Data processing method and device, information pushing method and device, medium and equipment

Abstract

The invention relates to the technical field of data processing, and provides a data processing method and device, an information pushing method and device, a computer readable storage medium, a computer program product and electronic equipment. The data processing method comprises the following steps: determining a second voiceprint feature similar to each first voiceprint feature, the first voiceprint feature corresponding to an identifier, and the second voiceprint feature corresponding to an identifier; determining a first number of different identifiers in the identifiers; and determining a risk identification result of the first voiceprint feature according to the first number. The risk identification accuracy and identification efficiency can be improved.

Description

Technical Field

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

[0002] When providing business services, many companies will authenticate users to ensure business security. However, due to the possibility of identity information being forged or stolen, some criminals will use forged or stolen identity information to conduct illegal activities, which poses a threat to the security of enterprise services. Therefore, there is an urgent need for a method that can identify risk objects to improve the security of enterprise services. Summary of the invention

[0003] The purpose of the present disclosure is to provide a data processing method and device, an information push method and device, a computer-readable storage medium, a computer program product and an electronic device, so as to improve the accuracy and efficiency of identifying risky users.

[0004] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0005] According to a first aspect of the present disclosure, a data processing method is provided, including: determining a second voiceprint feature similar to each first voiceprint feature, the first voiceprint feature corresponding to an identifier, and the second voiceprint feature corresponding to an identifier; determining a first number of different identifiers in the identifiers; and determining a risk identification result of the first voiceprint feature based on the first number.

[0006] According to the second aspect of the present disclosure, there is provided an information push method, including: determining a fifth voiceprint feature similar to a fourth voiceprint feature corresponding to each user's voice, wherein the fourth voiceprint feature corresponds to an identifier and the fifth voiceprint feature corresponds to an identifier; determining a third number of different identifiers in the identifiers; determining a risk identification result of the fourth voiceprint feature based on the third number; and, when the risk identification result is a target risk identification result, pushing prompt information to a terminal, the target risk identification result indicating that the fourth voiceprint feature has a risk, and the prompt information indicating that the user corresponding to the user's voice is a risky user.

[0007] According to a third aspect of the present disclosure, there is provided a data processing device, comprising: a first similar voiceprint determination module, configured to determine a second voiceprint feature similar to each first voiceprint feature, the first voiceprint feature corresponds to an identifier, and the second voiceprint feature corresponds to an identifier; a first quantity determination module, configured to determine a first number of different identifiers in the identifiers; and a first risk identification result determination module, configured to determine a risk identification result of the first voiceprint feature based on the first number.

[0008] According to a fourth aspect of the present disclosure, an information push device is provided, comprising: a second similar voiceprint determination module, configured to determine a fifth voiceprint feature similar to a fourth voiceprint feature corresponding to each user's voice, wherein the fourth voiceprint feature corresponds to an identifier and the fifth voiceprint feature corresponds to an identifier; a third quantity determination module, configured to determine a third quantity of different identifiers in the identifiers; a second risk identification result determination module, configured to determine a risk identification result of the fourth voiceprint feature according to the third quantity; an information push module, configured to push prompt information to a terminal when the risk identification result is a target risk identification result, the target risk identification result indicating that the fourth voiceprint feature is risky, and the prompt information indicating that the user corresponding to the user's voice is a risky user.

[0009] According to a fifth aspect of the present disclosure, a computer program product comprising instructions is provided, which, when executed on a computer, enables the computer to execute the steps of the data processing method described in the first aspect and / or the steps of the information pushing method described in the second aspect.

[0010] According to a sixth aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the data processing method described in the first aspect and / or the information push method described in the second aspect of the above embodiments is implemented.

[0011] According to the seventh aspect of the embodiments of the present disclosure, an electronic device is provided, comprising: a processor; and a storage device for storing one or more programs, which, when the one or more programs are executed by the one or more processors, enables the one or more processors to implement the data processing method described in the first aspect and / or the information push method described in the second aspect of the above embodiments.

[0012] It can be seen from the above technical solutions that the data processing method, information push method, data processing device, information push device, and computer program product, computer-readable storage medium, and electronic device implementing the data processing method and / or the information push method in the exemplary embodiments of the present disclosure have at least the following advantages and positive effects:

[0013] In the technical solutions provided by some embodiments of the present disclosure, by processing voiceprint features, the risk identification results of voiceprint features can be determined according to the number of different identifiers corresponding to similar voiceprint features. Compared with the related art, on the one hand, the present disclosure can automatically identify risks without human intervention by processing voiceprint features, thereby improving the efficiency of risk identification; on the other hand, since voiceprint features are difficult to forge, the accuracy of risk identification can be improved by performing risk identification through voiceprint feature processing.

[0014] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0016] Figure 1 A schematic diagram showing an exemplary system architecture to which embodiments of the present disclosure may be applied;

[0017] Figure 2 A schematic diagram showing a flow chart of a data processing method in an exemplary embodiment of the present disclosure;

[0018] Figure 3 A schematic diagram showing a flow chart of a method for determining a first call voice in an exemplary embodiment of the present disclosure;

[0019] Figure 4 A schematic diagram showing a flow chart of a method for determining a second voiceprint feature in an exemplary embodiment of the present disclosure;

[0020] Figure 5 A schematic flow chart showing another method for determining a second voiceprint feature in an exemplary embodiment of the present disclosure;

[0021] Figure 6 A schematic flow chart showing a method for determining a risk identification result of a first voiceprint feature in an exemplary embodiment of the present disclosure;

[0022] Figure 7 A schematic flow chart showing a method for determining a first intention similarity in an exemplary embodiment of the present disclosure;

[0023] Figure 8 A schematic diagram showing the structure of a data processing system in an exemplary embodiment of the present disclosure is shown;

[0024] Fig. 9 A schematic diagram showing a flow chart of an information push method in an exemplary embodiment of the present disclosure;

[0025] Fig.10 A schematic diagram showing a multi-party interaction process when information is pushed in an exemplary embodiment of the present disclosure;

[0026] Fig.11 A schematic diagram showing the composition of a data processing device in an exemplary embodiment of the present disclosure;

[0027] Fig.12 A schematic diagram showing the composition of an information push device in an exemplary embodiment of the present disclosure;

[0028] Fig.13 A schematic structural diagram of an electronic device in an exemplary embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0029] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present disclosure will be more comprehensive and complete, and the concepts of the example embodiments are fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0030] The terms "a", "an", "the" and "said" are used in this specification to indicate the presence of one or more elements / components / etc.; the terms "including" and "having" are used to express an open-ended inclusion and mean that additional elements / components / etc. may exist in addition to the listed elements / components / etc.; the terms "first" and "second" etc. are used only as labels and are not intended to limit the quantity of their objects.

[0031] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0032] First, some technical terms of the present disclosure are explained to facilitate a better understanding of the technical content of the present disclosure.

[0033] Voiceprint features: a general term for the voice features contained in speech that can characterize and identify the speaker, as well as the voice model established based on these features (parameters). Voiceprints are biological characteristics of people, and it is difficult to find two people with exactly the same voiceprints. Voiceprint features include but are not limited to the frequency value and trend of the resonance peak, and these features are highly specific and stable.

[0034] First voiceprint feature: the voiceprint feature extracted in real time based on the user's real-time voice.

[0035] Second voiceprint feature: an existing voiceprint feature, which may be a voiceprint feature extracted based on a generated user voice, such as a voiceprint feature in the first voiceprint library.

[0036] First voiceprint database: a database used to store existing voiceprint features.

[0037] Second voiceprint database: a database used to store risky voiceprint features.

[0038] First identifier: a user identifier corresponding to the first voiceprint feature. The first identifier can be determined based on the user identifier corresponding to the user voice that generates the first voiceprint feature.

[0039] Second identifier: the user identifier corresponding to the second voiceprint feature. The second identifier can be determined based on the user identifier corresponding to the user voice that generates the second voiceprint feature.

[0040] Candidate voiceprint feature pair: a voiceprint feature pair consisting of two voiceprint features with different identifiers in the first voiceprint feature and the second voiceprint feature.

[0041] In response to the problem in the related art that voiceprints in the blacklist voiceprint library need to rely on manual determination, and have low accuracy and efficiency, the inventors of the present disclosure discovered that the key point in identifying blacklist voiceprints is that the same person may handle business in the name of different users. In this way, by identifying similar voiceprints with multiple user identifiers, blacklist voiceprints, i.e., risk voiceprints, can be identified, thereby improving recognition accuracy and efficiency.

[0042] Next, combine Figure 1 The system architecture of an exemplary application environment of the present exemplary embodiment will be described.

[0043] like Figure 1 As shown, the system architecture 100 may include a terminal device 110 and a server 120. The terminal device 110 may be a terminal device such as a smart phone, a tablet computer, a desktop computer, a laptop computer, a smart wearable device, etc. The server 120 generally refers to a background system that provides services related to the data processing method and / or the information push method in this exemplary embodiment, and may be a server or a cluster formed by multiple servers. The terminal device 110 and the server 120 may be connected via a wired or wireless communication link to perform data exchange.

