Risk event identification method, device, computer equipment and storage medium

Abstract

The present invention relates to the fields of artificial intelligence technology and financial technology, and discloses a risk event identification method, device, computer equipment and storage medium. The method comprises the following steps: collecting service call recordings of all customers within a target time period and performing preprocessing to obtain recording data, which is then converted into text data; extracting voiceprint features of each recording data through a voiceprint model, and performing PLDA scoring on any two voiceprint features to obtain a scoring result; identifying multiple feature clusters based on the scoring result, calculating the average value of all scoring results in each feature cluster, and generating a voice risk value; selecting a target feature cluster based on each voice risk value and a cluster screening strategy, and using the voice risk value of the target feature cluster as a voice risk factor; generating a text risk factor for each text data through a large language model; and identifying whether the target service call recording is a risk event through a risk event identification model based on the voice risk factor and the text risk factor. The method can effectively improve the efficiency and accuracy of risk event identification.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence technology and financial technology, and specifically to a risk event identification method, device, computer equipment and computer-readable storage medium. Background Art

[0002] With the advancement of computer technology, more and more technologies are being applied in the financial sector. Traditional finance is gradually shifting towards financial technology (Fintech), and risk event identification technology is no exception. For example, in financial institutions such as banks, customer service and transaction processing are core aspects of their operations. With the advancement of digital transformation, a large number of customer interactions occur via telephone or online channels. However, these channels have also become hotbeds for financial risk activities. For example, in the insurance industry, behaviors such as agent surrenders and malicious complaints not only harm the interests of insurance companies but also affect customer experience and market order. In banking, behaviors such as fraudulent account openings and loan fraud similarly pose significant financial losses and reputational risks to financial institutions. Therefore, financial institutions can use risk event identification technology to promptly detect financial risk behaviors and avoid economic losses. However, the financial industry's security and real-time requirements also place higher demands on risk event identification technology.

[0003] With the development of the financial industry, banks and insurance companies are facing increasingly severe financial risks. Traditional risk event identification methods, such as rule-based systems and manual review, while able to identify potential risk events to a certain extent, have significant limitations. First, these methods rely on historical data and empirical rules, making them difficult to adapt to rapidly changing risk measures. Second, manual review is inefficient and susceptible to subjective judgment, limiting recognition accuracy and efficiency. Furthermore, existing technologies often struggle to process large amounts of data, especially complex voice data, making it difficult to effectively extract and analyze key information, which limits the identification of risk events.

[0004] Based on this, how to provide a risk event identification method, device, computer equipment and computer-readable storage medium that can effectively improve the efficiency and accuracy of risk event identification is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of the above-mentioned deficiencies in the prior art, the object of the present invention is to provide a risk event identification method, apparatus, computer equipment and computer-readable storage medium, aiming to solve the problem of how to effectively improve the efficiency and accuracy of risk event identification.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a risk event identification method, which includes:

[0008] Collecting service call recordings of all customers within a target time period, pre-processing each service call recording to obtain corresponding recording data, and converting all the recording data into corresponding text data;

[0009] Extracting voiceprint features of each recording data through a voiceprint model, and performing PLDA scoring on any two voiceprint features to obtain a scoring result;

[0010] Performing voiceprint cluster analysis based on the scoring results to identify multiple feature clusters, each of which has multiple similar voiceprint features, and calculating the average of all the scoring results in each feature cluster to generate a corresponding voice risk value;

[0011] Selecting a target feature cluster based on each of the speech risk values ​​and a cluster screening strategy, and using the speech risk value of the target feature cluster as a speech risk factor;

[0012] Generating a text risk factor corresponding to each of the text data using a large language model;

[0013] Based on the voice risk factor and the text risk factor, a risk event identification model is used to identify whether the target service call recording is a risk event.

[0014] In a second aspect, the present invention provides a risk event identification device, comprising:

[0015] A collection module is used to collect service call recordings of all customers within a target time period, pre-process each service call recording to obtain corresponding recording data, and convert all the recording data into corresponding text data;

[0016] An extraction module, configured to extract voiceprint features of each recording data through a voiceprint model, and perform PLDA scoring on any two voiceprint features to obtain a scoring result;

[0017] a calculation module, configured to perform voiceprint cluster analysis based on the scoring results, identify multiple feature clusters, each of which has multiple similar voiceprint features, and calculate the average of all the scoring results in each feature cluster to generate a corresponding voice risk value;

[0018] a selection module, configured to select a target feature cluster based on each of the speech risk values ​​and a cluster screening strategy, and use the speech risk value of the target feature cluster as a speech risk factor;

[0019] A generation module, configured to generate a text risk factor corresponding to each of the text data using a large language model;

[0020] The identification module is used to identify whether the target service call recording is a risk event based on the voice risk factor and the text risk factor through a risk event identification model.

[0021] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the risk event identification method as described above when executing the computer program.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the risk event identification method as described above.

