Artificial Intelligence-Based Consumer Protection Method, Device, and Storage Medium

Abstract

The present application discloses an artificial intelligence-based consumer protection method, device, and storage medium. Among them, the method includes: obtaining multi-modal customer communication corpus data; extracting features from the customer communication corpus data to obtain target features, and determining the emotion type corresponding to the customer and the emotion intensity corresponding to the emotion type based on the target features; determining the complaint probability corresponding to the customer according to the customer's personal user portrait, emotion type, and emotion intensity, where the personal user portrait is at least used to characterize the customer's historical complaint preferences; and sending a complaint warning message to the responsible department corresponding to the target service when the complaint probability exceeds the probability threshold corresponding to the target service handled by the customer. The present application solves the technical problem that the early warning effect for customer complaints is poor and the consumer needs cannot be met due to the fact that complaint management in related technologies still remains in the manual judgment and processing mode.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and more specifically, to an artificial intelligence-based consumer protection method, device, and storage medium. Background Art

[0002] Customer complaint management is not only a core concern of regulatory authorities but also an important breakthrough for banks to enhance their competitiveness. However, currently, the complaint management platforms in related technologies face many challenges, with deficiencies in aspects such as customer service, process design, and internal management. For example, the complaint management in related technologies still remains in the mode of manual judgment and processing, with problems such as low accuracy in early warning of customer complaints and poor complaint handling effects, unable to meet the needs of consumers and effectively protect the rights and interests of consumers.

[0003] In response to the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] Embodiments of this application provide an artificial intelligence-based consumer protection method, device, and storage medium to at least solve the technical problem that the early warning effect of customer complaints is poor due to the fact that the complaint management in related technologies still remains in the manual judgment and processing mode and cannot meet the needs of consumers.

[0005] According to one aspect of the embodiments of this application, an artificial intelligence-based consumer protection method is provided, including: obtaining multi-modal customer communication corpus data, where the customer communication corpus data includes at least one of the following: text data, video data, and audio data of customers collected during the business handling process in each service channel; performing feature extraction on the customer communication corpus data to obtain target features, and based on the target features, determining the emotional type corresponding to the customer and the emotional intensity corresponding to the emotional type; determining the complaint probability corresponding to the customer based on the personal user profile of the customer, the emotional type, and the emotional intensity, where the personal user profile is at least used to represent the historical complaint preferences of the customer; and in the case where the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer, sending a complaint early warning message to the responsible department corresponding to the target business.

[0006] Optionally, performing feature extraction on the customer communication corpus data includes: performing word segmentation on the text data in the customer communication corpus data and determining the part of speech corresponding to each word obtained after word segmentation; performing syntactic analysis on the text data based on the part of speech corresponding to each word to obtain semantic structure information corresponding to the text data, where the semantic structure information is used to represent the context association relationship between the words in the text data; determining keywords in the text data based on the semantic structure information and mapping the keywords to a preset vector space to obtain a first feature corresponding to the text data, where the keywords are the words in the text data used to represent the customer's emotion.

[0007] Optionally, feature extraction is performed on the customer communication corpus data, and obtaining the target features further includes: determining key frames of the video data, and determining the face region and limb region in the key frames; determining facial key points in the face region, and determining limb key points in the limb region; determining the facial expression features of the customer based on the position change information of the facial key points in adjacent key frames, and determining the body language features of the customer based on the position change information of the limb key points in adjacent key frames; determining the second feature corresponding to the video data based on the facial expression features and body language features; and / or, extracting the third feature for characterizing the customer's emotion from the audio data of the customer communication corpus data, where the third feature includes: acoustic features and prosody features, the acoustic features include at least one of the following: fundamental frequency feature, energy feature, Mel frequency cepstral coefficient, and the prosody features include at least one of the following: speech rate feature, pause feature; determining the target feature based on the first feature, the second feature, and the third feature.

[0008] Optionally, determining the emotion type and emotion intensity corresponding to the customer based on the target feature includes: using an emotion classification model to determine the emotion type and emotion intensity based on the target feature, where the emotion classification model is trained based on a training data set, the training data set contains multiple training samples, and each training sample includes: historical customer communication corpus data, and an emotion label corresponding to the historical customer communication corpus data for characterizing the emotion type and emotion intensity, the emotion type includes at least one of the following: anger, sadness, joy, neutral, and the emotion intensity includes: slight, moderate, strong; determining the complaint probability corresponding to the customer based on the customer's personal user profile and emotion type and emotion intensity includes: obtaining relevant information of the customer, where the relevant information includes: personal basic information, historical business handling records, historical complaint records; determining the personal user profile corresponding to the customer based on the relevant information; using a complaint probability prediction model to determine the complaint probability based on the personal user profile and emotion type and emotion intensity, where the complaint probability is used to characterize the probability that the customer initiates a complaint under the emotion type and emotion intensity.

[0009] Optionally, when the complaint probability exceeds the probability threshold corresponding to the target service handled by the customer, sending a complaint warning message to the responsible department corresponding to the target service includes: obtaining the complaint record data of the target service within a preset historical time period, where the complaint record data includes at least one of the following: the number of complaints, the type of complaint, and the complaint handling result; determining the complaint trend curve and the complaint handling satisfaction corresponding to the target service based on the complaint record data, where the complaint trend curve is at least used to characterize the trend of the number of complaints received by the target service over time; updating the probability threshold corresponding to the target service based on the complaint trend curve and the complaint handling satisfaction, and determining whether there is a policy update related to the target service within the target time period, where the target time period is the time period from the first moment to the second moment, the second moment is the latest time when the probability threshold corresponding to the target service is updated, and the first moment is the previous time when the probability threshold corresponding to the target service is updated; when there is a policy update related to the target service within the target time period, determining the policy impact coefficient corresponding to the target service according to the latest updated policy, and adjusting the probability threshold corresponding to the target service according to the policy impact coefficient; when the complaint probability exceeds the adjusted probability threshold corresponding to the target service, generating a complaint warning message and sending the complaint warning message to the responsible department corresponding to the target service.