[0044] In an exemplary implementation, the above data processing method may be performed by the terminal device 110. Accordingly, the data processing device may be provided in the terminal device 110 to implement the corresponding module functions. For example, the terminal device 110 may determine the second voiceprint feature similar to each first voiceprint feature according to the similarity, and then obtain the identifier of the first voiceprint feature and the identifier of the second voiceprint feature, determine the first number of different identifiers in the identifier, and determine the risk identification result of the first voiceprint feature according to the first number.

[0045] In an exemplary implementation, the above data processing method may be performed by the server 120. Accordingly, the data processing device may be set in the server 120 to implement the corresponding module functions. For example, the server 120 determines the second voiceprint feature similar to each first voiceprint feature according to the similarity, then obtains the identifier of the first voiceprint feature and the identifier of the second voiceprint feature, determines the first number of different identifiers in the identifier, and determines the risk identification result of the first voiceprint feature according to the first number.

[0046] In an exemplary embodiment, the above-mentioned information push method can be executed by the terminal device 110. Accordingly, the information push device can be set in the terminal device 110 to realize the corresponding module function. For example, through the terminal device 110 and the user communicating by phone, the terminal device 110 can obtain the user's voice during the telephone communication, extract the fourth voiceprint feature corresponding to the user's voice, determine the fifth voiceprint feature similar to the fourth voiceprint feature, obtain the identifier of the user's voice and the identifier of the fifth voiceprint feature, determine the third number of different identifiers in the identifier, and when it is determined that the fourth voiceprint feature is at risk according to the third number, a prompt message is displayed in the graphical user interface of the terminal device 110 to remind the user of the terminal device 110 that the user communicating with him by phone is a risky user.

[0047] In an exemplary embodiment, the above-mentioned information push method can be executed by the server 120. Accordingly, the information push device can be set in the server 120 to realize the corresponding module function. For example, in the process of communicating with the user by phone through the terminal device 110, the terminal device 110 can send the user voice in the real-time voice stream of the telephone communication to the server 120. After receiving the user voice, the server 120 can extract the fourth voiceprint feature of the user voice, determine the fifth voiceprint feature similar to the fourth voiceprint feature, obtain the identifier of the user voice and the identifier of the fifth voiceprint feature, determine the third number of different identifiers in the identifier, and push a prompt message to the terminal device 110 when it is determined that the fourth voiceprint feature is at risk according to the third number, so as to prompt the user of the terminal device 110 that the user with whom the phone communication is being conducted is a risky user.

[0048] It should be understood that Figure 1 The number of terminal devices and servers in the example is only illustrative. Any number of terminal devices and servers may be provided as required. For example, the server 120 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0049] It is easy for those skilled in the art to understand that the above application scenarios are only used as examples and are not limited to the present exemplary embodiment.

[0050] Figure 2 A data processing method in an exemplary embodiment of the present disclosure is shown, referring to Figure 2 , the method comprising:

[0051] Step S210, determining a second voiceprint feature similar to each first voiceprint feature, wherein the first voiceprint feature corresponds to an identifier and the second voiceprint feature corresponds to an identifier;

[0052] Step S220, determining a first number of different identifiers in the identifier;

[0053] Step S230: determining a risk identification result of the first voiceprint feature according to the first quantity.

[0054] exist Figure 2In the technical solution provided by the illustrated embodiment, by processing the voiceprint features, the risk identification result of the voiceprint features can be determined according to the number of different identifiers corresponding to similar voiceprint features. Compared with the related art, on the one hand, the present disclosure can automatically identify risks without human intervention by processing the voiceprint features, thereby improving the efficiency of risk identification; on the other hand, since voiceprint features are difficult to forge, the accuracy of risk identification can be improved by processing the voiceprint features for risk identification.

[0055] Below, first, the specific implementation method of step S210 of determining the second voiceprint feature similar to each first voiceprint feature is described in detail.

[0056] In an exemplary embodiment, the first voiceprint feature includes a real-time voiceprint feature, and the second voiceprint feature includes a generated voiceprint feature. The real-time generated voiceprint feature can be understood as a voiceprint feature obtained by performing voiceprint extraction on real-time user voice, such as performing voiceprint extraction on user voice in a real-time call voice stream to obtain the first voiceprint feature. The generated voiceprint feature can be understood as a voiceprint feature obtained by performing voiceprint extraction on historical user voice.

[0057] In another exemplary embodiment, the second voiceprint feature includes a voiceprint feature corresponding to the first call speech.

[0058] Next, combine Figure 3 An exemplary method for determining the first call voice is described. Figure 3 A flowchart of a method for determining a first call voice in an exemplary embodiment of the present disclosure is shown. Figure 3 , the method may include steps S310 to S330. Wherein:

[0059] In step S310, a first work order corresponding to a preset work order type is determined among the generated work orders.

[0060] In an exemplary embodiment, the generated work order can be determined according to the preset cycle and the first preset duration. For example, the work order within the first preset duration can be regularly obtained according to the preset cycle. The preset cycle and the first preset duration can be customized according to user needs or experience, such as obtaining the work order generated the previous day at 2 a.m. every day and treating the work order as the generated work order, or obtaining the work order generated the previous week at 3 a.m. every Sunday and treating the work order as the generated work order.

[0061] For example, at the moment corresponding to each preset period, the work orders whose time from the moment is within the first preset time period can be obtained, so as to determine the generated work orders. For example, the work orders created N days ago can be obtained at the moment corresponding to each preset period, and the work orders created N days ago can be determined as the generated work orders.

[0062] After determining the generated work orders, the work order corresponding to the preset work order type in the generated work orders can be determined as the first work order. For example, at 2 a.m. every day, the work order created the previous day is obtained, the work order created the previous day is determined as the generated work order, and then the work order of the preset work order type in the generated work order is determined as the first work order. It is also possible to obtain the work orders created two days ago at 3 a.m. every other day, determine the work orders created two days ago as the generated work orders, and then determine the work order of the preset work order type in the generated work orders as the first work order according to the preset work order type.

[0063] In an exemplary embodiment, a work order is a work document, which generally refers to a document that records a task, event, or operation process, such as a document that records the business processing process when an enterprise is processing a business. The enterprise can classify the work orders according to the business type, thereby obtaining different types of work orders.

[0064] In an exemplary embodiment, the preset work order type can be customized according to the needs. If it is found in actual business that users corresponding to certain types of work orders have serious suspicious behaviors, these types can be used as preset work order types. Of course, the preset work order type can also be determined according to other methods, and this exemplary embodiment does not specifically limit this.

[0065] By presetting the work order type, the work orders can be preliminarily screened, so that the work orders with a high possibility of suspicious objects can be processed in a targeted manner without processing all the work orders, thereby improving the risk identification efficiency of the first voiceprint feature.

[0066] In step S320, a third identifier of the first work order is determined.

[0067] In an exemplary embodiment of the present disclosure, a work order may include a business work order when an enterprise processes business over the phone. The business work order records the processed business information, such as business type, business content, business processing results, storage address of call recordings during the business processing process, voice identifier of the call voice during the business processing process, user identifier corresponding to the user conducting the telephone communication, work order type of the business work order, business processing method and other business information.

[0068] Based on this, in an exemplary implementation, the third identifier can be understood as the third user identifier recorded in the first work order. In other words, after determining the first work order, the user identifier in the first work order can be obtained to obtain the third identifier.

[0069] For example, in step S310, 100 first work orders are determined, and user identifiers in the 100 first work orders can be extracted, and then the extracted user identifiers are deduplicated, and the third identifier in step S320 is obtained based on the result of the deduplication process.

[0070] In step S330, the call speech corresponding to the third identifier in the generated call speech is determined as the first call speech.

[0071] In an exemplary embodiment, the generated call voice may include call voice when business is processed via telephone, call voice when business is processed via voice call, call voice when business is processed via video, and of course, call voice when online business is processed in other ways. This exemplary embodiment does not specifically limit this.

[0072] In an exemplary scenario, the generated call voice may include the voice generated by the user when handling business. For example, a user purchases a product on a platform and finds that there is a problem with the product and needs the platform to solve it. The user can communicate with the customer service of the platform for a solution, and the voice generated in this process can be understood as the generated call voice.

[0073] In an exemplary embodiment, the generated call voice includes the generated voice of the first object. The first object can be customized according to the needs. For example, in the business processing scenario, when the user and the staff of the enterprise providing the service communicate about the business, a communication voice will be generated. The generated call voice can include the user voice in the communication voice, that is, the first object is the user, not the staff of the enterprise.

[0074] In an exemplary embodiment, each work order may also record the voice identifier of the user's call voice when handling the business. After the business call is over, the call voice can be stored in the voice platform, and the voice platform generates a storage address for the call voice, and associates the storage address with the voice identifier of the current call voice. The call voice associated with the voice identifier can be obtained later through the storage address associated with the voice identifier. Based on this, in an exemplary embodiment, the generated call voice in step S330 may include the call voice associated with the voice identifier in the second generated work order.

[0075] In an exemplary implementation, the second generated work order may be determined according to a second preset duration, for example, a historical work order whose duration from the corresponding moment of the preset period is within the second preset duration may be determined as the second generated work order.

[0076] The second preset duration can be customized according to the needs, such as the second preset duration can be the last six months, the last three months, etc., and this exemplary embodiment does not specifically limit this. For example, taking the preset period as 3 a.m. every day and the second preset duration as half a year as an example, at 3 a.m. every day, the historical work order within half a year from today can be determined as the second generated work order.