[0023] Compared with the prior art, the present invention provides a risk event identification method, device, computer equipment and computer-readable storage medium, wherein, by collecting service call recordings of all customers within a target time period, each service call recording is preprocessed to obtain corresponding recording data, and all the recording data are converted into corresponding text data; the voiceprint features of each recording data are extracted through a voiceprint model, and any two of the voiceprint features are scored by PLDA to obtain a scoring result; based on the scoring result, voiceprint clustering analysis is performed to identify multiple feature clusters, each feature cluster has multiple similar voiceprint features, and the average value of all the scoring results in each feature cluster is calculated to generate a corresponding voice risk value; based on each voice risk value and cluster screening strategy, a target feature cluster is selected, and the voice risk value of the target feature cluster is used as a voice risk factor; a text risk factor corresponding to each text data is generated through a large language model; based on the voice risk factor and the text risk factor, a risk event identification model is used to identify whether the target service call recording is a risk event; thereby, the present invention can effectively improve the efficiency and accuracy of risk event identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0025] Figure 1 A schematic diagram of an application environment of a risk event identification method provided by one embodiment of the present invention.

[0026] Figure 2A flow chart of a risk event identification method provided by one embodiment of the present invention.

[0027] Figure 3 A schematic diagram of a program module of a risk event identification device provided by one embodiment of the present invention.

[0028] Figure 4 A schematic diagram of the structure of a computer device provided in one embodiment of the present invention.

[0029] Figure 5 Another structural diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0031] It should be understood that when used in the present specification and the appended claims, the term "comprising" indicates the presence of described features, integers, steps, operations, elements and / or components, but does not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof.

[0032] It will also be understood that the term "and / or" used in the present description and appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.

[0033] As used in the present specification and the appended claims, the term "if" may be interpreted as "when" or "upon" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrase "if it is determined" or "if [described condition or event] is detected" may be interpreted as meaning "upon determination" or "in response to determining" or "upon detection of [described condition or event]" or "in response to detecting [described condition or event]," depending on the context.

[0034] In addition, in the description of the present specification and the appended claims, the terms "first", "second", "third", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0035] References to "one embodiment" or "some embodiments" in the present specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present invention. Thus, phrases such as "in one embodiment," "in some embodiments," "in other embodiments," and "in yet other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "including," "comprising," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0036] It should be understood that the order of execution of the steps in the following embodiments does not necessarily mean the order in which they are executed. The order in which each process is executed should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0037] In order to illustrate the technical solution of the present invention, specific embodiments are provided below.

[0038] An embodiment of the present invention provides a risk event identification method that can be applied in the following situations: Figure 1 In the application environment shown, the client and server communicate via a network. The client includes, but is not limited to, PDAs, desktop computers, laptops, ultra-mobile personal computers (UMPCs), netbooks, cloud computing devices, personal digital assistants (PDAs), and other computer devices. The server can be a standalone server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0039] See also Figure 2 An embodiment of the present invention provides a method for identifying risk events, wherein the method comprises the following steps:

[0040] S100: Collecting service call recordings of all customers within a target time period, pre-processing each service call recording to obtain corresponding recording data, and converting all the recording data into corresponding text data;

[0041] S200, extracting voiceprint features of each recording data through a voiceprint model, and performing PLDA scoring on any two voiceprint features to obtain a scoring result;

[0042] S300: Perform voiceprint cluster analysis based on the scoring results to identify multiple feature clusters, each of which has multiple similar voiceprint features, and calculate the average of all the scoring results in each feature cluster to generate a corresponding voice risk value;

[0043] S400: selecting a target feature cluster based on each of the speech risk values ​​and a cluster screening strategy, and using the speech risk value of the target feature cluster as a speech risk factor;

[0044] S500: Generate text risk factors corresponding to each text data using a large language model;

[0045] S600: Based on the voice risk factor and the text risk factor, identify whether the target service call recording is a risk event through a risk event identification model.

[0046] In specific implementation, this embodiment achieves a comprehensive risk assessment of service call recordings by comprehensively utilizing data from both voiceprint analysis and text analysis modalities, thereby effectively improving the efficiency and accuracy of risk event identification. First, voiceprint features are extracted through preprocessing and voiceprint modeling, and the PLDA score is used to assess the similarity between voiceprints. This helps identify voiceprint groups with common characteristics. Second, voiceprint clustering analysis can screen out call recordings with similar voiceprint features, providing a targeted analysis scope for subsequent risk assessment. Next, a large language model is used to analyze the text content of all converted text data to generate text risk factors, which helps identify potential risky behaviors. Finally, by fusing voice risk factors and text risk factors and applying a risk event identification model, it is possible to more accurately determine whether the target service call recording is a risk event. This embodiment's method not only improves the coverage of risk detection, but also enhances the reliability of identification results through multimodal data analysis, making risk management more precise and efficient.

[0047] Furthermore, in one embodiment, the risk event identification method, wherein step S100, collecting service call recordings of all customers within a target time period, pre-processing each of the service call recordings to obtain corresponding recording data, and converting all of the recording data into corresponding text data, specifically includes the following steps:

[0048] Determine the customer groups and target time periods for the service call recordings collected in the target financial institution based on the task objectives of the risk event identification task;

[0049] Obtaining the service call recordings of all customers in the customer group within the target time period from the customer service system of the target financial institution;

[0050] After denoising, enhancing and normalizing each of the service call recordings, corresponding recording data is obtained, and all of the recording data is converted into corresponding text data through automatic speech recognition technology.

