A compliance identification method and identification device

Abstract

The present application discloses a compliance identification method and an identification device, which belong to the field of artificial intelligence technology. The compliance identification method includes: obtaining a first voice in a bank marketing process; when the speaker corresponding to the first voice is a bank employee, determining the target stage to which the first voice belongs based on a second voice in the bank marketing process and a customer portrait corresponding to the bank marketing process; comparing the semantics of the first voice with the semantics of the standard speech of the target stage, and outputting at least one of an alarm message and a prompt message when the semantics of the two do not match. The compliance identification method and the identification device provided by the present application detect whether the marketing behavior of bank employees complies with the standard operating procedures by combining the standard operating procedures of the marketing scenario, discover and correct non-compliant behaviors that do not comply with the standard operating procedures by outputting information, and can manage the marketing behavior of the bank more effectively and efficiently.

Description

Technical Field

[0001] The present application belongs to the field of artificial intelligence technology, and in particular, relates to a compliance identification method and an identification device. Background Art

[0002] The service quality of bank branch staff (such as lobby managers, tellers or financial managers, etc.) is very important to both banks and branches. However, the implementation process of marketing behaviors of bank branch staff (hereinafter referred to as "bank employees") is difficult to supervise and manage, and it is difficult to timely and accurately identify whether the marketing behaviors of bank employees are compliant. Therefore, how to effectively identify the compliance of bank employees' marketing behaviors has become a technical issue that needs to be solved urgently in the banking industry. Summary of the invention

[0003] The present application aims to solve at least one of the technical problems existing in the prior art. To this end, the present application proposes a compliance identification method and an identification device, which can identify the compliance of marketing behavior and manage the marketing behavior of the bank more effectively and efficiently.

[0004] In a first aspect, the present application provides a compliance identification method, the method comprising:

[0005] Get the first voice in the bank's marketing process;

[0006] In the case where the speaker corresponding to the first voice is a bank employee, based on the second voice in the bank marketing process and the customer portrait corresponding to the bank marketing process, determining the target stage to which the first voice belongs; the target stage is the stage in the standard operating procedure of the target marketing scenario; the target marketing scenario is the marketing scenario corresponding to the first voice; the second voice data is the voice before the first voice in the bank marketing process;

[0007] The semantics of the first speech are compared with the semantics of the standard speech of the target stage, and if the semantics of the two do not match, at least one of a warning message and a prompt message is output.

[0008] According to the compliance identification method of the present application, by combining the standard operating procedures of the marketing scenario, it is determined whether the semantics of the bank employee's voice clip matches the semantics of the standard speech of the stage of the standard operating procedures of the marketing scenario to which the voice clip belongs, and whether the marketing behavior of the bank employee complies with the standard operating procedures. It is identified whether the marketing behavior is compliant, and non-compliant behaviors that do not comply with the standard operating procedures are discovered and corrected by outputting information. The implementation process of the marketing behavior can be supervised and managed, the marketing behavior of the bank employees can be standardized, and the marketing behavior of the bank can be managed more effectively and efficiently.

[0009] According to an embodiment of the present application, before acquiring the first voice in the banking marketing process, the method further includes:

[0010] Acquire the second voice;

[0011] Performing text analysis on the text corresponding to the second voice to obtain characteristic information of the customer among the speakers corresponding to the second voice;

[0012] Based on the feature information, the customer portrait is obtained.

[0013] According to the compliance identification method of the present application, by obtaining customer portraits, the target marketing scenario and the target stage to which the first voice belongs can be determined more accurately, so that it is possible to more accurately detect whether the marketing behavior of bank employees complies with standard operating procedures, identify whether the marketing behavior is compliant, discover and correct non-compliant behaviors that do not comply with standard operating procedures by outputting information, supervise and manage the implementation process of marketing behaviors, standardize the marketing behaviors of bank employees, and manage the marketing behaviors of banks more effectively and efficiently.

[0014] According to an embodiment of the present application, after acquiring the first voice in the bank marketing process, the method further includes:

[0015] In the case where the speaker corresponding to the first voice is a bank employee, the first voice is detected based on at least one of keywords, regular expressions, semantics, emotional rules, and speech speed rules to determine whether the bank employee has any illegal speech behavior;

[0016] In the event of any illegal speech behavior, at least one of a warning message and a prompt message is output.

[0017] According to the compliance identification method of the present application, by detecting whether bank employees have any illegal speech behaviors in the marketing process based on at least one of keywords, regular expressions, semantics, sentiment rules and speech speed rules, it is identified whether the marketing behaviors are compliant, and non-compliant behaviors such as illegal speech behaviors are discovered and corrected by outputting information. It is possible to supervise and manage illegal speech behaviors in marketing behaviors, standardize the marketing behaviors of bank employees, and manage the marketing behaviors of banks more effectively and efficiently.

[0018] According to an embodiment of the present application, when the speaker corresponding to the first voice is a bank employee, the first voice is detected based on at least one of keywords, regular expressions, semantics, emotional rules, and speech speed rules to determine whether the bank employee has illegal speech behavior, including:

[0019] Performing text classification on the first speech based on the TextRCNN model to determine whether the speaker corresponding to the first speech is a bank employee;

[0020] In the case that the speaker corresponding to the first voice is a bank employee, the first voice is detected based on a rule template to determine whether the bank employee has any illegal speech behavior; the rule template is used to represent at least one of keywords, regular expressions, semantics, emotional rules and speech speed rules.

[0021] According to the compliance identification method of the present application, text classification and key information extraction are performed based on the TextRCNN model, and based on at least one of keywords, regular expressions, semantics, sentiment rules and speech speed rules, it is detected whether bank employees have any illegal speech behaviors in the marketing process, and whether the marketing behaviors are compliant is identified. Illegal speech behaviors and other non-compliant behaviors are discovered and corrected by outputting information. Illegal speech behaviors in marketing behaviors can be supervised and managed, the marketing behaviors of bank employees can be standardized, the marketing behaviors of banks can be managed more effectively and efficiently, the compliance level of bank outlets' operations can be improved, and the management capabilities of banks can be strengthened.

[0022] According to an embodiment of the present application, comparing the semantics of the first speech with the semantics of the standard speech of the target stage includes:

[0023] Performing speech recognition processing on the first speech to obtain a first text;

[0024] Based on a corpus in the field of banking and finance, correct the first text to obtain a second text;

[0025] The second text and the text of the standard speech are input into the Sentence-BERT model to obtain the similarity between the semantics of the first speech and the semantics of the standard speech of the target stage.

[0026] According to the compliance identification method of the present application, based on the Sentence-BERT model, the semantics of the first speech are compared with the semantics of the standard speech of the stage, which can more accurately determine whether the semantics of the first speech are similar to the semantics of the standard speech of the target stage, thereby detecting whether the marketing behavior of bank employees complies with the standard operating procedures, identifying whether the marketing behavior is compliant, discovering and correcting non-compliant behaviors that do not comply with the standard operating procedures by outputting information, and supervising and managing the implementation process of marketing behaviors, standardizing the marketing behaviors of bank employees, and managing the marketing behaviors of banks more effectively and efficiently. In addition, based on the fact that there are great differences between the terms in the banking and financial fields and those in other fields, automatic error correction through the corpus of the banking and financial fields can greatly improve the accuracy of speech recognition, thereby improving the accuracy of semantic analysis of speech data, and more accurately and effectively identifying the compliance of bank marketing behaviors and managing bank marketing.

[0027] According to an embodiment of the present application, the obtaining of the first voice in the banking marketing process includes:

[0028] When at least two persons are detected in the target area, or when abnormal sounds are detected in the target area, the recording function is started to obtain voice data of the bank marketing process; the voice data includes the first voice.

