Voice content compliance intelligent judgment and risk grading system based on AI big model

Abstract

The present invention provides a voice content compliance intelligent judgment and risk grading system based on AI big model, including: a voice acquisition module, a voice recognition module, a risk assessment module and a reminder module; wherein the voice acquisition module acquires the external voice from the audio acquisition device arranged in the call terminal; the voice recognition module inputs the external voice into the AI ​​big model; the risk assessment module determines the risk level according to the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information. The voice content compliance intelligent judgment and risk grading system based on AI big model of the present invention monitors the voice of the external call to ensure the accuracy and compliance of the salesperson's speech.

Description

Technical Field

[0001] The present invention relates to the technical field of language recognition supervision systems, and in particular to a voice content compliance intelligent judgment and risk grading system based on an AI big model. Background Art

[0002] There are many external services in the financial securities industry, such as external call services, WeChat promotion, and WeChat communication with customers. These external services are the main channels for financial securities companies to contact customers. They represent the corporate image. The accuracy and compliance of the salesperson's speech can better serve and communicate with customers. Therefore, there is an urgent need for a regulatory system to monitor the voice of external call services to ensure the accuracy and compliance of the salesperson's speech. Summary of the invention

[0003] One of the purposes of the present invention is to provide a voice content compliance intelligent judgment and risk grading system based on an AI big model to monitor the voice of external calls to ensure the accuracy and compliance of the salesperson's speech.

[0004] An embodiment of the present invention provides a voice content compliance intelligent judgment and risk grading system based on an AI big model, comprising: a voice acquisition module, a voice recognition module, a risk assessment module and a reminder module;

[0005] Among them, the voice acquisition module obtains external voice from the audio collection device set in the call terminal; the voice recognition module inputs the external voice into the AI ​​big model; the risk assessment module determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information.

[0006] Preferably, the risk assessment steps of the risk assessment module are as follows:

[0007] Based on the pre-configured first keyword library, keyword extraction is performed on the output data of the AI ​​large model.

[0008] Based on the results of the keyword extraction operation, a risk assessment vector set is constructed;

[0009] Based on the preset quantitative library, determine the quantitative value corresponding to each data in the risk assessment vector set;

[0010] The risk value is determined based on the quantitative value corresponding to each data and the weight coefficient of the corresponding position of each data.

[0011] Preferably, the risk assessment steps of the risk assessment module are as follows:

[0012] Based on the pre-configured second keyword library corresponding to each risk level, keyword extraction operation is performed on the output data of the AI ​​large model;

[0013] The risk level of the external voice is determined according to the highest risk level among the risk levels corresponding to the second keyword library from which the keywords are extracted.

[0014] Preferably, the voice content compliance intelligent judgment and risk grading system based on the AI ​​big model also includes: a voice retrieval module;

[0015] Among them, the voice retrieval module retrieves the voice files in the call records of the call system to obtain historical voice; the voice recognition module inputs the historical voice into the AI ​​big model; the risk assessment module determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information.

[0016] Preferably, the voice content compliance intelligent judgment and risk grading system based on the AI ​​big model further includes: a voice separation module;

[0017] Among them, the voice separation module is used to separate the voice files in the call records of the Internet instant messaging software to obtain separated voice; the voice recognition module inputs the separated voice into the AI ​​big model; the risk assessment module determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information.

[0018] Preferably, the voice content compliance intelligent judgment and risk grading system based on the AI ​​big model further includes: an association grouping module and an association analysis module;

[0019] Among them, the association grouping module is used to associate and group the output data of the separated voice after passing through the AI ​​large model with the text data in the call record; the association analysis module is used to analyze the associated grouped data according to the pre-configured association analysis library to obtain the association analysis result; when the association analysis result is abnormal, the reminder module outputs abnormal reminder information.

[0020] Preferably, the text data includes: text input by a person into the instant messaging software and data obtained by interpreting expressions using an expression interpretation library.

