Consumer right protection method based on artificial intelligence large model and cloud platform

Abstract

The invention discloses a consumer right protection method based on an artificial intelligence large model and a cloud platform. The method comprises the steps of obtaining consumer complaint data; preprocessing the consumer complaint data to obtain multi-modal data; performing embedding optimization processing on each piece of modal data in the multi-modal data to obtain a corresponding embedding feature vector; aggregating the obtained corresponding embedded feature vectors to generate a path; obtaining a path interaction vector based on the path; and according to the path interaction vector, optimizing a path by a path optimization objective function to obtain an optimized path, and taking the optimized path as an electronic commerce dispute handling path. The problems that in the prior art, during complaint processing, intelligent path guidance is lacked, manual rule setting and manual intervention are usually relied on, the efficiency is low, the response time is long in the processing process, and a personalized processing path is difficult to generate automatically are solved.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a consumer rights protection method and cloud platform based on an artificial intelligence big model. Background Art

[0002] With the rapid development of e-commerce and the popularity of social media, consumer rights protection faces unprecedented challenges. A large number of consumer complaints and feedback are presented in various forms, including but not limited to online platform reviews, social media discussions, telephone consultations, email complaints, etc. Traditional complaint handling methods usually rely on manual intervention, with complex processes and slow response speeds, making it difficult to promptly and effectively resolve the problems faced by consumers. Therefore, there is an urgent need for a more efficient and intelligent consumer rights protection system to meet consumer needs and improve the processing efficiency of enterprises.

[0003] At the same time, with the development of data mining and artificial intelligence technology, especially the advancement of natural language processing (NLP), machine learning, deep learning and other technologies, it has become possible to intelligently analyze and process consumer complaint data. Consumers' demands, emotions, needs and opinions are widely disseminated through online feedback, evaluation and social media. How to effectively extract valuable information from these huge data and then accurately identify consumer demands has become an important research direction in the current field of consumer rights protection. Through the integration and intelligent analysis of multi-source data, it can not only help companies respond to consumer needs in a timely manner, but also promote the improvement of product and service quality.

[0004] There are at least the following technical problems in the existing technology: the existing technology lacks intelligent path guidance when handling complaints, and usually relies on manually set rules and manual intervention, resulting in low efficiency and long response time in the processing process, and it is difficult to automatically generate personalized processing paths based on consumer complaint information, and fails to fully explore and utilize the potential value contained in the data; there is a lack of effective optimization mechanism, and it is impossible to achieve reasonable allocation of features and accurate modeling of nonlinear relationships, which affects the system's comprehensive processing capabilities for multi-dimensional features and cannot meet the efficient and intelligent needs of modern consumer rights protection. Summary of the invention

[0005] The embodiments of the present invention provide a consumer rights protection method and cloud platform based on an artificial intelligence big model to solve the above-mentioned technical problems in the prior art.

[0006] In order to have a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not a general review, nor is it intended to identify key / important components or to describe the scope of protection of these embodiments. Its only purpose is to present some concepts in a simple form as a prelude to the detailed description that follows.

[0007] According to a first aspect of an embodiment of the present invention, a method for protecting consumer rights and interests based on an artificial intelligence big model is provided.

[0008] In one embodiment, the consumer rights protection method based on the artificial intelligence big model includes: Obtaining consumer complaint data; and preprocessing the consumer complaint data to obtain multimodal data; Performing embedding optimization processing on each modal data in the multimodal data to obtain a corresponding embedding feature vector; Aggregating the corresponding embedded feature vectors to generate a path; and obtaining a path interaction vector based on the path; According to the path interaction vector, the path optimization objective function is used to optimize the path to obtain the optimized path, and the optimized path is used as the e-commerce dispute handling path.

[0009] In one embodiment, the consumer complaint data is obtained from: Online complaint records of e-commerce platforms, consumer user feedback records in social media, after-sales service log records and telephone records of physical stores, and chat records of online customer service.