[0010] Optionally, the method further includes: when receiving the customer complaint information, determining the target service corresponding to the customer complaint information, and backtracking the complained target service to obtain the historical service handling information corresponding to the complained target service, where the customer complaint information includes at least one of the following: the complaint content, the complaint event, and the complaint object, and the historical service handling information includes: the time, location, operator, service process, and customer information of the service handling; determining the complaint text feature corresponding to the customer complaint information and the service handling feature corresponding to the historical service handling information, where the complaint text feature is at least used to characterize the complaint keywords and sentiment tendency of the customer, and the service handling feature is at least used to characterize the processing time and result of each link in the service process and the service attitude of the operator; performing attribution analysis based on the complaint text feature, the service handling feature, and the personal user portrait of the customer to obtain the analysis result corresponding to the customer complaint information, where the analysis result is used to characterize the cause of the complaint corresponding to the customer complaint information; determining the complaint handling process according to the analysis result, generating a complaint handling work order corresponding to the customer complaint information, and sending the complaint handling work order to the person in charge corresponding to the initial node in the complaint handling process.

[0011] Optionally, the method further includes: according to the analysis results, statistically analyzing the causes of various types of complaints for various target services, and based on the statistical results, generating and displaying a visualization chart, where the causes of complaints include at least one of the following: product problems, service attitude problems, and business process problems, and the visualization chart includes at least one of the following: a pie chart for representing the proportion of the causes of various types of complaints in the total number of complaints, a bar chart for representing the number of causes of different types of complaints, and a heat map for representing the distribution of different causes of complaints in different time periods or different regions; and / or, determining the timestamps corresponding to each process node in the complaint handling process, where the timestamps are used to represent the moment when the process node is executed and the processing duration of the customer complaint information at the process node in the complaint handling process; according to the timestamps, determining the customer complaint information processed by each process node during the selected time period, and obtaining a process heat map corresponding to the selected time period, where in the process heat map, the position of each process node is represented by different heat values to indicate the quantity of the customer complaint information processed by the process node and / or the length of the processing time of the customer complaint information at the process node.

[0012] According to another aspect of the embodiments of the present application, there is also provided an artificial intelligence-based consumer protection device, including: a data acquisition module, configured to acquire multi-modal customer communication corpus data, where the customer communication corpus data includes at least one of the following: text data, video data, and audio data of customers collected during the business handling process in each service channel; a feature extraction module, configured to extract features from the customer communication corpus data to obtain target features, and determine the emotional type corresponding to the customer and the emotional intensity corresponding to the emotional type according to the target features; a probability prediction module, configured to determine the complaint probability corresponding to the customer according to the personal user profile of the customer, the emotional type, and the emotional intensity, where the personal user profile is at least used to represent the historical complaint preferences of the customer; a complaint warning module, configured to send a complaint warning message to the responsible department corresponding to the target service when the complaint probability exceeds the probability threshold corresponding to the target service handled by the customer.

[0013] According to yet another aspect of the embodiments of the present application, there is also provided a non-volatile storage medium, where the non-volatile storage medium includes a stored computer program, and the device where the non-volatile storage medium is located executes the artificial intelligence-based consumer protection method by running the computer program.

[0014] According to still another aspect of the embodiments of the present application, there is also provided a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of the artificial intelligence-based consumer protection method are implemented.

[0015] In the embodiments of the present application, multi-modal customer communication corpus data is obtained, where the customer communication corpus data includes at least one of the following: text data, video data, and audio data of customers collected during the process of handling services through various service channels; feature extraction is performed on the customer communication corpus data to obtain target features, and based on the target features, the corresponding emotion type of the customer and the emotion intensity corresponding to the emotion type are determined; based on the personal user portrait, emotion type, and emotion intensity of the customer, the corresponding complaint probability of the customer is determined, where the personal user portrait is at least used to characterize the historical complaint preferences of the customer; in the case where the complaint probability exceeds the probability threshold corresponding to the target service handled by the customer, a complaint warning message is sent to the responsible department corresponding to the target service. By comprehensively analyzing the multi-modal fusion data, the purpose of accurately understanding the emotional state and complaint tendency of customers, pre-warning potential complaint risks, and effectively improving customer satisfaction and enterprise service quality is achieved, thereby solving the technical problem that the warning effect of customer complaints in related technologies remains in the manual judgment and processing mode and cannot meet the needs of consumers. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments and descriptions thereof of the present application are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0017] Figure 1 is a hardware structure block diagram of a computer terminal (or electronic device) for implementing a method for consumer protection based on artificial intelligence according to an embodiment of the present application;

[0018] Figure 2 is a schematic diagram of a method flow for consumer protection based on artificial intelligence according to an embodiment of the present application;

[0019] Figure 3 is a schematic diagram of a process for policy acquisition and analysis according to an embodiment of the present application;

[0020] Figure 4 is a schematic diagram of the structure of a consumer protection device based on artificial intelligence according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] To enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts shall fall within the scope of protection of this application.

[0022] It should be noted that the terms "first", "second", etc. in the description and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of this application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or are inherent to these processes, methods, products, or devices.

[0023] The data processing and analysis capabilities of the complaint management platform in the related art are lacking. One of the core values of the complaint management platform is to gain insights into service problems and improvement directions through data analysis. If the data processing capabilities of the platform are insufficient, such as lacking real-time data analysis capabilities, being difficult to support large-scale data processing, or having poor data visualization effects, it will limit its decision-making support role. In addition, the complaint management platform in the related art lacks intelligent support and still stays in the traditional manual processing mode, lacking the support of intelligent technologies such as natural language processing, intelligent recommendation, decision analysis, and intelligent warning, and it is difficult to meet the automatic classification, rapid response, and accurate decision-making of business complaints.

[0024] To solve the above problems, relevant solutions are provided in the embodiments of this application, which will be described in detail below.

[0025] According to the embodiments of this application, a method embodiment for consumer protection based on artificial intelligence is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described here can be executed in a different order than here.

[0026] The method embodiments provided by the embodiments of this application can be executed on a mobile terminal, a computer terminal, or a similar computing device. Figure 1The following shows a hardware block diagram of a computer terminal (or electronic device) for implementing an artificial intelligence-based consumer protection method. As Figure 1 shown, the computer terminal 10 (or electronic device) may include one or more processors 102 (illustrated as 102a, 102b, ……, 102n in the figure) (the processor 102 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply, and / or a camera. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the computer terminal 10 may further include more or fewer components than Figure 1 shown therein, or have a different configuration from Figure 1 shown.

[0027] It should be noted that the above one or more processors 102 and / or other artificial intelligence-based consumer protection circuits may generally be referred to as "artificial intelligence-based consumer protection circuits" herein. The artificial intelligence-based consumer protection circuit may be embodied in whole or in part as software, hardware, firmware, or any other combination. In addition, the artificial intelligence-based consumer protection circuit may be a single independent processing module, or be incorporated in whole or in part into any one of other elements in the computer terminal 10 (or electronic device). As involved in the embodiments of the present application, the artificial intelligence-based consumer protection circuit is a processor control (such as the selection of a variable resistance terminal path connected to an interface).