[0029] According to the compliance identification method of the present application, by adopting at least one of the human body sensing to start the recording function scheme and the abnormal sound to start the recording function scheme, the voice data of the bank marketing process is obtained, which can solve the problem of unreasonable recording timing. Under the condition of collecting as much and complete voice data as possible, the utilization rate of the pickup equipment can be improved, and the software and hardware resources consumed by various algorithms in the later stage can be reduced.

[0030] In a second aspect, the present application provides a compliance identification device, the device comprising:

[0031] An acquisition module, used to acquire the first voice in the bank marketing process;

[0032] A determination module, configured to determine, when the speaker corresponding to the first voice is a bank employee, based on a second voice in the bank marketing process and a customer portrait corresponding to the bank marketing process, a target stage to which the first voice belongs; the target stage is a stage in a standard operating procedure for a target marketing scenario; the target marketing scenario is a marketing scenario corresponding to the first voice; and the second voice data is the voice before the first voice in the bank marketing process;

[0033] The output module is used to compare the semantics of the first speech with the semantics of the standard speech of the target stage, and output at least one of the warning information and the prompt information when the semantics of the two do not match.

[0034] According to the compliance identification device of the present application, by combining the standard operating procedures of the marketing scenario, it is determined whether the semantics of the bank employee's voice segment matches the semantics of the standard speech of the stage of the standard operating procedures of the marketing scenario to which the voice segment belongs, detects whether the marketing behavior of the bank employee complies with the standard operating procedures, identifies whether the marketing behavior is compliant, discovers and corrects non-compliant behaviors that do not comply with the standard operating procedures by outputting information, can supervise and manage the implementation process of the marketing behavior, standardize the marketing behavior of the bank employee, and can manage the marketing behavior of the bank more effectively and efficiently.

[0035] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the compliance identification method as described in the first aspect above is implemented.

[0036] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the compliance identification method as described in the first aspect above.

[0037] In a fifth aspect, the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the compliance identification method as described in the first aspect.

[0038] In a sixth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the compliance identification method as described in the first aspect above.

[0039] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0041] Figure 1 It is a flowchart of the compliance identification method provided in the embodiment of the present application;

[0042] Figure 2 It is a flowchart of the SOP for the complaint handling scenario in the compliance identification method provided in the embodiment of the present application;

[0043] Figure 3 It is a flow chart of the SOP for a large-amount transfer business scenario or a large-amount withdrawal business scenario in the compliance identification method provided in an embodiment of the present application;

[0044] Figure 4 It is a schematic diagram of the principle of detecting illegal speech behavior in the compliance identification method provided in the embodiment of the present application;

[0045] Figure 5 It is a schematic diagram of the principle of the Sentence-BERT model in the compliance identification method provided in an embodiment of the present application;

[0046] Figure 6 It is a flowchart of the sub-steps of configuring step 120 and step 130 in the compliance identification method provided in an embodiment of the present application;

[0047] Figure 7 is a schematic diagram of the structure of a compliance identification device provided in an embodiment of the present application;

[0048] Figure 8 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0049] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0050] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0051] The compliance identification method, compliance identification device, electronic device and readable storage medium provided in the embodiments of the present application are described in detail below with reference to the accompanying drawings through specific embodiments and their application scenarios.

[0052] The compliance identification method may be applied to a terminal, and may be specifically executed by hardware or software in the terminal.

[0053] The terminal includes, but is not limited to, a portable communication device such as a mobile phone or tablet computer with a touch-sensitive surface (e.g., a touch screen display and / or a touch pad). It should also be understood that in some embodiments, the terminal may not be a portable communication device, but a desktop computer with a touch-sensitive surface (e.g., a touch screen display and / or a touch pad).

[0054] In the following various embodiments, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, a mouse and a joystick.

[0055] The compliance identification method provided in the embodiment of the present application may be executed by an electronic device or a functional module or functional entity in the electronic device that can implement the compliance identification method. The electronic devices mentioned in the embodiment of the present application include but are not limited to mobile phones, tablet computers, computers, cameras, and wearable devices. The compliance identification method provided in the embodiment of the present application is described below using an electronic device as an example of an execution subject.

[0056] like Figure 1 As shown, the compliance identification method includes: step 110, step 120 and step 130.

[0057] Step 110: Acquire the first voice of the bank marketing process.

[0058] In actual implementation, the first voice in the bank marketing process can be obtained by recording the bank marketing process.

[0059] It should be noted that the bank marketing process in the embodiment of the present application can be applicable to the process of face-to-face on-site marketing activities conducted by bank employees at bank branches and customers, and can also be applicable to the process of marketing activities conducted by bank employees with customers over the phone.

[0060] According to whether the speaker changes and whether the time interval between sentences exceeds a preset time threshold, a speech segment of the same speaker in the bank marketing process can be obtained as the first speech.

[0061] It is understandable that the speakers in the banking marketing process may include at least one customer and at least one bank employee.

[0062] In some embodiments, when a change in the speaker is detected, a segment of speech by the previous speaker may be used as the first speech.

[0063] Detecting whether the speaker has changed may be based on speech features such as voiceprints. The embodiment of the present application does not specifically limit the specific method for detecting whether the speaker has changed.

[0064] In some embodiments, when it is detected that the speaker has not changed, but no new sentence is detected within a time threshold after the previous sentence, the speech including the previous sentence can be used as the first speech.

[0065] The duration threshold can be set according to actual conditions such as human speaking habits. The specific value of the duration threshold is not specifically limited in the embodiment of the present application.

[0066] Step 120: When the speaker corresponding to the first voice is a bank employee, based on the second voice in the bank marketing process and the customer portrait corresponding to the bank marketing process, determine the target stage to which the first voice belongs; the target stage is the stage in the standard operating procedure of the target marketing scenario; the target marketing scenario is the marketing scenario corresponding to the first voice; the second voice data is the voice before the first voice in the bank marketing process.

[0067] In actual implementation, for the first voice, it can be determined whether the speaker corresponding to the first voice is a bank employee or a customer.

[0068] The judgment of whether the speaker corresponding to the first voice is a bank employee or a customer may be based on the bank employee's voiceprint and other features. The embodiment of the present application does not specifically limit the specific method of judging whether the speaker corresponding to the first voice is a bank employee or a customer.

[0069] In the case where the speaker corresponding to the first voice is a bank employee, the context semantics of multiple consecutive sentences in the second voice that occurred before the first voice in the marketing process of the bank can be analyzed, and the marketing scenario corresponding to the first voice can be determined in combination with the customer portrait, and based on the standard operating procedure (SOP) of the marketing scenario, it can be determined which stage of the standard operating procedure the first voice belongs to. This stage is the target stage. The marketing scenario corresponding to the first voice is the target marketing scenario.

[0070] In some optional implementations, the semantics of a sentence can be obtained by any common semantic analysis method (such as a semantic analysis method based on a pre-trained language model or a semantic analysis method oriented to business modeling, etc.). The specific semantic analysis method used is not specifically limited in the embodiments of the present application.

[0071] The above-mentioned marketing scenarios may include but are not limited to complaint handling, financial management business, insurance business, loan business, credit card business, deposit business, large-value transfer business and large-value withdrawal business.

[0072] It is understandable that the bank marketing process may include at least one marketing scenario. For example, the bank marketing process is to recommend customers to buy financial products, and complaints are generated during the process. Therefore, the bank marketing process includes two marketing scenarios, namely, financial services and complaints. The voice data of the bank marketing process can be divided into two data segments, corresponding to the two marketing scenarios of financial services and complaint handling.