[0010] In one embodiment, the consumer complaint data is preprocessed to obtain multimodal data including: De-noising and formatting the consumer complaint data to obtain processed consumer complaint information; Using natural language processing technology to extract features from the processed consumer complaint data to obtain key features, wherein the key features include: complaint text information, type information and time information, the type information includes complaint type and product type, and the time information includes complaint time and processing time; Based on the obtained key features, modal storage processing is performed to obtain multimodal data, which includes modal data of text dimension, modal data of classification dimension and modal data of time dimension.

[0011] In one embodiment, embedding optimization processing is performed on each modal data in the multimodal data to obtain a corresponding embedding feature vector including: Embedding mapping is performed on each modal data in the multimodal data to generate corresponding embedding features; and embedding optimization processing is performed based on the loss function according to the embedding features to obtain a corresponding embedding feature vector.

[0012] In one embodiment, the embedding mapping formula is: ; In the formula, Indicates i The embedded features of samples, the output dimension is ; express Function, used to capture the complex nonlinear periodic relationship between modes, express Function, used to model the global moderating effects of high-dimensional features; Represents the time characteristic term, which is used to reflect the high-order dynamic changes of the time mode. represents the time feature dimension, It is i The sample j The weight of the time feature, It is i The sample j time characteristics; Represents the square term of the classification feature, which is used to enhance the influence of important classification features through nonlinear amplification. represents the classification feature dimension, It is i The sample k The weight of the classification feature, It is i The sample k Classification features; Represents the text feature part, which is used to logarithmically scale the feature value to avoid bias problems caused by a large value range. represents the text feature dimension, It is i The sample The weight of the text features, It is i The sample text features; Represents the Gaussian attenuation term, which is used to simulate the changing trend of the importance of different modal features in the spatial dimension with distance, and capture the global effect of different features by dynamically adjusting the weights. represents the Gaussian attenuation coefficient, which represents the global influence of different modal characteristics. Represents the standard deviation parameter of Gaussian attenuation, which is used to adjust the decay speed of the feature. Indicatesi The sample p Features, Represents the total feature dimension, which is the sum of the dimensions of all modal features.

[0013] In one embodiment, the formula for the loss function is: ; In the formula, represents the loss function; Indicates i Embedding features of samples; Indicates i The target embedding vector of each sample; Represents the hyperparameters in the learning process, controlling the strength of regularization; Represents the weight smoothing term The weight ratio of represents the weight matrix of the kth embedded feature; N represents the number of samples.

[0014] In one embodiment, the embedding optimization processing formula is: ; In the formula, Indicates i The sample in +Embedding features after 1 iteration; Indicates i The sample in Embedded features after iterations; Represents the learning rate, which is used to control the step size of each parameter update. Represents the loss function Embedding features gradient.

[0015] In one embodiment, the path generation formula is: ; In the formula, Represents the feature vector after path interaction, that is, m path interaction vectors; is the path weight coefficient, indicating the i Sample pair m The contribution of each path; Represents the adaptive Gaussian kernel term, which is used to simulate the similarity between the embedded features and the path center and capture the clustering relationship of important features; Indicates m The center vector of each path; represents the width parameter, i.e., the variance of the Gaussian kernel; Represents the integral term of the periodic signal, which is used to capture the high-frequency response in the time characteristics and optimize the fine-grained expression of the time modal characteristics. Indicates the frequency of the periodic signal and controls the high-frequency fluctuations in the path generation process; represents the time variable; N represents the number of samples.

[0016] In one embodiment, the formula of the path optimization objective function is: ; In the formula, represents the path optimization objective function; Indicates the number of paths; Indicates m The ideal path vector of the path; represents the regularization coefficient; represents the Frobenius norm; Represents the weight matrix of path optimization; Indicates m Path interaction vectors.

[0017] According to a second aspect of an embodiment of the present invention, a consumer rights protection cloud platform based on an artificial intelligence big model is provided.