[0051] In specific implementation, this embodiment determines the customer group and target time period for collecting service call recordings in the target financial institution based on the task objectives of the risk event identification task, and obtains the service call recordings of all customers in the customer group within the target time period from the customer service system of the target financial institution, and performs denoising, enhancement and normalization on them, which can significantly improve the accuracy and efficiency of subsequent voice analysis. At the same time, all the recording data are converted into corresponding text data through automatic speech recognition (ASR) technology. Among them, denoising helps to eliminate background noise and interference, making the voice signal clearer; enhancement processing can improve the quality of the voice signal, making important acoustic features more prominent; normalization ensures consistency between different recordings, which is convenient for comparison and analysis. These preprocessing steps comprehensively improve the availability of recording data, laying a solid foundation for further voiceprint recognition and risk event detection.

[0052] The specific implementation process of the steps in this embodiment is roughly as follows:

[0053] 1. Determine the scope of collection:

[0054] Target time period setting: Set the start and end time points according to the time range to be analyzed;

[0055] Customer group definition: clearly define the customer groups or service categories to be collected;

[0056] 2. Collect call recordings:

[0057] System interface connection: Establish a connection with the financial institution's customer service system to ensure access to the call recording database;

[0058] Data extraction: Use the interface or database query provided by the customer service system to extract service call recordings of all customers in the customer group within the target time period;

[0059] 3. Pre-process call recordings:

[0060] Format conversion: Convert service call recordings into a format suitable for processing, such as WAV or MP3;

[0061] Denoising: Apply digital signal processing techniques, such as spectral subtraction or wavelet transform, to remove background noise;

[0062] Enhancement: Enhance the speech signal through methods such as gain control and dynamic range compression;

[0063] Normalization: Normalize the audio signal to keep the volume at a stable level;

[0064] 4. Text conversion of recording data:

[0065] ASR system call: Use the automatic speech recognition system to convert the recorded data into corresponding text data;

[0066] Conversion processing: Processing the output of the ASR system, including correcting recognition errors, formatting text, etc., to improve the quality of text data;

[0067] Through the above process, this embodiment can ensure that high-quality recording data and corresponding text data are extracted from the original service call recording, providing accurate input for subsequent voiceprint recognition and risk analysis.

[0068] Furthermore, in one embodiment, the risk event identification method, wherein the step S200, extracting the voiceprint features of each recording data through a voiceprint model, and performing PLDA scoring on any two voiceprint features to obtain a scoring result, specifically includes the following steps:

[0069] Pre-constructing the voiceprint model and the PLDA model, and inputting each of the recording data into the voiceprint model for feature extraction to generate voiceprint features of each of the recording data;

[0070] Any two of the voiceprint features are input into the PLDA model for PLDA scoring to generate a preliminary result, and the preliminary result is standardized to obtain a scoring result.

[0071] In practice, this embodiment pre-builds the voiceprint model and PLDA model to automate the process of extracting key voiceprint features from recorded data and evaluating the similarity between these features. First, the voiceprint model extracts features from the recorded data, generating a data representation that represents the speaker's unique voice characteristics. Second, the PLDA model scores these voiceprint features to quantify the similarity between different voiceprints. This process not only improves the accuracy of voiceprint recognition but also provides standardized scoring results for subsequent voiceprint cluster analysis and risk assessment, thereby enhancing the reliability and efficiency of the entire system in identifying and distinguishing different speakers.

[0072] The specific implementation process of the steps in this embodiment is roughly as follows:

[0073] 1. Build a voiceprint model:

[0074] Select an algorithm: Choose a suitable voiceprint recognition algorithm, such as a Gaussian mixture model (GMM) or a deep learning model;

[0075] Training the model: Use a voiceprint dataset with known identities to train the voiceprint model so that the model can learn the characteristics that distinguish different speakers;

[0076] 2. Constructing the PLDA model:

[0077] Configure the model: set the parameters of the PLDA model, such as the regularization coefficient;

[0078] Model training: Use labeled voiceprint data to train the PLDA model so that it can evaluate the similarity between voiceprint features;

[0079] 3. Feature extraction:

[0080] Framing and windowing: Framing the recorded data into frames and applying a window function (such as a Hamming window) to each frame;

[0081] Feature calculation: Calculate the voiceprint features of each frame, such as Mel-frequency cepstral coefficients (MFCC);

[0082] 4. Voiceprint feature vector generation:

[0083] Feature vectorization: convert the extracted voiceprint features into voiceprint feature vectors;

[0084] Time series processing: Process the feature sequence of the entire recording data to generate a fixed-length voiceprint feature vector;

[0085] 5. PLDA score:

[0086] Pairwise comparison: compare voiceprint feature vectors pairwise;

[0087] PLDA model application: Each pair of voiceprint feature vectors is input into the PLDA model to calculate the similarity score between them;

[0088] Scoring output: The PLDA model outputs preliminary results, which reflect the similarity between different voiceprints;

[0089] 6. Standardization processing:

[0090] Standardize the preliminary results to obtain scoring results to eliminate the dimensionality effects between different scores so that the scoring results can be compared under the same standard. Standardization may include methods such as minimum-maximum standardization and Z-score standardization, which helps to improve the comparability of scores and the generalization ability of the PLDA model.