[0028] The memory 104 may be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the artificial intelligence-based consumer protection method in the embodiments of the present application. The processor 102 executes various functional applications and artificial intelligence-based consumer protection by running the software programs and modules stored in the memory 104, that is, implements the above-mentioned artificial intelligence-based consumer protection method. The memory 104 may include a high-speed random access memory, and may further include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories may be connected to the computer terminal 10 through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0029] The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can be connected to other network devices through a base station to communicate with the Internet. In one example, the transmission device 106 can be a Radio Frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0030] The display can be, for example, a touch-screen liquid crystal display (LCD), which enables users to interact with the user interface of the computer terminal 10 (or electronic device).

[0031] Under the above operating environment, the embodiments of the present application provide a method for consumer protection based on artificial intelligence. Figure 2 It is a schematic diagram of the method flow of a method for consumer protection based on artificial intelligence provided by the embodiments of the present application, as Figure 2 shown, the method includes the following steps:

[0032] Step S202, obtain multi-modal customer communication corpus data, where the customer communication corpus data includes at least one of the following: text data, video data, and audio data of customers collected during the process of handling business in each service channel;

[0033] Step S204, extract features from the customer communication corpus data to obtain target features, and based on the target features, determine the emotional type corresponding to the customer and the emotional intensity corresponding to the emotional type;

[0034] Step S206, determine the complaint probability corresponding to the customer based on the personal user portrait of the customer, the emotional type, and the emotional intensity, where the personal user portrait is at least used to characterize the historical complaint preferences of the customer;

[0035] Step S208, when the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer, send a complaint warning message to the responsible department corresponding to the target business.

[0036] Through the above steps, through the comprehensive analysis of multi-modal fusion data, the purpose of accurately understanding the emotional state and complaint tendency of customers, early warning of potential complaint risks, effectively improving customer satisfaction and enterprise service quality is achieved, and thus the technical problem that the early warning effect of customer complaints is poor due to the fact that the complaint management in related technologies still stays in the manual judgment and processing mode and cannot meet the needs of consumers is solved.

[0037] The following further introduces the artificial intelligence-based consumer protection method in steps S202 to S208 of the embodiments of the present application.

[0038] First, in the embodiments of the present application, feedback information of customers can be collected through various service channels of the bank (such as online banking, mobile banking, telephone banking, self-service terminals, branch counters, etc.), including text data such as customers' inquiries, feedback, suggestions, complaints, etc. when handling business, as well as video and audio data when customers handle business at the branch (for example, audio data when customers communicate with customer service by phone), etc., to obtain customer communication corpus data.

[0039] For example, through data acquisition interfaces, it can be docked with the bank's customer service system, social media monitoring system, telephone service center system, etc. respectively to automatically capture customers' communication records, such as online chat records, social media comments, call recording videos, etc., for subsequent analysis.

[0040] After that, feature extraction can be performed on the multi-modal customer communication corpus data, and the specific steps are as follows.

[0041] In some embodiments of the present application, feature extraction of customer communication corpus data includes the following steps: performing word segmentation on the text data in the customer communication corpus data and determining the part of speech corresponding to each word obtained after word segmentation; performing syntactic analysis on the text data based on the part of speech corresponding to each word to obtain semantic structure information corresponding to the text data, where the semantic structure information is used to represent the context association relationship between the words in the text data; determining keywords in the text data based on the semantic structure information and mapping the keywords to a preset vector space to obtain the first feature corresponding to the text data, where the keywords are words in the text data used to represent customers' emotions.

[0042] In some embodiments of the present application, feature extraction of customer communication corpus data to obtain the target feature further includes the following steps: determining the key frames of the video data and determining the face region and limb region in the key frames; determining the facial key points in the face region and determining the limb key points in the limb region; determining the facial expression feature of the customer based on the position change information of the facial key points in adjacent key frames, and determining the body language feature of the customer based on the position change information of the limb key points in adjacent key frames; determining the second feature corresponding to the video data based on the facial expression feature and the body language feature; and / or extracting the third feature used to represent customers' emotions in the audio data of the customer communication corpus data, where the third feature includes: acoustic features and prosodic features, and the acoustic features include at least one of the following: fundamental frequency feature, energy feature, Mel frequency cepstral coefficient, and the prosodic features include at least one of the following: speech rate feature, pause feature; determining the target feature based on the first feature, the second feature, and the third feature.

[0043] Specifically, for the text data in the customer communication corpus data, the text data can be cleaned first to remove irrelevant characters, noise information, etc. to ensure the accuracy and consistency of the data. The text is then processed by word segmentation and stop word removal to prepare for subsequent analysis. After that, natural language processing technology is used in combination with sentiment dictionaries to identify keywords such as sentiment words and intensity adverbs in the text, and pre-trained word embedding models (such as Word2Vec, BERT, etc.) are used to convert the text data into word vector representations, extract the semantic features of the keywords used to represent customer sentiment in the text, and obtain the first feature corresponding to the text data.

[0044] For example, the system will use the word segmentation tool to divide the text into vocabulary units, and then determine the grammatical role of each word through part-of-speech tagging, such as subject, predicate, object, etc. Then, use syntactic analysis technology to parse the sentence structure and understand the logical relationship between words. Through the mapping of sentiment words, the keywords are converted into vector representations to facilitate subsequent machine learning model processing. For example, the system identifies "bad service attitude" as a negative sentiment keyword and converts it into a vector for subsequent sentiment analysis.

[0045] For the video data in the customer communication corpus data, the facial expressions and body language in the video will be analyzed. Specifically, key frame images are first extracted from the video file. Frames with large motion changes can be extracted at fixed time intervals (such as one frame per second) or through motion detection algorithms. The extracted video frame images are cropped, scaled, normalized, and other processes are performed to facilitate subsequent feature extraction. After that, facial expression features can be extracted from the pre-processed video key frames, and face detection algorithms (such as Haar feature classifiers, etc.) can be used to locate the face area in the video frame images. Use a deep learning model to detect facial key points (such as eyebrows, eyes, nose, corners of the mouth, etc.) in the face area, and extract facial expression features based on the position changes of facial key points, such as the movement amplitude of the eyebrows, the degree of opening and closing of the eyes, the upward or downward movement of the corners of the mouth, etc., to obtain facial expression features; and, you can also use posture estimation algorithms (such as OpenPose, AlphaPose, etc.) to detect the body key points of the characters in the video, such as shoulders, elbows, wrists, knees, etc., and extract body movement features based on the movement trajectory and relative position changes of the body key points, such as the swing amplitude of the arms, the swing frequency of the legs, etc., to obtain body language features, so as to obtain the second feature corresponding to the video data based on the customer's facial expression features and body language features.

