Credit auditing method and device, equipment and medium

Abstract

The embodiment of the invention provides a credit auditing method and device, equipment and a medium, relates to the technical field of artificial intelligence, and is used for improving the determination efficiency and quality of a credit auditing result and improving the user experience. The method comprises the following steps: in response to an audit request for a to-be-audited object, performing multiple rounds of dialogue interaction with the to-be-audited object to generate a dialogue file, the dialogue file comprising a risk control decision key data set of the to-be-audited object; determining a first decision result of the to-be-audited object based on the dialogue file and the risk decision original data of the to-be-audited object; when the first decision result is that the to-be-audited object needs to be supplemented, generating and sending a supplementing notification to the to-be-audited object according to the object type of the to-be-audited object; and when a supplementary file uploaded by the to-be-audited object is received, analyzing the supplementary file, extracting risk control decision auxiliary data, and determining a second decision result of the to-be-audited object based on the risk control decision auxiliary data set, the dialogue file and the risk decision original data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and in particular to a credit auditing method and device, equipment and medium. BACKGROUND

[0002] Under the background of rapid development of Internet finance, in the process of handling credit business, the credit auditing link is an important task link for risk control.

[0003] At present, when the automatic risk control system cannot make a direct decision due to insufficient customer information or credit investigation doubts, the solution of manual auditing intervention is generally adopted. The specific process is as follows: (1) the auditor contacts the customer by manually dialing the phone; (2) the auditor manually records the key information of the customer (such as occupation, income, overdue reason, etc.) during the call, and makes a subjective judgment on the risk of the customer based on personal experience; (3) after the call, the auditor writes a report or enters the key information into the system, and another authorized person or another system makes the final decision.

[0004] However, the above-mentioned manual auditing method has the following problems: (1) low efficiency and high cost: completely relying on manual seats, long single call time, limited per capita processing capacity, and when the business volume is large, a large number of auditors need to be employed, and the labor cost and management cost are extremely high; (2) low standardization and large quality fluctuation: the auditing quality is highly dependent on the professional quality, experience and state of the auditor at that time, it is difficult to ensure the consistency of the service and the risk control standard, and different auditors may bring different decision results. SUMMARY The embodiments of the present application provide a credit auditing method, device, equipment and medium, to improve the determination efficiency and quality of the credit auditing result, and improve the user experience.

[0005] In a first aspect, the embodiments of the present application provide a credit auditing method, which comprises: In response to an auditing request for a to-be-audited object, a multi-round dialogue interaction is performed with the to-be-audited object, and a dialogue archive is generated, the dialogue archive comprising a risk control decision key data set of the to-be-audited object; Based on the dialogue archive and the risk decision original data of the to-be-audited object, a first decision result of the to-be-audited object is determined; When the first decision result is that the to-be-audited object needs to supplement, according to the object type of the to-be-audited object, a supplement notice is generated and sent to the to-be-audited object, the supplement notice comprising a material list required for supplement; When receiving the supplementary file uploaded by the to-be-audited object, the supplementary file is parsed, the risk control decision auxiliary data is extracted, and based on the risk control decision auxiliary data set, the dialogue archive and the risk decision original data, a second decision result of the to-be-audited object is determined, and the second decision result represents whether the to-be-audited object passes the audit.

[0006] In an optional embodiment, a multi-round dialogue interaction is performed with the to-be-audited object, and a dialogue archive is generated, including: A multi-round dialogue interaction is performed with the to-be-audited object, and in one round of dialogue interaction: a first interactive script fed back by the to-be-audited object is received, risk control decision key data of the to-be-audited object is extracted from the first interactive script, and based on the first interactive script, an emotional recognition result and a fraud risk recognition result of the to-be-audited object are determined, and based on the emotional recognition result, the fraud risk recognition result and the script generation basis data, a new second interactive script is generated and sent to the to-be-audited object.

[0007] In an optional embodiment, the risk control decision key data of the to-be-audited object is extracted from the first interactive script, including: From the first interactive script, a pre-defined key slot is extracted to obtain the risk control decision key data of the to-be-audited object, and the risk control decision key data of the to-be-audited object is stored in the dialogue archive.

[0008] In an optional embodiment, based on the first interactive script, the emotional recognition result and the fraud risk recognition result of the to-be-audited object are determined, including: The acoustic features and / or text content of the first interactive script are extracted, and based on the acoustic features and / or text content, the emotional recognition result of the to-be-audited object is determined; Based on the consistency between the first interactive script and the historical dialogue content and the risk decision original data, the fraud risk recognition result of the to-be-audited object is determined.

[0009] In an optional embodiment, the script generation basis data includes: historical dialogue content and a script template, the script template is used to inquire each risk control decision key data that needs to be filled in by the to-be-audited object, and the script template is generated based on the object type, the risk scenario of the to-be-audited object and the credit report of the to-be-audited object; Then, based on the emotional recognition result, the fraud risk recognition result and the script generation basis data, a new second interactive script is generated, including: According to the emotional recognition result, it is determined whether to generate a empathetic script, and the empathetic script is taken as the new second interactive script; According to the fraud risk recognition result, it is determined whether to generate a verification script, and the verification script is taken as the new second interactive script; According to the historical dialogue content and the dialogue template, a next to-be-asked risk control decision key data is determined, and an inquiry dialogue for the next to-be-asked risk control decision key data is generated, and the inquiry dialogue is taken as a new second interactive dialogue.

[0010] In an optional embodiment, after the first decision result of the to-be-audited object is determined, the method further includes: When the first decision result is an audit pass, a downstream system interface is called to trigger a loan process. When the first decision result is an audit fail, a rejection reason and a suggestion for the to-be-audited object are reported through voice.

[0011] In an optional embodiment, the supplement file is parsed to extract risk control decision auxiliary data, including: The supplement file is preprocessed. Text information in the supplement file is recognized, key information is extracted, and the key information is converted into structured data to obtain a risk control decision auxiliary data set.