[0073] For another example, in the marketing process of a bank, a customer first inquires about a large-value transfer service, and then purchases an insurance product and applies for a credit card on the recommendation of a bank employee. Therefore, the marketing process of the bank includes three marketing scenarios: large-value transfer service, insurance service, and credit card service. The voice data of the bank's marketing process can be divided into three data segments, corresponding to the three marketing scenarios of large-value transfer service, insurance service, and credit card service.

[0074] For another example, the bank marketing process is to recommend suitable loan products to customers when they have financial needs. Therefore, the bank marketing process includes a marketing scenario of loan business. The voice data of the bank marketing process as a whole is regarded as a data segment, and the corresponding marketing scenario is loan business.

[0075] The customer portrait corresponding to the bank marketing process refers to the portrait of the customer participating in the bank marketing process. The customer portrait can be obtained based on the historical behavior data of the customer, or can also be obtained based on the second voice data.

[0076] Historical behavior data refers to historical conversation data between bank employees and customers. The historical conversation data may come from at least one of the bank's surveillance video, audio recording, and text chat records based on communication programs.

[0077] In some embodiments, based on any customer profiling method, at least one item of the second voice data in the above historical behavior data can be analyzed to obtain a description of the customer's characteristics, thereby obtaining a customer profiling. The specific customer profiling method used is not specifically limited in the embodiments of the present application.

[0078] It should be noted that, according to different marketing scenarios of bank marketing, banks have formulated standard operating procedures for each marketing scenario in advance. The standard operating procedures for each marketing scenario can include multiple different stages.

[0079] Figure 2 is a flow chart of the SOP for the complaint handling scenario in the compliance identification method provided in the embodiment of the present application. For example, Figure 2 As shown, the SOP for complaint handling scenarios can include the following stages: customer emotional comfort stage, problem information collection stage, solution proposal stage, and customer opinion solicitation stage.

[0080] Figure 3 : is a flow chart of the SOP for a large amount transfer business scenario or a large amount withdrawal business scenario in the compliance identification method provided in the embodiment of the present application. Figure 3 As shown, the SOP for large-value transfer business scenarios or large-value withdrawal business scenarios may include the following stages: the stage of inquiring about the reason; when the reason is telecommunications fraud, the next stage of the stage of inquiring about the reason is the stage of announcing risk warnings and dissuading the customer; when dissuasion fails, the next stage of the stage of announcing risk warnings and dissuading the customer is the stage of contacting the police for assistance; when the reason is other, the next stage of the stage of inquiring about the reason is the stage of recommending other products or services of the Bank to customers; the next stage of the stage of recommending other products or services of the Bank to customers is the stage of establishing customer contact.

[0081] For the first speech, it can be determined according to the semantics of the second speech to which stage of the standard operating procedure the first speech belongs.

[0082] For example, based on the semantics of the second voice, it is determined that the customer emotional comfort stage of the SOP for the complaint handling scenario has been completed, then it can be determined that the first voice belongs to the problem information collection stage of the SOP for the complaint handling scenario.

[0083] For another example, based on the semantics of the second voice, it is determined that the inquiry reason stage of the SOP for large-value transfer business has been completed, and it is determined to be telecommunications fraud, then it can be determined that the first voice belongs to the risk warning and dissuasion stage of the SOP for large-value transfer business.

[0084] In some optional implementations, the above process may include the following steps:

[0085] First, based on whether the continuous speech belongs to the same speaker, a speech of a certain speaker can be obtained as the first speech;

[0086] Determining that the speaker corresponding to the first speech is a bank employee;

[0087] When the marketing scenario corresponding to the first voice is not determined, the semantics of the second voice can be analyzed and combined with the customer portrait to determine the marketing scenario corresponding to the first voice, that is, to determine the target marketing scenario;

[0088] When the marketing scenario corresponding to the first voice has been determined, it is possible to determine which stage of the SOP of the target marketing scenario the first voice belongs to based on the semantics of the second voice and the SOP of the target marketing scenario, that is, to determine the target stage.

[0089] Step 130: Compare the semantics of the first speech with the semantics of the standard speech in the target stage, and if the semantics of the two do not match, output at least one of a warning message and a prompt message.

[0090] In actual implementation, based on the pre-trained model, the semantics of each sentence in the first speech can be compared with the semantics of each sentence in the standard speech of the target stage to which the first speech belongs, so as to determine whether the semantics of the first speech match the semantics of the standard speech of the target stage to which the first speech belongs. Whether the semantics match refers to whether the semantics are similar.

[0091] The above model can be used to obtain the semantic similarity of two sentences. When the semantic similarity of the two sentences is greater than or equal to the first threshold, it can be determined that the semantics of the two sentences are matched; conversely, when the semantic similarity of the two sentences is less than the first threshold, it can be determined that the semantics of the two sentences are not matched.

[0092] It should be noted that the first threshold value may be predetermined according to actual conditions. The specific value of the first threshold value is not specifically limited in the embodiment of the present application.

[0093] It is determined whether the semantics of each sentence in the first speech match the semantics of each sentence in the standard speech of the target stage to which the first speech belongs.

[0094] In actual implementation, for each sentence in the standard speech of the target stage to which the first speech belongs, when the number of sentences without semantic matching in the first speech is greater than the second threshold or the proportion is greater than the third threshold, it can be determined that the semantics of the first speech do not match the semantics of the standard speech of the target stage to which the first speech belongs.

[0095] For each sentence in the standard speech of the target stage to which the first speech belongs, when the number of sentences without semantic matching in the first speech is less than the fourth threshold or the proportion is less than the fifth threshold, it can be determined that the semantics of the first speech matches the semantics of the standard speech of the target stage to which the first speech belongs.

[0096] For example, for each sentence in the standard speech of the target stage to which the first speech belongs, if there is a semantically matching sentence in the first speech, it can be determined that the semantics of the first speech matches the semantics of the standard speech of the target stage to which the first speech belongs.

[0097] It should be noted that the second threshold, the third threshold, the fourth threshold and the fifth threshold can be predetermined according to actual conditions. The specific values ​​of the second threshold, the third threshold, the fourth threshold and the fifth threshold are not limited in the embodiment of the present application.

[0098] After determining whether the semantics of the first voice matches the semantics of the standard speech of the target stage to which the first voice belongs, if the semantics of the two do not match, it can be determined that the marketing behavior corresponding to the first voice does not comply with the standard operating procedures and an alarm message is output to warn the bank employee, reminding him that his marketing behavior does not comply with the standard operating procedures and needs to be corrected immediately. If the semantics of the two do not match, a prompt message can also be output to remind relevant managers (such as the bank employee's leader or compliance department staff, etc.) that the bank employee's marketing behavior does not comply with the standard operating procedures.

[0099] It is understandable that the marketing behavior does not comply with standard operating procedures, which means that the compliance of the marketing behavior is identified as non-compliant and the marketing behavior is non-compliant.

[0100] Through the above process, a monitoring and feedback mechanism can be established to detect and correct possible behaviors that do not conform to standard operating procedures by monitoring the conversations of bank employees in real time. If it is found that bank employees may have behaviors that do not conform to standard operating procedures, appropriate measures can be taken immediately to deal with them and feedback can be given to relevant managers for timely improvement.

[0101] In some optional implementations, when the semantics of the first voice do not match the semantics of the standard speech of the target stage to which the first voice belongs, the first voice can be marked to facilitate subsequent manual verification and processing. In this case, there is no need to manually listen to the voice data of the entire marketing process, and only the marked voice segments are manually verified, which can greatly improve efficiency.