[0046] For example, the system will determine the key frames of the video and analyze the faces and body movements therein. Through facial expression recognition and body language analysis, the system can capture the emotional changes and non-verbal signals of the customer. For example, if the system recognizes that the customer frowns, shakes their head, etc. in the video, it can be judged that the customer may be dissatisfied with the service.

[0047] In addition, as an optional implementation manner, when analyzing video data, scene features can also be extracted. For example, the background environment in the video can be analyzed to extract scene features such as the layout of a bank branch, the relative positions of customers and staff, etc.; or object detection algorithms can be used to detect relevant objects in the video, such as counters, ATMs, seats, etc. The relative position changes between the customer and the staff or the facilities in the scene are used to assist in judging the customer's emotions.

[0048] For the audio data in the customer communication corpus data, it can be analyzed by identifying features such as speech rate, pitch, and emotional vocabulary. Specifically, first, the collected audio signal can be denoised to remove background noise, improve the clarity of the speech signal, and segment the continuous speech signal into multiple short-time speech segments for subsequent feature extraction. Then, acoustic feature extraction can be performed on the preprocessed audio data, including: using methods such as the autocorrelation method and the cepstrum method to extract the fundamental frequency (pitch) feature of the speech signal (the fundamental frequency under different emotional states will have obvious differences. For example, the fundamental frequency is usually higher when angry and lower when sad); calculating the short-time energy feature of the speech signal to reflect the intensity of the speech signal (indicating the loudness or intensity of the sound, usually related to the excitement level of the speaker. High energy may indicate excitement or anger, while low energy may indicate sadness or calm); extracting the Mel-frequency cepstral coefficient feature to capture the spectral information of the speech signal, etc.; in addition, prosody feature extraction can also be performed on the preprocessed audio data, including: calculating the speech rate of the speech signal, that is, the length of the speech per unit time, which can reflect the tension level or emotional state of the speaker. For example, the speech rate is usually faster when angry or excited, and slower when sad or contemplative; detecting the pause points in the speech signal and extracting the duration and frequency of the pauses (which can reflect the thinking process and emotional changes of the speaker. Frequent pauses may indicate hesitation or uncertainty, while long pauses may indicate silence or emotional suppression), etc., and finally obtaining the third feature corresponding to the audio data.

[0049] For example, the system will extract acoustic features and prosody features from the audio, such as fundamental frequency, energy, speech rate, and pauses, etc., which can reflect the emotional state of the customer. For example, identifying a high fundamental frequency, a fast speech rate, and frequent pauses from a call recording may indicate that the customer is in a tense or angry state.

[0050] Afterwards, based on the first feature, the second feature, and the third feature, the final target feature can be determined. For example, the text, video, and audio feature vectors can be concatenated to form a comprehensive feature vector; or different weights can be assigned to the features of different modalities and then weighted summation is performed to obtain the fused feature. The weights can be adjusted according to the importance of each modality feature for the sentiment analysis task.

[0051] After obtaining the target feature corresponding to the customer communication corpus data, the sentiment type corresponding to the customer and the sentiment intensity corresponding to the sentiment type can be determined based on the target feature. The specific steps are as follows.

[0052] In some embodiments of the present application, determining the sentiment type corresponding to the customer and the sentiment intensity based on the target feature includes the following steps: using a sentiment classification model to determine the sentiment type and the sentiment intensity based on the target feature, where the sentiment classification model is trained based on a training data set, and the training data set contains multiple training samples. Each training sample includes: historical customer communication corpus data, and a sentiment label corresponding to the historical customer communication corpus data for representing the sentiment type and the sentiment intensity. The sentiment type includes at least one of the following: anger, sadness, joy, neutral, and the sentiment intensity includes: slight, moderate, strong;

[0053] Specifically, a multi-modal sentiment classification model can be constructed, and the fused target feature is used to classify the sentiment type of the customer and quantitatively analyze the sentiment degree of the customer. In this embodiment, a deep learning model (such as Transformer, etc.) can be used for training. The model can identify the sentiment type of the customer, such as anger, sadness, joy, neutral, etc., and classify the sentiment intensity into different levels, such as slight, medium, strong, etc. Through the deep learning method, the system can automatically learn the feature representations under different sentiment types and improve the accuracy of classification.

[0054] Afterwards, based on the personal user portrait of the customer, the sentiment type, and the sentiment intensity, the complaint probability corresponding to the customer is determined. The specific steps are as follows.

[0055] In some embodiments of the present application, determining the complaint probability corresponding to the customer based on the personal user portrait of the customer, the sentiment type, and the sentiment intensity includes the following steps: obtaining the relevant information of the customer, where the relevant information includes: personal basic information, historical business handling records, historical complaint records; determining the personal user portrait corresponding to the customer based on the relevant information; using a complaint probability prediction model to determine the complaint probability based on the personal user portrait, the sentiment type, and the sentiment intensity, where the complaint probability is used to represent the probability that the customer initiates a complaint under the sentiment type and the sentiment intensity.

[0056] Specifically, basic information of customers (such as age, gender, occupation, etc.), historical transaction records, historical complaint records, service preferences and other data can be collected to construct a personal user profile of the customers. Specifically as follows: Encode basic information of customers such as age, gender, occupation, etc., such as using one-hot encoding or label encoding; Extract key features from the historical transaction records of customers, such as transaction frequency, transaction amount, transaction type, etc.; Extract features such as complaint frequency, complaint type, and complaint handling satisfaction from the historical complaint records; Analyze the service preferences of customers, such as preferred service channels, service times, etc.

[0057] Then, a complaint probability prediction model can be constructed by combining the personal user profile, emotion type and emotion intensity of the customer, as well as information such as the business type of the customer and historical complaint records. Machine learning algorithms (such as random forest, support vector machine, etc.) or deep learning models (such as neural network) can be used for training, so that the model can predict the probability of customer complaints according to the emotional state and business situation of the customer. For example, if a customer shows moderate dissatisfaction in a recent transaction and there are similar complaints in the historical records, the model will predict a higher probability of their complaint.