[0012] In a second aspect, the embodiments of the present application also provide a credit audit device, and the device includes: A generation module is configured to respond to an audit request for a to-be-audited object, to perform multi-round dialogue interaction on the to-be-audited object, and to generate a dialogue archive, the dialogue archive including a risk control decision key data set of the to-be-audited object. A first decision module is configured to determine a first decision result of the to-be-audited object based on the dialogue archive and risk decision original data of the to-be-audited object. A notification module is configured to, when the first decision result is that the to-be-audited object needs a supplement, generate and send a supplement notification to the to-be-audited object according to an object type of the to-be-audited object, the supplement notification including a list of materials required for the supplement. A second decision module is configured to, when a supplement file uploaded by the to-be-audited object is received, parse the supplement file to extract risk control decision auxiliary data, and determine a second decision result of the to-be-audited object based on the risk control decision auxiliary data set, the dialogue archive, and the risk decision original data, the second decision result representing whether the to-be-audited object passes the audit.

[0013] In an optional embodiment, when the multi-round dialogue interaction with the to-be-audited object is performed to generate the dialogue archive, the generation module is further configured to: The multi-round dialogue interaction with the to-be-audited object is performed, and in one round of dialogue interaction: a first interactive dialogue fed back by the to-be-audited object is received, risk control decision key data of the to-be-audited object is extracted from the first interactive dialogue, an emotional recognition result and a fraud risk recognition result of the to-be-audited object are determined based on the first interactive dialogue, and a new second interactive dialogue is generated based on the emotional recognition result, the fraud risk recognition result, and dialogue generation basis data, and the new second interactive dialogue is sent to the to-be-audited object.

[0014] In an optional embodiment, when extracting the risk control decision key data of the to-be-audited object from the first interactive dialogue, the generation module is further configured to: extract the pre-defined key slots from the first interactive dialogue to obtain the risk control decision key data of the to-be-audited object, and store the risk control decision key data of the to-be-audited object into the dialogue archive.

[0015] In an optional embodiment, when determining the emotion recognition result and the fraud risk recognition result of the to-be-audited object based on the first interactive dialogue, the generation module is further configured to: extract the acoustic features and / or the text content of the first interactive dialogue, and determine the emotion recognition result of the to-be-audited object based on the acoustic features and / or the text content; determine the fraud risk recognition result of the to-be-audited object based on the consistency between the first interactive dialogue and the historical dialogue content and the risk decision original data.

[0016] In an optional embodiment, the dialogue generation basis data includes: the historical dialogue content and a dialogue template, the dialogue template is used to inquire each risk control decision key data that needs to be filled in by the to-be-audited object, and the dialogue template is generated based on the object type, the risk scenario of the to-be-audited object and the credit report of the to-be-audited object; Then, when generating the new second interactive dialogue based on the emotion recognition result, the fraud risk recognition result and the dialogue generation basis data, the generation module is further configured to: determine whether to generate a sympathetic dialogue according to the emotion recognition result, and take the sympathetic dialogue as the new second interactive dialogue; determine whether to generate a verification dialogue according to the fraud risk recognition result, and take the verification dialogue as the new second interactive dialogue; determine the next risk control decision key data to be inquired according to the historical dialogue content and the dialogue template, and generate an inquiry dialogue for the next risk control decision key data to be inquired, and take the inquiry dialogue as the new second interactive dialogue.

[0017] In an optional embodiment, after determining the first decision result of the to-be-audited object, the notification module is further configured to: when the first decision result is to pass the audit, call a downstream system interface to trigger a loan process; when the first decision result is to fail the audit, broadcast the rejection reason and the suggestion for the to-be-audited object through voice.

[0018] In an optional embodiment, when parsing the supplement file to extract the risk control decision auxiliary data, the second decision module is further configured to: pre-process the supplement file; Identify the text information in the supplement file, extract the key information, and convert the key information into structured data to obtain a risk control decision auxiliary data set.

[0019] In a third aspect, the embodiments of the present application further provide an electronic device, comprising: a processor; and a memory storing programs, wherein the programs include instructions that, when executed by the processor, cause the processor to perform the credit review method according to the first aspect.

[0020] In a fourth aspect, the embodiments of the present application further provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to perform the credit review method according to the first aspect.

[0021] In a fifth aspect, the present application provides a computer program product, which, when invoked by a computer, causes the computer to perform the credit review method steps according to the first aspect.

[0022] The beneficial effects of the present application are as follows: In the credit review method provided by the embodiments of the present application, first, in response to a review request for a to-be-reviewed object, a plurality of rounds of dialogue interaction are performed with the to-be-reviewed object to generate a dialogue archive, then based on the dialogue archive and risk decision original data of the to-be-reviewed object, a first decision result of the to-be-reviewed object is determined, when the first decision result is that the to-be-reviewed object needs to supplement, according to the object type of the to-be-reviewed object, a supplement notice is generated and sent to the to-be-reviewed object, when the supplement file uploaded by the to-be-reviewed object is received, the supplement file is parsed, risk control decision auxiliary data is extracted, and finally based on the risk control decision auxiliary data set, the dialogue archive and the risk decision original data, a second decision result of the to-be-reviewed object is determined. In this way, the traditional broken artificial review process is integrated into an automatic closed-loop process, which does not require human intervention, reduces the user waiting time, improves the processing efficiency, determines the decision based on the risk control decision auxiliary data set, the dialogue archive and the risk decision original data, and guarantees the quality and accuracy of the review result, thus effectively solving the core pain points of the existing artificial mode, such as low efficiency, poor experience and poor quality, improving the determination efficiency and quality of the credit review result, and improving the user experience.

[0023] In addition, other features and advantages of the present application will be described in the following description, and some will become apparent from the description, or will be understood by those skilled in the art. The purpose and other advantages of the present application can be achieved and obtained by the structure specifically pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments description. Obviously, the drawings described here are used to provide further understanding of the present application and form a part of the present application, but not constitute improper limitation to the present application. In the drawings: Figure 1 An optional system architecture schematic diagram applicable to the embodiments of the present application; Figure 2 An implementation flow schematic diagram of a credit auditing method provided by the embodiments of the present application; Figure 3 A logic schematic diagram of a credit auditing method provided by the embodiments of the present application; Figure 4 A structure schematic diagram of a credit auditing apparatus provided by the embodiments of the present application; Figure 5 A structure schematic diagram of an electronic device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0025] The embodiments of the present application will be described in detail with reference to the drawings. Although some embodiments of the present application are shown in the drawings, it should be understood that the present application can be implemented in various forms, and should not be interpreted as being limited to the embodiments set forth herein, rather these embodiments are provided to make the present application more thorough and complete. It should be understood that the drawings and embodiments of the present application are only for exemplary purposes, and are not intended to limit the protection scope of the present application.