[0018] In one embodiment, the consumer rights protection cloud platform based on the artificial intelligence big model includes: A data processing module, used to obtain consumer complaint data; and pre-process the consumer complaint data to obtain multimodal data; An embedding processing module, used to perform embedding optimization processing on each modal data in the multimodal data to obtain a corresponding embedding feature vector; A path generation module is used to aggregate the corresponding embedded feature vectors to generate a path; and based on the path, obtain a path interaction vector; The path optimization module is used to optimize the path according to the path interaction vector to facilitate the path optimization objective function to obtain the optimized path, and use the optimized path as the e-commerce dispute handling path.

[0019] In one embodiment, the consumer complaint data is obtained from: Online complaint records of e-commerce platforms, consumer user feedback records in social media, after-sales service log records and telephone records of physical stores, and chat records of online customer service.

[0020] In one embodiment, when the data processing module pre-processes the consumer complaint data to obtain multimodal data, it denoises and formats the consumer complaint data to obtain processed consumer complaint information; uses natural language processing technology to extract features from the processed consumer complaint data to obtain key features, wherein the key features include: complaint text information, type information and time information, the type information includes complaint type and product type, and the time information includes complaint time and processing time; based on the obtained key features, it performs modal storage processing to obtain multimodal data, and the multimodal data includes modal data of text dimension, modal data of classification dimension and modal data of time dimension.

[0021] In one embodiment, when the embedding processing module performs embedding optimization processing on each modal data in the multimodal data to obtain a corresponding embedding feature vector, it performs embedding mapping on each modal data in the multimodal data to generate a corresponding embedding feature; and according to the embedding feature, performs embedding optimization processing based on a loss function to obtain a corresponding embedding feature vector.

[0022] In one embodiment, the embedding mapping formula is: ; In the formula, Indicates i The embedded features of samples, the output dimension is ; express Function, used to capture the complex nonlinear periodic relationship between modes, express Function, used to model the global moderating effects of high-dimensional features; Represents the time characteristic term, which is used to reflect the high-order dynamic changes of the time mode. represents the time feature dimension, It is i The sample j The weight of the time feature, It is i The sample j time characteristics; Represents the square term of the classification feature, which is used to enhance the influence of important classification features through nonlinear amplification. represents the classification feature dimension, It is i The sample k The weight of the classification feature, It is i The sample k Classification features; Represents the text feature part, which is used to logarithmically scale the feature value to avoid bias problems caused by a large value range. represents the text feature dimension, It is i The sample The weight of the text features, It is i The sample text features; Represents the Gaussian attenuation term, which is used to simulate the changing trend of the importance of different modal features in the spatial dimension with distance, and capture the global effect of different features by dynamically adjusting the weights. represents the Gaussian attenuation coefficient, which represents the global influence of different modal characteristics. Represents the standard deviation parameter of Gaussian attenuation, which is used to adjust the decay speed of the feature. Indicates i The sample p Features, Represents the total feature dimension, which is the sum of the dimensions of all modal features.

[0023] In one embodiment, the formula for the loss function is: ; In the formula, represents the loss function; Indicates i Embedding features of samples; Indicates i The target embedding vector of each sample; Represents the hyperparameters in the learning process, controlling the strength of regularization; Represents the weight smoothing term The weight ratio of represents the weight matrix of the kth embedded feature; N represents the number of samples.

[0024] In one embodiment, the embedding optimization processing formula is: ; In the formula, Indicates i The sample in +Embedding features after 1 iteration; Indicates i The sample in Embedded features after iterations; Represents the learning rate, which is used to control the step size of each parameter update. Represents the loss function Embedding features gradient.

