A page generation method, device, equipment and computer readable storage medium

Abstract

The application discloses a page generation method and device, equipment and computer readable storage medium, apply to computer technology field, including: according to demand theme generates including multiple sub argument demand landing page outline, according to each sub argument and template knowledge base in template label determine with each sub argument corresponding target card HTML template;According to each sub argument and target card HTML template generates with each sub argument corresponding search term, according to search term searches the content to be filled corresponding with each sub argument;According to the content to be filled and target card HTML template utilize rendering prompt word generates the card HTML after filling content, based on the card HTML after filling content utilize page visual prompt word generates target page.The application will determine target card HTML template and the text to be filled based on each sub argument, thereby directly generating a page, improve the accuracy and efficiency of target page generation.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a page generation method, apparatus, device, and computer-readable storage medium. Background Technology

[0002] Existing technologies typically include manual design, large language model-assisted generation, and simple template library application. Manual design requires designers to complete information hierarchy sorting, layout drawing, and style specification definition, resulting in a long overall delivery cycle. While large language model-assisted generation can generate line art descriptions or simple layouts based on set requirements, the generated results are prone to issues such as monotonous layout styles and incompatibility with the set requirements. Simple template library application typically relies on keywords or visual similarity for template selection, lacking an understanding of the required scenario, information hierarchy, and desired effect, leading to insufficient template recall accuracy.

[0003] It is evident that improving the accuracy and efficiency of page generation is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a page generation method, apparatus, device and computer-readable storage medium, which solves the technical problems of low accuracy and efficiency of page generation in the prior art.

[0005] To solve the above technical problems, the present invention provides a page generation method, comprising: A template knowledge base is determined; the construction process of the template knowledge base includes processing the landing page card component template image to obtain a card HTML template, tagging the card HTML template to obtain template tags, and constructing the template knowledge base based on the card HTML template and the template tags; the template tags include card description, adaptation text, adaptation scenario, card title, and requirement level; Generate a landing page outline for the requirements based on the requirements theme, including multiple sub-arguments, and determine the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base. Based on each sub-argument and the target card HTML template, generate search terms corresponding to each sub-argument, and retrieve the content to be filled corresponding to each sub-argument based on the search terms; Based on the content to be filled and the target card HTML template, rendering prompts are used to generate the card HTML with the filled content, and the target page is generated based on the card HTML with the filled content using page visual prompts.

[0006] Optionally, the landing page card component template image is processed to obtain a card HTML template, including: inputting the landing page card component template image into a multimodal large language model, enabling the multimodal large language model to recognize the text area, image area, icon area, button area, border style, and hierarchical structure in the landing page card component template image; generating the card HTML template based on the recognition results output by the multimodal large language model; wherein, the container structure in the card HTML template is generated according to a preset container naming rule, and the icon content, icon size, and icon style description are recorded; and the nested structure of the card HTML template is generated according to a preset hierarchical structure specification.

[0007] Optionally, before obtaining the template knowledge base, the method further includes: generating the card description based on the layout and purpose of the card HTML template; generating the adaptation text based on the text content of the card HTML template; generating the adaptation scenario based on the requirement scenario of the card HTML template; generating the card title based on the adaptation scenario and determining the requirement level of the card HTML template; and associating and storing the card HTML template, the card description, the adaptation text, the adaptation scenario, the card title, and the requirement level to obtain the template knowledge base for template retrieval.

[0008] Optionally, a landing page outline including multiple sub-arguments is generated based on the demand theme, and a target card HTML template corresponding to each sub-argument is determined based on each sub-argument and the template tags in the template knowledge base. This includes: generating the landing page outline including a visual anchoring layer, a logical construction layer, a threshold resolution layer, and an action conversion layer based on the demand theme; wherein, each sub-argument includes the visual anchoring layer, the logical construction layer, the threshold resolution layer, and the action conversion layer; generating a demand objective, demand strategy, and content description corresponding to each sub-argument; matching the demand objective, demand strategy, and content description of each sub-argument with the template tags in the template knowledge base; and determining the target card HTML template corresponding to each sub-argument based on the matching results.

[0009] Optionally, after generating the filled card HTML using rendering prompts based on the content to be filled and the target card HTML template, and generating the target page based on the page visual prompts, the method further includes: obtaining the page evaluation level, custom text evaluation, and page modification instructions of the target page; wherein, the page evaluation level is determined by visual design, layout structure, interactive experience, requirement transformation, and scene adaptability; performing semantic parsing on the page modification instructions and converting the page modification instructions into structured page evaluations; and adjusting the models involved in the page generation process based on the page evaluation level, custom text evaluation, and structured page evaluation.