[0058] In the case where the predicted customer complaint probability exceeds the corresponding probability threshold, trigger sending complaint threshold information to the responsible department corresponding to the target business, and the specific steps are as follows.

[0059] In some embodiments of the present application, when the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer, sending a complaint warning information to the responsible department corresponding to the target business includes the following steps: Obtain the complaint record data of the target business within a preset historical time period length, where the complaint record data includes at least one of the following: the number of complaints, complaint types, and complaint handling results; Determine the complaint trend curve and complaint handling satisfaction corresponding to the target business according to the complaint record data, where the complaint trend curve is at least used to characterize the trend of the number of complaints received by the target business changing with time; Update the probability threshold corresponding to the target business according to the complaint trend curve and complaint handling satisfaction, and determine whether there is a policy update related to the target business within the target time period, where the target time period is the time period from the first moment to the second moment, the second moment is the latest moment when the probability threshold corresponding to the target business is updated, and the first moment is the previous moment when the probability threshold corresponding to the target business is updated; In the case where there is a policy update related to the target business within the target time period, determine the policy impact coefficient corresponding to the target business according to the latest updated policy, and adjust the probability threshold corresponding to the target business according to the policy impact coefficient; In the case where the complaint probability exceeds the adjusted probability threshold corresponding to the target business, generate a complaint warning information and send the complaint warning information to the responsible department corresponding to the target business.

[0060] Specifically, in this embodiment, corresponding complaint probability thresholds can be set according to the characteristics of different business types of the target business and historical complaint data. For example, for high-risk businesses (such as loan approvals, large transfers, etc.), relatively low complaint probability thresholds can be set for timely early warning. At the same time, based on historical data, analyze the complaint trend to provide data support for complaint early warning. Specifically, the complaint record data of each target business within a certain historical period can be collected, including information such as the number of complaints, complaint types, and complaint handling results. Statistically analyze the data, calculate indicators such as the complaint frequency and complaint handling satisfaction of each business, and analyze the time series characteristics of the complaint data by establishing a complaint trend monitoring model to identify the upward or downward trend of the number of complaints. For example, time series analysis methods can be used to model and predict the number of complaints, and the complaint probability threshold of each target business can be dynamically adjusted according to the complaint trend monitoring results and complaint handling satisfaction. When the number of complaints of a certain business shows an upward trend or the complaint handling satisfaction is low, the complaint probability threshold of this business can be increased to more strictly monitor the complaint risk.

[0061] In addition, if there is a policy update within a specific time period, the system will consider the policy impact coefficient and adjust the probability threshold to adapt to the changes in the business environment for a more accurate assessment of complaint risks. Specifically, the latest financial policy release information can be tracked and collected in real time. Use natural language processing technology to parse the policy text, extract information such as keywords related to banking business and policy change points. Analyze the impact degree of financial policies on different banking businesses and construct a business impact assessment model. The model can evaluate the potential risks and complaint probabilities of policies on each target business according to factors such as policy change points, business types, and historical business data. Dynamically adjust the complaint probability threshold corresponding to each target business according to the business impact assessment results. For example, when the regulatory requirements for the loan business increase due to a certain financial policy, the complaint probability threshold for the loan business can be appropriately reduced to earlier warn of potential complaint risks.

[0062] For example, as Figure 3 shown, it is possible to analyze and calculate in real time by crawling the content of relevant regulatory websites, use big data analysis to predict potential complaint hotspots and trends, and discover unexpected events (such as policy adjustments) that may lead to concentrated complaints, so that relevant staff can make relevant preparations for the resolution of the events. For example, the adjustment of relevant mortgage policies may lead to a large complaint outbreak period for existing mortgage users.

[0063] On the other hand, for the complaints that have occurred, the system will automatically identify the complaint content, determine the relevant business of the complaint, trace back the historical business handling information, and conduct attribution analysis. The specific steps are as follows.

[0064] In some embodiments of the present application, the method further includes the following steps: in the case of receiving customer complaint information, determining the target service corresponding to the customer complaint information, and tracing back the target service being complained about to obtain historical service handling information corresponding to the target service being complained about, wherein the customer complaint information includes at least one of the following: complaint content, complaint event, complaint object, and the historical service handling information includes: time of service handling, location, handler, service process, customer information; determining the complaint text features corresponding to the customer complaint information and the service handling features corresponding to the historical service handling information, wherein the complaint text features are at least used to characterize the complaint keywords and sentiment tendency of the customer, and the service handling features are at least used to characterize the processing time and result of each link in the service process and the service attitude of the handler; performing attribution analysis based on the complaint text features, service handling features, and the personal user profile of the customer to obtain an analysis result corresponding to the customer complaint information, wherein the analysis result is used to characterize the cause of the complaint corresponding to the customer complaint information; determining a complaint handling process based on the analysis result, generating a complaint handling work order corresponding to the customer complaint information, and sending the complaint handling work order to the person in charge corresponding to the initial node in the complaint handling process.

[0065] For example, if a customer complains about a credit card bill problem, the system will automatically retrieve all service process records related to the credit card, analyze the root cause of the problem, such as errors in the service process or the attitude of the service staff. Based on the analysis result, the system will automatically initiate the complaint handling process, generate a work order, assign it to the corresponding processing personnel, and track the handling progress.

[0066] In addition, it is also possible to generate visual charts such as pie charts, bar charts, and heat maps by counting the causes of different types of complaints, so that the management can intuitively understand the complaint distribution, as follows.

[0067] In some embodiments of the present application, the method further includes: according to the analysis results, statistically analyzing the causes of various complaints for various target services, and generating and displaying a visualization chart based on the statistical results, where the causes of complaints include at least one of the following: product problems, service attitude problems, and business process problems, and the visualization chart includes at least one of the following: a pie chart for characterizing the proportion of the causes of various complaints in the total number of complaints, a bar chart for characterizing the number of causes of different types of complaints, and a heat map for characterizing the distribution of different causes of complaints in different time periods or different regions; and / or determining the timestamps corresponding to each process node in the complaint handling process, where the timestamps are used to characterize the moment when the process node is executed and the processing duration of the customer complaint information at the process node in the complaint handling process; according to the timestamps, determining the customer complaint information processed by each process node during the selected time period, and obtaining a process heat map corresponding to the selected time period, where in the process heat map, the position of each process node is characterized by different heat values to represent the quantity of the customer complaint information processed by the process node and / or the length of the processing time of the customer complaint information at the process node.

