RAG evaluation method, system and device based on user portrait and medium

Abstract

The invention provides an RAG evaluation method, system and device based on a user portrait and a medium, and the method comprises the steps: obtaining user interaction data, and constructing the user portrait of a current user according to the user interaction data; generating a search answer in combination with a search request and the user portrait under the condition that the search request is received; according to the method, the evaluation index associated with the user portrait is determined based on the pre-constructed mapping model, the search answer is evaluated according to the evaluation index, and the evaluation report is generated, so that the evaluation result better fits the user characteristics, the evaluation accuracy is improved, and more valuable feedback can be provided for continuous optimization of the RAG system.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and natural language processing, and in particular to a RAG evaluation method, system, device and medium based on user portrait. Background Art

[0002] With the rapid development of artificial intelligence technology, RAG (Retrieval-Augmented Generation) systems have been widely used in applications such as question-answering, dialogue, and content generation. However, existing RAG evaluation methods generally use fixed scoring criteria to evaluate the generated answers, resulting in a single evaluation index and ignoring the correlation between the answer and user characteristics, making the evaluation results lack of pertinence and low accuracy, affecting the continuous optimization of the RAG system and the improvement of user experience. Summary of the invention

[0003] The embodiment of the present invention provides a RAG evaluation method, system, device and medium based on user portrait to solve the problems existing in the related technology. The technical solution is as follows:

[0004] In a first aspect, an embodiment of the present invention provides a RAG evaluation method based on user portrait, including:

[0005] Obtain user interaction data and build a user profile of the current user based on the user interaction data;

[0006] Upon receiving a search request, generate a search answer by combining the search request and the user profile;

[0007] Based on the pre-built mapping model, the evaluation indicators associated with the user portrait are determined, the search answers are evaluated according to the evaluation indicators, and an evaluation report is generated.

[0008] In one embodiment, the user interaction data includes interaction behavior, interaction content, dwell time, and usage frequency; the interaction behavior includes query behavior, click behavior, and feedback behavior.

[0009] In one embodiment, constructing a user profile based on user interaction data includes:

[0010] Determine the multi-dimensional characteristics of the current user based on user interaction data, including user role, professional level, interest preferences, and interaction habits;

[0011] Compare the multi-dimensional features with the group features of different pre-classified user groups, and determine the target user group to which the current user belongs based on the comparison results;

[0012] Determine the user profile of the current user based on the behavior patterns of the target user group.

[0013] In one embodiment, it further includes:

[0014] Obtain sample data, train the pre-created decision tree model with the sample data, determine the correlation between user portrait features and evaluation indicators, and generate a mapping model.

[0015] In one embodiment, it further includes:

[0016] Filter out user feedback data from user interaction data, and determine the target evaluation indicators to be adjusted based on the user feedback data;

[0017] According to the target evaluation index, the weight of the correlation between the user portrait features and the evaluation index in the mapping model is adjusted to obtain an updated mapping model;

[0018] A new evaluation metric is determined based on the updated mapping model, the search answer is evaluated according to the new evaluation metric, and a new evaluation report is generated.

[0019] In one embodiment, it further includes:

[0020] Identify feedback features based on user feedback data;

[0021] The user portrait is updated according to the feedback features to obtain an updated portrait, the search answer is updated based on the updated portrait, and the evaluation report is updated.

[0022] In one embodiment, it further includes:

[0023] The evaluation report is displayed visually, including the dimension scores of each evaluation indicator and the comprehensive score calculated based on the scores of each dimension.

[0024] In a second aspect, an embodiment of the present invention provides a RAG evaluation system based on user portrait, which executes the RAG evaluation method based on user portrait as described above.

[0025] In a third aspect, an embodiment of the present invention provides an electronic device, the device comprising: a memory and a processor. The memory and the processor communicate with each other through an internal connection path, the memory is used to store instructions, the processor is used to execute the instructions stored in the memory, and when the processor executes the instructions stored in the memory, the processor executes the method in any one of the above-mentioned embodiments.

[0026] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a computer, the method in any one of the above-mentioned embodiments is executed.

[0027] The advantages or beneficial effects of the above technical solution include at least:

[0028] The present invention constructs a user profile of the current user based on user interaction data, and generates a search answer in combination with the search request and the user profile when a search request is received; determines the evaluation indicators associated with the user profile based on a pre-built mapping model, evaluates the generated search answer according to the evaluation indicators, and generates a corresponding evaluation report, so that the evaluation results are more in line with user characteristics, improves the evaluation accuracy, and can provide more valuable feedback for the continuous optimization of the RAG system.