[0091] Through the above process, this embodiment can effectively extract voiceprint features from the recorded data and use the PLDA model to score these features, providing data support for further voiceprint analysis and risk assessment.

[0092] Furthermore, in one embodiment, the risk event identification method, wherein step S300, performs voiceprint cluster analysis based on the scoring results, identifies multiple feature clusters having multiple similar voiceprint features, and calculates the average value of all the scoring results in each feature cluster to generate a corresponding voice risk value, specifically includes the following steps:

[0093] Determine the clustering algorithm based on the task requirements of the risk event identification task;

[0094] Based on the clustering algorithm, voiceprint clustering is performed according to each of the scoring results, and each of the voiceprint features is grouped into a plurality of feature clusters, wherein the feature clusters have a plurality of similar voiceprint features;

[0095] The average value of all the scoring results in each feature cluster is calculated to generate a speech risk value corresponding to each feature cluster.

[0096] In practice, this embodiment achieves effective grouping of voiceprint features by carefully selecting a clustering algorithm based on the requirements of the risk event identification task, clustering data points with similar voiceprint features into characteristic clusters. This clustering method not only reveals the inherent structure in the data, but also, by calculating the average score within each cluster, assesses the similarity of cluster members, thereby generating a voice risk score reflecting the characteristics of each cluster. This process significantly improves the accuracy of voiceprint recognition, enhances the ability to verify the speaker's identity, and provides a solid data foundation for further risk assessment or biometric authentication.

[0097] The specific implementation process of the steps in this embodiment is roughly as follows:

[0098] 1. Requirements analysis and algorithm selection:

[0099] Task requirement analysis: Analyze the specific requirements of the risk event identification task, such as recognition accuracy, computational efficiency, cluster shape, etc.

[0100] Clustering algorithm determination: select the appropriate clustering algorithm according to the needs;

[0101] 2. Implement voiceprint clustering:

[0102] Data preparation: Ensure that all voiceprint features and corresponding PLDA scoring results are ready;

[0103] Algorithm configuration: configure parameters according to the characteristics of the selected clustering algorithm;

[0104] 3. Feature cluster formation:

[0105] Perform clustering: Run the clustering algorithm to process the voiceprint features and group the voiceprint features according to the similarity of the scoring results;

[0106] Cluster verification: Evaluate the quality of the formed feature clusters to ensure that the voiceprint features within each feature cluster are sufficiently similar and that there are significant differences between clusters;

[0107] 4. Calculate the average score within the cluster:

[0108] Aggregate scoring: Aggregate the PLDA scores of all voiceprint features within each feature cluster;

[0109] Average value calculation: Calculate the average PLDA score of each feature cluster to obtain the similarity index of the feature cluster;

[0110] 5. Generate voice risk value:

[0111] Result formatting: format the average score results of each feature cluster for further analysis;

[0112] Output results: Output the voice risk value of each feature cluster, which can be used for subsequent risk assessment or identity verification;

[0113] Through the above process, this embodiment can effectively perform cluster analysis on voiceprint features and calculate the similarity index of each feature cluster, providing a basis for identifying speakers or assessing the risks of call recordings.

[0114] Furthermore, in one embodiment, the risk event identification method, wherein step S400, selecting a target feature cluster based on each of the voice risk values ​​and a cluster screening strategy, and using the voice risk value of the target feature cluster as a voice risk factor, specifically includes the following steps:

[0115] Determining a cluster screening strategy according to the task requirements of the risk event identification task;

[0116] Based on the cluster screening strategy, a target feature cluster is selected according to the speech risk value of each feature cluster, and the speech risk value of the target feature cluster is used as a speech risk factor.

[0117] In practice, this embodiment achieves accurate identification and screening of feature clusters by customizing a cluster screening strategy based on the task requirements of risk event identification. This process first involves determining a cluster screening strategy based on the predetermined task requirements. This strategy is then applied to evaluate the generated feature clusters, selecting target feature clusters with high-risk characteristics. By marking the voice risk values ​​of these target feature clusters as voice risk factors, this implementation method not only improves the targetedness and efficiency of voiceprint analysis but also enhances the ability to identify potential risk events, thereby providing financial institutions with an effective risk management and early warning mechanism.