[0102] It is understandable that the marking of the voice segment can be implemented in a variety of flexible forms, such as marking the start and end points of the voice segment, or extracting the voice segment from the voice data and adding a label as "to be verified".

[0103] According to the compliance identification method provided in the embodiment of the present application, by combining the standard operating procedures of the marketing scenario, it is determined whether the semantics of the bank employee's voice segment matches the semantics of the standard speech of the stage of the standard operating procedures of the marketing scenario to which the voice segment belongs, and whether the marketing behavior of the bank employee complies with the standard operating procedures. It is identified whether the marketing behavior is compliant, and non-compliant behaviors that do not comply with the standard operating procedures are discovered and corrected by outputting information. The implementation process of the marketing behavior can be supervised and managed, the marketing behavior of the bank employees can be standardized, and the marketing behavior of the bank can be managed more effectively and efficiently.

[0104] In some embodiments, before acquiring the first voice in the banking marketing process, the method further includes: acquiring a second voice.

[0105] In actual implementation, the method for obtaining the second voice is the same as the method for obtaining the first voice. Therefore, the specific process of obtaining the second voice can refer to the above embodiment and will not be repeated here.

[0106] Perform text analysis on the text corresponding to the second voice to obtain feature information of the customer among the speakers corresponding to the second voice.

[0107] In actual implementation, natural language processing technology can be used to perform text analysis on the text corresponding to the second voice to obtain customer information of each customer participating in the bank's marketing process; key information in the customer information can be further extracted as feature information of the customer.

[0108] In some optional implementations, the above-mentioned customer information may include but is not limited to information on dimensions such as customer preferences, loan preferences, investment preferences, life attributes, consumption characteristics, work attributes, family status, suspected customer complaints, company information, relationship maps, industry maps and corporate portraits.

[0109] In some optional implementations, the above-mentioned characteristic information may include the client's language style, topic focus, emotional state, etc.

[0110] Obtain customer portraits based on feature information.

[0111] In actual implementation, for each customer, the customer's portrait label can be determined based on the customer's characteristic information; based on all the customer's portrait labels, the customer's portrait can be obtained.

[0112] It should be noted that different algorithms can be used to learn and predict the characteristic information of sample customers according to different types of tasks, so as to establish a customer portrait label system that conforms to the characteristics of the banking and financial industry. The customer portrait label system can include portrait labels of multiple dimensions. The number of portrait labels in each dimension can be multiple.

[0113] The portrait labels of the customer preference dimension may include business preferences, business processing channels and customer intentions, etc.

[0114] The profile labels of the loan preference dimension may include funding requirements, loan type, and loan intention, etc.

[0115] The portrait labels of the investment preference dimension may include investment amount, deposit preference, fund preference, gold preference, foreign exchange preference, bond preference, investment period, investment years, investment experience and risk preference, etc.

[0116] The portrait labels of the life attribute dimension may include income source, vehicle condition, whether there is a car loan, living conditions and annual income, etc.

[0117] The portrait labels of the consumption characteristic dimension may include major economic expenditures, consumption channels, brand preferences, consumption hotspots, work areas and entertainment methods, etc.

[0118] The portrait labels of the work attribute dimension may include work unit, social security status, industry category and job position, etc.

[0119] The portrait labels of the family status dimension may include family status and children's status, etc.

[0120] The portrait tags of the suspected customer complaint dimension may include suspected customer complaints, etc.

[0121] The portrait tags of the company information dimension may include company name, business scale, company industry and annual income, etc.

[0122] The portrait labels of the relationship map dimension may include equity relationship, group relationship, transaction relationship, litigation relationship, guarantee relationship and actual controller, etc.

[0123] The portrait labels of the industrial map dimension can include upstream and downstream of the industry, policy planning, industrial status, industry structure, enterprise distribution and customer echelons, etc.

[0124] The portrait tags of the enterprise portrait dimension may include credit load, efficiency, industry classification, risk profile, ownership classification and financial status, etc.

[0125] It should be noted that customer portrait labels and customer portraits can improve the utilization and value of voice data, and can more accurately determine the secondary marketing scenario corresponding to the first voice.

[0126] According to the compliance identification method of the present application, by obtaining customer portraits, the target marketing scenario and the target stage to which the first voice belongs can be determined more accurately, so that it is possible to more accurately detect whether the marketing behavior of bank employees complies with standard operating procedures, identify whether the marketing behavior is compliant, discover and correct non-compliant behaviors that do not comply with standard operating procedures by outputting information, supervise and manage the implementation process of marketing behaviors, standardize the marketing behaviors of bank employees, and manage the marketing behaviors of banks more effectively and efficiently.

[0127] In some embodiments, after obtaining the first voice of the bank marketing process, the method also includes: when the speaker corresponding to the first voice is a bank employee, based on at least one of keywords, regular expressions, semantics, emotional rules and speech speed rules, detecting the first voice to determine whether the bank employee has any illegal speech behavior.

[0128] In actual implementation, in addition to checking whether the marketing behavior of bank employees complies with standard operating procedures, it is also possible to check whether bank employees have any illegal sales talk during the marketing process. The above illegal sales talk behaviors may include but are not limited to flying orders, operating on behalf of customers, illegal promises, avoiding double recording and coercing transactions.

[0129] It is understandable that marketing behaviors that do not comply with standard operating procedures are one type of non-compliant behaviors, and behaviors that use illegal speech techniques are another type of non-compliant behaviors.

[0130] After determining that the speaker corresponding to the first voice is a bank employee, at least one violation detection method can be flexibly adopted to detect the first voice. The above-mentioned violation detection method can include at least one of a keyword-based detection method, a regular expression-based detection method, a semantics-based detection method, an emotion rule-based detection method, and a speech speed rule-based detection method.

[0131] Detecting the first speech based on keywords means determining whether the bank employee has illegal speech behavior by detecting whether the bank employee's speech in the first speech includes specific keywords. If the bank employee's speech includes any specific keywords, it can be preliminarily determined that the bank employee has illegal speech behavior; if the bank employee's speech does not include any specific keywords, it can be determined that the bank employee has not violated the law.

[0132] For example, the keywords may be "cash back" and "absolutely safe", etc. If it is detected that the bank employee's speech includes "cash back" or "absolutely safe", it can be determined that the bank employee has violated the regulations in speech.

[0133] It is understandable that, in the step of detecting the voice data based on the keyword, at least one keyword may be pre-set. The specific keyword is not limited in the embodiment of the present application.

[0134] By building a customer-employee violation keyword library containing a certain number of keywords and detecting illegal speech behaviors, the compliance level of bank branches' operations can be improved and the bank's management capabilities can be strengthened.

[0135] Detecting the first speech based on a regular expression means searching the bank employee's speech in the first speech through a preset regular expression to determine whether the bank employee has any illegal speech behavior. If the bank employee's speech that matches the regular expression is found, it can be preliminarily determined that the bank employee has any illegal speech behavior; if the bank employee's speech that matches the regular expression is not found, it can be determined that the bank employee has no illegal speech behavior.

[0136] It is understandable that, in the step of detecting the voice data based on the regular expression, at least one regular expression may be pre-set. The specific regular expression is not limited in the embodiment of the present application.

[0137] Detecting the first speech based on semantics means analyzing the semantics of the bank employee's speech in the first speech to determine whether the bank employee has violated the law in speech. If the semantics of the bank employee's speech is a certain violation of the law in speech, it can be preliminarily determined that the bank employee has violated the law in speech; if the semantics of the bank employee's speech is not a violation of the law in speech, it can be determined that the bank employee has not violated the law in speech.

[0138] It is understandable that a commonly used semantic model can be used to analyze the semantics of the bank employee's speech in the first voice. The specific semantic model is not limited in the embodiment of the present application.