[0068] For example, the pie chart can show the proportion of service attitude problems, business process problems, and product problems in the total number of complaints; the bar chart can display the number of different types of complaints; the heat map can reveal the distribution of complaints at different times and in different regions. In addition, the system will also record and analyze the timestamps in the complaint handling process to generate a process heat map to evaluate the processing efficiency and the number of complaints at each node, helping to optimize the process, improve the processing speed, and enhance customer satisfaction.

[0069] Through the comprehensive analysis of multi-modal data, the solution of the present application can more accurately understand the emotional state and complaint tendency of customers, thereby early warning potential complaint risks, effectively enhancing customer satisfaction and enterprise service quality. It can timely adjust business processes and policies to prevent complaints from occurring. At the same time, after a complaint occurs, it can quickly locate the problem, improve the complaint handling efficiency, continuously improve the prevention, interception, and handling mechanisms for customer complaints, and ensure the accurate control of the entire process of complaint events from pre-warning, in-process response to post-improvement, enhancing the professionalism and efficiency of consumer protection work. At the same time, using big data analysis technology, deeply mining and analyzing complaint data to identify potential problems and trends. In a data-driven manner, it helps enterprises build a more efficient and user-friendly complaint management system, reduce operating costs, and enhance customer trust and loyalty.

[0070] According to an embodiment of the present application, an embodiment of a consumer protection device based on artificial intelligence is further provided. Figure 4 is a schematic structural diagram of a consumer protection device based on artificial intelligence provided according to an embodiment of the present application. As Figure 4 shown, the device includes:

[0071] A data acquisition module 40, configured to acquire multi-modal customer communication corpus data, where the customer communication corpus data includes at least one of the following: text data, video data, and audio data of customers collected during the business handling process in each service channel;

[0072] A feature extraction module 42, configured to extract features from the customer communication corpus data to obtain target features, and determine the emotion type corresponding to the customer and the emotion intensity corresponding to the emotion type based on the target features;

[0073] A probability prediction module 44, configured to determine the complaint probability corresponding to the customer based on the personal user profile of the customer, the emotion type, and the emotion intensity, where the personal user profile is at least used to characterize the historical complaint preference of the customer;

[0074] A complaint warning module 46, configured to send a complaint warning message to the responsible department corresponding to the target business when the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer.

[0075] Optionally, extracting features from the customer communication corpus data includes: performing word segmentation on the text data in the customer communication corpus data, and determining the part of speech corresponding to each word obtained after word segmentation; performing syntactic analysis on the text data based on the part of speech corresponding to each word to obtain semantic structure information corresponding to the text data, where the semantic structure information is used to characterize the context association relationship between the words in the text data; determining keywords in the text data based on the semantic structure information, and mapping the keywords to a preset vector space to obtain a first feature corresponding to the text data, where the keywords are words in the text data used to characterize the customer's emotion.

[0076] Optionally, extracting features from the customer communication corpus data to obtain target features further includes: determining key frames of the video data, and determining the face region and limb region in the key frames; determining facial key points in the face region, and determining limb key points in the limb region; determining the facial expression features of the customer based on the position change information of the facial key points in adjacent key frames, and determining the body language features of the customer based on the position change information of the limb key points in adjacent key frames; determining a second feature corresponding to the video data based on the facial expression features and the body language features; and / or, extracting a third feature used to characterize the customer's emotion from the audio data of the customer communication corpus data, where the third feature includes: acoustic features and prosody features, and the acoustic features include at least one of the following: fundamental frequency feature, energy feature, Mel frequency cepstral coefficient, and the prosody features include at least one of the following: speech rate feature, pause feature; determining the target features based on the first feature, the second feature, and the third feature.

[0077] Optionally, determining the emotional type and intensity corresponding to the customer based on the target features includes: using an emotion classification model to determine the emotional type and intensity based on the target features, where the emotion classification model is trained based on a training dataset. The training dataset contains multiple training samples, and each training sample includes: historical customer communication corpus data and an emotion label corresponding to the historical customer communication corpus data for representing the emotional type and intensity. The emotional type includes at least one of the following: anger, sadness, joy, neutral, and the emotional intensity includes: slight, moderate, strong; determining the complaint probability corresponding to the customer based on the customer's personal user profile and emotional type and intensity includes: obtaining relevant information of the customer, where the relevant information includes: personal basic information, historical business handling records, historical complaint records; determining the personal user profile corresponding to the customer based on the relevant information; using a complaint probability prediction model to determine the complaint probability based on the personal user profile and emotional type and intensity, where the complaint probability is used to represent the probability that the customer initiates a complaint under the emotional type and intensity.

[0078] Optionally, in the case where the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer, sending a complaint warning message to the responsible department corresponding to the target business includes: obtaining the complaint record data of the target business within a preset historical time period length, where the complaint record data includes at least one of the following: the number of complaints, complaint types, complaint handling results; determining the complaint trend curve and complaint handling satisfaction corresponding to the target business based on the complaint record data, where the complaint trend curve is at least used to represent the trend of the number of complaints received by the target business changing over time; updating the probability threshold corresponding to the target business based on the complaint trend curve and complaint handling satisfaction, and determining whether there is a policy update related to the target business within the target time period, where the target time period is the time period from the first moment to the second moment, the second moment is the latest moment when the probability threshold corresponding to the target business is updated, and the first moment is the previous moment when the probability threshold corresponding to the target business is updated; in the case where there is a policy update related to the target business within the target time period, determining the policy impact coefficient corresponding to the target business based on the latest updated policy, and adjusting the probability threshold corresponding to the target business based on the policy impact coefficient; in the case where the complaint probability exceeds the adjusted probability threshold corresponding to the target business, generating a complaint warning message and sending the complaint warning message to the responsible department corresponding to the target business.

[0079] Optionally, the artificial intelligence-based consumer protection device is further configured to: in the case of receiving customer complaint information, determine the target service corresponding to the customer complaint information, and trace back the target service being complained about to obtain the historical service handling information corresponding to the target service being complained about, where the customer complaint information includes at least one of the following: complaint content, complaint event, and complaint object, and the historical service handling information includes: the time, location, handler, service process, and customer information of the service handling; determine the complaint text features corresponding to the customer complaint information and the service handling features corresponding to the historical service handling information, where the complaint text features are at least used to represent the complaint keywords and sentiment tendency of the customer, and the service handling features are at least used to represent the processing time and result of each link in the service process, as well as the service attitude of the handler; perform attribution analysis based on the complaint text features, service handling features, and the personal user profile of the customer to obtain the analysis result corresponding to the customer complaint information, where the analysis result is used to represent the cause of the complaint corresponding to the customer complaint information; based on the analysis result, determine the complaint handling process, generate a complaint handling work order corresponding to the customer complaint information, and send the complaint handling work order to the person in charge corresponding to the initial node in the complaint handling process.