[0026] It should be understood that each step recorded in the method embodiments of the present application can be executed in different order, and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the execution of the steps shown. The scope of the present application is not limited in this respect.

[0027] The term "comprising" and its variants used herein are open-ended, i.e., "including but not limited to". The term "based on" is "at least in part based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Related definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present application are only used to distinguish different devices, modules or units, and are not intended to limit the order or interdependence of the functions performed by these devices, modules or units.

[0028] It should be noted that the modification of "one" and "multiple" mentioned in the present application is illustrative but not restrictive, and those skilled in the art should understand that unless otherwise explicitly indicated in the context, it should be understood as "one or more".

[0029] The names of the messages or information exchanged between the devices in the embodiments of the present application are only for illustrative purposes, and are not intended to limit the scope of the messages or information.

[0030] The following explains some terms in the embodiments of the present application to facilitate understanding by those skilled in the art.

[0031] (1) Intelligent outbound call: a system that automatically initiates a telephone call by a computer program and interacts with the user through voice synthesis, speech recognition and other technologies.

[0032] (2) Multi-round dialogue: refers to the multi-round, context-related question and answer interaction between the machine and the user, rather than a single question and answer.

[0033] (3) Natural Language Processing (NLP): a subfield of artificial intelligence, aiming to enable computers to understand, interpret and generate human language.

[0034] (4) Slot information: pre-defined key information unit in the dialogue system that needs to be extracted from the user's speech. For example, occupation, income, etc.

[0035] (5) Emotion recognition model: an artificial intelligence model that can identify the emotional state (e.g. happy, angry, anxious, etc.) of the subject by analyzing the content of the text and / or acoustic features.

[0036] (6) Fraud identification model: a data mining or machine learning model that identifies fraud risks by analyzing the user's dialogue content, risk decision raw data, etc.

[0037] (7) Optical Character Recognition (OCR): a technology that converts text in images into machine-encodable text. In the present scheme, it is used to analyze the bank flow screenshot submitted by the user and other supplementary files.

[0038] Based on the above explanations of the above terms and related terms, the design idea of the embodiments of the present application is briefly introduced as follows: Under the background of rapid development of Internet finance, in the process of handling credit business, the credit audit link is an important task link to control risks.

[0039] Currently, when the automated risk control system cannot make a direct decision due to insufficient customer information or credit investigation doubts, the solution of manual review intervention is generally adopted. The specific process is as follows: (1) the reviewer contacts the customer by manually dialing the phone; (2) the reviewer manually records the key information of the customer (such as occupation, income, overdue reason, etc.) during the call and makes a subjective judgment on the risk of the customer based on personal experience; (3) after the call, the reviewer writes a report or enters the key information into the system, and another authorized person or another system makes the final decision; (4) for customers who need to supplement materials, the reviewer needs to separately notify the customer to supplement materials through SMS or email, and after the customer submits the supplementary materials, the reviewer needs to manually download, view and supplement the information (such as salary account amount, transaction flow) and enter it into the risk control system.

[0040] However, the above manual review method has the following problems: (1) low efficiency and high cost: completely relying on manual agents, long single call time, limited per capita processing capacity, and when the business volume is large, a large number of auditors need to be hired, the labor cost and management cost are extremely high; (2) low standardization and large quality fluctuation: the quality of the audit is highly dependent on the professional quality, experience and state of the reviewer at the time, it is difficult to ensure the consistency of the service and the risk control standard, and different reviewers' operations may lead to different decision results; (3) long decision-making cycle and poor customer experience: from reaching the customer to the final decision, the process links are multiple and broken (such as calls and approvals belonging to different links), the customer needs to wait for a long time, and the experience is not smooth; (4) insufficient information utilization, easy to produce risk blind spots: manual dialogue is difficult to combine external data such as credit reports in real time, dynamically adjust the inquiry strategy, and lack of tools to deeply analyze potential risk points such as customer emotions and semantic contradictions, which may miss fraud clues; (5) cumbersome operation, easy to make mistakes: manual recording of information, manual entry of flow data and other operations not only have low efficiency, but also have the possibility of transcription errors and operation risks.

[0041] In view of this, in the embodiments of the present application, a credit audit method, device, equipment and medium are provided, which can specifically include the following steps: first, in response to an audit request for a to-be-audited object, a multi-round dialogue interaction is performed with the to-be-audited object to generate a dialogue archive, the dialogue archive including a risk control decision key data set of the to-be-audited object, then based on the dialogue archive and risk decision original data of the to-be-audited object, a first decision result of the to-be-audited object is determined, when the first decision result is that the to-be-audited object needs to supplement, a supplement notice is generated and sent to the to-be-audited object according to the object type of the to-be-audited object, the supplement notice including a list of materials required for supplement, then when the supplement file uploaded by the to-be-audited object is received, the supplement file is parsed, risk control decision auxiliary data is extracted, and based on the risk control decision auxiliary data set, the dialogue archive and the risk decision original data, a second decision result of the to-be-audited object is determined, the second decision result representing whether the to-be-audited object passes the audit. In this way, the traditional broken artificial audit process (reach, inquire, decide, supplement) is integrated into an automatic closed-loop process, which does not require human intervention, reduces user waiting time, improves processing efficiency, determines the decision based on the risk control decision auxiliary data set, the dialogue archive and the risk decision original data, and guarantees the quality and accuracy of the audit result, thus effectively solving the core pain points of the existing artificial mode, such as low efficiency, poor experience and poor quality, improving the determination efficiency and quality of the credit audit result, and improving the user experience.

[0042] In particular, the preferred embodiments of the present application are described below in conjunction with the accompanying drawings of the specification, it should be understood that the preferred embodiments described herein are only used to illustrate and explain the present application, and are not used to limit the present application, and the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0043] Reference Figure 1 As shown in the figure, it is an optional system architecture schematic diagram applicable to the embodiments of the present application, which can include: terminal equipment (101a, 101b) and server 102. Terminal equipment (101a, 101b) and server 102 can exchange information through a communication network, wherein the communication network can adopt a communication mode including wireless communication mode and wired communication mode. Exemplarily, terminal equipment (101a, 101b) can access the network through cellular mobile communication technology and communicate with server 102. Wherein, the cellular mobile communication technology, such as, includes the fifth generation mobile communication (5th generation mobile networks, 5G) technology or the next generation mobile communication technology. Optionally, terminal equipment (101a, 101b) can access the network through short-range wireless communication mode and communicate with server 102. Wherein, the short-range wireless communication mode, such as, includes wireless fidelity (wireless fidelity, Wi-Fi) technology.

