result is determined to be a negative emotion; wherein, the feature weighted value corresponding to the first preset interval is higher than the feature weighted value corresponding to the second preset interval, and the feature weighted value corresponding to the second preset interval is higher than the feature weighted value corresponding to the third preset interval.

[0141] In some optional embodiments, the intention recognition result includes the willingness level of the first object to execute the first business; the second recognition module 502 performs intention recognition based on the call voice data to obtain the intention recognition result of the first object, including: obtaining the first call text data corresponding to the call voice data; identifying the execution intention of the first object for the first business based on the first call text data; classifying the execution intention to obtain the willingness level corresponding to the execution intention.

[0142] In some optional embodiments, the business processing device 500 is also used to perform the following: obtain the second call text data and business execution results of multiple second objects in the process of processing the first business; cluster the second call text data based on text similarity to obtain multiple text clusters; for each text cluster, determine the target behavior execution rate of the text cluster according to the business execution results corresponding to the multiple second call text data in the text cluster; obtain a preset call library based on the text cluster and the target behavior execution rate of the text cluster.

[0143] In some optional embodiments, the matching module 503 performs similarity matching in a preset call library based on the call voice data to obtain a call matching result of the first object, including: obtaining the first call text data corresponding to the call voice data; determining a text cluster in the preset call library that matches the first call text data based on semantic similarity; and determining the call matching result based on the target behavior execution rate corresponding to the matched text cluster.

[0144] In some optional embodiments, the prediction module 504 performs behavior prediction based on the business execution status data of the first object corresponding to the first business to obtain a behavior prediction result of the first object, including: inputting the business execution status data of the first business into a preset behavior prediction model to obtain a behavior prediction result; wherein the behavior prediction model is trained based on multiple business execution status data within a preset time period and the business execution results corresponding to each business execution status data.

[0145] In some optional embodiments, the determination module 505 determines the processing strategy for the first object on the first business based on the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result, including: quantifying the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result respectively to obtain a first score corresponding to the emotion recognition result, a second score corresponding to the intention recognition result, a third score corresponding to the call matching result and a fourth score corresponding to the behavior prediction result; obtaining a fusion score based on the first score, the second score, the third score, the fourth score and a preset second weight coefficient; determining the processing strategy for the first object on the first business based on the correspondence between the fusion score and the preset score strategy, the score strategy correspondence being used to characterize the correspondence between the fusion score and the processing strategy.

[0146] The business processing device provided by the embodiment of the present disclosure uses a first recognition module to perform emotion recognition based on the call voice data of the first object during the first business process to obtain an emotion recognition result of the first object; uses a second recognition module to perform intent recognition based on the call voice data to obtain an intent recognition result of the first object; uses a matching module to perform call similarity matching in a preset call library based on the call voice data to obtain a call matching result of the first object; uses a prediction module to perform behavior prediction based on the business execution status data of the first object corresponding to the first business to obtain a behavior prediction result of the first object; uses a determination module to determine a processing strategy for the first object in the first business based on the emotion recognition result, intent recognition result, call matching result, and behavior prediction result. Due to multi-dimensional processing from the aspects of emotion recognition, intent recognition, call matching, and behavior prediction, emotion recognition results, intent recognition results, call matching results, and behavior prediction results can be obtained, and a processing strategy can be determined based on the above multiple results. Therefore, the determined processing strategy is more compatible with the first object, thereby improving the processing effect of the first business.

[0147] Each module in the above-mentioned business processing device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0148] Figure 6 A block diagram of an electronic device provided in an embodiment of the present disclosure.

[0149] Reference Figure 6 An embodiment of the present disclosure provides an electronic device, which includes: at least one processor 601; at least one memory 602, and one or more I / O interfaces 603, connected between the processor 601 and the memory 602; wherein the memory 602 stores one or more computer programs that can be executed by the at least one processor 601, and the one or more computer programs are executed by the at least one processor 601 to enable the at least one processor 601 to execute the above-mentioned business processing method.

[0150] Each module in the above-mentioned electronic device can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0151] The present disclosure also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the above-mentioned business processing method when executed by a processor. The computer-readable storage medium may be a volatile or non-volatile computer-readable storage medium.