[0025] In one embodiment, the path generation formula is: ; In the formula, Represents the feature vector after path interaction, that is, m path interaction vectors; is the path weight coefficient, indicating the i Sample pair m The contribution of each path; Represents the adaptive Gaussian kernel term, which is used to simulate the similarity between the embedded features and the path center and capture the clustering relationship of important features; Indicates m The center vector of each path; represents the width parameter, i.e., the variance of the Gaussian kernel; Represents the integral term of the periodic signal, which is used to capture the high-frequency response in the time characteristics and optimize the fine-grained expression of the time modal characteristics. Indicates the frequency of the periodic signal and controls the high-frequency fluctuations in the path generation process; represents the time variable; N represents the number of samples.

[0026] In one embodiment, the formula of the path optimization objective function is: ; In the formula, represents the path optimization objective function; Indicates the number of paths; Indicates m The ideal path vector of the path; represents the regularization coefficient; represents the Frobenius norm; Represents the weight matrix of path optimization; Indicates m Path interaction vectors.

[0027] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: 1. Through the embedding generation method of multimodal features, the platform can capture the complex correlations in consumer complaint data. The embedding technology not only improves the accuracy of data processing, but also enhances the expressiveness of features by jointly modeling time features, classification features, and text features, providing more accurate input for subsequent analysis and decision-making; 2. Through the path generation and path optimization mechanism, the optimal processing path can be intelligently formulated for each complaint event according to the specific circumstances of the consumer complaint. It not only considers the interaction between multimodal features, but also can adaptively adjust the importance of the path center and features, effectively improving the efficiency and accuracy of consumer rights protection; 3. It can identify high-risk complaints based on consumer complaint data and automatically generate personalized processing paths to help the platform predict potential disputes in advance, thereby minimizing conflicts between consumers and merchants; through data-driven optimization paths and decision-making processes, the platform can ensure that every consumer complaint can be handled in accordance with compliant procedures to ensure that the legitimate rights and interests of consumers are maximally protected; in addition, the intelligent path optimization mechanism can also ensure that merchants follow fair and just standards when resolving complaints, thereby enhancing the transparency of the processing process.

[0028] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0030] Figure 1 It is a flowchart of a method for protecting consumer rights and interests based on an artificial intelligence large model according to an exemplary embodiment; Figure 2 It is a structural block diagram of a consumer rights protection cloud platform based on an artificial intelligence big model according to an exemplary embodiment; Figure 3 It is a schematic diagram showing the structure of a computer device according to an exemplary embodiment. DETAILED DESCRIPTION

[0031] The following description and accompanying drawings fully illustrate the specific embodiments of this article so that those skilled in the art can practice them. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. The scope of the embodiments of this article includes the entire scope of the claims, as well as all available equivalents of the claims. Herein, the terms "first", "second", etc. are only used to distinguish one element from another, without requiring or implying any actual relationship or order between these elements. In fact, the first element can also be called the second element, and vice versa. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, so that the structure, device or equipment including a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also include elements inherent to such structure, device or equipment. In the absence of more restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the structure, device or equipment including the elements. Each embodiment is described in a progressive manner herein, and each embodiment focuses on the differences from other embodiments, and the same and similar parts between the embodiments can be referred to each other.

[0032] The terms "longitudinal", "lateral", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc. in this document indicate the orientation or position relationship based on the orientation or position relationship shown in the drawings, and are only for the convenience of describing this document and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operate in a specific orientation, and therefore cannot be understood as a limitation on the present invention. In the description of this document, unless otherwise specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a mechanical connection or an electrical connection, it can also be the internal communication of two elements, it can be a direct connection, or it can be an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0033] As used herein, the term "plurality" means two or more than two, unless otherwise specified.

[0034] In this document, the character " / " indicates that the preceding and following objects are in an "or" relationship. For example, A / B means: A or B.

[0035] In this article, the term "and / or" is a description of the association relationship between objects, indicating that three relationships may exist. For example, A and / or B means: A or B, or, A and B.

[0036] It should be understood that, although the various steps in the flow chart are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the figure may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0037] Each module in the device or system of the present application can be implemented in whole or in part by software, hardware, or a combination thereof. The above modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to the above modules.