[0080] Optionally, the artificial intelligence-based consumer protection device is further configured to: based on the analysis result, count the causes of various complaints for various target services, and generate a visual chart for display based on the statistical result, where the causes of complaints include at least one of the following: product problems, service attitude problems, and service process problems, and the visual chart includes at least one of the following: a pie chart used to represent the proportion of various causes of complaints in the total number of complaints, a bar chart used to represent the number of different types of causes of complaints, and a heat map used to represent the distribution of different causes of complaints in different time periods or different regions; and / or, determine the timestamps corresponding to each process node in the complaint handling process, where the timestamps are used to represent the moment when the process node is executed and the processing duration of the customer complaint information at the process node in the complaint handling process; based on the timestamps, determine the customer complaint information processed by each process node during the selected time period to obtain a process heat map corresponding to the selected time period, where in the process heat map, the position of each process node is represented by different heat values to indicate the quantity and / or the processing time length of the customer complaint information processed by the process node.

[0081] It should be noted that each module in the above artificial intelligence-based consumer protection device may be a program module (for example, a set of program instructions that implement a specific function), or a hardware module. For the latter, it may be presented in the following forms, but not limited to: the manifestation form of each of the above modules is a processor, or the functions of each of the above modules are implemented by a processor.

[0082] It should be noted that the artificial intelligence-based consumer protection device provided in this embodiment can be used to execute Figure 2 the artificial intelligence-based consumer protection method shown. Therefore, the relevant explanations of the above artificial intelligence-based consumer protection method also apply to the embodiments of this application and will not be elaborated here.

[0083] The embodiments of this application also provide a non-volatile storage medium, which includes a stored computer program. Among them, the device where the non-volatile storage medium is located executes the following artificial intelligence-based consumer protection method by running the computer program: obtaining multi-modal customer communication corpus data, where the customer communication corpus data includes at least one of the following: text data, video data, and audio data of customers collected during the business handling process in each service channel; extracting features from the customer communication corpus data to obtain target features, and based on the target features, determining the emotional type corresponding to the customer and the emotional intensity corresponding to the emotional type; determining the complaint probability corresponding to the customer based on the personal user profile of the customer, the emotional type, and the emotional intensity, where the personal user profile is at least used to characterize the historical complaint preferences of the customer; and sending a complaint warning message to the responsible department corresponding to the target business when the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer.

[0084] The embodiments of this application also provide a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the steps of the artificial intelligence-based consumer protection method described in each embodiment of this application: obtaining multi-modal customer communication corpus data, where the customer communication corpus data includes at least one of the following: text data, video data, and audio data of customers collected during the business handling process in each service channel; extracting features from the customer communication corpus data to obtain target features, and based on the target features, determining the emotional type corresponding to the customer and the emotional intensity corresponding to the emotional type; determining the complaint probability corresponding to the customer based on the personal user profile of the customer, the emotional type, and the emotional intensity, where the personal user profile is at least used to characterize the historical complaint preferences of the customer; and sending a complaint warning message to the responsible department corresponding to the target business when the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer.

[0085] The serial numbers of the above embodiments of this application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0086] In the above embodiments of this application, the descriptions of each embodiment have their own focuses. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0087] In several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, 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 displayed or discussed couplings or direct couplings or communication connections between each other can be through some interfaces. The indirect couplings or communication connections of units or modules can be in electrical or other forms.

[0088] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0089] In addition, in each embodiment of the present application, the functional units can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0090] If the above-mentioned 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. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present application. The foregoing storage medium includes: USB flash drives, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), mobile hard disks, magnetic disks, or optical discs and other various media that can store program codes.

[0091] The above is only the preferred embodiment of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A consumer protection method based on artificial intelligence, characterized in that: include: Acquire multimodal customer communication corpus data, wherein the customer communication corpus data includes at least one of the following: text data, video data, and audio data collected from customers during business handling in various service channels; Performing feature extraction on the customer communication corpus data to obtain target features, including: determining key frames of the video data, and determining face areas and limb areas in the key frames; determining facial key points in the face area, and determining limb key points in the limb area; determining facial expression features of the customer based on position change information of the facial key points in adjacent key frames, and determining body language features of the customer based on position change information of the limb key points in adjacent key frames; determining a second feature corresponding to the video data based on the facial expression features and the body language features, wherein the second feature is used to determine the target feature; and determining the emotion type corresponding to the customer and the emotion intensity corresponding to the emotion type according to the target feature; Determining the complaint probability corresponding to the customer based on the customer's personal user portrait and the emotion type and emotion intensity, wherein the personal user portrait is at least used to characterize the customer's historical complaint preference; In the case where the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer, sending complaint warning information to the responsible department corresponding to the target business includes: obtaining complaint record data of the target business within a preset historical time period, wherein the complaint record data includes at least one of the following: the number of complaints, the type of complaint, and the result of complaint handling; determining the complaint trend curve and complaint handling satisfaction corresponding to the target business based on the complaint record data, wherein the complaint trend curve is at least used to characterize the trend of the number of complaints received by the target business over time; updating the probability threshold corresponding to the target business based on the complaint trend curve and the complaint handling satisfaction, and judging whether there is a complaint related to the target business within the target time period. Related policy updates, wherein the target time period is the time period from the first moment to the second moment, the second moment is the moment when the probability threshold corresponding to the target business is most recently updated, and the first moment is the moment when the probability threshold corresponding to the target business is last updated; in the case where there is a policy update related to the target business within the target time period, the policy impact coefficient corresponding to the target business is determined based on the latest updated policy, and the probability threshold corresponding to the target business is adjusted based on the policy impact coefficient; in the case where the complaint probability exceeds the adjusted probability threshold corresponding to the target business, the complaint warning information is generated and sent to the responsible department corresponding to the target business.

2. The artificial intelligence-based consumer protection method according to claim 1, characterized in that: Extracting features from the customer communication corpus data includes: Performing word segmentation on the text data in the customer communication corpus data, and determining the part of speech corresponding to each word obtained after the word segmentation; According to the part of speech corresponding to each word, the text data is subjected to syntactic analysis to obtain semantic structure information corresponding to the text data, wherein the semantic structure information is used to characterize the contextual association relationship between the words in the text data; Based on the semantic structure information, keywords in the text data are determined, and the keywords are mapped to a preset vector space to obtain a first feature corresponding to the text data, wherein the keywords are words in the text data used to characterize customer emotions.