[0010] Optionally, a search term corresponding to each sub-argument is generated based on each sub-argument and the target card HTML template, and the content to be filled corresponding to each sub-argument is retrieved based on the search term, including: calling a search tool to retrieve information based on the search term; and using a multimodal large language model to assemble the retrieved information into the content to be filled, consisting of image descriptions and image links.

[0011] Optionally, based on the content to be filled and the target card HTML template, a card HTML with filled content is generated using rendering prompts, and a target page is generated based on page visual prompts. This includes: processing the content in the target card HTML template based on the content to be filled using the rendering prompts to obtain the card HTML with filled content; and designing the page based on the page visual prompts to obtain the target page. The design content includes a header image, card borders, and page color scheme.

[0012] The present invention also provides a page generation apparatus, comprising: The template knowledge base determination module is used to determine the template knowledge base. The construction process of the template knowledge base includes processing the landing page card component template image to obtain the card HTML template, performing tagging processing on the card HTML template to obtain template tags, and constructing the template knowledge base based on the card HTML template and the template tags. The template tags include card description, adaptation text, adaptation scenario, card title, and requirement level. The target card HTML template building module is used to generate a landing page outline including multiple sub-arguments based on the demand theme, and to determine the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base. The module for determining the content to be filled is used to generate search terms corresponding to each sub-argument based on each sub-argument and the target card HTML template, and to retrieve the content to be filled corresponding to each sub-argument based on the search terms; The target page generation module is used to generate a card HTML with filled content based on the content to be filled and the target card HTML template using rendering prompts, and to generate a target page based on the card HTML with filled content using page visual prompts.

[0013] The present invention also provides a page generation device, comprising: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of the page generation method described above.

[0014] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the page generation method described above.

[0015] As can be seen, this invention generates a landing page outline including multiple sub-arguments based on the required theme, and determines the target card HTML template based on each sub-argument and template tags, so that each sub-argument can be matched with a page card structure that is suitable for its required expression goal; by generating search terms based on the sub-arguments and target card HTML templates and retrieving the content to be filled, the content to be filled is associated with the sub-arguments and template structure; by generating target pages through rendering prompts and page visual prompts, the card content filling and page visual design can be completed continuously. Therefore, this invention can improve the adaptability between page templates and required content, reduce the cost of manually selecting templates and organizing materials, and improve the efficiency and accuracy of target page generation.

[0016] In addition, the present invention also provides a page generation apparatus, device, and computer-readable storage medium, which also have the above-mentioned beneficial effects. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 A flowchart of a page generation method provided in an embodiment of the present invention; Figure 2 A framework diagram for determining the content to be filled, provided in an embodiment of the present invention; Figure 3 A feedback adjustment framework diagram provided for an embodiment of the present invention; Figure 4 A system architecture diagram of a page generation method provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a page generation device provided in an embodiment of the present invention; Figure 6 This is a schematic diagram of the structure of a page generation device provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0020] Some terms that appear in the description of the embodiments of this application are subject to the following interpretation: Large Language Models (LLMs), also known as large-scale language models, are artificial intelligence models designed to understand and generate human language. They are trained on massive amounts of text data and can perform a wide range of tasks, including text summarization, translation, and sentiment analysis. LLMs are characterized by their massive scale, containing billions of parameters that help them learn complex patterns in language data. These models are typically based on deep learning architectures, such as transformers, which contributes to their impressive performance on various NLP tasks.

[0021] Knowledge base retrieval refers to the process by which a system extracts relevant information in real time from a structured / unstructured data pool using vectorization or indexing techniques during task execution. In this project's architecture, the retrieval occurs within the template and resource libraries. It provides pre-defined design specifications, industry knowledge, and requirement logic as background context to the model, ensuring that the generated results are industry compliant and brand consistent, thereby eliminating the randomness and illusions inherent in the native generation of large models.

[0022] A marketing landing page is a specific web page that carries a specific user intent and aims to guide users to complete a conversion action (such as depositing funds or seeking information). It is a key node in the demand funnel, characterized by highly concentrated information logic and visual impact. In automated production scenarios, the landing page is deconstructed into a modular combination of visual structure (HTML), content (copywriting), and brand style (styling), representing the final digital asset delivered by the collaborative system.