[0038] In the absence of conflict, the embodiments of the present invention and the features of the embodiments may be combined with each other.

[0039] Figure 1 An embodiment of a consumer rights protection method based on an artificial intelligence big model of the present invention is shown.

[0040] In this optional embodiment, the consumer rights protection method based on the artificial intelligence big model includes: Step S101, obtaining consumer complaint data; and preprocessing the consumer complaint data to obtain multimodal data; Step S102, performing embedding optimization processing on each modal data in the multimodal data to obtain a corresponding embedding feature vector; Step S103, aggregating the corresponding embedded feature vectors to generate a path; and obtaining a path interaction vector based on the path; Step S104, optimizing the path according to the path interaction vector in a way that facilitates the path optimization objective function to obtain an optimized path, and using the optimized path as an e-commerce dispute resolution path.

[0041] Figure 2 An embodiment of a consumer rights protection cloud platform based on an artificial intelligence big model of the present invention is shown.

[0042] In this optional embodiment, the consumer rights protection cloud platform based on the artificial intelligence big model includes: The data processing module 201 is used to obtain consumer complaint data; and pre-process the consumer complaint data to obtain multi-modal data; The embedding processing module 202 is used to perform embedding optimization processing on each modality data in the multimodal data to obtain a corresponding embedding feature vector; A path generation module 203 is used to aggregate the corresponding embedded feature vectors to generate a path; and obtain a path interaction vector based on the path; The path optimization module 204 is used to optimize the path according to the path interaction vector and the path optimization objective function to obtain the optimized path, and use the optimized path as the e-commerce dispute handling path.

[0043] In the above optional embodiment, the sources of obtaining the consumer complaint data include: online complaint records of the e-commerce platform, consumer user feedback records in social media, after-sales service log records and telephone records of physical stores, and chat records of online customer service.

[0044] Each data form corresponds to different characteristics. For example, text data (such as complaint descriptions) is unstructured and requires the use of natural language processing (NLP) technology to parse out key fields; time series data (such as complaint time, processing time) has continuity characteristics and needs to maintain its original order; classification features (such as complaint type, product category) are structured data.

[0045] Therefore, when preprocessing the consumer complaint data to obtain multimodal data, the consumer complaint data can be denoised and formatted to obtain processed consumer complaint information; natural language processing technology is used to extract features of the processed consumer complaint data to obtain key features, wherein the key features include: complaint text information, type information and time information, the type information includes complaint type and product type, and the time information includes complaint time and processing time; based on the obtained key features, modal storage processing is performed to obtain multimodal data, and the multimodal data includes modal data of text dimension, modal data of classification dimension and modal data of time dimension.

[0046] Among them, text, time series and classification features are stored in independent dimensions to ensure that the subsequent embedding generation can accurately capture the interaction and potential correlation between modalities. In practical application, it can be expressed as ,in, is the number of samples, each sample Contains three types of modal information, namely Dimensional time characteristics (such as the time series and frequency of complaints), Dimensional classification features (such as complaint type, product type, etc.) and Dimensional text features (such as text descriptions of user complaints).

[0047] In addition, since multimodal features differ significantly in distribution and physical sense, they must be uniformly mapped into a high-dimensional feature space through an embedding generation step. The core goal of embedding generation is to capture the potential correlation between multimodal features while enhancing the weight expression of key features to better provide high-quality input for subsequent modeling. Therefore, in the above optional embodiments, when embedding optimization processing is performed on each modal data in the multimodal data to obtain the corresponding embedded feature vector, embedding mapping is performed on each modal data in the multimodal data to generate corresponding embedded features; and according to the embedded features, embedding optimization processing is performed based on the loss function to obtain the corresponding embedded feature vector.