[0044] The number of communication devices involved in the system architecture is not limited in the embodiments of the present application. For example, the system architecture can include more terminal devices, or can include fewer terminal devices, or can include other network devices. As shown in the above description, only terminal devices (101a, 101b) and server 102 are described as examples, and the following briefly introduces each communication device and its respective function. Figure 1

[0045] The terminal device (101a, 101b) is a device that can provide voice and / or data connectivity to a user, and can be a device that supports wired and / or wireless connection modes.

[0046] For example, the terminal device (101a, 101b) can include, but is not limited to, a mobile phone, a tablet computer, a notebook computer, a palm computer, a mobile internet device (MID), a wearable device, a virtual reality (VR) device, an augmented reality (AR) device, a wireless terminal device in industrial control, a wireless terminal device in unmanned driving, a wireless terminal device in smart grid, a wireless terminal device in transportation safety, a wireless terminal device in smart city, or a wireless terminal device in smart home, etc.

[0047] In addition, the terminal device (101a, 101b) can be installed with a related client, which can be software, such as an application (APP), a browser, a short video software, etc., or a webpage, an applet, etc. It should be noted that the terminal device (101a, 101b) in the embodiments of the present application can be the above-mentioned client related to credit audit.

[0048] The server 102 can be a standalone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms, etc. basic cloud computing services.

[0049] ​It is worth mentioning that the execution subject of the credit audit method provided by the embodiments of the present application can be one or more electronic devices, and the present application does not limit this; wherein the electronic device can be a terminal device or a server, so when the execution subject includes multiple electronic devices and the multiple electronic devices include at least one terminal device and at least one server, the credit audit method provided by the embodiments of the present application can be executed by the terminal device and the server together. For the convenience of description, the credit audit method executed by the electronic device is taken as an example for description hereinafter.

[0050] The credit audit method provided by the exemplary embodiments of the present application will be described below in combination with the above-mentioned system architecture and in reference to the accompanying drawings. It should be noted that the above-mentioned system architecture is only shown for the convenience of understanding the spirit and principles of the present application, and the embodiments of the present application are not limited in this respect.

[0051] Referring to Figure 2 As shown in the figure, it is an implementation flow diagram of a credit audit method provided by the embodiments of the present application, and the specific implementation flow of the method is as follows: S20: In response to the audit request for the to-be-audited object, a multi-round dialogue interaction is performed with the to-be-audited object to generate a dialogue archive.

[0052] Wherein, the dialogue archive includes a set of key data for risk control decision of the to-be-audited object, and each key data for risk control decision in the set of key data for risk control decision is key data (such as work, income, overdue reason, etc.) for credit audit of the to-be-audited object. The to-be-audited object is a user who applies for credit, but the automated risk control system cannot make a direct decision on the to-be-audited object due to insufficient customer information or suspicious credit investigation.

[0053] In the embodiments of the present application, when the automated risk control system cannot make a direct decision on the to-be-audited object, an audit request for the to-be-audited object is triggered, and in response to the audit request for the to-be-audited object, a multi-round dialogue interaction is performed with the to-be-audited object by intelligent outbound call or online chat to generate a dialogue archive. Wherein, the intelligent outbound call is to initiate a telephone call to the to-be-audited object to realize voice call, to synthesize the generated voice text into an audio stream and send it to the to-be-audited object, and to receive the audio stream of the to-be-audited object. The online chat is to initiate an online text chat to the to-be-audited object, and to complete the multi-round dialogue interaction through the intelligent chat robot interface.

[0054] In the embodiments of the present application, when interacting with the to-be-audited object in multiple rounds of dialogue, in one round of dialogue interaction: a first interactive dialogue is received from the to-be-audited object, key data of the to-be-audited object for risk control decision is extracted from the first interactive dialogue, and based on the first interactive dialogue, an emotional recognition result and a fraud risk recognition result of the to-be-audited object are determined, and based on the emotional recognition result, the fraud risk recognition result and dialogue generation basic data, a new second interactive dialogue is generated and sent to the to-be-audited object.

[0055] In this way, through the emotional recognition result, the fraud risk recognition result and the dialogue generation basic data, the new second interactive dialogue is generated, and the best dialogue is generated, which guarantees the accuracy of information collection and improves the user experience.

[0056] Optionally, in the embodiments of the present application, when the key data of the to-be-audited object for risk control decision is extracted from the first interactive dialogue, the pre-defined key slot is extracted from the first interactive dialogue to obtain the key data of the to-be-audited object for risk control decision, and the key data of the to-be-audited object for risk control decision is stored in the dialogue archive.

[0057] In the embodiments of the present application, the semantic understanding of the first interactive dialogue is to extract the pre-defined key slot (such as [job: teacher], [monthly_income: 15000], [overdue_reason: forget to repay]) from the first interactive dialogue to obtain structured data, that is, the key data of the to-be-audited object for risk control decision, and store the key data of the to-be-audited object for risk control decision in the dialogue archive corresponding to the to-be-audited object. The dialogue archive is associated with the identifier of the to-be-audited object.

[0058] In this way, the pre-defined key slot ensures the standardization and consistency of the collected information, eliminates the subjective differences of manual recording, and provides machine-readable structured data for subsequent risk control decision, which is convenient for subsequent generation of decision results.

[0059] Optionally, in the embodiments of the present application, when the emotional recognition result of the to-be-audited object is determined based on the first interactive dialogue, the acoustic features and / or text content of the first interactive dialogue are extracted, and the emotional recognition result of the to-be-audited object is determined based on the acoustic features and / or text content.

[0060] The emotional recognition result is: positive, neutral, negative, or angry emotion.

[0061] In the embodiments of the present application, when the first interactive dialogue is text (the form of multi-round dialogue interaction is online chat), the text content of the first interactive dialogue is extracted, and the semantics of the text content is analyzed through NLP to determine the emotional recognition result of the to-be-audited object. When the first interactive dialogue is voice (the form of multi-round dialogue interaction is intelligent outbound call), the acoustic features and / or text content of the first interactive dialogue are extracted, and the tone, speech rate and energy of the acoustic features are analyzed, and / or the semantics of the text content is analyzed through NLP to determine the emotional recognition result of the to-be-audited object.