[0021] Preferably, the association grouping rules of the association grouping module include:

[0022] Perform group association based on the time difference between the time corresponding to the separated voice and the time corresponding to the text data;

[0023] and / or,

[0024] Perform group association based on the pre-configured association statement library.

[0025] Preferably, the association analysis step of the association analysis module includes:

[0026] According to the associated data query sub-library in the associated analysis library, the standard associated data corresponding to each data in the associated group is determined;

[0027] Matching the standard associated data with the associated data corresponding to each data in the associated group to determine the degree of match;

[0028] Calculate the association value of the association group based on the matching degree between the standard association data and the association data;

[0029] When the correlation value is less than or equal to the preset abnormal threshold, it is determined that the correlation analysis result is abnormal.

[0030] Preferably, the voice content compliance intelligent judgment and risk grading system based on the AI ​​big model also includes: a coherent association module and a compliance analysis module.

[0031] Among them, the coherent association module associates the voices of different voice speakers according to the voice receiving object and the communication time to form coherent association data; the compliance analysis module analyzes the coherent association data to determine whether it meets the pre-configured compliance requirements; when it does not meet the requirements, the reminder module sends a reminder message.

[0032] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0033] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0035] Figure 1 It is a schematic diagram of a voice content compliance intelligent judgment and risk grading system based on an AI big model in an embodiment of the present invention;

[0036] Figure 2 This is a schematic diagram of another voice content compliance intelligent judgment and risk grading system based on the AI ​​big model in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0038] The embodiment of the present invention provides a voice content compliance intelligent judgment and risk grading system based on an AI big model, such as Figure 1 As shown, it includes: a voice acquisition module 1, a voice recognition module 2, a risk assessment module 3 and a reminder module 4;

[0039] Among them, the voice acquisition module 1 obtains the external voice from the audio collection device set in the call terminal; the voice recognition module 2 inputs the external voice into the AI ​​big model; the risk assessment module 3 determines the risk level according to the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module 4 outputs abnormal reminder information.

[0040] Among them, the risk assessment steps of the risk assessment module are as follows:

[0041] According to the pre-configured first keyword library, a keyword extraction operation is performed on the output data of the AI ​​big model, and the parameter name in the output data of the AI ​​big model that needs to be extracted is pre-configured in the first keyword library (the row and column in the output data set of the AI ​​big model can be used as the parameter name), and the parameter value corresponding to the parameter name is extracted;

[0042] Based on the results of the keyword extraction operation, a risk assessment vector set is constructed; parameter values ​​are arranged in a predetermined order to form a risk assessment vector set;

[0043] Based on the preset quantization library, determine the quantization value corresponding to each data in the risk assessment vector set; the specific operation is to match the risk assessment vector set with the call data set corresponding to each quantization template in the quantization library, call the corresponding quantization template according to the matching result, and then determine the quantization value corresponding to each data in the risk assessment vector set according to the quantization template; the data in the quantization template is correspondingly associated with the quantization value, and the quantization value corresponding to the data can be determined by searching;

[0044] The risk value is determined based on the quantization value corresponding to each data and the weight coefficient of the corresponding position of each data. The weight coefficient is pre-configured and stored in the quantization template, and is associated with each data; the risk value is the sum of the products of the quantization value corresponding to each data and the weight coefficient of the corresponding position of each data.

[0045] The AI ​​big model-based voice content compliance intelligent judgment and risk grading system provided in this embodiment monitors the voice of the call center personnel when they use the call center terminal to make outbound calls through the voice acquisition module, and uses advanced natural language processing technology and big model algorithms during monitoring to ensure the accuracy and compliance of voice information. At the same time, it performs intelligent grading according to the risk level, provides effective risk management and monitoring means for financial institutions, and ensures the stability and healthy development of the financial market; further provides intelligent compliance monitoring, analysis, early warning and control for the financial securities industry, and performs intelligent compliance judgment on various content resources (such as: voice files of call system call records, image, video and text files of communication chat conversations, etc.) through the AI ​​financial big model and expert knowledge base, analyzes the summary of the resources, and classifies the compliance risks of the resources by level, provides statistical analysis, sampling analysis, and early warning analysis in various dimensions, and provides corresponding intervention and control measures for resources of each compliance risk level.