[0029] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments disclosed in the present invention and should not be regarded as limiting the scope of the present invention.

[0031] Figure 1 It is a flowchart of the RAG evaluation method based on user portrait of the present invention;

[0032] Figure 2 It is a timing diagram of the RAG evaluation method based on user portrait of the present invention;

[0033] Figure 3 This is a logical architecture diagram of the RAG evaluation system based on user portraits of the present invention;

[0034] Figure 4 The structure block diagram of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0035] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0036] With the rapid development of artificial intelligence technology, RAG (Retrieval-Augmented Generation) systems have been widely used in applications such as question answering, dialogue, and content generation. However, existing RAG evaluation methods usually have the following problems:

[0037] 1. Single evaluation criteria: Most evaluators use simple label matching or fixed scoring criteria;

[0038] 2. Lack of personalization: Existing evaluation systems often use the same evaluation criteria for all users, failing to consider differences in system user roles, knowledge backgrounds, preferences, and usage scenarios;

[0039] 3. Static evaluation mechanism: Evaluation criteria and weights are usually preset and lack the ability to dynamically adjust based on user feedback and behavior;

[0040] 4. Single evaluation metric: Most evaluators only focus on the correctness of the answer, ignoring other important dimensions such as the depth, innovation, and clarity of expression of the answer;

[0041] 5. Lack of context awareness: Existing evaluation methods usually do not consider users’ query history and interaction patterns, resulting in a lack of consistency and personalization in the evaluation results.

[0042] In order to solve the above problems, an embodiment of the present invention proposes a RAG evaluation method based on user portrait, which can improve the pertinence and accuracy of the evaluation results, and is conducive to the continuous optimization of the RAG system and the improvement of user experience.

[0043] like Figure 1 to Figure 3 As shown, the RAG evaluation method based on user portrait proposed in this embodiment specifically includes the following steps:

[0044] Step S1: Obtain user interaction data, and build a user profile of the current user based on the user interaction data.

[0045] It should be noted that user interaction data includes interaction behavior, interaction content, dwell time, and usage frequency; interaction behavior includes query behavior, click behavior, and feedback behavior.

[0046] Specifically, all the user's previous user interaction data is obtained, such as recording the user's various query behaviors in the RAG system, capturing the query content of the query behavior, and counting the maintenance time of the query behavior as the dwell time, and counting the number of times the query behavior is used within a specified time to calculate the usage frequency, thereby obtaining the user interaction data corresponding to the query behavior.

[0047] Similarly, identify the click actions performed by users in the RAG system, capture the click content, identify the page to which the click leads, calculate the time the user stays on the page, and obtain user interaction data corresponding to the click behavior by counting the click frequency of the same content.

[0048] In addition, users can also provide feedback based on the answers generated by the RAG system. The feedback can be provided in the form of comments or ratings. The user's comment content or rating scores can be captured to obtain the corresponding feedback content, and the frequency of users using the feedback function can be counted. The maintenance time of the feedback behavior is counted as the residence time, thereby obtaining the user interaction data corresponding to the feedback behavior.

[0049] The user profile of the current user is constructed based on the user interaction data of the above-mentioned various interaction behaviors. The method is as follows:

[0050] Step S11: Determine the multi-dimensional characteristics of the current user based on the user interaction data, where the multi-dimensional characteristics include user role, professional level, interest preferences, and interaction habits.

[0051] Multidimensional features are features that can represent user behavior patterns and preferences. Specifically: user roles can be inferred by analyzing user behavior patterns, such as usage frequency, interaction content, etc.; user professionalism can be inferred by analyzing the quality of user contributions on professional forums or question-and-answer platforms and the frequency and accuracy of professional terminology used in user language; user interest preferences can be reflected by analyzing the content that users browse and search; interaction habits can be determined by analyzing the frequency of user interaction with the system and the duration of each interaction.

[0052] The extraction of multi-dimensional features can be achieved through a deep learning model. The deep learning model for multi-dimensional feature analysis is obtained by training with a large number of data samples. After obtaining the user interaction data, the deep learning model is used to analyze the text corresponding to the user interaction data, and the multi-dimensional features of each user are determined by learning the deep features in the text.