[0152] An embodiment of the present disclosure also provides a computer program product, including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code. When the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the above-mentioned business processing method.

[0153] It will be understood by those skilled in the art that all or some of the steps, systems, and functional modules / units in the methods disclosed above may be implemented as software, firmware, hardware, and appropriate combinations thereof. In a hardware implementation, the division between the functional modules / units mentioned in the above description does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed by several physical components in cooperation. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software may be distributed on a computer-readable storage medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or temporary medium).

[0154] As is well known to those skilled in the art, the term computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information (such as computer-readable program instructions, data structures, program modules or other data). Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technology, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those skilled in the art, communication media typically contains computer-readable program instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

[0155] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions to be stored in the computer-readable storage medium in each computing / processing device.

[0156] The computer program instructions for performing the operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, and conventional procedural programming languages ​​such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., utilizing an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions. The electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present disclosure.

[0157] The computer program product described herein may be implemented in hardware, software, or a combination thereof. In one embodiment, the computer program product is implemented as a computer storage medium. In another embodiment, the computer program product is implemented as a software product, such as a software development kit (SDK).

[0158] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0159] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0160] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0161] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and the part of the module, program segment or instruction contains one or more executable instructions for realizing the prescribed logical function. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the prescribed function or action, or can be implemented by a combination of dedicated hardware and computer instructions.

[0162] Example embodiments have been disclosed herein, and although specific terms are employed, they are used and should be interpreted only in a general illustrative sense and not for purposes of limitation. In some instances, it will be apparent to those skilled in the art that, unless otherwise expressly indicated, features, characteristics, and / or elements described in conjunction with a particular embodiment may be used alone or in combination with features, characteristics, and / or elements described in conjunction with other embodiments. Therefore, it will be understood by those skilled in the art that various changes in form and detail may be made without departing from the scope of the present disclosure as set forth in the appended claims.

Claims

1. A business processing method, characterized in that: include: performing emotion recognition based on the call voice data of the first subject during the first service process to obtain an emotion recognition result of the first subject; Performing intent recognition based on the call voice data to obtain an intent recognition result of the first object; Performing similarity matching in a preset call database based on the call voice data to obtain a call matching result of the first object; performing behavior prediction based on the business execution status data of the first object corresponding to the first business, to obtain a behavior prediction result of the first object; Determine a processing strategy for the first object on the first business based on the emotion recognition result, the intention recognition result, the call matching result, and the behavior prediction result.

2. The method according to claim 1, characterized in that The performing emotion recognition based on the call voice data of the first subject in the first service process to obtain the emotion recognition result of the first subject includes: Acquire speech features and target word detection results of the call voice data, wherein the speech features are used to characterize speech state information of the first subject during the call, and the target word detection results are used to characterize hit information of the call voice data for a preset target word; Emotion recognition is performed based on the speech features and the target word detection result to obtain the emotion recognition result.

3. The method according to claim 2, characterized in that The speech features include intonation sub-features, speech speed sub-features, and volume sub-features; The performing emotion recognition according to the speech feature and the target word detection result to obtain the emotion recognition result includes: Determining a tone recognition sub-result corresponding to the tone sub-feature; Determine a speech rate identification sub-result corresponding to the speech rate sub-feature; Determining a volume identification sub-result corresponding to the volume sub-feature; Fusing the intonation recognition sub-result, the speech rate recognition sub-result, and the volume recognition sub-result to obtain a feature recognition result; The emotion recognition result is determined according to the feature recognition result and the target word detection result.

