Public service platform suitable for aging multi-modal data intelligent processing system and method

Abstract

The application discloses a kind of public service platform old-adaptable multi-modal data intelligent processing system and method, the method includes the following steps: S1.data collection module is collected from data terminal the image data of public service platform needing to carry out old-adaptable evaluation;S2.data reporting module inputs the image data collected to multi-modal big model, and the image content is matched, analyzed and structured by the preset natural language prompt word and structured report;S3.data calculation module calculates and analyzes the report data according to the evaluation dimension and scoring rule set, generates the quantitative score of each dimension and judgment basis;S4.data storage module stores the result after calculation and analysis in storage medium;S5.data visualization module shows the result on display device by chart mode, to intuitively show the public service platform old-adaptable level. Can improve the processing efficiency, consistency and availability of old-adaptable evaluation data.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence and big data processing technology, specifically to a multimodal data intelligent processing system and method for assessing the age-friendliness level of a lifelong learning public service platform. Background Technology

[0002] In advancing the development of digitally age-friendly systems, it is crucial to assess the age-friendliness of numerous lifelong learning public service platforms, such as the National Smart Education Platform, the National University for the Elderly, and local lifelong education networks. Currently, this assessment primarily relies on human expert review, which essentially involves subjective interpretation and scoring of the platform interface—a multimodal data source. However, this human expert assessment approach is not only inefficient in data processing and unable to handle the dynamic evaluation of large-scale platforms, but also suffers from significant fluctuations in results due to subjective differences among experts regarding factors such as "font size" and "button clickability." Furthermore, the data generated by expert assessments is unstructured and difficult to directly use for statistical analysis, cross-sectional comparisons, or database import. Therefore, there is an urgent need for an intelligent data processing system that can simulate expert thinking, execute standardized rules, and output structured data. Summary of the Invention

[0003] To address the problems existing in current technologies, the present invention aims to provide an intelligent multimodal data processing system and method for age-friendly features in a lifelong learning public service platform. This system can automatically identify, process, and analyze age-friendly features in platform page images, reducing manual operation and improving the processing efficiency, consistency, and usability of age-friendly assessment data. The system can analyze visual and textual elements of platform page images using a multimodal large model and generate structured, quantifiable age-friendly assessment data according to preset assessment rules, thereby achieving automation, standardization, and intelligence in the assessment process.

[0004] To achieve the above objectives, the present invention provides an age-friendly multimodal data intelligent processing method for public service platforms, the method comprising the following steps: S1. The data collection module collects image data from the data terminal of the public service platform that needs to be assessed for age-friendliness. The collected image data includes screenshots of the platform page. S2. The data entry module inputs the collected image data into the multimodal large model, and uses preset natural language prompts to match, parse and structure the image content to obtain the entry data; S3. The data calculation module calculates and analyzes the submitted data according to the set evaluation dimensions and scoring rules, and generates quantitative scores and judgment criteria for each dimension for subsequent visualization. S4. The data storage module stores the results of the calculation and analysis in the storage medium for subsequent querying, retrieval and use; S5. The data visualization module displays the results on the display device in the form of charts, making it easy to intuitively understand the platform's aging-friendly level.

[0005] Furthermore, in step S1, the page images of the target platform are periodically collected according to the set collection time points and frequencies.

[0006] Furthermore, in step S1, the collected image data is in PNG or JPG format; the data collection module is used to collect images of various core pages of the lifelong public education platform, including at least one of the following: homepage, course details page, course list page, personal center page, help and support page, and search results page.

[0007] Furthermore, in step S2, the multimodal large model adopts Qwen-VL-Max.

[0008] Furthermore, in step S2, during the data entry process, the system performs JSON format validation on the model output to ensure data consistency and accuracy.

[0009] Furthermore, in step S2, the system's data entry module adopts a two-stage query mechanism to balance the depth of analysis with the standardization of output: Phase 1: Load analytical prompts to guide the multimodal large model to perform free analysis on the image; Phase 2: Load structured output prompts and force the multimodal large model to output evaluation results according to a predefined JSON Schema.