[0077] After the second generated work order is determined, the call voice associated with the voice identifier in the second generated work order can be understood as the call voice generated in step S330. At the same time, since the work order records the user identifier and the voice identifier, based on this, a specific implementation of step S330 may include: after the second generated work order is determined, a second work order whose user identifier is the third identifier is determined from the second generated work order, and the call voice associated with the voice identifier in the second work order is determined as the first call voice.

[0078] As mentioned above, the call voice generated during the business handling process can be stored in the voice platform, and the voice platform generates a storage address for the call voice, and associates the storage address with the voice identifier of the current call voice. The voice identifier of the call voice generated during the business handling process can be recorded in the work order. In this way, after determining the second work order, the storage address associated with the voice identifier can be obtained from the voice platform according to the voice identifier in the second work order, and then the call voice associated with the voice identifier can be obtained according to the storage address, thereby obtaining the first call voice.

[0079] In an exemplary implementation, the first call voice in step S330 includes the first call voice associated with multiple user identifiers. Taking a third identifier as B, a preset period of 2 a.m. every day, and a second preset duration of the last six months as an example, an exemplary implementation of step S330 can be understood as obtaining a work order within six months from today at 2 a.m. every day, determining it as the second generated work order, and determining the second generated work order with user identifier B as the second work order corresponding to the third identifier B. According to the voice identifier in the second work order corresponding to the third identifier B, the storage address of the call voice associated with the voice identifier is obtained from the voice platform, so as to obtain the first call voice corresponding to the third identifier B according to the storage address. This processing is performed for each third identifier to obtain the first call voice corresponding to each third identifier, and the first call voice corresponding to each third identifier is aggregated and merged, so that the first call voice associated with multiple user identifiers can be obtained in step S330.

[0080] Since the first call speech in step S330 may include first call speech associated with multiple user identifiers, and the second voiceprint feature includes the voiceprint feature corresponding to the first call speech, based on this, in an exemplary embodiment, the second voiceprint feature in step S210 may be multiple, each second voiceprint feature corresponds to a second identifier, and the second identifier can be determined according to the user identifier associated with the first call speech that generates the second voiceprint feature.

[0081] In an exemplary embodiment, the second voiceprint feature in step S210 can also be one. For example, if a first call voice is determined according to the above steps S310 to S330, then the second voiceprint feature is the voiceprint feature corresponding to the first call voice, and the second identifier corresponding to the second voiceprint feature is the user identifier associated with the first call voice. This exemplary embodiment does not make any special limitations on this.

[0082] In an exemplary embodiment, a user identifier can indicate a unique user, that is, the user identifier can distinguish different users. The user identifier can be customized according to the needs, such as when the user logs in or registers on the service platform or when the user handles the service for the first time, a unique string can be randomly generated for the user to represent the user identifier of the user. Of course, the user identifier can also be generated in other ways, such as the user's mobile phone number and other user-unique information can be used as the user identifier, so that when the user subsequently handles the service, the user identifier of the user can be determined based on his mobile phone number, etc.

[0083] Exemplarily, the user voice generated when handling a service may be associated with the user identifier, such as adding the user identifier to the user voice when handling a service to associate the two.

[0084] In another exemplary embodiment, the generated work orders may not be screened according to the preset work order type, and the generated work order determined in the above step S310 may be directly determined as the first work order, thereby determining the first call voice according to the above steps S320 and S330.

[0085] After the first call voice is determined through the above steps S320 and S330, the second voiceprint feature can be determined based on the voiceprint feature of the first call voice.

[0086] For example, Figure 4 A schematic diagram of a flow chart showing a method for determining a second voiceprint feature in an exemplary embodiment of the present disclosure is shown. Figure 4 , the method may include steps S410 to S430.

[0087] In step S410, a reference voiceprint feature is determined based on the voiceprint feature of the first call speech that has been marked as having passed the user identity review.

[0088] For example, when a user handles a business, such as after-sales rights protection, the user's identity can be verified, such as through account passwords, video face recognition, etc., and then according to the verification results, the user's identity verification mark is added to the work order corresponding to the current business. If the verification is passed, the identity verification pass mark is added. If the verification fails or no verification is conducted, the unverified identity mark is added. If an identity verification pass mark is added to a work order, the call voice corresponding to the voice identifier in the work order can be understood as the call voice that has been marked as the user identity verification passed. It is also possible to directly add a verification mark to the call voice according to the user identity verification situation. For example, after the call ends, the user identity verification mark recorded in the work order can be obtained, and then the user identity verification mark can be added to the call voice, and the call voice with the user identity verification mark added can be uploaded to the voice platform.

[0089] In other words, the call voice associated with the second work order with the identity review passed mark in the above-mentioned second work order can be determined as the first call voice marked as the user identity review passed; the first call voice with the identity review passed mark can also be determined as the first call voice marked as the user identity review passed.

[0090] After determining that the first call voice has been marked as having passed the user identity review, the voiceprint feature of the first call voice is extracted to obtain the reference voiceprint feature.

[0091] In step S420, a first similarity between the voiceprint feature of the first call speech and the reference voiceprint feature is determined.

[0092] In an exemplary embodiment, the first similarity can be calculated using any similarity measurement method, such as the cosine similarity between the voiceprint feature of the first call speech and the reference voiceprint feature, or the Euclidean distance between the voiceprint feature of the first call speech and the reference voiceprint feature, which is not particularly limited in this exemplary embodiment.

[0093] In step S430, the voiceprint feature of the first call speech whose first similarity is less than the first similarity threshold is determined as the second voiceprint feature.

[0094] For example, after determining the first similarity, the voiceprint feature of the first call speech whose first similarity is less than the first similarity threshold can be determined as the second voiceprint feature. The first similarity threshold can be customized, but the first similarity threshold should not be set too small. The first similarity threshold should be sufficient to reflect that the voiceprint feature that is not determined as the second voiceprint feature has a high similarity with the reference voiceprint feature, so that the voiceprint feature with low similarity to the reference voiceprint feature is determined as the second voiceprint feature.

[0095] If the user identity verification is passed, it usually indicates that the business process is legal, and the call voice in this business process is the real voice of a normal user, or the real voice of a legal user who has been legally authorized by the normal user. Through the above steps S410 to S430, the voiceprint features corresponding to the voice of the legal user can be determined from the voiceprint features of the first call voice without the identity verification mark according to the reference voiceprint features and the first similarity threshold, thereby deleting the voiceprint features corresponding to the voice of the legal user to the greatest extent, thereby avoiding to the greatest extent possible the determination of the legal voiceprint features as risky voiceprint features in subsequent steps, so as to improve the accuracy of risk identification.

[0096] In an exemplary implementation, for each third identifier, a reference voiceprint feature of the third identifier can be determined based on the voiceprint feature of the first call voice corresponding to the third identifier that has been marked as having passed the user identity review. Then, for each third identifier, the similarity between the voiceprint feature of the first call voice corresponding to the third identifier and the reference voiceprint feature corresponding to the third identifier is calculated, and the voiceprint feature of the first call voice with a similarity less than a first similarity threshold is determined as the second voiceprint feature corresponding to the third identifier, and the second voiceprint features corresponding to each third identifier are combined to obtain the second voiceprint feature in step S210.

[0097] Exemplarily, when there are multiple first call voices corresponding to the third identifier that have been marked as having passed the user identity review, the voiceprint feature of one of the call voices that has passed the identity review can be randomly selected as a reference voiceprint feature, or the voiceprint feature of the first call voice that has passed the most recent identity review can be selected as the reference voiceprint feature. This exemplary embodiment does not impose any special limitation on this.

[0098] For example, Figure 5 A flowchart showing another method for determining a second voiceprint feature in an exemplary embodiment of the present disclosure is shown. Figure 5 , the method may include steps S510 to S530.

[0099] In step S510, the intelligent robot conversation voice in the first conversation voice is identified.

[0100] Exemplarily, the method of identifying the intelligent robot conversation voice in the first call voice includes at least one of the following methods: identifying the intelligent robot conversation voice in the first call voice according to the duration of the first call voice; identifying the intelligent robot conversation voice in the first call voice according to the voice tag in the work order corresponding to the first call voice; identifying the intelligent robot conversation voice in the first call voice according to the similarity between the voiceprint features of the first call voice and the voiceprint features in the intelligent robot voiceprint library; identifying the voice content of the first call voice, and identifying the intelligent robot conversation voice in the first call voice according to the matching degree between the voice content and preset keywords associated with the preset business scenario.

[0101] Below, each of the above-mentioned methods for recognizing the intelligent robot's conversation voice is further explained in detail.

[0102] Exemplarily, when the duration of the first call voice is less than a duration threshold, it can be determined that the first call voice is an intelligent robot call voice. The duration threshold can be customized according to needs or experience, such as 1 minute.

[0103] Because in the actual business processing process, multiple communications are required to handle business, and the call duration is usually longer. If an intelligent robot assistant answers the call for business communication, the intelligent robot assistant has a fixed and simple speech mode, so there are fewer rounds of interaction, resulting in a shorter call duration. Therefore, when the call duration is less than the duration threshold, it can be determined that the first call voice is the intelligent robot call voice. This method can quickly determine the intelligent robot call voice through the call duration, which can improve the risk identification efficiency of voiceprint features.