[0139] Detecting the first speech based on the emotion rule means obtaining the emotions of both parties in the conversation by analyzing the first speech, and then judging whether the bank employee has violated the rules of speech according to whether the emotions of the customer and / or employee meet the preset first rule. If the first rule is met, it can be preliminarily judged that the bank employee has violated the rules of speech; if the first rule is not met, it can be judged that the bank employee has not violated the rules of speech.

[0140] It is understandable that a commonly used conversation emotion recognition model can be used to analyze the first speech to obtain the emotions of the two parties in the conversation. The specific conversation emotion recognition model is not limited in the embodiment of the present application.

[0141] It is understandable that the first rule is a rule about the speaker's emotions. The first rule can be predetermined according to actual conditions. The specific first rule is not limited in the embodiment of the present application.

[0142] Detecting the first speech based on the speech speed rule means analyzing the first speech to obtain the speech speed of both parties in the conversation, and then judging whether the bank employee has violated the rules of speech skills according to whether the speech speed of the customer and / or bank employee meets the preset second rule. If the second rule is met, it can be preliminarily judged that the bank employee has violated the rules of speech skills; if the second rule is not met, it can be judged that the bank employee has not violated the rules of speech skills.

[0143] It is understandable that a commonly used speech rate estimation method or speech rate detection method can be used to analyze the first speech to obtain the speech rates of the two parties in the conversation. The specific speech rate estimation method and speech rate detection method used are not limited in the embodiments of the present application.

[0144] It is understandable that the second rule is a rule about the speaker's speaking speed. The second rule can be predetermined according to actual conditions. The specific second rule is not limited in the embodiment of the present application.

[0145] In actual implementation, the first speech can be detected based on at least two of keywords, regular expressions, semantics, emotional rules and speech speed rules to improve the accuracy of detecting illegal speech behaviors.

[0146] In the event of any illegal speech behavior, at least one of a warning message and a prompt message is output.

[0147] In actual implementation, if a bank employee has engaged in illegal speech, an alarm message can be output to warn the bank employee, reminding him / her that he / she has engaged in illegal speech and needs to stop immediately. If a bank employee has engaged in illegal speech, a prompt message can also be output to remind relevant managers (such as the bank employee's leader or compliance department staff, etc.) that the bank employee has engaged in illegal speech.

[0148] It is understandable that determining that the bank employee did not engage in any illegal speech behavior means that the compliance of the marketing behavior corresponding to the recognition of the first voice is non-compliant, and the marketing behavior is a non-compliant behavior.

[0149] In some optional implementations, in the case of illegal speech behavior, the first speech may also be marked.

[0150] It is understandable that marking the illegal part can be implemented in a variety of flexible ways, such as marking the starting and ending points of the illegal part, or cutting out the illegal part from the voice data and adding a label as "illegal", etc.

[0151] The first voice that is initially flagged as having illegal speech behavior can be manually verified later.

[0152] According to the compliance identification method provided in the embodiment of the present application, by detecting whether bank employees have any illegal speech behaviors in the marketing process based on at least one of keywords, regular expressions, semantics, emotional rules and speech speed rules, identifying whether the marketing behavior is compliant, discovering and correcting non-compliant behaviors such as illegal speech behaviors by outputting information, it is possible to supervise and manage illegal speech behaviors in marketing behaviors, standardize the marketing behaviors of bank employees, and manage the marketing behaviors of banks more effectively and efficiently. Compared with the method of manually detecting illegal speech behaviors, the compliance identification method provided in the embodiment of the present application can greatly improve the efficiency of detection.

[0153] In some embodiments, when the speaker corresponding to the first voice is a bank employee, the first voice is detected based on at least one of keywords, regular expressions, semantics, emotional rules and speech speed rules to determine whether the bank employee has any illegal speech behavior, including: performing text classification on the first voice based on the TextRCNN model to determine whether the speaker corresponding to the first voice is a bank employee.

[0154] In actual implementation, the first speech can be classified based on the trained TextRCNN model to distinguish different speakers and determine whether the speaker corresponding to the first speech is a bank employee.

[0155] It should be noted that, if the speakers in the first voice and the second voice include m bank employees, then m sets of first speech scripts can be obtained. Where m is a positive integer. The speech script of each bank employee is a set of first speech scripts. Similarly, if the speakers in the voice data include n customers, then n sets of second speech scripts can be obtained. Where n is a positive integer. The speech script of each customer is a set of second speech scripts.

[0156] The above-mentioned TextRCNN model can be a text classification model established based on TextRCNN combined with a convolutional neural network (CNN). The model can first use CNN to capture local features in a sentence, and then use a recurrent neural network (RNN) to capture global features, thereby improving the performance of the model.

[0157] The above TextRCNN model can classify the words spoken by bank employees and customers. By introducing convolution operations on the basis of RNN, the above TextRCNN model can better capture local information in the text, so it can be applied to situations where there are multiple relationships between entities.

[0158] It should be noted that the above-trained TextRCNN model can be obtained by training on the basis of a pre-trained model in the banking and finance field improved based on TextRCNN.

[0159] During the training process, model training can be performed after parameter setting and preprocessing of training data.

[0160] The parameters that need to be set may include: the number of trainings, the frequency of evaluating the model, the frequency of saving the model, high-dimensional sparse features, the dimension of the embedding vector, the number of feature maps, the size of the convolution kernel, the probability of retaining neurons, the weight of the regularization term, the mean of all sequence lengths, the number of samples in the model, classification settings (the value of this parameter can be set to 1 for binary classification and the value of this parameter can be set to the number of categories for multi-classification) and the proportion of the training set, etc.

[0161] The specific values ​​of the above parameters are not limited in this embodiment of the present application.

[0162] Preprocessing the training data may include but is not limited to the following steps: loading data, segmenting sentences into word representations, and removing low-frequency words and stop words; generating training sets and test sets based on the training data; reading word vectors from a pre-trained word vector model and inputting them into the model as initialization values; outputting data sets for each batch; building a TextRCNN model for text classification; and calculating performance indicators of the model.

[0163] The specific steps of training the model may include but are not limited to defining a computational graph.

[0164] In the case where the speaker corresponding to the first voice is a bank employee, the first voice is detected based on a rule template to determine whether the bank employee has any illegal speech behavior; the rule template is used to represent at least one of keywords, regular expressions, semantics, emotional rules and speech speed rules.

[0165] In actual implementation, key information of the first speech may be extracted based on a rule template for representing at least one of a keyword, a regular expression, a semantics, an emotion rule, and a speech speed rule. The key information may include a detection result of whether the first speech contains any illegal speech behavior of the bank employee.

[0166] Figure 4 Schematic diagram of the principle of detecting illegal speech behavior in the compliance identification method provided in the embodiment of the present application. Figure 4As shown, TextRCNN text classification is to classify the sentences in the first voice based on the TextRCNN model, determine whether it belongs to the first speech or the second speech, and can match the rule template based on the text classification. The rule template may include at least one of the template of the first rule and the template of the second rule. Semi-pointer relationship extraction is to extract relationships using a semi-pointer method. LSTM (Long Short-Term Memory) + CRF (Conditional Random Fields) named entity recognition is to identify named entities through the LSTM + CRF model; semantic matching is to perform semantic matching based on the extracted relationships and the identified named entities to detect illegal speech behaviors. Structured key information is to structure the key information obtained.

[0167] It should be noted that the business rules in the banking and financial field are very different from those in other fields. Therefore, it is necessary to customize and train the model for the data in the banking and financial field, so as to obtain a comprehensive bank employee illegal speech behavior detection model. The historical dialogue data of bank marketing can be used as training samples for the bank employee illegal speech behavior detection model.