[0118] The specific implementation process of the steps in this embodiment is roughly as follows:

[0119] 1. Demand analysis to determine cluster screening strategy:

[0120] Task requirements analysis: Detailed understanding of the specific requirements of the risk event identification task, including risk definition, expected identification accuracy, and response time;

[0121] Cluster screening strategy formulation: Based on the results of the task requirements analysis, a cluster screening strategy is formulated, which may include selecting a specific feature cluster size, average PLDA score threshold, or other relevant indicators;

[0122] 2. Evaluate feature clusters:

[0123] Speech risk value collection: collect the speech risk value of each feature cluster generated in the clustering process;

[0124] Cluster feature analysis: Analyze the features of each feature cluster to determine whether they meet the criteria of the cluster screening strategy;

[0125] 3. Select the target feature cluster:

[0126] Apply screening rules: Select the target feature cluster that meets the conditions from all feature clusters according to the rules set in the cluster screening strategy;

[0127] Target cluster confirmation: Confirm the selected target feature clusters and ensure they are representative and relevant so as to serve as the basis for risk assessment;

[0128] 4. Generate Voice Risk Factors:

[0129] Calculate risk index: For the selected target feature cluster, further calculate or confirm its speech risk value, which will be used as the speech risk factor;

[0130] Risk factor standardization: Perform necessary standardization on voice risk factors to facilitate integration with other types of risk factors or models;

[0131] Through the above process, this embodiment can ensure that target feature clusters with potential risks are effectively screened out according to task requirements, and the speech risk values ​​of these clusters are used as speech risk factors for further risk management and decision support.

[0132] Furthermore, in one embodiment, the risk event identification method, wherein the step S500 of generating a text risk factor corresponding to each text data by using a large language model, specifically comprises the steps of:

[0133] Pre-building the large language model;

[0134] Each of the text data is processed by word segmentation, stop word removal, and part-of-speech tagging, and then input into the large language model;

[0135] The large language model performs text analysis on each of the text data after word segmentation, stop word removal, and part-of-speech tagging to obtain corresponding risk text data, and generates a text risk factor corresponding to each of the risk text data.

[0136] In specific implementation, this embodiment achieves efficient conversion from voiceprint features to text risk assessment through a pre-built large language model. First, the text data obtained by converting the aforementioned recording data is segmented, stop words are removed, and part-of-speech tagging is performed; then, the text data after segmentation, stop words are removed, and part-of-speech tagging is input into the large language model for in-depth text analysis. The large language model can identify risk patterns, abnormal words, or potential risk signs in the text to obtain corresponding risk text data and generate text risk factors corresponding to each risk text data. This process not only improves the coverage of risk detection, but also enhances the ability to identify potential risk events, providing financial institutions with a comprehensive and efficient risk assessment method.

[0137] The specific implementation process of the steps in this embodiment is roughly as follows:

[0138] 1. Build a large language model:

[0139] Model selection: Select or develop a large language model suitable for the task, such as BERT or GPT based on Transformer.

[0140] Training data preparation: Collect and prepare text data for model training, which may include historical risk event texts, normal call texts, etc.

[0141] Model training: Use the prepared data to train the model so that it can understand and analyze risk factors in the text;

[0142] 2. Text risk factor analysis:

[0143] Input processing: The text data obtained from the conversion of the aforementioned recording data is organized (by word segmentation, stop word removal, part-of-speech tagging, etc.) into a format suitable for input into the large language model;

[0144] Model analysis: Input text data into the large language model, which analyzes the text content and identifies risk patterns or risk signs to generate analysis results;

[0145] 3. Generate text risk factors:

[0146] Obtain corresponding risk text data based on the analysis results of the large language model, and generate a risk score or risk label, i.e., a text risk factor, for each corresponding text data based on the risk text data;

[0147] Through the above process, this embodiment can effectively use the large language model to perform text risk analysis and obtain the text risk factor of each text data, thereby providing a more comprehensive perspective for risk management.

[0148] Furthermore, in one embodiment, the risk event identification method, wherein step S600, identifying whether the target service call recording is a risk event using a risk event identification model based on the voice risk factor and the text risk factor, specifically includes the following steps:

[0149] Pre-building the risk event identification model;

[0150] When the target service call recording has the corresponding voice risk factor and the text risk factor, integrating the voice risk factor and the text risk factor corresponding to the target service call recording to obtain a mixed risk factor for the target service call recording;

[0151] The mixed risk factor is input into the risk event identification model to generate a risk event identification result of the target service call recording.

[0152] In specific implementation, this embodiment achieves a comprehensive risk assessment of service call recordings through a pre-built risk event identification model. First, by integrating the risk factors of both voice and text modalities, a hybrid risk factor is generated. This process fully utilizes information from different data sources and improves the comprehensiveness and accuracy of risk assessment. Then, the hybrid risk factor is input into the risk event identification model. The model can comprehensively consider multiple risk factors to more accurately identify and predict whether the target service call recording is a risk event. This implementation method not only improves the efficiency of risk identification, but also enhances the ability to identify potential risk events through comprehensive analysis of multiple data types, providing financial institutions with an effective risk management and early warning mechanism.

[0153] The specific implementation process of the steps in this embodiment is roughly as follows:

[0154] 1. Build a risk event identification model:

[0155] Model selection: Choose a suitable machine learning or deep learning framework, such as random forest, support vector machine, neural network, etc.