[0023] Bocha Search is a web-based enhanced search service designed specifically for AI agents and large language models. Unlike traditional search engines aimed at general users, Bocha focuses on semantic ranking and high-value information extraction. In this project architecture, Bocha Search acts as the system's real-time sensing organ, responsible for capturing the latest financial information and market trends from across the web. Through its structured data interface, the system can transform the real-time dynamics obtained from searches (such as the latest industry reports or stock market fluctuations) into context that the model can understand.

[0024] ReAct (Reasoning) is a large-scale intelligent decision-making framework that combines logical reasoning with practical action, breaking the limitations of models that merely think and output. Its operating mode is a cyclical process of reasoning and action: first, it analyzes and determines the root cause of the problem based on information analysis, then formulates optimization strategies; next, it implements corresponding adjustments; after completion, it reviews and evaluates the results, and analyzes and corrects them again based on feedback.

[0025] Please refer to Figure 1 , Figure 1 A flowchart illustrating a page generation method provided in an embodiment of the present invention. The method may include: S101, Determine the template knowledge base; wherein, the construction process of the template knowledge base includes processing the landing page card component template image to obtain the card HTML template, tagging the card HTML template to obtain template tags, and constructing the template knowledge base based on the card HTML template and template tags; the template tags include card description, adaptation text, adaptation scenario, card title and requirement level.

[0026] Each step in this embodiment can be executed by a designated electronic device, which can be a server, a portable terminal, or other form. The electronic device contains memory modules, the specific number of which is not limited. The landing page card component template image in this embodiment can be a screenshot of a single card component in a pre-designed landing page, or a partial component image extracted from an existing landing page. This embodiment does not limit the source of the landing page card component template image; for example, the landing page card component template image can be an image exported from a design tool; or it can be a screenshot of a historical landing page; or it can be a manually uploaded card component illustration. The card HTML (HyperText Markup Language) template in this embodiment refers to an HTML template that can represent the page structure of the card component. The card HTML template can include text containers, image containers, icon containers, button containers, border structures, and hierarchical nesting structures. This embodiment does not limit the specific format of the card HTML template; for example, the card HTML template can be a template containing HTML tags and style class names; or it can be an HTML code snippet containing placeholders; or it can be a structured template formed by combining HTML code snippets and style descriptions. In this embodiment, the template tags are used to describe the applicable characteristics of the card HTML template. Specifically, the card description describes the layout and intended purpose of the card HTML template; the adaptation text describes the text type that the card HTML template is suitable to carry; the adaptation scenario describes the application scenario that the card HTML template is suitable for; the card title summarizes the theme that the card HTML template is suitable for; and the requirement level describes the display level of the card HTML template in the landing page. This embodiment does not specify a specific requirement level type; for example, it can be a marketing level, a display level, or other levels specific to vertical businesses.

[0027] It should be further explained that, based on any of the above embodiments, the above processing of the landing page card component template image to obtain a card HTML template may include: inputting the landing page card component template image into a multimodal large language model, enabling the multimodal large language model to recognize the text area, image area, icon area, button area, border style, and hierarchical structure in the landing page card component template image; generating a card HTML template based on the recognition results output by the multimodal large language model; wherein, the container structure in the card HTML template is generated according to a preset container naming rule, and the icon content, icon size, and icon style description are recorded; and the nested structure of the card HTML template is generated according to a preset hierarchical structure specification. In this embodiment, the multimodal large language model refers to a large language model capable of processing and understanding multiple modal information such as text, images, audio, and video. This embodiment does not limit the specific type of multimodal large language model, as long as the multimodal large language model can recognize the results based on the image input and output areas. For example, the multimodal large language model may be a cloud-based model supporting image understanding and text generation; or it may be a visual language model deployed on a local server. This embodiment can generate the container structure in the card HTML template according to a preset container naming rule, and record the icon content, icon size, and icon style description; this embodiment can also generate the nested structure of the card HTML template according to a preset hierarchical structure specification. This embodiment is not limited to the preset container naming rule. For example, the preset container naming rule can be to generate container names according to region type; or it can be to generate container names according to region type and display order.

[0028] It should be further noted that, based on any of the above embodiments, the process of obtaining the template knowledge base may further include: Step 1: Generate card descriptions based on the layout of the card HTML template and the intended purpose.

[0029] For example, it can be for marketing purposes.

[0030] Step 2: Generate adapted text based on the text content of the card's HTML template.

[0031] Step 3: Generate an adaptation scenario based on the requirements of the card HTML template.

[0032] The requirement scenario in this embodiment can be a marketing scenario.

[0033] Step 4: Generate card titles based on the appropriate scenario and determine the required hierarchy of the card HTML template.

[0034] Step 5: Link and store the card HTML template, card description, matching text, matching scenario, card title and requirement hierarchy to obtain a template knowledge base for template retrieval.