[0104] For example, after each business is completed, the staff, such as customer service, can generate a voice tag corresponding to the business call based on the business communication situation, and add the voice tag to the work order as one of the work order contents. The voice tag can be determined based on whether the call is answered by an intelligent robot during the business processing, that is, the voice tag can include two types, whether the call is answered by an intelligent robot and whether the call is not answered by an intelligent robot. If the call is answered by an intelligent robot, a voice tag indicating that the call is answered by an intelligent robot can be added to the work order. If not, a voice tag indicating that the call is not answered by an intelligent robot can be added to the work order.

[0105] Exemplarily, an intelligent robot voiceprint library can also be pre-generated, for example, the intelligent robot voiceprint library can be pre-generated based on the voiceprint features of the call voice in the work order with the voice tag of the intelligent robot answering the phone. Of course, the intelligent robot voiceprint library can also be generated in other ways, and this exemplary embodiment does not specifically limit this. In this way, the similarity between the voiceprint features of the first call voice and the voiceprint features in the intelligent robot voiceprint library can be calculated, and when the similarity is greater than a certain similarity threshold, it is determined that the first call voice is the call voice of the intelligent robot.

[0106] For example, for intelligent robot calls, due to their fixed call mode, they usually cannot accurately express the user's demands, that is, they cannot accurately describe the business content to be handled. Therefore, preset keywords associated with preset business scenarios can be generated in advance, such as preset keywords associated with after-sales rights protection business scenarios, preset keywords associated with pre-sales consultation business scenarios, etc. Then, the voice content of the first call voice is recognized, and the recognized voice content is matched with the preset keywords associated with the preset business scenario. When the matching degree is less than a certain matching threshold, it is determined that the first call voice is the intelligent robot call voice.

[0107] Of course, the intelligent robot's call voice may also be recognized in other ways, and this exemplary embodiment does not specifically limit this.

[0108] In an exemplary embodiment, when the method of recognizing the first robot conversation voice includes multiple methods described above, a weight can be configured for each recognition method, and the results of each recognition method can be fused according to the weight, so as to determine the recognition result of the intelligent robot conversation voice according to the fusion result.

[0109] In step S520, the intelligent robot call voice in the first call voice is filtered out to obtain candidate call voices.

[0110] For example, after identifying the intelligent robot conversation voice in the first call voice, the intelligent robot conversation voice in the first call voice can be deleted, and the remaining first call voice can be determined as the candidate call voice.

[0111] In step S530, the voiceprint feature of the candidate call speech is determined as the second voiceprint feature.

[0112] For example, the voiceprint feature of the candidate call speech may be extracted to obtain the second voiceprint feature.

[0113] Through the above steps S510 to S530, the intelligent robot voice in the first call voice can be identified, and after the identified intelligent robot call voice is deleted, the remaining voiceprint features of the first call voice are processed to perform risk identification of the voiceprint features. In this way, it is possible to avoid determining the voiceprint features of the intelligent answering assistant as risk voiceprint features, further improving the accuracy of risk voiceprint feature identification.

[0114] In an exemplary embodiment, the method of determining the second voiceprint feature based on the first call voice may also include: deleting the first call voice that has been marked as having passed the user identity review in the first call voice to obtain a third call voice; and determining the voiceprint feature of the third call voice as the second voiceprint feature.

[0115] As mentioned above, if the user identity review is passed, it usually indicates that the business process is legal, and the call voice in this business process is the real voice of the legal user, or the legal voice authorized by the legal user. By deleting the call voice that has passed the user identity review in the first call voice, and obtaining the second voiceprint feature based on the remaining first call voice that has not passed the user identity review, it is possible to avoid determining the voiceprint feature of the legal voice as the risk voiceprint feature, thereby improving the accuracy of risk voiceprint feature recognition.

[0116] In an exemplary implementation, after determining the second voiceprint feature according to the first call speech, the second voiceprint feature can be stored in the first voiceprint library. That is, the second voiceprint feature can include the voiceprint feature in the first voiceprint library.

[0117] In an exemplary embodiment, a new first call voice can be periodically determined according to the above-mentioned preset period, so that a new second voiceprint feature can be periodically generated according to the new first call voice, and then the new second voiceprint feature is also stored in the first voiceprint library to update the second voiceprint feature in the first voiceprint library.

[0118] Among them, storing the new second voiceprint feature in the first voiceprint library may include replacing the existing second voiceprint feature in the first voiceprint library with the new second voiceprint feature; storing the new second voiceprint feature in the first voiceprint library may also include adding the new second voiceprint feature to the first voiceprint library, and using it together with the existing second voiceprint feature in the first voiceprint library as the second voiceprint feature in the first voiceprint library.

[0119] In an exemplary implementation, before storing the second voiceprint feature in the first voiceprint library, the second voiceprint feature may be processed for similarity, and similar second voiceprint features may be stored as a set in the first voiceprint library. That is, the first voiceprint library may include a sixth voiceprint feature set and a seventh voiceprint feature set, wherein the sixth voiceprint feature set includes multiple similar second voiceprint features, and the seventh voiceprint feature set includes a single second voiceprint feature.

[0120] The method of performing similarity processing on the second voiceprint features may include: clustering the second voiceprint features to obtain similar second voiceprint features.

[0121] In an exemplary implementation, any clustering method may be used to cluster the second voiceprint features to obtain one or more sixth voiceprint feature sets. The purpose of clustering is to determine highly similar voiceprint features from the second voiceprint features.

[0122] For example, the second voiceprint features may be clustered using a hidden Markov model and a K-means clustering algorithm. For example, the similarity between two second voiceprint features may be calculated based on the hidden Markov model, and then the second voiceprint features may be clustered based on the K-means clustering algorithm according to the similarity to obtain one or more sixth voiceprint feature sets.

[0123] Exemplarily, in step S210, only the similarity between the first voiceprint feature and the second voiceprint feature in the sixth voiceprint feature set may be calculated to determine the second voiceprint feature similar to the first voiceprint feature from the sixth voiceprint feature set.

[0124] As mentioned above, the present disclosure automatically identifies risk voiceprints by identifying similar voiceprints with multiple user identifiers. Since the second voiceprint features in the sixth voiceprint feature set are similar to each other, if only the second voiceprint feature similar to the first voiceprint feature is determined from the sixth voiceprint feature set to determine the risk identification result of the first voiceprint feature, it is only necessary to calculate the similarity between the first voiceprint feature and the second voiceprint feature, and there is no need to calculate the similarity between the second voiceprint features, thereby improving the risk identification efficiency of the first voiceprint feature.

[0125] In addition, when a second voiceprint feature similar to the first voiceprint feature is determined from the sixth voiceprint feature set, the sixth voiceprint feature set in the first voiceprint library can be traversed. When the second voiceprint feature in the traversed sixth voiceprint feature set is similar to the first voiceprint feature, the traversal can be stopped, that is, there is no need to compare whether the second voiceprint feature in other sixth voiceprint feature sets is similar to the first voiceprint feature, because the similarity of voiceprint features in different sixth voiceprint feature sets is extremely low, and it is impossible for the first voiceprint feature to be similar to voiceprint features in multiple sixth voiceprint feature sets at the same time, thereby further improving the risk identification efficiency of the first voiceprint feature.

[0126] Exemplarily, in step S210, a second voiceprint feature similar to the first voiceprint feature may also be determined from the sixth voiceprint feature set and the seventh voiceprint feature set. That is, the second voiceprint feature similar to the first voiceprint feature may include the second voiceprint feature in the sixth voiceprint feature set or the second voiceprint feature in the seventh voiceprint feature set, and this exemplary embodiment does not specifically limit this.

[0127] In an exemplary embodiment, determining the second voiceprint feature similar to the first voiceprint feature includes: calculating the second similarity between the first voiceprint feature and the second voiceprint feature, and determining the second voiceprint feature whose second similarity is greater than the second similarity threshold as the second voiceprint feature similar to the first voiceprint feature. The second similarity is measured in the same way as the first similarity, such as cosine similarity, Euclidean distance, etc., which will not be described in detail here.

[0128] In another exemplary embodiment, determining a second voiceprint feature similar to each first voiceprint feature includes: when the duration of the user voice in each real-time call voice stream is greater than the target preset duration, extracting the first voiceprint feature of the user voice in each real-time call voice stream, and determining a second voiceprint feature similar to the first voiceprint feature.

[0129] Next, a specific implementation of determining the first number of different identifiers in the identifiers in step S220 is described.

[0130] As mentioned above, the first voiceprint feature corresponds to an identifier, and the second voiceprint feature corresponds to an identifier. Taking the identifier corresponding to the first voiceprint feature as the first identifier and the identifier corresponding to the second voiceprint feature as the second identifier as an example, based on this, the different identifiers in step S220 may include that the second identifier is different, and may also include that the first identifier and the second identifier are different.

[0131] For example, the first identifier and the second identifier may be deduplicated, and the remaining identifiers after the deduplication process are all different identifiers, so the number of the remaining identifiers after the deduplication process may be counted and used as the first number.