[0156] Feature engineering: Determine the features required for risk event identification, including features extracted from voice and text risk factors;

[0157] Model training: Use historical data and labeled risk event samples to train the model until the model achieves satisfactory accuracy;

[0158] 2. Collect risk factors:

[0159] Voice risk factor collection: obtain voice risk factors from the voice risk values ​​of all scoring results in the feature cluster;

[0160] Text risk factor collection: Obtain text risk factors from text analysis of large language models;

[0161] 3. Integrate risk factors:

[0162] Data preprocessing: preprocess the collected voice risk factors and text risk factors, such as standardization and normalization;

[0163] Feature integration: Integrate voice risk factors and text risk factors according to certain weights or models to generate hybrid risk factors;

[0164] 4. Risk event identification:

[0165] Model application: Hybrid risk factors are input into the risk event identification model, and the model performs risk event assessment on target service call recordings with hybrid risk factors.

[0166] Threshold determination: Based on the risk score output by the risk event identification model and the set threshold, determine whether the target service call recording is a risk event;

[0167] 5. Output risk event identification results:

[0168] Result analysis: Analyze the identification results of the risk event identification model to determine the type and level of the risk event;

[0169] Report generation: Generate risk event identification reports and provide them to risk management personnel or automatically trigger early warning mechanisms;

[0170] Through the above process, this embodiment can effectively implement risk assessment of target service call recordings and generate accurate risk event identification results, providing decision support for risk management of financial institutions.

[0171] As can be seen from the above method embodiments, the risk event identification method provided by the present invention includes: collecting service call recordings of all customers within a target time period, preprocessing each service call recording to obtain corresponding recording data, and converting all the recording data into corresponding text data; extracting the voiceprint features of each recording data through a voiceprint model, and performing PLDA scoring on any two voiceprint features to obtain a scoring result; performing voiceprint cluster analysis based on the scoring result to identify multiple feature clusters, each of which has multiple similar voiceprint features, and calculating the average value of all the scoring results in each feature cluster to generate a corresponding voice risk value; selecting a target feature cluster based on each voice risk value and a cluster screening strategy, and using the voice risk value of the target feature cluster as a voice risk factor; generating a text risk factor corresponding to each text data through a large language model; and identifying whether the target service call recording is a risk event through a risk event identification model based on the voice risk factor and the text risk factor. In this way, the method of the present invention can effectively improve the efficiency and accuracy of risk event identification.

[0172] It can be understood that the risk event identification method provided by the embodiment of the present invention can be applied to risk event identification scenarios related to the field of financial technology. Specifically, the risk event identification function can be integrated into the corresponding customer service and transaction processing scenarios of financial institutions (such as banks) to timely discover financial risk events and behaviors and avoid economic losses caused to financial institutions.

[0173] It should be understood that although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on conventional or non-creative work, and these operation steps are not necessarily performed in the order of the embodiments or flowcharts. The order of steps listed in the embodiments or flowcharts is only one way of executing the steps among many steps and does not represent the only execution order. It should be noted that there is not necessarily a certain order between the above steps. Those of ordinary skill in the art can understand from the description of the embodiments of the present invention that in different embodiments, the above steps may have different execution orders, that is, they may be executed in parallel, or they may be executed in an interchangeable manner, etc. Moreover, at least a portion of the steps in the embodiments or flowcharts may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but may be executed at different times. The execution order of these sub-steps or stages is not necessarily to be performed in sequence, but may be executed in turn, alternately or synchronously with other steps or at least a portion of the sub-steps or stages of other steps.

[0174] Based on the above method embodiment, please refer to Figure 3Another embodiment of the present invention further provides a risk event identification device, wherein the device includes:

[0175] A collection module 11 is used to collect service call recordings of all customers within a target time period, pre-process each service call recording to obtain corresponding recording data, and convert all the recording data into corresponding text data;

[0176] An extraction module 12 is configured to extract voiceprint features of each recording data through a voiceprint model, and perform PLDA scoring on any two voiceprint features to obtain a scoring result;

[0177] A calculation module 13 is configured to perform voiceprint cluster analysis based on the scoring results, identify multiple feature clusters, each of which has multiple similar voiceprint features, and calculate the average of all the scoring results in each feature cluster to generate a corresponding voice risk value;

[0178] A selection module 14 is configured to select a target feature cluster based on each of the speech risk values ​​and a cluster screening strategy, and use the speech risk value of the target feature cluster as a speech risk factor;

[0179] A generating module 15, configured to generate a text risk factor corresponding to each of the text data using a large language model;

[0180] The identification module 16 is configured to identify whether the target service call recording is a risk event based on the voice risk factor and the text risk factor using a risk event identification model.

[0181] Furthermore, in one embodiment, the risk event identification device, wherein the collection module 11 is specifically configured to:

[0182] Determine the customer groups and target time periods for the service call recordings collected in the target financial institution based on the task objectives of the risk event identification task;

[0183] Obtaining the service call recordings of all customers in the customer group within the target time period from the customer service system of the target financial institution;

[0184] After denoising, enhancing and normalizing each of the service call recordings, corresponding recording data is obtained, and all of the recording data is converted into corresponding text data through automatic speech recognition technology.