[0010] Furthermore, in step S3, the data calculation module evaluates the platform's aging-in-time performance based on the following four dimensions: A. Perceptibility: This includes font size, color contrast, icon recognizability, and page clutter. B. Operability: including button size, click feedback, navigation hierarchy, and protection against accidental operation; C. Understandability: This includes terminology usage, clarity of instructions, and logical organization of information; D. Supportability: This includes the visibility of the help entry point, the usability of the help content, and the friendliness of error messages.

[0011] Furthermore, in step S4, the data storage module adopts cloud storage to aggregate the four-dimensional structured evaluation records of each page into a comprehensive evaluation data entry and store it in a relational database or NoSQL database.

[0012] Furthermore, in step S5, the data visualization module uses Power BI tools to present the data, connects to the database through a data connector, creates an age-friendly assessment dashboard, and displays heatmaps, radar charts, and trend line charts of scores for each platform across four dimensions.

[0013] On the other hand, this invention provides a multimodal intelligent data processing system for age-friendly public service platforms. The system includes a data collection module, a data entry module, a data calculation module, a data storage module, and a data visualization module. The system is used to implement the multimodal intelligent data processing method for age-friendly public service platforms according to this invention. The beneficial effects of this invention are as follows: The multimodal intelligent data processing system and method of this invention automatically identify and quantify age-friendly features by collecting, parsing, structuring, and displaying platform page image data, and utilizing multimodal large-scale modeling and prompt engineering techniques. This invention achieves automated production of assessment data, greatly improves data processing efficiency, reduces manual operation, and lowers subjective bias. The output structured data can be directly used for regulatory decision-making, platform optimization, and academic research, providing core technical support for large-scale, standardized monitoring of age-friendly levels. Attached Figure Description

[0014] Figure 1 A flowchart of the age-friendly multimodal data intelligent processing system for a public service platform according to the present invention is shown. Figure 2 This diagram illustrates a four-dimensional breakdown of the age-appropriate assessment according to the present invention. Figure 3 An example of an aging-friendly assessment result is shown on a platform page (SME_HP_01.png). Detailed Implementation

[0015] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0016] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0017] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0018] The following combination Figures 1-3 Specific embodiments of the present invention will be described in detail below. It should be understood that the specific embodiments described herein are for illustrative and explanatory purposes only and are not intended to limit the present invention.

[0019] This invention provides an age-friendly multimodal data intelligent processing system for a lifelong learning public service platform. The system includes a data collection module, a data entry module, a data calculation module, a data storage module, and a data visualization module. This system can automatically identify, process, and analyze age-friendly features in platform page images, reducing manual operation and improving evaluation efficiency.

[0020] Figure 1 The operation flow of the age-friendly multimodal data intelligent processing method for a public service platform according to the present invention is shown. The method includes the following steps: S1. First, the data collection module collects image data from the data terminal. The collected image data includes screenshots of platform pages (PNG / JPG format). During the collection process, the images must be clear and complete. The collection frequency (e.g., once a week) and page type (e.g., mandatory homepage, course pages, etc.) can be set according to regulatory requirements. Page images of the target platform are collected periodically according to the set collection time and frequency. The data collection module also collects image data from the lifelong learning public service platform from the data terminal using the Playwright browser's automated tool. This is done by setting the target platform URL and the page type to be evaluated (e.g., homepage HP, course list CL, course details CD, etc.). Playwright's `chromium.launch(headless=True)` is called to launch a headless browser instance, setting the viewport size to 1920×1080 to simulate a desktop access environment. The target page is precisely reached through the explicitly configured navigation path (as shown in Table 1). On the target page, Playwright's full-page screenshot API is called, and the image is saved as a PNG format image. The image is named according to the rule "platform abbreviation_page type_serial number.png", for example, SME_HP_01.png. After taking the screenshot, the image file is saved to the local temporary storage directory.