[0168] First, the LSTM+CRF model can be used in sequence labeling tasks.

[0169] The bank employee's illegal speech behavior detection model is trained based on the LSTM+CRF model. Through deep learning of training samples, it can effectively identify the bank employees' illegal speech behaviors.

[0170] Furthermore, there are a large number of professional terms and terminology in the field of banking and finance, so in the embodiment of the present application, noun parts of speech can be selected as important features for training to reduce the number of model parameters and improve the training efficiency of the model.

[0171] In the process of training the bank employee's irregular speech behavior detection model, in addition to using basic feature engineering methods such as bag-of-words model or N-gram model, named entity recognition technology can also be introduced to effectively identify entity information such as professional terms and terminology in the conversation, so as to improve the accuracy of the bank employee's irregular speech behavior detection model.

[0172] Second, a large-scale unsupervised corpus can be used for pre-training to obtain the parameters of the language model, which are then used as initial parameters and fine-tuned through training. This training method can use the pre-trained model to capture language rules and context information, which can improve the training efficiency and accuracy of the model.

[0173] Third, the attention mechanism can be introduced after LSTM, and the Softmax function can be used to convert the attention score into attention weights so as to weight the elements in the input sequence. Considering the context information in a weighted manner can improve the accuracy of the model and other effects.

[0174] It should be noted that in actual implementation, training can be performed for specific tasks. The text classification task is to classify a given sentence into a specified category. The clustering task is to divide a given text set into several clusters so that the text similarity within the same cluster is high and the text similarity between different clusters is low. The recommendation task is to recommend texts related to the user's interests based on the given user and context information. By training for specific tasks separately, it can better adapt to the text characteristics and distribution in the banking and financial field.

[0175] It should be noted that the key information extracted based on the TextRCNN model can also include the aforementioned customer portrait labels. The above TextRCNN model can also extract key information from the bank employees' and customers' speeches. The above TextRCNN model can better capture local information in the text by introducing convolution operations based on RNN, thereby solving the situation where there are multiple relationships between entities.

[0176] In some optional implementations, the detection results of the bank employee illegal speech behavior detection model (including the LSTM+CRF model and the TextRCNN model therein) can be continuously evaluated and fed back to continuously improve and optimize the bank employee illegal speech behavior detection model to improve its accuracy and reliability.

[0177] According to the compliance identification method provided in the embodiment of the present application, text classification and key information extraction are performed based on the TextRCNN model, and based on at least one of keywords, regular expressions, semantics, sentiment rules and speech speed rules, it is detected whether there are any illegal speech behaviors of bank employees in the marketing process, and whether the marketing behavior is compliant is identified. Illegal speech behaviors such as illegal speech behaviors are discovered and corrected by outputting information. Illegal speech behaviors in marketing behaviors can be supervised and managed, the marketing behaviors of bank employees can be standardized, the marketing behaviors of banks can be managed more effectively and efficiently, the compliance level of bank outlets' operations can be improved, and the management capabilities of banks can be strengthened.

[0178] In some embodiments, the semantics of the first speech are compared with the semantics of the standard speech in the target stage, including: performing speech recognition processing on the first speech to obtain a first text.

[0179] In actual implementation, the first voice can be recognized by any speech recognition method (such as a speech recognition method based on template matching or a speech recognition method based on a hidden Markov model, etc.), and the speech can be converted into text to obtain a first text. The first text is the text corresponding to the first voice. The specific speech recognition method used is not specifically limited in the embodiments of the present application.

[0180] Based on the corpus in the field of banking and finance, the first text is corrected to obtain the second text.

[0181] In actual implementation, the corpus in the field of banking and finance may store a large number of terms and words in the field of banking and finance. The above-mentioned terms and words in the field of banking and finance may be obtained through the Internet or professional dictionaries.

[0182] In some optional implementations, the corpus in the banking and finance field may store terms and vocabulary commonly used in the banking and finance field.

[0183] The corpus in the field of banking and finance can be an unstructured database (such as HBase, etc.). The above-mentioned terms and vocabulary in the field of banking and finance can be stored in the corpus in the field of banking and finance in the form of text, which is convenient for searching and other processing.

[0184] According to the corpus in the banking and finance field, the "wrong words" in the text data converted from the voice data can be automatically corrected, that is, the first text can be automatically corrected to obtain the second text.

[0185] For example, based on the corpus in the banking and finance field, the word "bandwidth" in the first text may be corrected to "loan", or the word "heterogeneous" "card" in the first text may be corrected to "a type of card".

[0186] The corpus in the banking and financial field can also be updated based on the Internet, which can enrich the corpus in the banking and financial field to the greatest extent possible, update hot words in a timely manner, and improve the accuracy of error correction.

[0187] The second text and the text of the standard speech are input into the Sentence-BERT model to obtain the similarity between the semantics of the first speech and the semantics of the standard speech in the target stage.

[0188] In actual implementation, the trained Sentence-BERT (also known as SBERT) model can be used to compare the semantics of each sentence in the second text corresponding to the first speech with the semantics of each sentence in the standard speech of the target stage to which the first speech belongs, so as to determine whether the semantics of the first speech match the semantics of the standard speech of the target stage to which the first speech belongs, that is, whether the semantic similarity is greater than or equal to the first threshold.

[0189] The above-mentioned trained Sentence-BERT model can be obtained by training on the basis of a pre-trained model in the banking and finance field improved based on Sentence-BERT.

[0190] In the embodiment of the present application, a semantic-based SOP speech detection system is established, and the sentences in the marketing scenario are first row-vectorized. In actual implementation, the bank employee's speech is matched with the standard speech for similarity. The Sentence-BERT model can be used to pre-encode each sentence in the standard speech offline to obtain a sentence vector. In the SOP speech detection, it is only necessary to encode the bank employee's speech and the customer's speech, which can greatly improve the efficiency of similarity comparison.

[0191] Figure 5 Schematic diagram of the principle of the Sentence-BERT model in the compliance identification method provided in the embodiment of the present application. Figure 5 As shown in Figure 1, the sentences in the speech fragment can be input into the BERT of the left branch, and the sentences in the standard speech can be input into the BERT of the right branch. The representation vectors of these two sentences are calculated to obtain the corresponding CLS values, that is, Figure 5 u and v in ; then concatenate u, v and |uv| and classify them through the Softmax layer to obtain the semantic similarity of the above two sentences.

[0192] It should be noted that Figure 5 The pooling step in can select different pooling methods according to the actual situation. For example, the maximum pooling is adopted in the embodiment of the present application.

[0193] It can be understood that this step includes a semantic similarity calculation task. The semantic similarity calculation task refers to a task of calculating the semantic similarity between two sentences given two sentences.

[0194] It should be noted that in actual execution, training can be performed for specific tasks. For the semantic similarity calculation task in the embodiment of the present application, in the training phase of the Sentence-BERT model, data augmentation technology is used to randomly insert, delete, replace, replace synonyms, reverse and merge the original data collected and preprocessed in the early stage, and at least one of the multiple processing methods is used to generate new sentences, thereby improving the generalization and adaptability of the Sentence-BERT model.

[0195] In some optional implementations, the detection results of the Sentence-BERT model can be continuously evaluated and fed back to continuously improve and optimize the Sentence-BERT model to improve its accuracy and reliability.