[0132] In another exemplary embodiment of the present disclosure, other first voiceprint features similar to each first voiceprint feature may also be determined, and then a first number of different identifiers in the identifiers of similar first voiceprint features may be determined. At this time, the different identifiers include different first identifiers. Similarly, the identifiers of similar first voiceprint features may also be deduplicated, and the number of identifiers remaining after deduplication may be counted, and the number may be used as the first number. Below, a specific implementation method of determining the risk identification result of the first voiceprint feature according to the first number in step S230 is described.

[0133] Exemplarily, a specific implementation of step S230 may include: in response to the first number being greater than a first preset threshold, determining that the risk identification result of the first voiceprint feature is a target risk identification result, wherein the target risk identification result indicates that the first voiceprint feature has a risk.

[0134] For example, when the first number is greater than the first preset threshold, the risk identification result of the first voiceprint feature can be directly determined as the target risk identification result.

[0135] In an exemplary embodiment, when the first number is less than or equal to the first preset threshold, the risk identification result of the first voiceprint feature is determined to be pending risk. That is, when the first number is less than or equal to the first preset threshold, it is determined that the first voiceprint feature temporarily does not have a risk, but it cannot be directly determined that the first voiceprint feature does not have a risk. This is because new voiceprint features are constantly generated during business processing. The first voiceprint feature is an existing voiceprint feature for the subsequent newly generated voiceprint features. Since the subsequent newly generated voiceprint features are unknown at this time, it is impossible to determine whether the first voiceprint feature is similar to the subsequent newly generated voiceprint features. If the first voiceprint feature is similar to the subsequent newly generated voiceprint feature, and the number of different identifiers in the identifier of the first voiceprint feature and the identifier of the subsequent newly generated voiceprint feature similar to the first voiceprint feature is also greater than the first preset threshold, then the first voiceprint feature has a risk, so it cannot be determined at this time that the first voiceprint feature really does not have a risk.

[0136] In other words, in the present disclosure, it is only possible to determine that a certain voiceprint feature has a risk when processing the voiceprint feature, and for a voiceprint feature that has not been determined to have a risk, it can only be determined that its risk is pending or unknown, and it cannot be directly determined that it does not have a risk. For a voiceprint feature that has not been determined to have a risk, it is possible to continue to determine whether its risk identification result is the target risk identification result based on the newly generated voiceprint feature in the subsequent business processing according to the method in the present disclosure.

[0137] Exemplarily, another specific implementation of step S230 may include: when the first number is greater than a first preset threshold, determining the risk identification result of the first voiceprint feature based on the first voiceprint feature, the first identifier of the first voiceprint feature, the second voiceprint feature, and the second identifier of the second voiceprint feature.

[0138] For example, when the first number is greater than the first preset threshold, the risk identification result of the first voiceprint feature can be further determined according to the first voiceprint feature, the first identifier of the first voiceprint feature, the second voiceprint feature and the second identifier of the second voiceprint feature. Figure 6 A flow chart showing a method for determining a risk identification result of a first voiceprint feature in an exemplary embodiment of the present disclosure is shown. Figure 6 , the method may include steps S610 to S640. Wherein:

[0139] In step S610, a plurality of candidate voiceprint feature pairs are determined according to the first voiceprint feature, the first identifier, the second voiceprint feature, and the second identifier.

[0140] In this step, the two voiceprint features of the candidate voiceprint feature pair correspond to different identifiers.

[0141] For example, a plurality of candidate voiceprint feature pairs may be determined based on any two voiceprint features with different identifiers in the first voiceprint feature and the second voiceprint feature.

[0142] In an exemplary embodiment, two voiceprint features can be randomly selected from the first voiceprint feature and the second voiceprint feature to be combined and cannot be repeated to determine multiple voiceprint feature pairs, and then the voiceprint feature pairs with two voiceprint features having different identifiers in the multiple voiceprint feature pairs are determined as candidate voiceprint feature pairs, thereby obtaining multiple candidate voiceprint feature pairs. For example, any two voiceprint features from the first voiceprint feature and the second voiceprint feature can form 20 voiceprint feature pairs, and the two voiceprint features in 18 of the 20 voiceprint feature pairs have different identifiers, so the number of candidate voiceprint feature pairs is 18.

[0143] In another exemplary embodiment, the first voiceprint feature and the second voiceprint feature may be deduplicated according to the identifier corresponding to the voiceprint feature, that is, the voiceprint features with the same identifier in the first voiceprint feature and the second voiceprint feature may be deduplicated to obtain candidate voiceprint features, and then any two voiceprint features in the candidate voiceprint features may be combined without duplication to obtain multiple candidate voiceprint feature pairs. For example, there are 12 first voiceprint features and second voiceprint features in total, and after deduplication of the voiceprint features with the same identifier in these 12 voiceprint features, there are 10 voiceprint features left, and any two voiceprint features in these 10 voiceprint features may be combined without duplication to obtain 5 candidate voiceprint feature pairs.

[0144] In step S620, the first intention similarity between the two voiceprint features of the candidate voiceprint feature pair is determined.

[0145] For example, Figure 7 A flowchart of a method for determining a first intention similarity in an exemplary embodiment of the present disclosure is shown. Figure 7 , the method may include steps S710 to S730. Wherein:

[0146] In step S710, the work orders corresponding to the two voiceprint features are determined according to the identifiers of the two voiceprint features of the candidate voiceprint feature pair.

[0147] For example, each candidate voiceprint feature has a corresponding user ID, and the work order corresponding to the candidate voiceprint feature can be determined based on the user ID. For example, the user ID corresponding to the candidate voiceprint feature can be matched with the user ID in the work order, and the work order corresponding to the candidate voiceprint feature can be obtained based on the successfully matched work order.

[0148] In step S720, the level label of each work order is obtained.

[0149] In an exemplary embodiment, each work order may also record the work order label of the work order. The work order label may be a multi-level work order label, and the multi-level work order label may represent the business level to which the business involved in the work order belongs. The higher the business level, the wider the business scope. For example, for the multi-level work order label ABCD, the business involved in level A includes level B, the business involved in level B includes level C, and the business involved in level C includes level D, that is, the label of the next level is a branch of the label of the previous level.

[0150] For example, the label of the lowest business level to which the business of each work order belongs can be marked. According to the preset multi-level label structure of the multi-branch tree, the label of the lowest business level of the work order can be used as the leaf node, and the label of the upper business level of the business of the work order can be searched upward until the label of the business level of the root node is found, that is, the hierarchical label of the work order is obtained, that is, the hierarchical label of the work order includes all labels involved in the path from the root node to the leaf node. For example, there is a work order 1, and the label of the business level to which the business involved belongs is D, that is, the work order label recorded in the work order 1 is D. In the preset multi-level label structure based on the multi-branch tree, the upper label of D is C, the upper label of C is B, the upper label of B is A, and A is the label of the first level, that is, A is the label of the highest level (that is, the business level of the root node), and it has no upper label, then, the hierarchical label of the work order 1 is ABCD. The work order label of a work order 2 is E. In the preset multi-level label structure based on a multi-branch tree, the parent label of E is also C mentioned above, the parent label of C is B, and the parent label of B is A. Then the hierarchical label of work order 2 is: ABCE.

[0151] In step S730 , the first intent similarity is determined according to the levels of the hierarchical tags and the tag information of each level.

[0152] Exemplarily, a specific implementation of step S730 may include: calculating the second intent similarity between tag information at the same level; and determining the first intent similarity according to the second intent similarity and the weight of the level corresponding to the second intent similarity.

[0153] For example, for two work orders corresponding to two voiceprint features in the candidate voiceprint feature pair, the second intent similarity between the label information of the same level in the hierarchical labels of the two work orders can be calculated. If the hierarchical labels of the two work orders are the same, both are 4 levels, then the second intent similarity 1 between the label information of the first level in the hierarchical labels of the two work orders, the second intent similarity 2 between the label information of the second level, the second intent similarity 3 between the label information of the third level, and the second intent similarity 4 between the label information of the fourth level are calculated respectively. If the hierarchical labels of the two work orders are different, such as one is 5 levels and the other is 4 levels, then only the second intent similarity between the hierarchical label information of the same level in the first 4 levels can be calculated.

[0154] Exemplarily, the implementation method of calculating the second intention similarity between tag information of the same level may include: obtaining a feature vector of the tag information according to any text feature vector extraction model, and calculating the similarity between the feature vectors of the tag information to obtain the second intention similarity. The similarity between the feature vectors of the tag information can also be determined according to the above-mentioned cosine similarity, or Euclidean distance, or other similarity measurement methods. For example, if the work orders corresponding to the two voiceprint features in the candidate voiceprint feature pair are work order 1 and work order 2 respectively, for calculating the second intention similarity between the first-level label information of work order 1 and work order 2, the first-level level label A of work order 1 can be first input into the BERT (Bidirectional Encoder Representations from Transformers) model to obtain the feature vector A1 of the first-level level label A of work order 1, and the first-level level label A of work order 2 is also input into the BERT model to obtain the feature vector A2 of the first-level level label 1 of work order 2, and then the cosine similarity between the feature vector A1 and the feature vector A2 is calculated, and the cosine similarity is determined as the second intention similarity between the first-level label information of work order 1 and work order 2. The second intention similarity between the other levels of labels of work order 1 and work order 2 can also be calculated in the same way.

[0155] A weight may be configured in advance for each level in a preset multi-level label structure based on a multi-branch tree. After obtaining the second intent similarity between the level label information of the same level, the first intent similarity may be determined based on the second intent similarity between the level label information of the same level and the weight of the level corresponding to the second intent similarity.