[0185] Furthermore, in one embodiment, the risk event identification device, wherein the extraction module 12 is specifically configured to:

[0186] Pre-constructing the voiceprint model and the PLDA model, and inputting each of the recording data into the voiceprint model for feature extraction to generate voiceprint features of each of the recording data;

[0187] Any two of the voiceprint features are input into the PLDA model for PLDA scoring to generate a preliminary result, and the preliminary result is standardized to obtain a scoring result.

[0188] Furthermore, in one embodiment, the risk event identification device, wherein the calculation module 13 is specifically configured to:

[0189] Determine the clustering algorithm based on the task requirements of the risk event identification task;

[0190] Based on the clustering algorithm, voiceprint clustering is performed according to each of the scoring results, and each of the voiceprint features is grouped into a plurality of feature clusters, wherein the feature clusters have a plurality of similar voiceprint features;

[0191] The average value of all the scoring results in each feature cluster is calculated to generate a speech risk value corresponding to each feature cluster.

[0192] Furthermore, in one embodiment, the risk event identification device, wherein the selection module 14 is specifically configured to:

[0193] Determining a cluster screening strategy according to the task requirements of the risk event identification task;

[0194] Based on the cluster screening strategy, a target feature cluster is selected according to the speech risk value of each feature cluster, and the speech risk value of the target feature cluster is used as a speech risk factor.

[0195] Furthermore, in one embodiment, the risk event identification device, wherein the generating module 15 is specifically configured to:

[0196] Pre-building the large language model;

[0197] Each of the text data is processed by word segmentation, stop word removal, and part-of-speech tagging, and then input into the large language model;

[0198] The large language model performs text analysis on each of the text data after word segmentation, stop word removal, and part-of-speech tagging to obtain corresponding risk text data, and generates a text risk factor corresponding to each of the risk text data.

[0199] Furthermore, in one embodiment, the risk event identification device, wherein the identification module 16 is specifically configured to:

[0200] Pre-building the risk event identification model;

[0201] When the target service call recording has the corresponding voice risk factor and the text risk factor, integrating the voice risk factor and the text risk factor corresponding to the target service call recording to obtain a mixed risk factor for the target service call recording;

[0202] The mixed risk factor is input into the risk event identification model to generate a risk event identification result of the target service call recording.

[0203] It should be noted that, in the embodiment of the device of the present invention, the information interaction, execution process and other contents between the above modules are based on the same concept as the embodiment of the method of the present invention. Their specific functions and technical effects can be found in the aforementioned method embodiment part and will not be repeated here.

[0204] Based on the above method embodiment, another embodiment of the present invention further provides a computer device, which can be a server, and its internal structure diagram can be as follows: Figure 4 As shown. The computer device includes a processor, memory, a network interface, and a database connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When executed by the processor, the computer program implements the functions or steps of the risk event identification method service side of any of the above-mentioned method embodiments.

[0205] Based on the above method embodiment, another embodiment of the present invention further provides a computer device, which can be a client, and its internal structure diagram can be as follows: Figure 5 As shown. The computer device includes a processor, memory, a network interface, a display screen, and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When executed by the processor, the computer program implements the client-side functions or steps of the risk event identification method in any of the above-mentioned method embodiments.

[0206] Those skilled in the art will understand that Figure 4 and Figure 5The structural diagram shown in the figure is only a schematic diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more components than shown in the figure, or combine certain components, or have a different component arrangement.

[0207] The processor referred to herein may be a CPU, other general-purpose processors, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or any conventional processor, etc.

[0208] The memory includes a readable storage medium, an internal memory, etc., wherein the internal memory can be the memory of a computer device, and the internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The readable storage medium can be a hard disk of the computer device, and in other embodiments, it can also be an external storage device of the computer device, for example, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the computer device. Furthermore, the memory can also include both an internal storage unit of the computer device and an external storage device. The memory is used to store an operating system, an application program, a boot loader (BootLoader), data, and other programs, such as the program code of a computer program. The memory can also be used to temporarily store data that has been output or is about to be output.

[0209] Based on the above method embodiments, another embodiment of the present invention further provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the risk event identification method described in any of the above method embodiments. The computer-readable storage medium may be non-volatile or volatile.

[0210] It should be noted that the above-mentioned functions or steps that can be implemented by computer-readable storage media or computer devices, and the technical effects brought about by the functions / steps, can be found in the relevant descriptions in the aforementioned method embodiments. To avoid repetition, they will not be described one by one here.

[0211] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM). The disclosed memory components or memories of the operating environments described herein are intended to comprise one or more of these and / or any other suitable types of memory.