[0021] Table 1 Target Page Navigation Path

[0022] S2. Next, the data entry module inputs the collected image data into the multimodal large model, and matches, parses, and structures the image content using preset natural language prompts. Two prompt engineering processes are performed. The first uses analytical prompts, with highly specialized analytical prompts (q1, q2, q3, and q4) pre-defined for each of the four evaluation dimensions (perceptibility A, operability B, comprehensibility C, and supportability D). The system reads a screenshot, calls the large model API, takes the image and the corresponding analytical prompt (e.g., q1) as input, and returns a text description. Based on this, a second query is performed, inputting the corresponding structured output prompts, denoted as rq1, rq2, rq3, and rq4. The JSON is parsed using the json.loads() function and validated. Taking q1 (perceptibility) as an example, its core content is: specifying the model as an "aging-friendly user experience evaluation expert," clearly providing a 1-5 point scoring standard for four indicators (font size, color contrast, icon recognizability, and page clutter) from A1 to A4, and requiring the model to "provide a score and specific reasons for each indicator." The second use is structured prompts. The corresponding structured output prompts are input, denoted as rq1, rq2, rq3, and rq4. Taking rq1 as an example, it strictly instructs the model: "Please reorganize and output your evaluation results for the 'perceptibility' dimensions (A1–A4) strictly according to the following JSON format...", and provides a complete JSON example, explicitly requiring key names, scores (integers), and remarks (strings) fields.

[0023] During the data entry process, the system performs JSON format validation on the model output to ensure that the fields are complete and the value range is valid; it also performs format validation on the model output to ensure the consistency and accuracy of the data.

[0024] S3. Next, the data calculation module performs calculations and analyses on the submitted data according to the set evaluation dimensions and scoring rules, generating quantitative scores and judgment criteria for each dimension for subsequent visualization. The data calculation module performs calculations and analyses on the submitted data according to the four-dimensional evaluation rules. Each dimension includes multiple indicators, generating a score and judgment criteria for each indicator.

[0025] S4. Next, the data storage module stores the complete results after calculation and analysis in the cloud database. Each record corresponds to a single evaluation of a page image, stored in a structured table format (this embodiment uses a MySQL relational database). Each record contains 36 specific fields, including page metadata fields, secondary indicator score fields, secondary indicator judgment basis fields, and comprehensive difficulty score fields. Among them, page metadata includes a unique page identifier, platform name, page type, and evaluation timestamp; secondary indicators include font size, color contrast, icon recognizability, page clutter, button size, click feedback, navigation hierarchy, accidental operation protection, terminology usage, instruction clarity, information organization logic, help entry visibility, help content usability, and error message friendliness; the comprehensive dimensional score includes the total score for perceptibility dimension, the total score for operability dimension, the total score for comprehensibility dimension, and the total score for supportability dimension. The perceptibility dimension is the sum of scores for four items: font size, color contrast, icon recognizability, and page clutter. The operability dimension is the sum of scores for four items: button size, click feedback, navigation hierarchy, and accidental operation protection. The operability dimension is also the sum of scores for three items: terminology usage, instruction clarity, and information organization logic. The supportability dimension is the sum of scores for three items: help entry visibility, help content usability, and error message friendliness. This invention comprehensively covers four levels: page metadata, secondary indicator scores, secondary indicator judgment criteria, and overall dimension scores. Specific field information is shown in Table 2.

[0026] Table 2. Age-Friendliness Assessment Data Fields

[0027]

[0028] S5. Finally, the data visualization module displays the results on the display device in the form of charts, making it easier for regulators to intuitively understand the age-friendliness level of each platform.

[0029] Specifically, in step S1, the data collection module is used to collect images of various core pages of the lifelong public education platform, including the homepage (HP), course details page (CD), course list page (CL), personal center page (PC), help and support page (HS), search results page (SR), etc.

[0030] In step S2, the system's data entry module employs a two-stage query mechanism to balance analysis depth and output standardization: I. First stage: Load analytical prompts (q1–q4) (e.g., q1) to guide the multimodal large model to perform free analysis of the image; where q1 is the perceptibility dimension, q2 is the operability dimension, q3 is the total score of the comprehensibility dimension, and q4 is the supportability dimension.