4. The method according to claim 3, characterized in that The intonation recognition sub-result includes an intonation score, the speaking rate recognition sub-result includes a speaking rate score, and the volume recognition sub-result includes a volume score. The feature recognition result is a feature weighted value determined based on a preset first weight coefficient, the intonation score, the speaking rate score, and the volume score. The target word detection result is determined based on the call voice data and a preset hit rule, wherein the hit rule is used to characterize the number of preset target words contained in the call voice data and the number of occurrences of the preset target words in a hit situation; Determining the emotion recognition result according to the feature recognition result and the target word detection result includes: When the target word detection result indicates that the call voice data meets the preset hit rule, determining that the emotion recognition result is a negative emotion; When the target word detection result indicates that the call voice data does not meet the preset hit rule and the feature weight value is within a first preset interval, determining that the emotion recognition result is a positive emotion; When the target word detection result indicates that the call voice data does not meet the preset hit rule and the feature weight value is within a second preset interval, determining that the emotion recognition result is a neutral emotion; When the target word detection result indicates that the call voice data does not meet the preset hit rule, and the feature weight value is within a third preset interval, determining that the emotion recognition result is the negative emotion; The feature weighted value corresponding to the first preset interval is higher than the feature weighted value corresponding to the second preset interval, and the feature weighted value corresponding to the second preset interval is higher than the feature weighted value corresponding to the third preset interval.

5. The method according to claim 1, wherein The intention recognition result includes the willingness level of the first subject to perform the first business; The performing intent recognition according to the call voice data to obtain an intent recognition result of the first object includes: Acquire first call text data corresponding to the call voice data; identifying, based on the first call text data, an execution intention of the first object for the first service; The execution intentions are classified to obtain willingness levels corresponding to the execution intentions.

6. The method according to claim 1, characterized in that The method further comprises: Acquire second call text data and service execution results of multiple second objects in the process of processing the first service; Clustering the second call text data based on text similarity to obtain multiple text clusters; For each text cluster, determining a target behavior execution rate of the text cluster according to the service execution results corresponding to the plurality of second call text data in the text cluster; The preset call library is obtained based on the text cluster and the target behavior execution rate of the text cluster.

7. The method according to claim 6, characterized in that The performing similarity matching in a preset call database according to the call voice data to obtain a call matching result of the first object includes: Acquire first call text data corresponding to the call voice data; Determining a text cluster in the preset call library that matches the first call text data based on semantic similarity; The call matching result is determined according to the target behavior execution rate corresponding to the matched text cluster.

8. The method according to claim 1, characterized in that The performing behavior prediction based on the business execution status data of the first object corresponding to the first business to obtain a behavior prediction result of the first object includes: Inputting the business execution status data of the first business into a preset behavior prediction model to obtain the behavior prediction result; The behavior prediction model is obtained by training based on a plurality of business execution status data within a preset time period and a business execution result corresponding to each business execution status data.

9. The method according to claim 1, characterized in that The determining, based on the emotion recognition result, the intention recognition result, the call matching result, and the behavior prediction result, a processing strategy for the first object in the first service includes: quantifying the emotion recognition result, the intention recognition result, the call matching result, and the behavior prediction result respectively to obtain a first score corresponding to the emotion recognition result, a second score corresponding to the intention recognition result, a third score corresponding to the call matching result, and a fourth score corresponding to the behavior prediction result; Obtaining a fusion score based on the first score, the second score, the third score, the fourth score, and a preset second weight coefficient; A processing strategy for the first object in the first service is determined according to a correspondence between the fusion score and a preset score strategy, wherein the score strategy correspondence is used to characterize a correspondence between the fusion score and the processing strategy.

10. A business processing device, characterized in that: include: a first recognition module, configured to perform emotion recognition based on call voice data of a first subject during a first service process, and obtain an emotion recognition result of the first subject; a second recognition module, configured to perform intent recognition based on the call voice data to obtain an intent recognition result of the first object; a matching module, configured to perform call similarity matching in a preset call library based on the call voice data to obtain a call matching result for the first object; A prediction module, configured to perform behavior prediction based on the business execution status data of the first object corresponding to the first business, to obtain a behavior prediction result of the first object; A determination module is used to determine a processing strategy for the first object in the first business based on the emotion recognition result, the intention recognition result, the call matching result and the behavior prediction result.

11. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores one or more computer programs that can be executed by the at least one processor. The one or more computer programs are executed by the at least one processor to enable the at least one processor to execute the business processing method according to any one of claims 1 to 9.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program implements the service processing method according to any one of claims 1 to 9.

13. A computer program product, characterized in that A computer-readable code or a non-volatile computer-readable storage medium carrying a computer-readable code, wherein when the computer-readable code runs in a processor of an electronic device, the processor in the electronic device executes the business processing method according to any one of claims 1 to 9.

Images