[0196] According to the compliance identification method provided in the embodiment of the present application, based on the Sentence-BERT model, the semantics of the first speech are compared with the semantics of the standard speech of the target stage, which can more accurately determine whether the semantics of the first speech are similar to the semantics of the standard speech of the target stage, thereby detecting whether the marketing behavior of bank employees complies with the standard operating procedures, identifying whether the marketing behavior is compliant, discovering and correcting non-compliant behaviors that do not comply with the standard operating procedures by outputting information, and supervising and managing the implementation process of marketing behaviors, standardizing the marketing behaviors of bank employees, and managing the marketing behaviors of banks more effectively and efficiently. In addition, based on the fact that there are great differences between the terms in the banking and financial fields and the terms in other fields, automatic error correction through the corpus of the banking and financial fields can greatly improve the accuracy of speech recognition, thereby improving the accuracy of semantic analysis of speech data, and more accurately and effectively identifying the compliance of bank marketing behaviors and managing bank marketing.

[0197] In some embodiments, the standard operating procedure can be configured in a visual manner. The process of the standard operating procedure for each marketing scenario can be generated by configuring the personnel to drag and drop.

[0198] Figure 6 1 is a flow chart of the sub-steps of configuring step 120 and step 130 in the compliance identification method provided in the embodiment of the present application. Figure 6 As shown, the configuration process may include: step 610 , step 620 , step 630 , step 640 , step 650 , step 660 and step 670 .

[0199] Step 610: Write a BPMN file.

[0200] A Business Process Modeling Notation (BPMN) file may be written first to generate various nodes of the SOP speech detection process based on the BPMN file, wherein the various nodes may include a start node, an end node, a judgment node, and a detection node.

[0201] It can be understood that the SOP speech detection process includes step 120 and step 130.

[0202] Step 620: Create a process engine.

[0203] The process of creating a process engine will create multiple core tables, including SOP_GE_*, SOP_HI_*, and SOP_RE_*.

[0204] The core table SOP_GE_* can be used to store the basic information of the SOP speech detection process.

[0205] The core table SOP_HI_* can be used to save historical data and SOP detection speech instances.

[0206] The core table SOP_RE_* can be used to store detection rule information. A detection rule can be composed of one or more detection conditions. For example, the "Civility in Customer Service" detection rule can be configured to meet one of the two detection conditions: whether the bank employee uses prohibited words or whether the bank employee's speaking speed is not within the set range.

[0207] Step 630: Get the core class.

[0208] The core class SOPRepositoryService can be used to deploy and manage process resources, such as BPMN files and attachments.

[0209] Step 640, deployment process.

[0210] Based on the core class SOPRepositoryService, deploy the SOP process (referring to the various stages included in the SOP).

[0211] Step 650, start the process.

[0212] Create the SOPRuntimeService class to start the SOP speech detection runtime process.

[0213] Step 660: Create a task.

[0214] Create the SOPTaskService class to save the task information in the SOP speech detection process.

[0215] Step 670, complete the task.

[0216] Create the SOPHistoryService class to save the historical information of the SOP speech detection process.

[0217] Create the SOPManagerService class to save the process engine information.

[0218] According to the compliance identification method provided in the embodiment of the present application, the SOP speech detection process is configured through a visualization method, and the configuration of the SOP and the SOP speech detection process is more flexible and convenient, which can improve the operational capacity of bank branches.

[0219] In the related art, the recording timing is unreasonable. The sound pickup equipment installed in the bank branches is turned on every morning and turned off at night, and it is always recording. Therefore, there are a lot of non-human voices and invalid recordings in the recording data, which will cause a great burden on the subsequent data processing and consume a lot of software and hardware resources.

[0220] In some embodiments, obtaining the first voice of the banking marketing process includes: when at least two people are detected in the target area, or when abnormal sounds are detected in the target area, starting a recording function to obtain voice data of the banking marketing process; the voice data includes the first voice.

[0221] In actual implementation, the voice data collection in step 110 can adopt at least one of the human body sensing and abnormal sound recording function schemes. Through the recording function, all voice data of the bank marketing process including the first voice and the second voice can be obtained.

[0222] In the actual implementation of the human body induction wake-up sound pickup device solution, human body induction is added to the traditional sound pickup device. The human body sensor can analyze the temperature of the sensed target to identify the human body in the target and count the number of people. When there are people and the number of people is greater than or equal to 2, the recording function can be automatically started to wake up the hardware sound pickup device.

[0223] In the actual implementation of the abnormal sound recording function solution, an abnormal sound detection device is installed on the basis of the traditional sound pickup device. When at least one of the abnormal sounds such as screaming, crying, loud talking, continuous hitting and glass breaking occurs at the bank branch, the recording function can be automatically activated to wake up the hardware sound pickup device.

[0224] In some optional implementations, the recording function may be automatically turned off after a period of time has passed since the abnormal sound ended.

[0225] It is understandable that the sound pickup device is electrically connected to the compliance identification device. If the sound pickup device has a recording function, the sound pickup device is awakened and the recording function of the sound pickup device is turned on. After the sound pickup device starts recording, the compliance identification device can obtain the voice data of the bank marketing process obtained by the recording from the sound pickup device. If the sound pickup device does not have a recording function, the recording software installed on the compliance identification device can be turned on to record the sound signal collected by the sound pickup device to obtain the voice data of the bank marketing process.

[0226] It is understandable that the above-mentioned sound pickup device can be set in a target area. The target area can be a business handling area or a financial management room of a bank branch, etc., where marketing activities may occur.

[0227] In some optional implementations, after acquiring the voice data of the bank's marketing process, the voice data can also be automatically uploaded to the server to establish a data warehouse for later tracing.

[0228] In some optional implementations, different specifications of pickup devices can be used in different target areas, and the human body sensing or abnormal sound sensing recording function can be flexibly adopted.

[0229] According to the compliance identification method provided in the embodiment of the present application, by adopting at least one of the human body sensing and abnormal sound to start the recording function solution to obtain the voice data of the bank marketing process, the problem of unreasonable recording timing can be solved. Under the condition of collecting as much and complete voice data as possible, the utilization rate of the pickup device can be improved, and the software and hardware resources consumed by various algorithms in the later stage can be reduced.

[0230] The compliance identification method provided in the embodiment of the present application can be executed by a compliance identification device. In the embodiment of the present application, the compliance identification method is executed by a compliance identification device as an example to illustrate the compliance identification device provided in the embodiment of the present application.

[0231] An embodiment of the present application also provides a compliance identification device.

[0232] like Figure 7 As shown, the compliance identification device includes: an acquisition module 710 , a determination module 720 and an output module 730 .

[0233] An acquisition module 710 is used to acquire a first voice in a bank marketing process;

[0234] The determination module 720 is used to determine the target stage to which the first voice belongs based on the second voice in the bank marketing process and the customer portrait corresponding to the bank marketing process when the speaker corresponding to the first voice is a bank employee; the target stage is the stage in the standard operating procedure of the target marketing scenario; the target marketing scenario is the marketing scenario corresponding to the first voice; and the second voice data is the voice before the first voice in the bank marketing process;

[0235] The output module 730 is used to compare the semantics of the first speech with the semantics of the standard speech in the target stage, and output at least one of the warning information and the prompt information when the semantics of the two do not match.

[0236] According to the compliance identification device provided in the embodiment of the present application, by combining the standard operating procedures of the marketing scenario, it is determined whether the semantics of the bank employee's voice segment matches the semantics of the standard speech of the stage of the standard operating procedures of the marketing scenario to which the voice segment belongs, detects whether the marketing behavior of the bank employee complies with the standard operating procedures, identifies whether the marketing behavior is compliant, discovers and corrects non-compliant behaviors that do not comply with the standard operating procedures by outputting information, can supervise and manage the implementation process of the marketing behavior, standardize the marketing behavior of the bank employee, and can manage the marketing behavior of the bank more effectively and efficiently.