[0062] In this way, the acoustic features capture emotional clues at the voice level, the text content identifies risk signals at the semantic level, multi-dimensional cross verification improves the accuracy of emotional recognition, and reduces the risk of misjudgment.

[0063] Optionally, in the embodiments of the present application, when determining the fraud risk recognition result of the to-be-audited object based on the first interactive dialogue, the fraud risk recognition result of the to-be-audited object is determined based on the consistency between the first interactive dialogue and the historical dialogue content and the risk decision original data.

[0064] The risk decision original data includes: object original application information and a credit report of the to-be-audited object obtained from a third party.

[0065] In the embodiments of the present application, whether there is a logical contradiction between the object feedback dialogue in the historical dialogue content, the object original application information and the credit report is compared through a preset consistency verification rule or a machine learning model to obtain a fraud risk score, thereby obtaining the fraud risk recognition result.

[0066] For example, the object feedback dialogue in the historical dialogue content says "unemployed", and the first interactive dialogue says "monthly income of 30,000", so it is determined that the fraud risk recognition result of the to-be-audited object is that there is a fraud risk.

[0067] In addition, it is worth noting that the fraud risk recognition result of the to-be-audited object can also be determined comprehensively by identifying the emotional fluctuations of the to-be-audited object in combination with consistency and emotional fluctuations.

[0068] In this way, the answers of the object are analyzed in real time during the dialogue interaction process, and fraud features such as inconsistent information and contradictory semantics are detected, thereby realizing the instant recognition and prevention and control of fraud risks.

[0069] In the embodiments of the present application, the dialogue generation basis data includes: historical dialogue content and dialogue templates, the dialogue templates are used to inquire about each risk control decision key data that needs to be filled in by the to-be-audited object, and the dialogue templates are generated based on the object type, the risk scene of the to-be-audited object and the credit report of the to-be-audited object.

[0070] Optionally, in the embodiment of the present application, before the start of the multi-round dialogue interaction, based on the object type, the risk scenario of the to-be-audited object, and the credit report of the to-be-audited object, the key data of each risk control decision that needs to be filled in by the to-be-audited object is determined, and based on the key data of each risk control decision that needs to be filled in by the to-be-audited object, a dialogue template is generated.

[0071] Among them, the object type is divided into consumer users and small and micro users, and the risk scenario is divided into: multiple heads and overdue.

[0072] For example, if it is detected that there is an overdue record in the credit report, the key data of the risk control decision of "overdue reason" is automatically inserted, and the dialogue of asking "overdue reason" is added in the dialogue template.

[0073] In addition, it is worth noting that the dialogue template not only includes the dialogue of asking the key data of each risk control decision that needs to be filled in by the to-be-audited object, but also includes dialogue branches. For example, the current inquiry dialogue is "whether there is a job", and the dialogue branch includes "monthly salary". When the to-be-audited object replies "yes" to "whether there is a job", the to-be-audited object is asked "monthly salary".

[0074] Optionally, in the embodiment of the present application, based on the emotion recognition result, the fraud risk identification result and the dialogue generation basic data, a new second interaction dialogue is generated, and the following operations are specifically performed: S200: According to the emotion recognition result, it is determined whether to generate a sympathetic dialogue, and the sympathetic dialogue is taken as a new second interaction dialogue.

[0075] In the embodiment of the present application, after obtaining the emotion recognition result, it is judged whether to generate a sympathetic dialogue according to the emotion recognition result, and the sympathetic dialogue is taken as a new second interaction dialogue.

[0076] For example, after obtaining the emotion recognition result, if the emotion recognition result is a negative emotion, a sympathetic dialogue is generated as a new second interaction dialogue to appease the to-be-audited object.

[0077] In addition, it is worth noting that the emotion recognition result can not only be used as a basis for generating a new second interaction dialogue, but also be used as a basis for determining the tone of the new second interaction dialogue. For example, if the emotion recognition result is a positive emotion, the tone of the new second interaction dialogue is determined to be formal.

[0078] S201: According to the fraud risk identification result, it is determined whether to generate a verification dialogue, and the verification dialogue is taken as a new second interaction dialogue.

[0079] Among them, the verification dialogue is a dialogue for politely verifying whether the to-be-audited object is fraudulent.

[0080] For example, after obtaining the fraud risk identification result, if the fraud risk identification result is that there is a fraud risk, a verification dialogue is generated as a new second interaction dialogue to verify whether the to-be-audited object is fraudulent. For example, the verification dialogue is "You just said that your monthly income is 30,000 yuan. Previously, you said that you are unemployed. Is it because you have recently found a job?"

[0081] In addition, it is worth noting that if the verification dialogue is generated as a new second interaction dialogue according to the fraud risk identification result, and the empathetic dialogue is generated through the emotion identification result before the verification dialogue is generated as a new second interaction dialogue, the empathetic dialogue is used as a new second interaction dialogue, and the fraud risk identification result of this round is retained. In the next round, if no empathetic dialogue is generated, the fraud risk identification result is used to determine whether to generate a verification dialogue, and the verification dialogue is used as a new second interaction dialogue.

[0082] S202: According to the historical dialogue content and the dialogue template, determine the next to-be-asked risk control decision key data, and generate an inquiry dialogue for the next to-be-asked risk control decision key data, and use the inquiry dialogue as a new second interaction dialogue.

[0083] For example, assuming that the dialogue template includes the risk control decision key data to be asked, such as "work", "income", and "overdue reason", and the "work" has been asked and answered in the historical dialogue content, then "income" is the next to-be-asked risk control decision key data, and an inquiry dialogue for the "income" is generated, and the inquiry dialogue is used as a new second interaction dialogue.