[0156] In an exemplary embodiment, when the second intent similarity between tag information at the same level is greater than a third preset threshold, the intent similarity value of the level can be configured as 1, otherwise, the intent similarity of the level can be configured as 0. Then, the intent similarity of each level is weighted according to the weight to obtain the first intent similarity.

[0157] For example, in a preset multi-level label structure based on a multi-branch tree, the weight of the first level is configured as 0.1, the weight of the second level is configured as 0.2, the weight of the third level is configured as 0.3, the weight of the fourth level is configured as 0.6, and the work orders corresponding to the two voiceprint features in the candidate voiceprint feature pair have 4 levels of labels. The second intention similarity between the label information of the first level is greater than the third preset threshold, the intention similarity of the first level is configured as 1, the second intention similarity between the label information of the second level is greater than the third preset threshold, and the second intention similarity of the label information of the second level is configured as 1. The similarity is also configured to 1. If the second intention similarity between the label information of the third level is less than the third preset threshold, the intention similarity of the third level is configured to 0. If the second intention similarity between the label information of the fourth level is less than the third preset threshold, the intention similarity of the fourth level is configured to 0. After weighting the intention similarity of each level according to the weight of each level, it can be obtained that the first intention similarity between the two voiceprint features in the candidate voiceprint feature pair is 0.3, that is, 0.1×1+0.2×1+0.3×0+0.4×0=0.3.

[0158] In another exemplary implementation, the second intent similarities at each level may be weighted according to the weight of the level corresponding to the second intent similarities to determine the first intent similarity.

[0159] If we continue with the above-mentioned multi-level label structure based on a multi-branch tree, in which the weight of the first level is configured as 0.1, the weight of the second level is configured as 0.2, the weight of the third level is configured as 0.3, the weight of the fourth level is configured as 0.6, and the work orders corresponding to the two voiceprint features in the candidate voiceprint feature pair both have 4-level labels, the second intention similarity between the label information of the first level is 0.8, the second intention similarity between the label information of the second level is 0.9, the second intention similarity between the label information of the third level is 0.3, and the second intention similarity between the label information of the fourth level is 0.4. Then, the first intention similarity between the two voiceprint features in the candidate voiceprint feature pair is 0.51, that is, 0.1×0.8+0.2×0.9+0.3×0.3+0.4×0.4=0.51.

[0160] In step S630, a second number of candidate voiceprint feature pairs whose first intention similarity is greater than or equal to a first similarity threshold is counted.

[0161] For example, after determining the first intention similarity, the number of candidate voiceprint feature pairs whose first intention similarity is greater than or equal to the first intention similarity can be counted, and the number is used as the second number. The first intention similarity can be customized according to needs or experience, and this exemplary embodiment does not specifically limit this.

[0162] In step S640, when the second number is greater than or equal to a second preset threshold, the risk identification result of the first voiceprint feature is determined as a target risk identification result.

[0163] In an exemplary embodiment, the target risk identification result indicates that the first voiceprint feature has a risk.

[0164] For example, when the first number is greater than the first preset threshold, the second number can be determined according to the above steps S610 to S640. When the second number is greater than or equal to the second preset threshold, the first voiceprint feature is determined to be risky.

[0165] In an exemplary embodiment, determining the risk identification result of the first voiceprint feature as the target risk identification result may include: adding a risk mark to the first voiceprint feature and storing the first voiceprint feature in the first voiceprint library; or directly storing the first voiceprint feature in the second voiceprint library. The risk mark may be in any form as long as it can distinguish the risky voiceprint feature from other voiceprint features, and this exemplary embodiment does not specifically limit this.

[0166] In other words, the voiceprint features stored in the first voiceprint database include risky voiceprint features and other voiceprint features with unknown risks, and the risky voiceprint features and other voiceprint features with unknown risks can be distinguished in the first voiceprint database according to whether risk tags are added. The second voiceprint database only stores risky voiceprint features, so no additional tags need to be added to them.

[0167] Through the above-mentioned steps S610 to S640, the blacklist voiceprint can be determined by identifying the intentions between similar voiceprints with multiple user identifiers. When it is identified that the same person handles the same or similar business in the name of different users, the voiceprint of this person is determined to be risky, thereby improving the accuracy of risk voiceprint recognition.

[0168] In an exemplary embodiment, when the first number is greater than the first preset threshold, the risk identification result of the second voiceprint feature similar to the first voiceprint feature may also be determined as the above-mentioned target risk identification result.

[0169] In an exemplary implementation, when the first number is less than or equal to the first preset threshold, the risk identification result of the second voiceprint feature similar to the first voiceprint feature can also be determined as the above-mentioned risk pending. The explanation of the risk pending can refer to the explanation of the above-mentioned relevant parts, which will not be repeated here.

[0170] In another exemplary embodiment, when the second number is greater than or equal to the second preset threshold, the risk identification result of the second voiceprint feature similar to the first voiceprint feature may also be determined as the above-mentioned target risk identification result.

[0171] In another exemplary embodiment, when the second number is less than the second preset value, the risk identification result of the second voiceprint feature similar to the first voiceprint feature can also be determined as a pending risk. The explanation of pending risk can also refer to the above-mentioned relevant part, which will not be repeated here.

[0172] Exemplarily, a method for determining that the risk identification result of a second voiceprint feature similar to a first voiceprint feature is the above-mentioned target risk identification result may include: adding a risk mark to the second voiceprint feature similar to the first voiceprint feature in the above-mentioned first voiceprint library, or storing the second voiceprint feature similar to the first voiceprint feature in the second voiceprint library.

[0173] In an exemplary application scenario, the first voiceprint library or the second voiceprint library mentioned above can be used to identify risky users. For example, when a user and an enterprise service staff communicate by phone to handle business, the user's voice in the call voice stream can be obtained in real time. When the voice duration of the user's voice is greater than the preset duration, the target voiceprint feature of the user's voice is extracted, and the third similarity between the target voiceprint feature of the user's voice and the voiceprint feature with risk tags added in the first voiceprint library is calculated, or the fourth similarity between the target voiceprint feature of the user's voice and the voiceprint feature in the second voiceprint library is calculated. When the third similarity or the fourth similarity is greater than the fourth preset threshold, the user is determined to be a risky user. Since voiceprint features are difficult to forge, the accuracy of risk object identification can be improved through voiceprint features.

[0174] For example, Figure 8 A schematic diagram of the structure of a data processing system in an exemplary embodiment of the present disclosure is shown. Figure 8 The data processing system may include a business processing platform 81 , a first database 82 , a voice platform 83 , a message queue 84 , a voiceprint library service platform 85 and a second database 86 .

[0175] In an exemplary embodiment, the business processing platform 81 can store the created work orders in the first database 82, obtain the user identifier in the work order of a preset type created the previous day from the first database 82 at dawn every day, determine the obtained user identifier as the user identifier to be processed, and then for each user identifier to be processed, query the contact history table for the third historical voice identifier corresponding to the contact history of the user identifier to be processed in the past six months, and determine the third historical voice identifier with an identity review mark as the fourth historical voice identifier, wherein the contact history table can record the user identifier and the historical voice identifier of the historical call voice associated with the user identifier.

[0176] The business processing platform 81 obtains the third storage address associated with the third historical voice identifier and the fourth storage address associated with the fourth historical voice identifier from the voice platform 83 according to the queried third historical voice identifier and the fourth storage address, and sends the third storage address, the fourth storage address, the user identifier and the work order number corresponding to the third historical voice identifier, and the user identifier and the work order number corresponding to the fourth historical voice identifier to the message queue 84.

[0177] The voiceprint library service platform 85 obtains the third historical voice corresponding to the third historical voice identifier and the fourth historical voice corresponding to the fourth historical voice identifier through the voice platform 83 according to the storage address in the message queue 84, and extracts the eighth voiceprint feature of the third historical voice and the ninth voiceprint feature of the fourth historical voice. For each eighth voiceprint feature, the similarity between the eighth voiceprint feature and the ninth voiceprint feature is calculated. If the similarity is less than a threshold, the eighth voiceprint feature is retained. Otherwise, the eighth voiceprint feature is eliminated to determine the eighth voiceprint feature set.

[0178] The voiceprint library service platform 85 can cluster the voiceprint features in the eighth voiceprint feature set at a fixed time every day, such as 7 o'clock, and obtain the sixth voiceprint feature set and the seventh voiceprint feature set according to the clustering results, and send the sixth voiceprint feature set and the seventh voiceprint feature set to the second database 86. The second database 86 stores the sixth voiceprint feature set and the seventh voiceprint feature set in the first voiceprint library of the second database 86.

[0179] When a user and an enterprise service staff communicate by phone to handle business, the business processing platform 81 can obtain the user voice in the call voice stream in real time, and extract the target voiceprint feature corresponding to the user voice, and then calculate the similarity between the target voiceprint feature and the voiceprint feature in the first voiceprint library of the second database 86, determine the voiceprint feature in the first voiceprint library that is similar to the target voiceprint feature, and then determine the first number of different user identifiers in the user identifier of the target voiceprint feature and the user identifier in the voiceprint feature in the first voiceprint library that is similar to the target voiceprint feature, and determine whether the target voiceprint feature of the user is risky based on the first number.