[0046] The AI ​​big model used in this embodiment is trained and converged in advance. In order to make the model targeted to the financial securities industry, when constructing the training data, the voice data stored in the outbound call storage system in the industry is selected and processed, and then the training data is formed; to ensure the applicability of the model to the financial securities industry. In addition, the output of the model is the data of a data set, and the value corresponding to each parameter in the data set indicates whether the voice meets the regulations corresponding to the parameter. Furthermore, in order to implement intervention when risks occur, the intervention strategy can be retrieved from the preset intervention strategy determination library based on the risk assessment vector set and the intervention strategy can be displayed; the intervention strategy in the intervention strategy determination library is associated with the retrieved parameter set in a one-to-one correspondence, and the retrieval can be achieved by matching the retrieved parameter set with the wind direction assessment vector set. In addition, before the voice is input into the AI ​​big model, it needs to go through the ASR voice transcription technology to convert the voice file into text, and then the text content is combined with the prompt project to give it to the AI ​​big model.

[0047] In one embodiment, a voice content compliance intelligent judgment and risk grading system based on an AI big model includes: a voice acquisition module, a voice recognition module, a risk assessment module, and a reminder module;

[0048] Among them, the voice acquisition module obtains external voice from the audio collection device set in the call terminal; the voice recognition module inputs the external voice into the AI ​​big model; the risk assessment module determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information.

[0049] Among them, the risk assessment steps of the risk assessment module are as follows:

[0050] According to the pre-configured second keyword library corresponding to each risk level, a keyword extraction operation is performed on the output data of the AI ​​large model; the extraction operation is to determine whether the value corresponding to the corresponding keyword (parameter name) in the second keyword library is a preset value. When it is, it is determined that the keyword can be extracted;

[0051] The risk level of the external voice is determined according to the highest risk level among the risk levels corresponding to the second keyword library from which the keywords are extracted.

[0052] The risk assessment of this embodiment uses a pre-configured second keyword library corresponding to each risk level, that is, each level is configured with a corresponding second keyword library, analyzes the external voice, and determines the corresponding risk level based on which second keyword library the extracted keywords are located in, thereby achieving fast and accurate risk analysis.

[0053] In order to analyze the voice in the historical call records, in one embodiment, a voice content compliance intelligent judgment and risk grading system based on an AI big model is provided, such as Figure 2 As shown, it includes: a voice acquisition module 1, a voice recognition module 2, a risk assessment module 3, a reminder module 4 and a voice retrieval module 5;

[0054] Among them, the voice retrieval module 5 retrieves the voice files in the call records of the call system to obtain historical voices; the voice recognition module 2 inputs the historical voices into the AI ​​big model; the risk assessment module 3 determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module 4 outputs abnormal reminder information.

[0055] In order to realize the analysis of voice in Internet instant messaging software, in one embodiment, a voice content compliance intelligent judgment and risk grading system based on an AI big model includes: a voice acquisition module, a voice recognition module, a risk assessment module, a reminder module and a voice separation module;

[0056] Among them, the voice separation module is used to separate the voice files in the call records of the Internet instant messaging software to obtain separated voice; the voice recognition module inputs the separated voice into the AI ​​big model; the risk assessment module determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information.

[0057] In order to improve the accuracy and effectiveness of the analysis; in one embodiment, a voice content compliance intelligent judgment and risk grading system based on an AI big model includes: a voice acquisition module, a voice recognition module, a risk assessment module, a reminder module, a voice separation module, an association grouping module and an association analysis module;

[0058] Among them, the association grouping module is used to associate and group the output data of the separated voice after passing through the AI ​​large model with the text data in the call record; the association analysis module is used to analyze the associated grouped data according to the pre-configured association analysis library to obtain the association analysis result; when the association analysis result is abnormal, the reminder module outputs abnormal reminder information.