[0048] The embedding mapping formula is: ; In the formula, Indicates i The embedded features of samples, the output dimension is ; express Function, used to capture the complex nonlinear periodic relationship between modes, express function, Modeling the global moderating effects of high-dimensional features; Represents the time characteristic term, which is used to reflect the high-order dynamic changes of the time mode. represents the time feature dimension, It is i The sample j The weight of the time feature, It is i The sample j time characteristics; Represents the square term of the classification feature, which is used to enhance the influence of important classification features through nonlinear amplification. represents the classification feature dimension, It is i The sample k The weight of the classification feature, It is i The sample k Classification features; Represents the text feature part, which is used to logarithmically scale the feature value to avoid bias problems caused by a large value range. represents the text feature dimension, It is i The sample The weight of the text features, It is i The sample text features; Represents the Gaussian attenuation term, which is used to simulate the changing trend of the importance of different modal features in the spatial dimension with distance, and capture the global effect of different features by dynamically adjusting the weights. represents the Gaussian attenuation coefficient, which represents the global influence of different modal characteristics. Represents the standard deviation parameter of Gaussian attenuation, which is used to adjust the decay speed of the feature. Indicates i The sample p Features, Represents the total feature dimension, which is the sum of the dimensions of all modal features.

[0049] In order to ensure that the embedding generated vector can effectively represent multimodal features while meeting the semantic consistency and structured requirements in the feature space, embedding optimization is performed through a loss function. Specifically, the loss function consists of three parts: the first part is used to measure the distance between the current embedded feature and the target embedding. The target embedding is a reference vector set according to the task requirements, and its dynamic update ensures that the embedding process gradually approaches the ideal distribution; the second part performs 𝐿2 regularization on the embedded features of all samples to ensure that the embedded features are not too large, thereby avoiding overfitting; the third part is the weight smoothing term, which takes the form of a ratio of a cubic term to a quadratic term, and is used to encourage smooth updates of weights to prevent drastic fluctuations during the optimization process. Through the combination of these three parts, the loss function achieves stability, convergence, and accuracy in dynamic embedding generation. The formula for the loss function is: ; In the formula, represents the loss function; Indicates i Embedding features of samples; Indicates i The target embedding vector of each sample is initialized as a random vector generated by task-specific labels or features; Represents the hyperparameters in the learning process, controlling the strength of regularization; Represents the weight smoothing term The weight ratio is obtained through experiments; represents the weight matrix of the kth embedded feature; N represents the number of samples.

[0050] In the embedded optimization process, the embedded optimization process formula is: ; In the formula, Indicates i The sample in +Embedding features after 1 iteration; Indicates i The sample in Embedded features after iterations; Represents the learning rate, which is used to control the step size of each parameter update and is selected through experience or parameter adjustment methods; Represents the loss function Embedding features gradient.

[0051] In addition, in the above optional embodiments, the aggregated representation of all optimized embedded features is integrated to capture the complex nonlinear relationship between multiple features; the aggregated representation of the optimized embedded features based on consumer complaint data is used as the path, and the implicit interaction pattern between modalities is extracted through the matching degree between the optimized embedded features and the path center, as well as the dynamic weight distribution between features. The local weight of the embedded features is enhanced by the square term, so that the part with a larger contribution of the features occupies a higher proportion during path aggregation. The adaptive Gaussian kernel term is used to simulate the similarity between each embedded feature and the path center. The clustering relationship of the embedded features is captured by controlling the center and width dynamically, so that important embedded features are more easily aggregated by the path model. The path generation formula is: ; In the formula, Represents the feature vector after path interaction, that is, m path interaction vectors; is the path weight coefficient, indicating the i Sample pair m The contribution of each path; Represents the adaptive Gaussian kernel term, which is used to simulate the similarity between the embedded features and the path center and capture the clustering relationship of important features; Indicates m The center vector of each path; It represents the width parameter, i.e., the variance of the Gaussian kernel, which affects the decay speed of the similarity; Represents the integral term of the periodic signal, which is used to capture the high-frequency response in the time characteristics and optimize the fine-grained expression of the time modal characteristics. Indicates the frequency of the periodic signal and controls the high-frequency fluctuations in the path generation process; represents the time variable; N represents the number of samples. Path weight , Path Center and width parameters Initialized to random values ​​and then optimized using the path loss function.