[0180] For example, Fig. 9 A schematic diagram showing a flow chart of an information push method in an exemplary embodiment of the present disclosure. Fig. 9 , the method may include steps S910 to S940. Wherein: in step S910, a fifth voiceprint feature similar to the fourth voiceprint feature corresponding to each user voice is determined, wherein the fourth voiceprint feature corresponds to an identifier and the fifth voiceprint feature corresponds to an identifier; in step S920, a third number of different identifiers in the identifiers is determined; in step S930, a risk identification result of the fourth voiceprint feature is determined according to the third number; in step S940, when the risk identification result is a target risk identification result, a prompt message is pushed to the terminal, wherein the target risk identification result indicates that the fourth voiceprint feature has a risk, and the prompt message indicates that the user corresponding to the user voice is a risky user.

[0181] In an exemplary embodiment, each user voice in step S910 may include a user voice generated in real time in each real-time call voice stream, or may include a user voice that has been generated in an existing call voice stream but has not been risk identified. This exemplary embodiment does not specifically limit this.

[0182] In an exemplary implementation, taking the fourth identifier corresponding to the fourth voiceprint feature as the fourth identifier and the fifth identifier corresponding to the fifth voiceprint feature as the fifth identifier as an example, the fourth identifier can be understood as the fourth user identifier corresponding to the fourth voiceprint feature, and the fifth identifier can be understood as the fifth user identifier corresponding to the fifth voiceprint feature. The fifth voiceprint feature is determined in the same manner as the second voiceprint feature described above. Correspondingly, the implementation of step S910 can also refer to the specific implementation of step S210 described above, which will not be repeated here.

[0183] In an exemplary implementation, the implementation of step S920 may refer to the implementation of step S220 described above, and the implementation of step S930 may refer to the implementation of step S230 described above, which will not be described in detail here.

[0184] In an exemplary embodiment, the third number in step S920 and the first number in the above-mentioned step S220 may be the same or different, and this exemplary embodiment does not specifically limit this.

[0185] In an exemplary embodiment, the terminal in step S940 may include another terminal for online communication by a terminal generating user voice. For example, taking the example of a user and an enterprise service staff communicating and processing business online, the terminal in step S940 may include a terminal used by the enterprise service staff for communication.

[0186] Through the above steps S910 to S940, risky users can be automatically identified based on voiceprint features without manual intervention or manual determination of blacklists, thereby improving the efficiency of risk identification and avoiding risk identification errors caused by manual errors, thereby improving the accuracy of risk identification.

[0187] For example, Fig.10 A schematic diagram showing a multi-party interaction process during information push in an exemplary embodiment of the present disclosure.

[0188] refer to Fig.10 When the user 101 answers a call or makes a call to handle a business, the intelligent call center 102 will transfer the call of the user 101 to the business processing platform 103. The business processing platform 103 will allocate an idle seat 104 to the user 101. After the seat 103 answers the call, the intelligent call center 102 will send the voice of the user 101 in the real-time voice stream of the user 101 and the seat 104 to the voiceprint library service platform 105. After the voiceprint library service platform 105 accumulates the voice of the user 101 for a sufficient period of time, such as 1 minute, it extracts the voiceprint features of the user 101 voice, calculates the similarity between the extracted voiceprint features of the user 101 voice and the fourth voiceprint features, and determines the third number of voiceprint features similar to the voiceprint features of the user 101 voice. When it is determined that the voiceprint features of the user 101 voice are risky according to the third number, the voiceprint library service platform 105 will send a prompt message that the user 101 is a risky user to the seat 104 through the message queue 106 and the business processing platform 103.

[0189] In addition, it should be noted that the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0190] Furthermore, Fig.11FIG. 1 is a schematic diagram showing the composition of a data processing device in an exemplary embodiment of the present disclosure. Fig.11 , the data processing device 1100 includes the following program modules:

[0191] The first similar voiceprint determination module 1110 is configured to determine the second voiceprint feature similar to each first voiceprint feature, the first voiceprint feature corresponds to an identifier, and the second voiceprint feature corresponds to an identifier; the first quantity determination module 1120 is configured to determine the first number of different identifiers in the identifiers; the first risk identification result determination module 1130 is configured to determine the risk identification result of the first voiceprint feature based on the first number.

[0192] In an exemplary embodiment, the first risk identification result determination module 1130 can be specifically configured to: in response to the first number being greater than a first preset threshold, determine that the risk identification result of the first voiceprint feature is a target risk identification result, and the target risk identification result indicates that the first voiceprint feature is risky.

[0193] In an exemplary embodiment, the first risk identification result determination module 1130 can also be specifically configured as: when the first number is greater than a first preset threshold, determining the risk identification result of the first voiceprint feature according to the first voiceprint feature, the first identifier of the first voiceprint feature, the second voiceprint feature, and the second identifier of the second voiceprint feature.

[0194] In an exemplary embodiment, the method of determining the risk identification result of the first voiceprint feature based on the first voiceprint feature, the first identifier of the first voiceprint feature, the second voiceprint feature and the second identifier of the second voiceprint feature includes: determining multiple candidate voiceprint feature pairs based on the first voiceprint feature, the first identifier, the second voiceprint feature and the second identifier, wherein the two voiceprint features of the candidate voiceprint feature pair correspond to different identifiers; determining a first intention similarity between the two voiceprint features of the candidate voiceprint feature pair; counting a second number of candidate voiceprint feature pairs whose first intention similarity is greater than or equal to a first similarity threshold; and when the second number is greater than or equal to a second preset threshold, determining that the risk identification result of the first voiceprint feature is a target risk identification result, and the target risk identification result indicates that the first voiceprint feature is risky.

[0195] In an exemplary embodiment, the method of determining the first intention similarity between two voiceprint features of a candidate voiceprint feature pair includes: determining the work orders corresponding to the two voiceprint features according to the identifiers of the two voiceprint features of the candidate voiceprint feature pair; obtaining the hierarchical label of each of the work orders; and determining the first intention similarity according to the hierarchy of the hierarchical label and the label information of each of the hierarchical levels.

[0196] In an exemplary embodiment, determining the first intention similarity based on the hierarchy of the hierarchical labels and the label information of each hierarchy includes: calculating the second intention similarity between the label information of the same hierarchy; and determining the first intention similarity based on the second intention similarity and the weight of the hierarchy corresponding to the second intention similarity.

[0197] In an exemplary embodiment, the second voiceprint feature includes a voiceprint feature corresponding to the first call voice; wherein, the method for determining the first call voice includes: determining a first work order corresponding to a preset work order type in the generated work order; determining a third identifier of the first work order; and determining the call voice corresponding to the third identifier in the generated call voice as the first call voice.

[0198] For example, Fig.12 FIG. 1 is a schematic diagram showing the composition of an information push device in an exemplary embodiment of the present disclosure. Fig.12 , the information push device 1200 may include the following program modules: a second similar voiceprint determination module 1210, configured to determine a fifth voiceprint feature similar to the fourth voiceprint feature corresponding to each user's voice, wherein the fourth voiceprint feature corresponds to an identifier and the fifth voiceprint feature corresponds to an identifier; a third quantity determination module 1220, configured to determine a third quantity of different identifiers in the identifiers; a second risk identification result determination module 1230, configured to determine the risk identification result of the fourth voiceprint feature according to the third quantity; an information push module 1240, configured to push prompt information to the terminal when the risk identification result is a target risk identification result, the target risk identification result indicating that the fourth voiceprint feature has a risk, and the prompt information indicating that the user corresponding to the user voice is a risky user.

[0199] The specific details of each part of the above-mentioned device have been described in detail in the corresponding method part implementation method. The undisclosed details can be found in the implementation method content of the method part, so they will not be repeated here.

[0200] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the exemplary embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0201] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0202] The exemplary embodiments of the present disclosure also provide a computer program product. The computer program product includes a computer program, and when the computer program is executed by a processor, the above-mentioned voiceprint library construction method and / or object recognition method is implemented.

[0203] In one embodiment, the computer program product may be a tangible product containing a computer program, such as a computer-readable storage medium storing a computer program. The readable storage medium may be a storage medium based on electrical, magnetic, optical, electromagnetic, infrared, or other signals, including but not limited to: random access memory (RAM), read-only memory (ROM), magnetic tape, floppy disk, flash memory (Flash), mechanical hard disk (HDD), solid-state drive (SSD), and the like. Exemplarily, the computer program product may be implemented as a non-volatile storage medium storing a computer program, such as a read-only memory, a NAND flash memory, and the like.

[0204] In one embodiment, the computer program product may be an intangible product including a computer program. Exemplarily, the computer program product may be implemented as a virtual digital product, such as a digital file storing an executable file, an installation package, etc. of the computer program.

[0205] The code of the computer program can be written in one or more programming languages. Programming languages ​​such as C language, Java, C++, Python, etc. The program code can be executed entirely on the user computing device, or partially on the user computing device, or as a separate software package, or partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device can be connected to the user computing device through any type of network, such as a local area network (LAN), a wide area network (WAN), etc., or can be connected to an external computing device (e.g., an Internet connection provided by an operator).