[0053] Step S12: Compare the multi-dimensional features with the group features of pre-classified different user groups, determine the target user group to which the current user belongs based on the comparison result, and determine the user profile of the current user based on the behavior pattern of the target user group.

[0054] It should be noted that this embodiment adopts the principle of cluster analysis, collects user characteristics of different users in advance, distinguishes the characteristics of different users through cluster analysis, and divides multiple users with the same typical characteristics and similar user behavior patterns into a user group to understand the behavior patterns and preferences of different user groups.

[0055] In this embodiment, after determining the multi-dimensional characteristics of the current user, a target user group with the same typical characteristics and similar user behavior patterns as the current user is found, the current user is divided into the target user group, and a corresponding user portrait is constructed for the current user based on the behavior patterns and characteristics of the target user group, including the typical characteristics of the cluster, the general patterns of user behavior, and possible interests and needs.

[0056] Step S2: upon receiving a search request, generate a search answer by combining the search request and the user portrait.

[0057] It should be noted that when a user initiates a search request in the RAG system, the RAG system uses a retrieval algorithm to find documents or information fragments related to the search request from a pre-built knowledge base or database. The RAG system uses the retrieved information to construct an answer through natural language generation technology, thereby outputting the search answer, which includes text content or recommended applications generated by natural language.

[0058] Step S3: Determine the evaluation indicators associated with the user portrait based on the pre-built mapping model, evaluate the search answer according to the evaluation indicators, and generate an evaluation report.

[0059] Previously, it was necessary to obtain a large amount of sample data, including user data, evaluation data, etc., and preprocess the sample data, that is, convert non-numeric data into numerical data, or process missing values ​​and outliers; divide the processed sample data into training sets and test sets; create a model instance of the decision tree model, use the training set data to train the model, use the test set data to evaluate the model performance, analyze the contribution of each feature to the prediction evaluation index, and optimize the model performance by adjusting the model parameters. The final trained decision tree model is called a mapping model, which can be deployed in the production environment for actual prediction.

[0060] The decision tree model is trained through a large amount of sample data to learn and determine the relationship between user portrait features and evaluation indicators, thereby generating a corresponding mapping model. In the mapping model, each evaluation indicator has its corresponding indicator weight. The higher the indicator weight, the higher the correlation between the evaluation indicator and the user portrait feature, and the lower the indicator weight, the lower the correlation between the evaluation indicator and the user portrait feature. Among them, the user portrait usually includes the user's basic information, behavioral habits, preference settings and other characteristics; and the evaluation indicators can be answer accuracy, relevance, depth and simplicity, etc.

[0061] In this embodiment, based on the user portrait and mapping model, the evaluation indicators related to the current user are dynamically selected and combined to determine the evaluation criteria for evaluating the search answer. At this time, the evaluation criteria are associated with the user portrait characteristics of the current user. For example, for user roles that are more relevant to the industry, the system can increase the indicator weight of the depth and professionalism of the answer in this field, while reducing the indicator dimensions of explanatory and basic.

[0062] The search answer is evaluated through a preset evaluation execution engine set of evaluation indicators and dynamic weights associated with the current user, and a detailed evaluation report is generated to evaluate the quality of the answer generated by the RAG system based on the current user's search request.

[0063] For example, if the user profile is an operation manager in a smart park, the applications or generated content recommended by the RAG system should directly meet the professional characteristics of the operation manager in the smart park, support the core functions of park operation management, such as tenant management, property service coordination, etc., to ensure the efficient operation and service quality of the park. The evaluation execution engine evaluates the search answers according to the user role of the operation manager and the professional evaluation indicators, and the evaluation results may be:

[0064] Exact match: The recommended search answers focus on park operation management, providing functions such as vehicle identification and equipment positioning to help operation managers optimize park resources and services; or,

[0065] Partial match: The recommended search answer covers some functions of park operations, but may be more focused on other aspects, such as the asset management module included in the financial management system; or,

[0066] No match: The recommended search answers are completely unrelated to operations management, such as developer tools or document management systems.