[0052] Path interaction forms a new feature vector, namely the path interaction vector, by combining the nonlinear relationship of different optimized embedded features. In order to adjust the path interaction vector so that it can more accurately reflect the actual data distribution in the target task, path optimization needs to be performed to maintain reasonable smoothness and avoid overfitting during the optimization process, so as to be more robust and accurate in practical applications. The formula of the path optimization objective function is: ; In the formula, represents the path optimization objective function; Indicates the number of paths; Indicates m The ideal path vector of each path represents the best path or target path under ideal conditions and is set based on expert knowledge; represents the regularization coefficient; represents the Frobenius norm; The weight matrix for path optimization represents the parameters related to path optimization, such as the consumer's behavior weight, the correlation between different paths, etc. Indicates m Path interaction vectors.

[0053] The path optimization objective function is solved by the existing technology gradient descent method, and the optimized path is output. The optimized path can be used as the actual path in the consumer rights protection process. It is the optimal decision and action path based on the consumer complaint data when the platform handles consumer complaints, disputes, demand feedback, etc. For example: when consumers purchase goods or services, they may encounter various rights and interests issues, such as quality problems, substandard after-sales service, etc. After such problems occur, consumers usually complain to the platform or merchants. By optimizing the path, the platform can intelligently guide the dispute resolution process. For example, through data analysis, high-risk cases that may involve disputes are identified, and personalized processing paths are automatically formulated for high-risk cases to ensure that the dispute resolution steps are more standardized and compliant, and can effectively balance the interests of consumers and merchants.

[0054] Figure 3An embodiment of a computer device of the present invention is shown. The computer device may be a server, and the computer device includes a processor, a memory, and a network interface 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 the computer program in the non-volatile storage medium. The database of the computer device is used to store static information and dynamic information data. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, the steps in the above method embodiment are implemented.

[0055] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block 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 or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0056] In addition, the present invention also provides a computer device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiment when executing the computer program.

[0057] In addition, the present invention further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiment are implemented.

[0058] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and 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 by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0059] The present invention is not limited to the structures which have been described above and shown in the drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for protecting consumer rights based on an artificial intelligence big model, characterized in that: include: Obtain consumer complaint data; and preprocessing the consumer complaint data to obtain multimodal data; Performing embedding optimization processing on each modal data in the multimodal data to obtain a corresponding embedding feature vector; Aggregating the corresponding embedded feature vectors to generate a path; and obtaining a path interaction vector based on the path; According to the path interaction vector, the path optimization objective function is used to optimize the path to obtain the optimized path, and the optimized path is used as the e-commerce dispute handling path.

2. The method for protecting consumer rights based on artificial intelligence big model according to claim 1 is characterized in that: The sources of consumer complaint data include: Online complaint records of e-commerce platforms, consumer user feedback records in social media, after-sales service log records and telephone records of physical stores, and chat records of online customer service.

3. The method for protecting consumer rights based on artificial intelligence big model according to claim 1 is characterized in that: The consumer complaint data is preprocessed to obtain multimodal data including: De-noising and formatting the consumer complaint data to obtain processed consumer complaint information; Using natural language processing technology to extract features from the processed consumer complaint data to obtain key features, wherein the key features include: complaint text information, type information and time information, the type information includes complaint type and product type, and the time information includes complaint time and processing time; Based on the obtained key features, modal storage processing is performed to obtain multimodal data, which includes modal data of text dimension, modal data of classification dimension and modal data of time dimension.

4. The method for protecting consumer rights based on artificial intelligence big model according to claim 1 is characterized in that: Performing embedding optimization processing on each modal data in the multimodal data to obtain a corresponding embedding feature vector includes: Each modal data in the multimodal data is embedded and mapped to generate corresponding embedding features; and according to the embedding features, embedding optimization processing is performed based on a loss function to obtain a corresponding embedding feature vector.