[0084] In the embodiments of the present application, the new second interaction dialogue is generated according to the priority of the emotion identification result, the fraud risk identification result, and the dialogue generation basis data. The priority of the emotion identification result is higher than the priority of the fraud risk identification result, and the priority of the fraud risk identification result is higher than the priority of the dialogue generation basis data. The steps of generating the new second interaction dialogue according to the priority are as follows: according to the emotion identification result, it is determined whether to generate an empathetic dialogue, if it is determined to generate an empathetic dialogue, the empathetic dialogue is used as a new second interaction dialogue; if it is determined not to generate an empathetic dialogue, according to the fraud risk identification result, it is determined whether to generate a verification dialogue, if it is determined to generate a verification dialogue, the verification dialogue is used as a new second interaction dialogue; if it is determined not to generate a verification dialogue, according to the historical dialogue content and the dialogue template, the next to-be-asked risk control decision key data is determined, and an inquiry dialogue for the next to-be-asked risk control decision key data is generated, and the inquiry dialogue is used as a new second interaction dialogue.

[0085] S21: Based on the dialogue file and the risk decision original data of the to-be-audited object, a first decision result of the to-be-audited object is determined.

[0086] The first decision result represents whether the to-be-audited object is directly passed through the audit, and the second decision result is: audit pass (representing passing the credit application of the to-be-audited object), audit fail (representing rejecting the credit application of the to-be-audited object), and the to-be-audited object needs to supplement.

[0087] In the embodiment of the application, when the dialogue archive meets the decision determination condition, the dialogue archive and the risk decision original data of the to-be-audited object are integrated to obtain first decision data, and the first decision data is evaluated through a preset risk control rule set or a machine learning model to obtain the first decision result of the to-be-audited object.

[0088] Optionally, when the risk control decision key data of the to-be-audited object has been successfully extracted in full, it is determined that the dialogue archive meets the decision determination condition, the interaction is ended, and the final first decision result is determined.

[0089] Optionally, when there is new data in the dialogue archive, it is determined that the dialogue archive meets the decision determination condition. That is, after each round of dialogue interaction, the first decision result of the current round is determined based on the dialogue archive in the current state and the risk decision original data of the to-be-audited object. When the first decision result of the current round is audit fail, the interaction is ended and the final first decision result is determined to be fail. When the first decision result of the current round is not audit fail, the next round of dialogue interaction is performed, and the interaction is ended when the first decision result of the next round is fail or the risk control decision key data of the to-be-audited object has been successfully extracted in full.

[0090] Further, in the embodiment of the application, when the final first decision result is audit pass, a downstream system interface is called to trigger a loan process. In this way, the downstream interface is automatically called to realize instant processing of audit pass, shorten the business flow time, and avoid operation risks and efficiency bottlenecks introduced by manual intervention. When the final first decision result is audit fail, the rejection reason and suggestion for the to-be-audited object are broadcasted through voice, for example: “It is suggested that you maintain a good credit record”. The personalized rejection reason and suggestion improve user experience rather than abrupt hang up.

[0091] S22: When the first decision result is that the to-be-audited object needs to supplement, a supplement notice is generated and sent to the to-be-audited object according to the object type of the to-be-audited object.

[0092] The supplement notice includes a list of materials required for supplement, such as WeChat flow, bank flow, etc. The supplement notice also includes a supplement upload path, such as a supplement link.

[0093] In the embodiment of the present application, when the first decision result is that the to-be-audited object needs to supplement, because different object types need different supplements, according to the object type of the to-be-audited object, the material list needed by the to-be-audited object is determined, and a unique supplement uploading path is generated, thereby generating a supplement notice and sending the supplement notice to the to-be-audited object. After receiving the supplement notice, the to-be-audited object can upload the supplement file through the supplement uploading path.

[0094] The supplement notice sending mode can be voice broadcast or short message notification mode.

[0095] S23: When receiving the supplement file uploaded by the to-be-audited object, the supplement file is parsed, the risk control decision auxiliary data is extracted, and the second decision result of the to-be-audited object is determined based on the risk control decision auxiliary data set, the dialogue archive and the risk decision original data.

[0096] The second decision result represents whether the to-be-audited object passes the audit. The second decision result is that the audit is passed (representing passing the credit application of the to-be-audited object) or the audit is not passed (representing rejecting the credit application of the to-be-audited object). The second decision result is an irreversible decision result.

[0097] In the embodiment of the present application, when receiving the supplement file uploaded by the to-be-audited object, the supplement file is parsed, the risk control decision auxiliary data is extracted, the risk control decision auxiliary data set, the dialogue archive and the risk decision original data are integrated into second decision data, the second decision data is evaluated through a preset risk control rule set or a machine learning model, the second decision result of the to-be-audited object is obtained, and finally the second decision result is synchronized to the core business system, thereby completing the entire credit audit process.

[0098] Optionally, in the embodiment of the present application, in order to parse the supplement file and extract the risk control decision auxiliary data, a possible embodiment is provided, which specifically performs the following operations: S230: Preprocessing the supplement file.

[0099] The supplement file is in the form of a picture file or a PDF file, and the preprocessing includes but is not limited to image correction, denoising and the like.

[0100] S231: Recognizing the text information in the supplement file, extracting the key information, and converting the key information into structured data to obtain the risk control decision auxiliary data set.

[0101] In the embodiment of the present application, the OCR technology is used to recognize the text information in the supplement file, the NLP or rule template is used to extract the key information (such as extracting “transaction amount”, “transaction object” and “balance” from a bank flow screenshot), and the key information is converted into structured data to obtain the risk control decision auxiliary data set.

[0102] Referring to Figure 3 As shown in FIG. 1, it is a logic diagram of a credit audit method provided in the embodiments of the present application. First, the intelligent outbound call reaches the to-be-audited object, then the multi-round dialogue interaction is carried out by using the dialogue core processing layer, the real-time decision is made at the end of the interaction, and the first decision result is determined; when the first decision result is pass, the downstream system interface is called to trigger the loan process; when the first decision result is not pass, the reason and the suggestion are informed; when the first decision result is that the to-be-audited object needs to supplement, the supplement notice is generated and sent to the to-be-audited object according to the object type of the to-be-audited object, when the supplement file uploaded by the to-be-audited object is received, the supplement file is parsed, the risk control decision auxiliary data is extracted, and the second decision result of the to-be-audited object is determined based on the risk control decision auxiliary data set, the dialogue archive and the risk decision original data. In one round of dialogue interaction, the dialogue core processing layer: receives the first interaction script fed back by the to-be-audited object, extracts the risk control decision key data of the to-be-audited object from the first interaction script, and determines the emotion recognition result and the fraud risk recognition result of the to-be-audited object based on the first interaction script, and generates a new second interaction script based on the emotion recognition result, the fraud risk recognition result and the script generation basis data, and sends it to the to-be-audited object.