[0031] The specific contents of the analytical prompts q1–q4 are as follows; The system has pre-set analytical prompts for each of the four evaluation dimensions. Each prompt includes three parts: expert role setting, detailed scoring criteria, and specific task requirements.

[0032] Taking q1 (perceptibility dimension) as an example: You are a professional expert in age-friendly user experience evaluation, participating in an academic research project on "Construction and Empirical Research of Age-Friendly Intelligent Evaluation Model for China's Lifelong Learning Public Service Platform." Based on the provided screenshots of the platform interface, and using the following scoring criteria for the "Perceptibility" dimension, please conduct an objective and detailed analysis and score of the interface's visual readability.

[0033] Scoring criteria: - A1. Font size: 1 point (≤14px), 2 points (15–16px), 3 points (17–18px), 4 points (≥18px), 5 points (≥20px) - A2. Color contrast: 1 point (<3:1), 2 points (3:1–4:1), 3 points (=4.5:1), 4 points (≥5:1), 5 points (≥7:1) - A3. Icon recognizability: 1 point (pure icon with no text), 2 points (abstract icon, text too small or missing), 3 points (main icons have text descriptions), 4 points (all icons have text labels), 5 points (icons are intuitive and easy to understand, no text required). - A4. Page Clutter: 1 point (severe pop-up ads), 2 points (multiple promotional entry points), 3 points (limited promotional content), 4 points (focused content), 5 points (minimalist design) Task requirements: 1. Give a score of 1-5 points for each indicator in A1-A4; 2. Each rating should be followed by a detailed explanation of the basis for drawing that conclusion from the image; 3. Summarize the total score and analyze the platform's overall performance, strengths, and areas for improvement in terms of "perceptibility"; 4. If the image is blurry or difficult to judge, please indicate "Unable to judge accurately" and explain the reason.

[0034] Please produce your work in the style of an academic report, using accurate language and clear organization.

[0035] Q2 (operability dimension), Q3 (understandability dimension), and Q4 (supportability dimension) adopt the exact same structure, but are replaced with the scoring criteria of B1-B4, C1-C3, and D1-D3 and the corresponding expert role settings, respectively (see the appendix of this manual for the full details).

[0036] The results obtained from the free analysis are as follows: The output of the first stage is unstructured natural language text, presented in an academic report style. Below is a real output example using q1 for SME_HP_01.png: 1. A1. Font size: 3 points Reason: The body text font size is measured to be approximately 16-17px, falling between 15-18px, which meets the 3-point standard. The title font size is approximately 24px, which is good.

[0037] 2. A2. Color contrast: 5 points Reason: The contrast between the black text (#333333) and the white background (#FFFFFF) is estimated to be approximately 12.6:1, which is far higher than the WCAG AAA level requirement of 7:1, ensuring clear readability.

[0038] 3. A3. Icon recognizability: 4 points Reason: The icons in the bottom navigation bar, such as "Home," "Courses," and "Personal Center," all have text labels; the search icon at the top is a magnifying glass graphic, without text description, but the graphic's meaning is clear. Overall, it meets the 4-point standard of "all icons have text labels."

[0039] 4. A4. Page clutter: 3 points Reason: There are two course promotion banners on the first screen, but they do not obscure the main content area; there are no pop-up ads, and the overall layout is orderly, so it is rated as 3 points.

[0040] The results of the first phase are used as follows: The free analysis text output in the first stage is not directly used for subsequent data calculations. Instead, it serves as part of the large model's dialogue context, providing an informational basis for the structured output in the second stage. Specifically, the system stores the complete dialogue history of the first query in the conversation context; when a second query is initiated, the large model can "remember" the detailed observations and analysis conclusions made about the image beforehand, ensuring that the structured output completely corresponds to the scores and reasons in the original analysis report.