[0067] In addition, an evaluation report is generated based on the evaluation results. The evaluation report mainly presents the dimensional scores of each evaluation indicator. Taking the above-mentioned park operation and management personnel as an example, assuming that their evaluation results are completely matched, the dimensional score of the "professionalism" evaluation indicator is relatively high, while the dimensional scores of other evaluation indicators that are irrelevant or weakly relevant to operation and management personnel are relatively low. A comprehensive score can also be calculated based on the dimensional scores of all evaluation indicators, and the dimensional scores and comprehensive scores of each evaluation indicator are visualized, so that the current user can intuitively understand the evaluation situation.

[0068] At the same time, users can make corresponding feedback based on the evaluation results, thereby obtaining user feedback data. The user feedback data reflects the degree of user recognition of the evaluation results. The user feedback data can collect user opinions through questionnaires, dialogue windows, etc. Users can score the accuracy, relevance and other indicators of the evaluation results through questionnaires, dialogue windows, etc. Users can also make feedback in the form of text comments. The scores or comments obtained from user feedback are called user feedback data.

[0069] Determine user feedback data based on user feedback behavior, determine evaluation indicators associated with user feedback based on user feedback data, call the evaluation indicators the target evaluation indicators to be adjusted, and determine the weight adjustment strategy of the target evaluation indicators based on the degree of user feedback. For example, determine the quality of the answer generated by the RAG system based on the score number or the semantics represented by the user feedback comments, so as to determine the target evaluation indicators that need to be adjusted, the indicator weights, and the weight adjustment range.

[0070] The correlation between each evaluation indicator and the user portrait in the mapping model is adjusted, and the indicator weight of the target evaluation indicator is adjusted according to the weight adjustment range to obtain an updated mapping model.

[0071] New evaluation indicators associated with the user portrait are determined according to the updated mapping model, and the search answers are re-evaluated according to the new evaluation indicators to obtain a new evaluation report.

[0072] In addition, the user feedback data fed back by the user can also be used to update the user portrait of the current user. Specifically, the user's feedback features are identified based on the user feedback data. For example, the text of the user's comments is analyzed through semantic analysis, and the user's interaction habits, professional level, interest preferences and other feedback features are re-determined. The user portrait content is updated using the feedback features to update the portrait. The search answer is updated based on the updated portrait, and the evaluation report can also be updated based on the updated portrait, so that the search answer and the evaluation report are more in line with the characteristics of the current user.

[0073] In some embodiments, based on the real-time behavior analysis results, the weight parameters of the evaluation system are adjusted to ensure the sensitivity of the system to changes in user behavior. To ensure stability during the optimization process, the Adam optimizer is used in combination with adaptive learning rate control and momentum adjustment. The steps are as follows:

[0074] Initialize learning rate and momentum parameters: Set the initial learning rate value (such as 0.001) and momentum parameter (such as 0.9 or 0.99), which are used to control the size of gradient updates.

[0075] Calculate gradients and update weights: In each backpropagation, the gradient values ​​are calculated and the weights are updated based on the momentum and the accumulation of past gradients.

[0076] Adaptive adjustment of learning rate: According to the changing trend of the gradient, adjust the current learning rate. For example, if the gradient is large, you can reduce the learning rate to prevent jumping out of the local optimum; if the gradient is small, increase the learning rate to improve the convergence speed.

[0077] In the process of combining gradient adjustment, the change of loss function can be monitored at the end of each epoch. If the loss decreases slowly or stagnates, consider reducing the learning rate to achieve the optimization purpose.

[0078] The beneficial effects of this embodiment are as follows:

[0079] Highly personalized: Dynamically adjust the evaluation criteria according to user characteristics to provide tailored evaluation results;

[0080] Strong adaptability: Ability to dynamically update evaluation strategies based on user feedback and behavior changes;

[0081] Context-aware: Considers the user's query history and interaction patterns to provide a consistent evaluation experience;

[0082] Improve user satisfaction: Improve the quality of RAG system output through personalized evaluation and significantly enhance user experience;

[0083] System scalability: The modular design makes the system easy to expand and supports the rapid integration of new assessment dimensions and strategies;

[0084] Continuous optimization capability: Through self-monitoring and feedback analysis, continuous improvement of the evaluation system can be achieved.

[0085] In another embodiment of the present invention, a RAG evaluation system based on user portrait is also provided. Figure 3 As shown, the system includes a storage layer, a collection layer, a service layer and a RAG system. The storage layer is used for data storage, configuration management and log management. The collection layer is used for user data collection and user portrait construction. The service layer is used for dynamic weight adjustment, evaluation indicator determination, evaluation report generation and feedback analysis.