3. The artificial intelligence-based consumer protection method according to claim 2, characterized in that: Extracting features from the customer communication corpus data to obtain target features also includes: Extracting a third feature for characterizing customer emotions from the audio data of the customer communication corpus data, wherein the third feature includes: an acoustic feature and a prosodic feature, the acoustic feature includes at least one of the following: a fundamental frequency feature, an energy feature, and a Mel frequency cepstrum coefficient, and the prosodic feature includes at least one of the following: a speech rate feature and a pause feature; The target feature is determined based on the first feature, the second feature and the third feature.

4. The artificial intelligence-based consumer protection method according to claim 1, characterized in that: Determining the emotion type and emotion intensity corresponding to the customer according to the target feature includes: using an emotion classification model to determine the emotion type and emotion intensity according to the target feature, wherein the emotion classification model is obtained by training based on a training data set, the training data set includes a plurality of training samples, each of the training samples includes: historical customer communication corpus data, and an emotion label corresponding to the historical customer communication corpus data for characterizing the emotion type and emotion intensity, the emotion type includes at least one of the following: anger, sadness, joy, neutrality, and the emotion intensity includes: mild, moderate, and strong; Determining the complaint probability corresponding to the customer based on the customer's personal user portrait and the emotion type and emotion intensity includes: obtaining relevant information of the customer, wherein the relevant information includes: personal basic information, historical business handling records, and historical complaint records; determining the personal user portrait corresponding to the customer based on the relevant information; and using a complaint probability prediction model to determine the complaint probability based on the personal user portrait and the emotion type and emotion intensity, wherein the complaint probability is used to characterize the probability of the customer initiating a complaint under the emotion type and emotion intensity.

5. The artificial intelligence-based consumer protection method according to claim 1, characterized in that: The method further comprises: Upon receiving customer complaint information, determine the target business corresponding to the customer complaint information, and trace back the target business being complained about to obtain historical business handling information corresponding to the target business being complained about, wherein the customer complaint information includes at least one of the following: complaint content, complaint event, and complaint object, and the historical business handling information includes: business handling time, location, person handling the business, business process, and customer information; Determine the complaint text features corresponding to the customer complaint information and the business handling features corresponding to the historical business handling information, wherein the complaint text features are at least used to characterize the complaint keywords and sentiment tendencies of the customer, and the business handling features are at least used to characterize the processing time and results of each link in the business process, as well as the service attitude of the handling personnel; Performing attribution analysis based on the complaint text features, business handling features, and the personal user portrait of the customer to obtain analysis results corresponding to the customer complaint information, wherein the analysis results are used to characterize the cause of the complaint corresponding to the customer complaint information; Based on the analysis results, a complaint handling process is determined, and a complaint handling work order corresponding to the customer complaint information is generated, and the complaint handling work order is sent to the person in charge corresponding to the initial node in the complaint handling process.

6. The artificial intelligence-based consumer protection method according to claim 5, characterized in that: The method further comprises: According to the analysis results, statistics are collected on the causes of various types of complaints for various target businesses, and based on the statistical results, visualization charts are generated for display, wherein the causes of complaints include at least one of the following: product problems, service attitude problems, and business process problems, and the visualization charts include at least one of the following: a pie chart representing the proportion of various causes of complaints in the total number of complaints, a bar chart representing the number of different types of causes of complaints, and a heat map representing the distribution of different causes of complaints in different time periods or different regions; And / or, determine the timestamp corresponding to each process node in the complaint handling process, wherein the timestamp is used to characterize the moment when the process node is executed and the processing time of the customer complaint information at the process node in the complaint handling process; based on the timestamp, determine the customer complaint information processed by each process node in the selected time period, and obtain the process heat map corresponding to the selected time period, wherein the position of each process node in the process heat map is characterized by different thermal values ​​to characterize the amount of customer complaint information processed by the process node and / or the processing time of the customer complaint information at the process node.

7. A consumer protection device based on artificial intelligence, characterized in that: include: A data acquisition module, used to acquire multimodal customer communication corpus data, wherein the customer communication corpus data includes at least one of the following: text data, video data, and audio data collected from customers during the business handling process of various service channels; A feature extraction module is used to extract features from the customer communication corpus data to obtain target features, including: determining key frames of the video data, and determining face areas and limb areas in the key frames; determining facial key points in the face area, and determining limb key points in the limb area; determining facial expression features of the customer based on position change information of the facial key points in adjacent key frames, and determining body language features of the customer based on position change information of the limb key points in adjacent key frames; determining a second feature corresponding to the video data based on the facial expression features and the body language features, wherein the second feature is used to determine the target feature; and determining the emotion type corresponding to the customer and the emotion intensity corresponding to the emotion type based on the target feature; A probability prediction module, used to determine the complaint probability corresponding to the customer based on the customer's personal user portrait and the emotion type and emotion intensity, wherein the personal user portrait is at least used to characterize the customer's historical complaint preference; A complaint warning module is used to send complaint warning information to the responsible department corresponding to the target business when the complaint probability exceeds the probability threshold corresponding to the target business handled by the customer, including: obtaining complaint record data of the target business within a preset historical time period, wherein the complaint record data includes at least one of the following: number of complaints, type of complaint, and complaint handling results; determining a complaint trend curve and complaint handling satisfaction level corresponding to the target business based on the complaint record data, wherein the complaint trend curve is used at least to characterize the trend of the number of complaints received by the target business over time; updating the probability threshold corresponding to the target business based on the complaint trend curve and the complaint handling satisfaction level, and judging whether there is a complaint related to the target business within the target time period. The target business-related policy update, wherein the target time period is a time period from a first moment to a second moment, the second moment is the moment when the probability threshold corresponding to the target business is most recently updated, and the first moment is the moment when the probability threshold corresponding to the target business is last updated; in the case where there is a policy update related to the target business within the target time period, the policy impact coefficient corresponding to the target business is determined according to the most recently updated policy, and the probability threshold corresponding to the target business is adjusted according to the policy impact coefficient; in the case where the complaint probability exceeds the adjusted probability threshold corresponding to the target business, the complaint warning information is generated, and the complaint warning information is sent to the responsible department corresponding to the target business.

8. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the artificial intelligence-based consumer protection method described in any one of claims 1 to 6 by running the computer program.

9. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the artificial intelligence-based consumer protection method described in any one of claims 1 to 6 are implemented.

Images