[0237] In some embodiments, the acquisition module 710 may also be used to acquire a second voice;

[0238] The device may further include:

[0239] A feature extraction module, used to perform text analysis on the text corresponding to the second voice, and obtain feature information of the customer among the speakers corresponding to the second voice;

[0240] The portrait module is used to obtain customer portraits based on feature information.

[0241] In some embodiments, the device may further include:

[0242] The detection module is used to detect the first voice based on at least one of keywords, regular expressions, semantics, emotional rules and speech speed rules when the speaker corresponding to the first voice is a bank employee, to determine whether the bank employee has any illegal speech behavior; if there is any illegal speech behavior, output at least one of an alarm message and a prompt message.

[0243] In some embodiments, the detection module may be specifically used to:

[0244] Performing text classification on the first speech based on the TextRCNN model to determine whether the speaker corresponding to the first speech is a bank employee;

[0245] In the case where the speaker corresponding to the first voice is a bank employee, the first voice is detected based on a rule template to determine whether the bank employee has any illegal speech behavior; the rule template is used to represent at least one of keywords, regular expressions, semantics, emotional rules and speech speed rules.

[0246] In some embodiments, the output module 730 may be specifically used to:

[0247] Performing speech recognition processing on the first speech to obtain a first text;

[0248] Based on the corpus in the field of banking and finance, the first text is corrected to obtain the second text;

[0249] The second text and the text of the standard speech are input into the Sentence-BERT model to obtain the similarity between the semantics of the first speech and the semantics of the standard speech in the target stage.

[0250] In some embodiments, the acquisition module 710 can be specifically used to start the recording function and acquire voice data of the bank marketing process when at least two people are detected in the target area, or when abnormal sounds are detected in the target area; the voice data includes the first voice.

[0251] The compliance identification device in the embodiment of the present application can be an electronic device or a component in the electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal or other devices other than a terminal. Exemplarily, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc. It can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., and the embodiment of the present application is not specifically limited.

[0252] The compliance identification device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0253] The compliance identification device provided in the embodiment of the present application can achieve Figures 1 to 6 To avoid repetition, the various processes implemented by the method embodiment are not described here.

[0254] In some embodiments, Figure 8 As shown, an embodiment of the present application also provides an electronic device 800, including a processor 801, a memory 802, and a computer program stored in the memory 802 and executable on the processor 801. When the program is executed by the processor 801, the various processes of the above-mentioned compliance identification method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0255] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0256] An embodiment of the present application also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the above-mentioned compliance identification method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0257] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0258] An embodiment of the present application also provides a computer program product, including a computer program, which implements the above-mentioned compliance identification method when executed by a processor.

[0259] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0260] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned compliance identification method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0261] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0262] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0263] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0264] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

[0265] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0266] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the claims and their equivalents.

Claims

1. A compliance identification method, characterized in that: include: Get the first voice in the bank's marketing process; In the case where the speaker corresponding to the first voice is a bank employee, determining the target stage to which the first voice belongs based on the second voice in the bank marketing process and the customer portrait corresponding to the bank marketing process; The target stage is a stage in the standard operating procedure of the target marketing scenario; the target marketing scenario is the marketing scenario corresponding to the first voice; the second voice is the voice before the first voice in the bank marketing process; Comparing the semantics of the first speech with the semantics of the standard speech of the target stage, and outputting at least one of a warning message and a prompt message when the semantics of the two do not match; The determining, based on the second voice in the bank marketing process and the customer portrait corresponding to the bank marketing process, the target stage to which the first voice belongs includes: Analyze the context semantics of multiple consecutive sentences in the second voice, and determine the marketing scenario corresponding to the first voice in combination with the customer portrait, as the target marketing scenario, and determine the stage in the standard operating procedure of the target marketing scenario to which the first voice belongs based on the standard operating procedure of the target marketing scenario, as the target stage; The step of determining, based on the standard operating procedure of the target marketing scenario, the stage in the standard operating procedure of the target marketing scenario to which the first voice belongs as the target stage, includes: When the marketing scenario corresponding to the first voice has been determined, the stage in the standard operating procedure of the target marketing scenario to which the first voice belongs is determined based on the semantics of the second voice and the standard operating procedure of the target marketing scenario.

2. The compliance identification method according to claim 1, characterized in that: Before acquiring the first voice in the banking marketing process, the method further includes: Acquire the second voice; Performing text analysis on the text corresponding to the second voice to obtain characteristic information of the customer among the speakers corresponding to the second voice; Based on the feature information, the customer portrait is obtained.

3. The compliance identification method according to claim 1, characterized in that: After acquiring the first voice in the bank marketing process, the method further includes: In the case where the speaker corresponding to the first voice is a bank employee, the first voice is detected based on at least one of keywords, regular expressions, semantics, emotional rules, and speech speed rules to determine whether the bank employee has any illegal speech behavior; In the event of any illegal speech behavior, at least one of a warning message and a prompt message is output.

4. The compliance identification method according to claim 3, characterized in that: In the case where the speaker corresponding to the first voice is a bank employee, the first voice is detected based on at least one of keywords, regular expressions, semantics, emotional rules, and speech speed rules to determine whether the bank employee has illegal speech behavior, including: Performing text classification on the first speech based on the TextRCNN model to determine whether the speaker corresponding to the first speech is a bank employee; In the case that the speaker corresponding to the first voice is a bank employee, the first voice is detected based on a rule template to determine whether the bank employee has any illegal speech behavior; the rule template is used to represent at least one of keywords, regular expressions, semantics, emotional rules and speech speed rules.

5. The compliance identification method according to claim 1, characterized in that: The comparing the semantics of the first speech with the semantics of the standard speech of the target stage includes: Performing speech recognition processing on the first speech to obtain a first text; Based on a corpus in the field of banking and finance, correct the first text to obtain a second text; The second text and the text of the standard speech are input into the Sentence-BERT model to obtain the similarity between the semantics of the first speech and the semantics of the standard speech of the target stage.

6. The compliance identification method according to any one of claims 1 to 5, characterized in that: The obtaining of the first voice in the banking marketing process includes: When at least two persons are detected in the target area, or when abnormal sounds are detected in the target area, the recording function is started to obtain voice data of the bank marketing process; the voice data includes the first voice.

7. A compliance identification device, characterized in that: include: An acquisition module, used to acquire the first voice in the bank marketing process; A determination module, configured to determine the target stage to which the first voice belongs based on the second voice in the bank marketing process and the customer portrait corresponding to the bank marketing process when the speaker corresponding to the first voice is a bank employee; The target stage is a stage in the standard operating procedure of the target marketing scenario; the target marketing scenario is the marketing scenario corresponding to the first voice; the second voice is the voice before the first voice in the bank marketing process; An output module, configured to compare the semantics of the first speech with the semantics of the standard speech of the target stage, and output at least one of a warning message and a prompt message when the semantics of the two do not match; The determination module is specifically configured to analyze the context semantics of a plurality of consecutive sentences in the second speech, and determine the marketing scenario corresponding to the first speech in combination with the customer portrait, as the target marketing scenario, and determine the stage in the standard operating procedure of the target marketing scenario to which the first speech belongs based on the standard operating procedure of the target marketing scenario, as the target stage; The determination module is specifically used to determine, when the marketing scenario corresponding to the first voice has been determined, the stage in the standard operating procedure of the target marketing scenario to which the first voice belongs based on the semantics of the second voice and the standard operating procedure of the target marketing scenario.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the compliance identification method as described in any one of claims 1-6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the compliance identification method as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the compliance identification method according to any one of claims 1 to 6 is implemented.

Images