[0059] Among them, text data includes: text entered by people into instant messaging software and data obtained by interpreting emoticons through the emoticon interpretation library.

[0060] The association grouping rules of the association grouping module include:

[0061] Perform group association based on the time difference between the time corresponding to the separated voice and the time corresponding to the text data;

[0062] and / or,

[0063] Perform group association based on the pre-configured association statement library.

[0064] The association analysis steps of the association analysis module include:

[0065] According to the associated data query sub-database in the associated analysis database, the standard associated data corresponding to each data in the associated group is determined; the data in the associated data query sub-database is associated with the standard associated data, and the correspondence between the data and the standard associated data can be one-to-one or one-to-many;

[0066] Match the standard associated data with the associated data corresponding to each data in the associated group to determine the matching degree; when there are multiple associated data for a data, the largest matching value between the standard associated data and the associated data is used as the matching degree between the standard associated data and the data in the associated group; the matching value can be determined by analyzing and calculating the similarity between the two data;

[0067] Based on the matching degree between the standard associated data and the associated data, the associated value of the associated group is calculated; the associated value is the sum of the matching degrees of each data;

[0068] When the correlation value is less than or equal to the preset abnormal threshold, it is determined that the correlation analysis result is abnormal.

[0069] In order to analyze relevant data, in one embodiment, the voice content compliance intelligent judgment and risk grading system based on the AI ​​big model further includes: a relevant association module and a compliance analysis module.

[0070] Among them, the coherent association module associates the voices of different voice speakers according to the voice receiving object and the communication time to form coherent association data; the compliance analysis module analyzes the coherent association data to determine whether it meets the pre-configured compliance requirements; when it does not meet the requirements, the reminder module sends a reminder message.

[0071] The association rules of the relevant association data include:

[0072] When the voice receiving objects are the same, the communication time corresponding to the two voices is within the preset time range threshold, and the voice senders are different, the two voices are regarded as coherent related data;

[0073] The analysis steps of relevant correlation data include:

[0074] Based on the preset keyword library, keywords are extracted from the two voices respectively, and the portraits of the two voices are constructed based on the keyword extraction results;

[0075] Calculate the similarity between the two speech images; the similarity can be calculated using the cosine similarity calculation method;

[0076] When the similarity is greater than a preset threshold, it is judged as not meeting the pre-configured compliance requirements;

[0077] Among them, the keywords in the keyword library are associated with the business classification codes involved; by extracting the keywords of the voice, a portrait is constructed according to the business classification codes corresponding to the extracted keywords, and the portrait is a data set understood as the business classification codes involved in the voice;

[0078] Furthermore, the machine codes of the sending end and the receiving end of the voice interaction can be used as the basis for extracting coherent associated data; that is, the same machine codes are used as coherent associated data.

[0079] In order to achieve real-time monitoring and reminder, in one embodiment, the voice content compliance intelligent judgment and risk grading system based on the AI ​​big model also includes: a real-time monitoring and reminder module;

[0080] Analyze the staff's historical non-compliant speech and build a real-time reminder library;

[0081] The staff's current voice is matched with the trigger voice in the real-time reminder library. When there is a match, the reminder information corresponding to the match is retrieved and displayed on the staff's corresponding display terminal.

[0082] Among them, the triggering voice and reminder information in the real-time reminder library are associated one by one.

[0083] During matching, the trigger items in the real-time reminder library are analyzed and sorted to improve the efficiency of reminders. The specific analysis and sorting are as follows: real-time reminder items are constructed based on non-compliant voices monitored on the platform within a preset time period before the current moment, and the similarity between each real-time reminder item and each trigger item is calculated (equal to the similarity between the voice in the real-time reminder item and the trigger voice); based on the pre-configured time-weight comparison table, the weight coefficient corresponding to each real-time reminder item is determined; the sum of the products of the similarity of each trigger item and the weight coefficient is used as the first value; the number of historical non-compliant voices corresponding to the trigger item when it is constructed is used to query the preset second value mapping table to determine the second value; the sum of the first value and the second value is used as the basis for sorting, that is, the sum values ​​are sorted in descending order and then called in sequence for trigger analysis.