[0041] II. Second Stage: Structured Output of Prompt Words (rq1–rq4) 1. The specific content of rq1–rq4 Taking rq1 (perceptibility dimension) as an example, the details are as follows: Please rewrite your evaluation results for the "Perceptibility" dimensions (A1–A4) strictly according to the following JSON format. Each indicator should be a top-level key (preserving the original Chinese name), and its value should be an object containing two fields: "Score" (integer) and "Notes" (string).

[0042] { "A1. Font Size": { "Score": 3, Note: The body text size is 16-17px, which meets the 3-point standard. }, A2. Color Contrast: { "Score": 5, Note: The contrast ratio between the text and background is approximately 12.6:1, far exceeding the 7:1 threshold. }, "A3. Icon Recognition": { "Score": 4, Note: All main function icons have text labels; the search icon has no text, but its graphic meaning is clear. }, "A4. Page Clutter": { "Score": 3, Note: A small number of course promotion banners exist, but they do not significantly obscure the main content area. } } - Output only the JSON content; do not include any additional explanations, code block markers, or descriptions. - The key name must be exactly the same as the original indicator name (including the explanation in parentheses); - "Fraction" is an integer from 1 to 5; if it cannot be determined, set it to null. - The "Notes" should be concise, objective, and based on observations of the images.

[0043] Output structured JSON according to the above requirements.

[0044] RQ2 (operability dimension), RQ3 (understandability dimension), and RQ4 (supportability dimension) use the exact same structure, but are replaced with indicator key names B1-B4, C1-C3, and D1-D3 respectively, along with corresponding note examples.

[0045] 2. Structured output data format The output of the second stage is a JSON-formatted string. An example is shown below: json { "A1. Font Size": { "Score": 3, Note: The body text size is 16-17px, which meets the 3-point standard. }, A2. Color Contrast: { "Score": 5, Note: The contrast ratio between the text and background is approximately 12.6:1, far exceeding the 7:1 threshold. }, "A3. Icon Recognition": { "Score": 4, Note: All main function icons have text labels; the search icon has no text, but its graphic meaning is clear. }, "A4. Page Clutter": { "Score": 3, Note: A small number of course promotion banners exist, but they do not significantly obscure the main content area. } } 3. How to use the results of the second phase The JSON data output from the second stage is the final output of the data entry module and is directly consumed by the subsequent S3 data calculation module. The specific usage process is as follows: Step 1: Parsing and Verification # Parse a JSON string into a Python dictionary result_dict = json.loads(response[0]['text']) # Validation: Each metric must include "Score" and "Remarks" fields. for indicator, value in result_dict.items(): assert "score" in value, f"{indicator} Missing score field" assert "Note" in value, f"{indicator} Missing note field" assert 1 <= value["score"] <= 5, f"{indicator} score out of range" assert value["Note"].strip() != "", f"{indicator} Note is empty" Step 2: Field Mapping Map Chinese key names in JSON to database field names, for example: “A1. Font Size”.Score → A1_font_score “A1. Font Size”. Note → A1_font_remark “B1. Button Size”.Score → B1_button_score “B1. Button Size”. Notes → B1_button_remark Step 3: Transfer to the calculation module The parsed structured data (one JSON object for each of the four dimensions) along with the page metadata (page_id, platform name, page type, timestamp) are passed into the S3 data calculation module.