[0103] In this way, a highly intelligent interactive risk control hub is constructed. First, the user is reached by intelligent outbound call, and in the multi-round dialogue interaction, instead of mechanical questioning, personalized scripts are dynamically generated based on object type, credit report, emotion recognition result and fraud risk recognition result. The NLP model is used to accurately extract the key slots (risk control decision key data) in the user's answer, and the emotion recognition model is used to optimize the interaction experience, and the fraud identification model is called to identify fraud risks in real time during the dialogue process. After the information collection is completed, the system makes a decision to pass, not to pass or to supplement in real time, and directly informs the user of the result and the subsequent operation path in the current call. For the supplement user, the system automatically parses the proof materials submitted subsequently, and fuses the extracted information with the call data to automatically complete the final decision, so as to realize the full-process, automatic closed-loop processing from reaching to the final decision.

[0104] Further, based on the same technical concept, the embodiments of the present application provide a credit audit device for implementing the above method process of the embodiments of the present application. For example, referring to Figure 4 As shown in FIG. 4, the credit audit device 400 can include a generation module 401, a first decision module 402, a notification module 403 and a second decision module 404, wherein: The generating module 401 is configured to, in response to an audit request for an object to be audited, perform multi-round dialogue interaction with the object to be audited, and generate a dialogue archive, the dialogue archive including a risk control decision key data set of the object to be audited. The first decision module 402 is configured to determine a first decision result of the object to be audited based on the dialogue archive and risk decision original data of the object to be audited. The notification module 403 is configured to, when the first decision result is that the object to be audited needs to supplement, generate and send a supplement notice to the object to be audited according to an object type of the object to be audited, the supplement notice including a material list required for supplement. The second decision module 404 is configured to, when receiving a supplement file uploaded by the object to be audited, parse the supplement file, extract risk control decision auxiliary data, and determine a second decision result of the object to be audited based on the risk control decision auxiliary data set, the dialogue archive and the risk decision original data, the second decision result representing whether the object to be audited passes the audit.

[0105] In an optional embodiment, when the multi-round dialogue interaction with the object to be audited is performed and the dialogue archive is generated, the generating module 401 is further configured to: In one round of dialogue interaction, the multi-round dialogue interaction with the object to be audited is performed, the first interaction script fed back by the object to be audited is received, the risk control decision key data of the object to be audited is extracted from the first interaction script, the emotion recognition result and the fraud risk recognition result of the object to be audited are determined based on the first interaction script, and the new second interaction script is generated and sent to the object to be audited based on the emotion recognition result, the fraud risk recognition result and the script generation basis data.

[0106] In an optional embodiment, when the risk control decision key data of the object to be audited is extracted from the first interaction script, the generating module 401 is further configured to: The pre-defined key slot is extracted from the first interaction script to obtain the risk control decision key data of the object to be audited, and the risk control decision key data of the object to be audited is stored in the dialogue archive.

[0107] In an optional embodiment, when the emotion recognition result and the fraud risk recognition result of the object to be audited are determined based on the first interaction script, the generating module 401 is further configured to: The acoustic features and / or text content of the first interaction script are extracted, and the emotion recognition result of the object to be audited is determined based on the acoustic features and / or text content. The fraud risk recognition result of the object to be audited is determined based on the consistency between the first interaction script and the historical dialogue content and the risk decision original data.

[0108] In an optional embodiment, the dialogue generation basis data comprises historical dialogue content and dialogue templates, the dialogue templates are used to inquire about each risk control decision key data to be filled in by the to-be-audited object, and the dialogue templates are generated based on the object type, the risk scenario of the to-be-audited object, and the credit report of the to-be-audited object. Then, based on the emotion recognition result, the fraud risk recognition result, and the dialogue generation basis data, when generating the new second interactive dialogue, the generation module 401 is further configured to: determine whether to generate a sympathetic dialogue based on the emotion recognition result, and use the sympathetic dialogue as the new second interactive dialogue; determine whether to generate a verification dialogue based on the fraud risk recognition result, and use the verification dialogue as the new second interactive dialogue; determine the next to-be-inquired risk control decision key data based on the historical dialogue content and the dialogue templates, and generate an inquiry dialogue for the next to-be-inquired risk control decision key data, and use the inquiry dialogue as the new second interactive dialogue.

[0109] In an optional embodiment, after determining the first decision result of the to-be-audited object, the notification module 403 is further configured to: when the first decision result is to pass the audit, call a downstream system interface to trigger a loan process; when the first decision result is to fail the audit, broadcast a rejection reason and a suggestion for the to-be-audited object through voice.

[0110] In an optional embodiment, when the supplementary file is parsed to extract the risk control decision auxiliary data, the second decision module 404 is further configured to: preprocess the supplementary file; recognize the text information in the supplementary file, extract key information, and convert the key information into structured data to obtain a risk control decision auxiliary data set.

[0111] Based on the description of the above method embodiments and device embodiments, the exemplary embodiments of the present application further provide an electronic device, comprising at least one processor, and a memory communicatively connected to the at least one processor. The memory stores a computer program capable of being executed by the at least one processor, and the computer program, when executed by the at least one processor, is configured to cause the electronic device to perform the method according to the embodiments of the present application.

[0112] The embodiments of the present application further provide a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor of a computer, is configured to cause the computer to perform the method according to the embodiments of the present application.

[0113] The embodiment of the present application further provides a computer program product comprising a computer program, wherein the computer program is used for causing a computer to execute the method according to the embodiment of the present application when the computer program is executed by a processor of the computer.

[0114] Referring now to the Figure 5 Referring now to the The electronic device is intended to represent a variety of digital electronic computer devices, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computer devices. The electronic device can also represent a variety of mobile devices, such as personal digital assistants, cellular telephones, smart phones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not meant to limit implementations of the present application described and / or claimed in this document.