[0084] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A voice content compliance intelligent judgment and risk grading system based on an AI big model, characterized by: include: Speech acquisition module, speech recognition module, risk assessment module, reminder module, speech separation module, association grouping module and association analysis module; Among them, the voice acquisition module obtains the external voice from the audio collection device set in the call terminal; the voice recognition module inputs the external voice into the AI ​​big model; the risk assessment module determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information; The voice separation module is used to separate the voice files in the call records of the Internet instant messaging software to obtain separated voice; the voice recognition module inputs the separated voice into the AI ​​big model; the risk assessment module determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information; The association grouping module is used to associate and group the output data of the separated voice after passing through the AI ​​large model with the text data in the call record; the association analysis module is used to analyze the associated grouped data according to the pre-configured association analysis library to obtain the association analysis result; when the association analysis result is abnormal, the reminder module outputs abnormal reminder information; The association analysis steps of the association analysis module include: According to the associated data query sub-library in the associated analysis library, the standard associated data corresponding to each data in the associated group is determined; Matching the standard associated data with the associated data corresponding to each data in the associated group to determine the degree of match; Calculate the association value of the association group based on the matching degree between the standard association data and the association data; When the correlation value is less than or equal to the preset abnormal threshold, it is determined that the correlation analysis result is abnormal.

2. The voice content compliance intelligent judgment and risk grading system based on AI big model according to claim 1 is characterized in that: The risk assessment steps of the risk assessment module are as follows: Based on the pre-configured first keyword library, keyword extraction is performed on the output data of the AI ​​large model. Based on the results of the keyword extraction operation, a risk assessment vector set is constructed; Based on the preset quantitative library, determine the quantitative value corresponding to each data in the risk assessment vector set; The risk value is determined based on the quantitative value corresponding to each data and the weight coefficient of the corresponding position of each data.

3. According to claim 1, the voice content compliance intelligent judgment and risk grading system based on AI big model is characterized in that: The risk assessment steps of the risk assessment module are as follows: Based on the pre-configured second keyword library corresponding to each risk level, keyword extraction operation is performed on the output data of the AI ​​large model; The risk level of the external voice is determined according to the highest risk level among the risk levels corresponding to the second keyword library from which the keywords are extracted.

4. The voice content compliance intelligent judgment and risk grading system based on AI big model according to claim 1 is characterized in that: Also includes: Voice retrieval module; Among them, the voice retrieval module retrieves the voice files in the call records of the call system to obtain historical voice; the voice recognition module inputs the historical voice into the AI ​​big model; the risk assessment module determines the risk level based on the output of the AI ​​big model and the pre-configured risk assessment rules; when the risk level is greater than the preset level threshold, the reminder module outputs abnormal reminder information.

5. According to claim 1, the voice content compliance intelligent judgment and risk grading system based on AI big model is characterized in that: Text data includes: text entered by people into instant messaging software and data obtained by interpreting emoticons through the emoticon interpretation library.

6. The voice content compliance intelligent judgment and risk grading system based on AI big model according to claim 1 is characterized in that: The association grouping rules of the association grouping module include: Perform group association based on the time difference between the time corresponding to the separated voice and the time corresponding to the text data; and / or, Perform group association based on the pre-configured association statement library.

7. The voice content compliance intelligent judgment and risk grading system based on AI big model according to claim 1 is characterized in that: Also includes: coherent association module and compliance analysis module, Among them, the coherent association module associates the voices of different voice speakers according to the voice receiving object and the communication time to form coherent association data; the compliance analysis module analyzes the coherent association data to determine whether it meets the pre-configured compliance requirements; when it does not meet the requirements, the reminder module sends a reminder message.

Images