[0046] Step 4: Code Implementation Example def deal_a_pic(image_path, query, rquery): agent = LLM_Agent(model='qwen-vl-max') # Phase 1: Free Analysis (Output as a text report, stored in the dialogue context) agent.query([imgpath2dict(image_path), {"text": query}], mode='image') # Second stage: Structured output (output as a JSON string) response = agent.query([{"text": rquery}], mode='image') # Parse JSON and return return json.loads(response[0]['text']) # Complete the four-dimensional evaluation of an image A_results = deal_a_pic("SME_HP_01.png", q1, rq1) # Perceptibility B_results = deal_a_pic("SME_HP_01.png", q2, rq2) # Operability C_results = deal_a_pic("SME_HP_01.png", q3, rq3) # Understandability D_results = deal_a_pic("SME_HP_01.png", q4, rq4) # Supportability In step S3, the data calculation module evaluates the platform's aging-in-time performance based on the following four dimensions: The data calculation module receives structured data from S2, performs calculations and analysis on the data according to preset, refined quantitative scoring rules, and generates objective evaluation results. The scoring rules specify a scoring standard of 1-5 points for 14 secondary indicators across four dimensions: perceptibility (A), operability (B), comprehensibility (C), and supportability (D). Automated scoring is performed based on these scoring standards. For example, in calculating the score for icon recognition (A3), the module parses the "Remarks" field of this indicator in the S2 data and automatically maps the score using keyword matching rules. If the remarks contain "all have text labels," the score is 4 points; if they contain "pure icons with no text," the score is 1 point. For example, the scoring of button size (B1) is achieved by extracting pixel values ​​(width W and height H) from the "Remarks" using a regular expression (e.g., r'(\d+)\s*[×x]\s*(\d+)\s*pixels'), and calculating an objective score based on preset pixel threshold rules. After confirming the scores of all secondary indicators, dimensional scores are calculated: the comprehensive score for each dimension (S_A, S_B, S_C, S_D) is the sum of the scores of all secondary indicators under that dimension, calculated as: S_dimensional = Σ(scores of each indicator under that dimension). Calculation example: Suppose that the scores of the four indicators under the "Perceptibility (A)" dimension of a certain page are: A1=3 points, A2=4 points, A3=4 points, and A4=3 points.

[0047] The overall score for the "perceptibility" dimension of this page is: S_A = 3 + 4 + 4 + 3 = 14 points.

[0048] In step S3, the system generates a corresponding judgment basis field for each scoring item and records the specific observation results of the model to obtain the score, ensuring that the evaluation process is interpretable and traceable.

[0049] Each scoring item corresponds to a "judgment basis" field, storing the specific observation and reasoning process for obtaining the score in text format. The data is structured text, and the content integrates rule analysis from the data acquisition and data processing modules of the multimodal large model. For example, the indicator "B1. Button Size" identifies the visual size of the "Login" button as approximately 50×20 pixels; after rule extraction and calculation, the minimum side (20px) is less than the 30px threshold, and it is rated as 1 point according to the standard.

[0050] The indicator scores, judgment criteria, dimensional comprehensive scores, and page metadata are aggregated into a complete structured evaluation record. This record is the final output of the system and can be directly stored in the database for visualization. A record specifically includes: page metadata fields (page_id (e.g., SME_HP_01), platform_name, evaluation_time), score fields for 14 secondary indicators (A1_font_score, A2_contrast_score, ..., D3_error_score), judgment criteria fields for 14 secondary indicators (A1_font_remark, A2_contrast_remark, ..., D3_error_remark), and comprehensive score fields for 4 dimensions (A_total_score, B_total_score, C_total_score, D_total_score).

[0051] In step S4, the data storage module adopts cloud storage to aggregate the four-dimensional structured evaluation records of each page into a comprehensive evaluation data entry and store it in a relational database or NoSQL database.

[0052] In step S5, the data visualization module uses Power BI tools to present the data, connects to the database through the data connector, creates an age-friendly assessment dashboard, and displays heatmaps, radar charts, and trend line charts of scores for each platform across four dimensions.

[0053] Figure 2 The paper presents a four-dimensional decomposition structure for the age-friendly assessment of public service platforms for lifelong learning. Each dimension contains 3–4 secondary indicators, for a total of 14 quantifiable indicators.

[0054] Figure 3 The following is an example of the evaluation results for the platform's homepage (SME_HP_01.png): The page scored 12 points for "Perceptibility", 10 points for "Operability", 10 points for "Understandability", and 6 points for "Supportability". However, the "Button Size" only scored 2 points, and the color cues need further optimization.

[0055] In terms of data storage, the system stores the evaluation results of each page in the MySQL database table evaluation_results, as shown in Table 3.

[0056] Table 3 Data Structure Table

[0057] In terms of data visualization, the system connects to the database through Tableau to create interactive dashboards that support filtering and viewing by platform, page type, time, and other dimensions.