[0212] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, in the embodiment of the device of the present invention, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual application, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into a processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention. The specific working process of the units and modules in the above-mentioned device can refer to the corresponding process in the above-mentioned method embodiment, which will not be repeated here. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0213] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0214] In the embodiments provided by the present invention, it should be understood that the disclosed apparatus / computer equipment and methods can be implemented in other ways. For example, the apparatus / computer equipment embodiments described above are merely illustrative. For example, the division of modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0215] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0216] It should be noted that if software tools or components other than those of the Company appear in the embodiments of this application, they are merely for illustration and do not represent actual use. The above embodiments are intended only to illustrate the technical solutions of the present invention, not to limit them. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some of the technical features therein with equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. A risk event identification method, characterized in that: include: Collecting service call recordings of all customers within a target time period, pre-processing each service call recording to obtain corresponding recording data, and converting all the recording data into corresponding text data; Extracting voiceprint features of each recording data through a voiceprint model, and performing PLDA scoring on any two voiceprint features to obtain a scoring result; Performing voiceprint cluster analysis based on the scoring results to identify multiple feature clusters, each of which has multiple similar voiceprint features, and calculating the average of all the scoring results in each feature cluster to generate a corresponding voice risk value; Selecting a target feature cluster based on each of the speech risk values ​​and a cluster screening strategy, and using the speech risk value of the target feature cluster as a speech risk factor; Generating a text risk factor corresponding to each of the text data using a large language model; Based on the voice risk factor and the text risk factor, a risk event identification model is used to identify whether the target service call recording is a risk event.

2. The risk event identification method according to claim 1, characterized in that: The collecting of service call recordings of all customers within a target time period, pre-processing of each service call recording to obtain corresponding recording data, and converting all the recording data into corresponding text data includes: Determine the customer groups and target time periods for the service call recordings collected in the target financial institution based on the task objectives of the risk event identification task; Obtaining the service call recordings of all customers in the customer group within the target time period from the customer service system of the target financial institution; After denoising, enhancing and normalizing each of the service call recordings, corresponding recording data is obtained, and all of the recording data is converted into corresponding text data through automatic speech recognition technology.

3. The risk event identification method according to claim 1, characterized in that: The voiceprint features of each recording data are extracted by the voiceprint model, and any two voiceprint features are scored by PLDA to obtain a scoring result, including: Pre-constructing the voiceprint model and the PLDA model, and inputting each of the recording data into the voiceprint model for feature extraction to generate voiceprint features of each of the recording data; Any two of the voiceprint features are input into the PLDA model for PLDA scoring to generate a preliminary result, and the preliminary result is standardized to obtain a scoring result.

4. The risk event identification method according to claim 1, characterized in that: The voiceprint cluster analysis is performed based on the scoring results to identify multiple feature clusters, each of which has multiple similar voiceprint features, and the average of all the scoring results in each feature cluster is calculated to generate a corresponding voice risk value, including: Determine the clustering algorithm based on the task requirements of the risk event identification task; Based on the clustering algorithm, voiceprint clustering is performed according to each of the scoring results, and each of the voiceprint features is grouped into a plurality of feature clusters, wherein the feature clusters have a plurality of similar voiceprint features; The average value of all the scoring results in each feature cluster is calculated to generate a speech risk value corresponding to each feature cluster.

5. The risk event identification method according to claim 4, characterized in that: The step of selecting a target feature cluster based on each of the speech risk values ​​and a cluster screening strategy, and using the speech risk value of the target feature cluster as a speech risk factor, includes: Determining a cluster screening strategy according to the task requirements of the risk event identification task; Based on the cluster screening strategy, a target feature cluster is selected according to the speech risk value of each feature cluster, and the speech risk value of the target feature cluster is used as a speech risk factor.

6. The risk event identification method according to claim 1, characterized in that: Generating the text risk factor corresponding to each of the text data using the large language model includes: Pre-building the large language model; Each of the text data is processed by word segmentation, stop word removal, and part-of-speech tagging, and then input into the large language model; The large language model performs text analysis on each of the text data after word segmentation, stop word removal, and part-of-speech tagging to obtain corresponding risk text data, and generates a text risk factor corresponding to each of the risk text data.

7. The risk event identification method according to any one of claims 1 to 6, characterized in that: The identifying whether the target service call recording is a risk event by using a risk event identification model based on the voice risk factor and the text risk factor includes: Pre-building the risk event identification model; When the target service call recording has the corresponding voice risk factor and the text risk factor, integrating the voice risk factor and the text risk factor corresponding to the target service call recording to obtain a mixed risk factor for the target service call recording; The mixed risk factor is input into the risk event identification model to generate a risk event identification result of the target service call recording.

8. A risk event identification device, characterized in that: include: A collection module is used to collect service call recordings of all customers within a target time period, pre-process each service call recording to obtain corresponding recording data, and convert all the recording data into corresponding text data; An extraction module, configured to extract voiceprint features of each recording data through a voiceprint model, and perform PLDA scoring on any two voiceprint features to obtain a scoring result; a calculation module, configured to perform voiceprint cluster analysis based on the scoring results, identify multiple feature clusters, each of which has multiple similar voiceprint features, and calculate the average of all the scoring results in each feature cluster to generate a corresponding voice risk value; a selection module, configured to select a target feature cluster based on each of the speech risk values ​​and a cluster screening strategy, and use the speech risk value of the target feature cluster as a speech risk factor; A generation module, configured to generate a text risk factor corresponding to each of the text data using a large language model; The identification module is used to identify whether the target service call recording is a risk event based on the voice risk factor and the text risk factor through a risk event identification model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the risk event identification method according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the risk event identification method according to any one of claims 1 to 7 is implemented.

Images