[0206] Computer programs can be carried or transmitted through electrical, magnetic, optical, electromagnetic, infrared and other signals. Electronic devices can convert signals carrying computer programs into digital signals to run the computer programs. When a computer program runs on an electronic device, its code is used to enable the electronic device to execute (more specifically, it can enable the processor of the electronic device to execute) the method steps of various exemplary embodiments of the present disclosure, such as the above-mentioned data processing method, which includes the following steps: determining a second voiceprint feature similar to a first voiceprint feature, the first voiceprint feature corresponding to an identifier, and the second voiceprint feature corresponding to an identifier; determining a first number of different identifiers in the identifier; and determining a risk identification result of the first voiceprint feature based on the first number.

[0207] By executing the above method steps through a computer program, on the one hand, risk identification can be automatically performed without human intervention through voiceprint feature processing, thereby improving the efficiency of risk identification; on the other hand, since voiceprint features are difficult to forge, risk identification can be performed through voiceprint feature processing to improve the accuracy of risk identification.

[0208] The exemplary embodiments of the present disclosure also provide an electronic device, such as the terminal device 110 or the server 120 described above. The electronic device may include a processor and a memory. The memory stores executable instructions of the processor, such as a computer program. The processor executes the method steps of various exemplary embodiments of the present disclosure by executing the executable instructions. In addition, the electronic device may also include a display for displaying a graphical user interface.

[0209] Reference below Fig.13 , the electronic device is exemplarily described in the form of a general-purpose computing device. It should be understood that Fig.13 The electronic device 1300 shown is only an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0210] like Fig.13As shown, the electronic device 1300 may include: a processor 1310 , a memory 1320 , a bus 1330 , an I / O (input / output) interface 1340 , a network adapter 1350 , and a display 1360 .

[0211] The memory 1320 may include a volatile memory, such as a RAM 1321, a cache unit 1322, and may also include a non-volatile memory, such as a ROM 1323. The memory 1320 may also include one or more program modules 1324, such program modules 1324 include but are not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or a combination thereof may include the implementation of a network environment. For example, the program module 1324 may include each module in the above-mentioned device.

[0212] The processor 1310 may include one or more processing units. For example, the processor 1310 may include an AP (Application Processor), a modem processor, a GPU (Graphics Processing Unit), an ISP (Image Signal Processor), a controller, an encoder, a decoder, a DSP (Digital Signal Processor), a baseband processor and / or an NPU (Neural-Network Processing Unit), etc.

[0213] The processor 1310 can be used to execute executable instructions stored in the memory 1320, such as executing the following method: determining a second voiceprint feature that is similar to a first voiceprint feature, the first voiceprint feature corresponds to an identifier, and the second voiceprint feature corresponds to an identifier; determining a first number of different identifiers in the identifiers; and determining a risk identification result of the first voiceprint feature based on the first number.

[0214] The above method is implemented through a computer program. On the one hand, by processing the voiceprint features, risk identification can be automatically performed without human intervention, thereby improving the efficiency of risk identification. On the other hand, since voiceprint features are difficult to forge, risk identification can be performed by processing voiceprint features to improve the accuracy of risk identification.

[0215] The bus 1330 is used to realize the connection between different components of the electronic device 1300, and may include a data bus, an address bus, and a control bus.

[0216] The electronic device 1300 can communicate with one or more external devices 1400 (eg, a keyboard, a mouse, an external controller, etc.) through the I / O interface 1340 .

[0217] The electronic device 1300 can communicate with one or more networks through the network adapter 1350. For example, the network adapter 1350 can provide mobile communication solutions such as 3G / 4G / 5G, or wireless communication solutions such as wireless LAN, Bluetooth, near field communication, etc. The network adapter 1350 can communicate with other modules of the electronic device 1300 through the bus 1330.

[0218] The electronic device 1300 may display a graphical user interface through the display 1360 , such as an interface for displaying pushed prompt information.

[0219] although Fig.13 Not shown, other hardware and / or software modules may also be provided in the electronic device 1300, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0220] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0221] As can be seen from the above, the technical solution of the present disclosure can be implemented as a method, an apparatus, a system, a computer program product, a storage medium, an electronic device, etc. Those skilled in the art can understand that various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or an implementation combining hardware and software, such as being respectively referred to as a "circuit", "module" or "system".

[0222] It should be understood that the present disclosure is not limited to the specific method steps or structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from the scope thereof. Those skilled in the art will easily think of other embodiments based on the specific embodiments provided by the present disclosure. Therefore, the specific embodiments provided by the present disclosure are only exemplary, and the scope and spirit of the present disclosure are indicated by the claims, and any variations, uses or adaptive changes of the present disclosure should be covered, which follow the general principles of the present disclosure and include common knowledge or customary technical means in the technical field that are not disclosed in the present disclosure.

Claims

1. A data processing method, characterized in that: include: Determine a second voiceprint feature similar to each first voiceprint feature, each of the first voiceprint features corresponds to an identifier, and each of the second voiceprint features corresponds to an identifier; determining a first number of different identifiers in the identifiers; A risk identification result of the first voiceprint feature is determined according to the first number.

2. The method according to claim 1, characterized in that The step of determining a risk identification result of the first voiceprint feature according to the first quantity includes: When the first number is greater than a first preset threshold, a risk identification result of the first voiceprint feature is determined according to the first voiceprint feature, the first identifier of the first voiceprint feature, the second voiceprint feature, and the second identifier of the second voiceprint feature.

3. The method according to claim 2, characterized in that The determining the risk identification result of the first voiceprint feature according to the first voiceprint feature, the first identifier of the first voiceprint feature, the second voiceprint feature, and the second identifier of the second voiceprint feature includes: Determine a plurality of candidate voiceprint feature pairs according to the first voiceprint feature, the first identifier, the second voiceprint feature, and the second identifier, wherein two voiceprint features of the candidate voiceprint feature pairs correspond to different identifiers; Determine a first intention similarity between two voiceprint features of the candidate voiceprint feature pair; Counting a second number of candidate voiceprint feature pairs whose first intention similarity is greater than or equal to a first similarity threshold; When the second number is greater than or equal to a second preset threshold, the risk identification result of the first voiceprint feature is determined to be a target risk identification result, and the target risk identification result indicates that the first voiceprint feature has a risk.

4. The method according to claim 3, characterized in that The determining the first intention similarity between two voiceprint features of the candidate voiceprint feature pair includes: Determining, according to the identifiers of the two voiceprint features of the candidate voiceprint feature pair, the work orders corresponding to the two voiceprint features respectively; Get the level label of each of the work orders; The first intent similarity is determined according to the levels of the hierarchical labels and the label information of each of the levels.

5. The method according to claim 4, characterized in that The determining the first intention similarity according to the hierarchy of the hierarchical labels and the label information of each hierarchy includes: Calculate the second intent similarity between tag information at the same level; The first intent similarity is determined according to the second intent similarity and the weight of the level corresponding to the second intent similarity.

6. The method according to claim 1, characterized in that The second voiceprint feature includes the voiceprint feature corresponding to the first call voice; The method for determining the first call voice includes: Determine a first work order corresponding to a preset work order type among the generated work orders; Determining a third identifier of the first work order; The call voice corresponding to the third identifier in the generated call voice is determined as the first call voice.

7. The method according to claim 1, characterized in that The step of determining a risk identification result of the first voiceprint feature according to the first quantity includes: In response to the first number being greater than a first preset threshold, the risk identification result of the first voiceprint feature is determined to be a target risk identification result, and the target risk identification result indicates that the first voiceprint feature has a risk.

8. An information push method, characterized in that: include: Determine a fifth voiceprint feature similar to a fourth voiceprint feature corresponding to each user's voice, wherein the fourth voiceprint feature corresponds to an identifier and the fifth voiceprint feature corresponds to an identifier; determining a third number of different identifications in the identifications; Determining a risk identification result of the fourth voiceprint feature according to the third number; In the case where the risk identification result is a target risk identification result, prompt information is pushed to the terminal, the target risk identification result indicates that the fourth voiceprint feature has a risk, and the prompt information indicates that the user corresponding to the user voice is a risky user.

9. A data processing device, characterized in that: include: A first similar voiceprint determining module is configured to determine a second voiceprint feature similar to each first voiceprint feature, wherein the first voiceprint feature has an identifier and the second voiceprint feature corresponds to an identifier; A first quantity determination module is configured to determine a first quantity of different identifiers in the identifiers; The first risk identification result determination module is configured to determine a risk identification result of the first voiceprint feature according to the first quantity.

10. An information push device, characterized in that: include: A second similar voiceprint determination module is configured to determine a fifth voiceprint feature similar to a fourth voiceprint feature corresponding to each user's voice, wherein the fourth voiceprint feature corresponds to an identifier and the fifth voiceprint feature corresponds to an identifier; A third quantity determination module is configured to determine a third quantity of different identifiers in the identifiers; a second risk identification result determination module, configured to determine a risk identification result of the fourth voiceprint feature according to the third number; The information push module is configured to push prompt information to the terminal when the risk identification result is a target risk identification result, wherein the target risk identification result indicates that the fourth voiceprint feature has a risk, and the prompt information indicates that the user corresponding to the user voice is a risky user.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

12. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

13. An electronic device, characterized in that: include: one or more processors; A storage device for storing one or more programs, which, when executed by the one or more processors, enables the one or more processors to implement the method according to any one of claims 1 to 8.

Images