[0058] The advantages of this invention are: by using a multimodal large model and structured prompting engineering, it realizes the automatic identification, processing and analysis of age-friendly assessment data, which greatly improves the processing efficiency of age-friendly data on the platform interface, reduces the inconsistency of objective factors, supports large-scale age-friendly assessment of lifelong learning public service platforms, and significantly improves the age-friendly governance efficiency of lifelong learning public service platforms in my country.

[0059] Any process or method described in the flowcharts of this invention or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing a particular logical function or process, which can be implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device. The computer-readable medium can be any medium containing a program for storage, communication, propagation, or transmission for use by the execution system, apparatus, or device, including read-only memory, magnetic disks, or optical disks.

[0060] In the description of this specification, the references to terms such as "embodiment," "example," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, those skilled in the art can combine or integrate different embodiments or examples and features described in this specification without creating contradiction. Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A multimodal data intelligent processing method for age-friendly public service platforms, characterized in that: The method includes the following steps: S1. The data collection module collects image data from the data terminal of the public service platform that needs to be assessed for age-friendliness. The collected image data includes screenshots of the platform page. S2. The data entry module inputs the collected image data into the multimodal large model, and uses preset natural language prompts to match, parse and structure the image content to obtain the entry data; S3. The data calculation module calculates and analyzes the submitted data according to the set evaluation dimensions and scoring rules, and generates quantitative scores and judgment criteria for each dimension for subsequent visualization. S4. The data storage module stores the results of the calculation and analysis in the storage medium for subsequent querying, retrieval and use; S5. The data visualization module displays the results on the display device in the form of charts, which makes it easy to intuitively show the age-friendly level of the public service platform.

2. The method according to claim 1, characterized in that, In step S1, the page images of the target platform are periodically collected according to the set collection time points and frequencies.

3. The method according to claim 2, characterized in that, In step S1, the collected image data is in PNG or JPG format; the data collection module is used to collect images of various core pages of the lifelong public education platform, including at least one of the following: homepage, course details page, course list page, personal center page, help and support page, and search results page.

4. The method according to claim 1, characterized in that, In step S2, the multimodal large model adopts Qwen-VL-Max.

5. The method according to claim 4, characterized in that, In step S2, during the data entry process, the system performs JSON format validation on the model output to ensure data consistency and accuracy.

6. The method according to claim 5, characterized in that, In step S2, the system's data entry module employs a two-stage query mechanism to balance analysis depth and output standardization: Phase 1: Load analytical prompts to guide the multimodal large model to perform free analysis on the image; Phase 2: Load structured output prompts and force the multimodal large model to output evaluation results according to a predefined JSON Schema.

7. The method according to claim 1, characterized in that, In step S3, the data calculation module evaluates the platform's aging-in-time performance based on the following four dimensions: A. Perceptual aspects: including font size, color contrast, icon recognizability, and page clutter; B. Operability: including button size, click feedback, navigation hierarchy, and protection against accidental operation; C. Comprehensibility: This includes terminology usage, clarity of instructions, and logical organization of information; D. Supportive: This includes the visibility of the help entry point, the usability of the help content, and the friendliness of error messages.

8. The method according to claim 7, characterized in that, In step S4, the data storage module adopts cloud storage to aggregate the four-dimensional structured evaluation records of each page into a comprehensive evaluation data entry and store it in a relational database or NoSQL database.

9. The method according to claim 1, characterized in that, In step S5, the data visualization module uses Power BI tools to present the data, connects to the database through the data connector, creates an age-friendly assessment dashboard, and displays heatmaps, radar charts, and trend line charts of scores for each platform across four dimensions.

10. An age-friendly multimodal data intelligent processing system for public service platforms, characterized in that: The system includes a data collection module, a data entry module, a data calculation module, a data storage module, and a data visualization module; the system is used to implement the age-friendly multimodal data intelligent processing method for public service platforms according to any one of claims 1-9.

Images