[0086] It should be noted that the functions of each module in the system embodiment of the present invention can be found in the corresponding description in the above-mentioned RAG evaluation method based on user portrait, and will not be repeated here.

[0087] In another embodiment of the present invention, an electronic device is provided. Figure 4 FIG. 2 shows a structural block diagram of an electronic device according to an embodiment of the present invention. Figure 4 As shown, the electronic device includes: a memory 100 and a processor 200, wherein the memory 100 stores a computer program that can be run on the processor 200. When the processor 200 executes the computer program, the RAG evaluation method based on user portrait in the above embodiment is implemented. The number of the memory 100 and the processor 200 can be one or more.

[0088] The electronic device also includes:

[0089] The communication interface 300 is used to communicate with external devices and perform data exchange transmission.

[0090] If the memory 100, the processor 200 and the communication interface 300 are implemented independently, the memory 100, the processor 200 and the communication interface 300 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0091] Optionally, in a specific implementation, if the memory 100, the processor 200 and the communication interface 300 are integrated on a chip, the memory 100, the processor 200 and the communication interface 300 can communicate with each other through an internal interface.

[0092] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present invention when executed by a processor.

[0093] An embodiment of the present invention further provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present invention.

[0094] An embodiment of the present invention also provides a chip, including: an input interface, an output interface, a processor and a memory, wherein the input interface, the output interface, the processor and the memory are connected via an internal connection path, and the processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided by the embodiment of the invention.

[0095] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the advanced RISC machines (ARM) architecture.

[0096] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may also include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available. For example, static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct rambus RAM (DR RAM).

[0097] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0098] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0099] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0100] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A RAG evaluation method based on user portrait, characterized in that: include: Acquire user interaction data, and construct a user profile of the current user based on the user interaction data; Upon receiving a search request, generating a search answer by combining the search request and the user profile; An evaluation index associated with the user portrait is determined based on a pre-built mapping model, the search answer is evaluated according to the evaluation index, and an evaluation report is generated.

2. The RAG evaluation method based on user portrait according to claim 1, characterized in that: The user interaction data includes interaction behavior, interaction content, dwell time and usage frequency; the interaction behavior includes query behavior, click behavior and feedback behavior.

3. The RAG evaluation method based on user portrait according to claim 1, characterized in that: The constructing a user portrait according to the user interaction data comprises: Determine the multi-dimensional characteristics of the current user according to the user interaction data, wherein the multi-dimensional characteristics include user role, professional level, interest preference and interaction habits; Compare the multi-dimensional features with the group features of different pre-classified user groups, and determine the target user group to which the current user belongs according to the comparison result; The user portrait of the current user is determined according to the behavior pattern of the target user group.

4. The RAG evaluation method based on user portrait according to claim 1, characterized in that: Also includes: Sample data is obtained, and a pre-created decision tree model is trained using the sample data to determine the correlation between user portrait features and evaluation indicators, thereby generating the mapping model.

5. The RAG evaluation method based on user portrait according to claim 1, characterized in that: Also includes: Filtering user feedback data from the user interaction data, and determining a target evaluation indicator to be adjusted according to the user feedback data; According to the target evaluation index, weight adjustment is performed on the correlation relationship between the user portrait features and the evaluation index in the mapping model to obtain an updated mapping model; A new evaluation index is determined based on the updated mapping model, the search answer is evaluated according to the new evaluation index, and a new evaluation report is generated.

6. The RAG evaluation method based on user portrait according to claim 5 is characterized in that: Also includes: identifying feedback features according to the user feedback data; The user portrait is updated according to the feedback feature to obtain an updated portrait, and the search answer and the evaluation report are updated based on the updated portrait.

7. The RAG evaluation method based on user portrait according to claim 1, characterized in that: Also includes: The evaluation report is visually displayed, and the evaluation report includes the dimensional scores of each evaluation indicator and the comprehensive score calculated based on each dimensional score.

8. A RAG evaluation system based on user portrait, characterized in that: Execute the RAG evaluation method based on user portrait as described in any one of claims 1 to 7.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores instructions, and the instructions are loaded and executed by the processor to implement the RAG evaluation method based on user portrait as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the RAG evaluation method based on user portrait as described in any one of claims 1 to 7 is implemented.

Images