[0115] As shown in Figure 5 The electronic device 500 includes a computing unit 501 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. Various programs and data required for the operation of the device 500 can also be stored in the RAM 503. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0116] A plurality of components in the electronic device 500 are connected to the I / O interface 505, including an input unit 506, an output unit 507, a storage unit 508, and a communication unit 509. The input unit 506 can be any type of device that can input information to the electronic device 500, and can receive inputted digital or character information, and generate key signal inputs related to user settings and / or function controls of the electronic device. The output unit 507 can be any type of device that can present information, and can include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 508 can include, but is not limited to, a magnetic disk, an optical disk. The communication unit 509 allows the electronic device 500 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks, and can include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a Bluetooth device, a WiFi device, a worldwide interoperability for microwave access (WiMax) device, a cellular communication device, and / or the like.

[0117] The computing unit 501 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 501 performs various methods and processes described above. For example, in some embodiments, the credit review method described above can be implemented as a computer software program, which is tangibly embodied in a machine-readable medium, such as the storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 500 via the ROM 502 and / or the communication unit 509. In some embodiments, the computing unit 501 can be configured to perform the credit review method described above by any other appropriate means, such as by means of firmware.

[0118] Program code for carrying out methods of the present application can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the program code, when executed by the processor or controller, causes the functions / acts specified in the flowcharts and / or block diagrams to be implemented. The program code can be entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine or entirely on a remote machine or server.

[0119] In the context of the present application, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include one or more lines of electrical conductors, a portable computer disk, a hard disk, RAM, ROM, erasable programmable read-only memory (EPROM) or flash memory, an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0120] As used in the present application, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, apparatus and / or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0121] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a cathode ray tube (CRT) or a liquid crystal display (LCD) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0122] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0123] The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.

[0124] And, it should be understood that all the disclosed herein are only preferred embodiments of the application, and cannot be used to limit the scope of the application, therefore, any equivalent changes made on the basis of the claims of the application, still belong to the scope of the application.

Claims

1. A credit review method, characterized in that, include: In response to an audit request for an object to be audited, the system engages in multiple rounds of dialogue with the object to be audited to generate a dialogue file, which includes: the risk control decision key dataset of the object to be audited. Based on the dialogue file and the original risk decision data of the subject to be reviewed, the first decision result of the subject to be reviewed is determined; When the first decision result indicates that the object to be reviewed needs to provide supplementary documents, a supplementary document notification is generated and sent to the object to be reviewed based on the object type of the object to be reviewed. The supplementary document notification includes: a list of materials required for supplementary documents. When the supplementary documents uploaded by the object to be reviewed are received, the supplementary documents are parsed, risk control decision support data is extracted, and based on the risk control decision support dataset, the dialogue file and the original risk decision data, the second decision result of the object to be reviewed is determined. The second decision result indicates whether the object to be reviewed passes the review.

2. The method as described in claim 1, characterized in that, The process of engaging in multiple rounds of dialogue with the object to be reviewed and generating a dialogue profile includes: The system engages in multiple rounds of dialogue with the subject under review. In each round of dialogue, the system receives a first interactive script from the subject under review, extracts key risk control decision data of the subject under review from the first interactive script, determines the emotion recognition result and fraud risk recognition result of the subject under review based on the first interactive script, and generates a new second interactive script based on the emotion recognition result, the fraud risk recognition result and the script generation base data, and sends it to the subject under review.

3. The method as described in claim 2, characterized in that, The step of extracting key risk control decision data of the subject to be reviewed from the first interactive dialogue includes: From the first interactive dialogue, predefined key slots are extracted to obtain the risk control decision key data of the object to be reviewed, and the risk control decision key data of the object to be reviewed is stored in the dialogue file.

4. The method as described in claim 2, characterized in that, The step of determining the emotion recognition result and fraud risk recognition result of the subject to be reviewed based on the first interactive dialogue includes: Extract the acoustic features and / or text content of the first interactive dialogue, and determine the emotion recognition result of the subject to be reviewed based on the acoustic features and / or the text content; Based on the consistency between the first interactive dialogue and the historical dialogue content and the original risk decision data, the fraud risk identification result of the object to be reviewed is determined.

5. The method as described in claim 2, characterized in that, The basic data for generating the script includes: historical dialogue content and script templates. The script templates are used to inquire about the key risk control decision data that the subject to be reviewed needs to fill in. The script templates are generated based on the subject type, the risk scenario of the subject to be reviewed, and the credit report of the subject to be reviewed. Then, based on the emotion recognition result, the fraud risk recognition result, and the basic data for script generation, a new second interactive script is generated, including: Based on the emotion recognition results, determine whether to generate an empathic message and use the empathic message as a new second interactive message; Based on the fraud risk identification results, determine whether to generate a verification script, and use the verification script as a new second interactive script; Based on the historical dialogue content and the script template, determine the next key data for risk control decision to be asked, and generate an inquiry script for the next key data for risk control decision to be asked, and use the inquiry script as a new second interaction script.

6. The method as described in claim 1, characterized in that, After determining the first decision result of the object to be reviewed, the method further includes: When the first decision result is approval, the downstream system interface is called to trigger the loan disbursement process; When the first decision result is that the review is not approved, the reasons for rejection and suggestions for the object to be reviewed are announced via voice.

7. The method as described in claim 1, characterized in that, The process of parsing the supplementary document and extracting risk control decision support data includes: The replacement file is preprocessed; The text information in the supplementary document is identified, key information is extracted, and the key information is converted into structured data to obtain the risk control decision support dataset.

8. A credit review device, characterized in that, include: The generation module is used to respond to the audit request for the auditee, conduct multiple rounds of dialogue interaction with the auditee, and generate a dialogue file, which includes: the risk control decision key dataset of the auditee; The first decision module is used to determine the first decision result of the object to be reviewed based on the dialogue file and the original risk decision data of the object to be reviewed; The notification module is used to generate and send a supplementary document notification to the object to be reviewed based on the object type of the object to be reviewed when the first decision result indicates that the object to be reviewed needs to provide supplementary documents. The supplementary document notification includes a list of materials required for the supplementary documents. The second decision module is used to parse the supplementary document file uploaded by the object to be reviewed when it receives the supplementary document file, extract the risk control decision auxiliary data, and determine the second decision result of the object to be reviewed based on the risk control decision auxiliary dataset, the dialogue file and the original risk decision data. The second decision result indicates whether the object to be reviewed passes the review.

9. An electronic device, comprising: processor; as well as Stored program memory, The program includes instructions that, when executed by the processor, cause the processor to perform the method as described in any one of claims 1-7.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-7.

Images