5. The method for protecting consumer rights based on artificial intelligence big model according to claim 4 is characterized in that: The embedding mapping formula is: ； In the formula, Indicates i The embedded features of samples, the output dimension is ; express Function, used to capture the complex nonlinear periodic relationship between modes, express Function, used to model the global moderating effects of high-dimensional features; Represents the time characteristic term, which is used to reflect the high-order dynamic changes of the time mode. represents the time feature dimension, It is i The sample j The weight of the time feature, It is i The sample j time characteristics; Represents the square term of the classification feature, which is used to enhance the influence of important classification features through nonlinear amplification. represents the classification feature dimension, It is i The sample k The weight of the classification feature, It is i The sample k Classification features; Represents the text feature part, which is used to logarithmically scale the feature value to avoid bias problems caused by a large value range. represents the text feature dimension, It is i The sample The weight of the text features, It is i The sample text features; Represents the Gaussian attenuation term, which is used to simulate the changing trend of the importance of different modal features in the spatial dimension with distance, and capture the global effect of different features by dynamically adjusting the weights. represents the Gaussian attenuation coefficient, which represents the global influence of different modal characteristics. Represents the standard deviation parameter of Gaussian attenuation, which is used to adjust the decay speed of the feature. Indicates i The sample p Features, Represents the total feature dimension, which is the sum of the dimensions of all modal features.

6. The method for protecting consumer rights based on artificial intelligence big model according to claim 5 is characterized in that: The formula of the loss function is: ； In the formula, represents the loss function; Indicates i Embedding features of samples; Indicates i The target embedding vector of each sample; Represents the hyperparameters in the learning process, controlling the strength of regularization; Represents the weight smoothing term The weight ratio of represents the weight matrix of the kth embedded feature; N represents the number of samples.

7. The method for protecting consumer rights based on artificial intelligence big model according to claim 6 is characterized in that: The embedding optimization processing formula is: ; In the formula, Indicates i The sample in +Embedding features after 1 iteration; Indicates i The sample in Embedded features after iterations; Represents the learning rate, which is used to control the step size of each parameter update. Represents the loss function Embedded features gradient.

8. The method for protecting consumer rights based on artificial intelligence big model according to claim 7 is characterized in that: The path generation formula is: ； In the formula, Represents the feature vector after path interaction, that is, m path interaction vectors; is the path weight coefficient, indicating the i Sample pair m The contribution of each path; Represents the adaptive Gaussian kernel term, which is used to simulate the similarity between the embedded features and the path center and capture the clustering relationship of important features; Indicates m The center vector of each path; represents the width parameter, i.e., the variance of the Gaussian kernel; Represents the integral term of the periodic signal, which is used to capture the high-frequency response in the time characteristics and optimize the fine-grained expression of the time modal characteristics. Indicates the frequency of the periodic signal and controls the high-frequency fluctuations in the path generation process; represents the time variable; N represents the number of samples.

9. The method for protecting consumer rights based on artificial intelligence big model according to claim 8 is characterized in that: The formula of the path optimization objective function is: ； In the formula, represents the path optimization objective function; Indicates the number of paths; Indicates m The ideal path vector of the path; represents the regularization coefficient; represents the Frobenius norm; Represents the weight matrix of path optimization; Indicates m Path interaction vectors.

10. A consumer rights protection cloud platform based on artificial intelligence big model, characterized in that: include: Data processing module, used to obtain consumer complaint data; and preprocessing the consumer complaint data to obtain multimodal data; An embedding processing module, used to perform embedding optimization processing on each modal data in the multimodal data to obtain a corresponding embedding feature vector; A path generation module is used to aggregate the corresponding embedded feature vectors to generate a path; and based on the path, obtain a path interaction vector; The path optimization module is used to optimize the path according to the path interaction vector to facilitate the path optimization objective function to obtain the optimized path, and use the optimized path as the e-commerce dispute handling path.

Images