[0035] This embodiment does not limit the storage method of the template knowledge base. For example, the template knowledge base can be stored in a local database; or in a remote database; or as a structured file. The key is to associate and store the card HTML templates and template tags together, and support subsequent retrieval and matching based on template tags. This embodiment transforms visual card templates collected from various landing pages, automatically tags them, and converts them into HTML templates usable in subsequent processes. This embodiment provides a method for generating the template knowledge base, improving the accuracy of template knowledge base generation.

[0036] Specifically, the process of generating a template knowledge base can include: collecting landing page card component templates, most of which come from the open-source community, are in image format, and are provided by operations personnel. First, the component images are reproduced using HTML using a multimodal large language model. The reproduced HTML needs to follow certain paradigms, specifically in container naming and hierarchical structure standardization. For example, for images and icons appearing in the template, they need to be placed in containers (unit elements in the HTML code) named with fixed characters, namely `image_placehold` and `icon_placehold`, and the specific content, size, and style description of the image should be commented within them. The nested structure of all containers also needs to be designed, as detailed in the prompts. This facilitates precise location and modification of the HTML code during subsequent in-depth page optimization, allowing for localized modifications to elements, making the subsequent prompt optimization project easier. Specific prompts are as follows: Please reproduce this card page screenshot using HTML code according to the following requirements.

[0037] The card is laid out directly on the scale-container without a main-container; All units must use rem units, 1rem equals 20px. The scale-container width is 750px, the width adapts to 100% of the scale-container, and the height must be variable to adapt to the card content without exceeding it. It adapts to various mobile device sizes, ensuring that the relative layout is not changed on various width devices such as 750px, 375px, and 414px. After adjusting the font size, the icons, margins, and spacing should also be adjusted proportionally. If the original page contains images or icons that you cannot generate, use image-placehold for the image container and icon-placehold for the icon container. If not, please reproduce the entire page. Note that you should not change the layout. You need to replicate this page 100%, and you cannot use online styles. It must be implemented entirely locally.

[0038] The cards are comprehensively tagged, encompassing the following dimensions: card description, matching text, matching context, card title, and the level of requirement to which it belongs. The following are the prompts: Please summarize the usage instructions for this template according to the following three points, and keep the summary to no more than 500 words. Output in Markdown format.

[0039] Output example: Card description; 1. Please describe the content of this card, including its layout and desired effect; Adapted text; 2. Please explain what kind of text is suitable for using this template; Adapt to the scenario; 3. List the scenarios in which this layout is suitable for use; Card title; 4. Choose the most suitable scenario and condense it to no more than 15 characters as the title; The existing general market data conversion page strategy outline (Marketing Objective Framework) is as follows: Phase 1: The Visual Anchoring Layer (The Trigger); Target objectives: [Cognitive arousal] and [Emotional resonance]; Core task: Within 0-3 seconds of a user opening a page, instantly confirm the user's click intent and create a strong visual impact, making the user feel tense or excited that "something big is happening here"; Key strategic points: Trend visualization: Utilizing color psychology (such as a high-saturation background with red indicating rises and green indicating falls) to directly convey market sentiment; Status quo confirmation: Visually displays the current extreme state (such as breakthrough, extreme value, anomaly) to verify the authenticity of external advertisements / push information; Phase Two: The Conviction Layer; Target objectives: [Impulse rationalization] and [Multidimensional conviction]; Core task: To transform users' initial emotions into rational decision-making basis. Through arguments from different dimensions, to resolve users' inner doubts about "why now" and "how certain is the outcome?" Modular strategy (multiple combinations possible): Module A - Attribution Dimensions: Objective: To establish causal relationships. To identify the core driving forces behind the current fluctuations (whether macroeconomic, supply and demand, or other events), making the market movements appear "reasonable." Module B - Game Theory Dimension: Purpose: To create a herd mentality (Social Proof). This involves showcasing fund flows, changes in holdings, or the balance of power between buyers and sellers to demonstrate that a market consensus has been formed. Module C - Timing Dimension: Objective: To confirm entry points. Through pattern recognition or key price level indicators, it suggests that the current period is within the optimal risk-reward window. Module D - Authority Dimension: Objective: Credit endorsement. Utilizing the judgment of third parties (institutions, experts, models) to reduce the psychological burden of decision-making for users; Phase 3: The Feasibility Layer; Target objectives: [Mental accounting anchoring] and [Admission downgrading]; Core mission: To address the question of "Can I participate?" This involves transforming complex futures contract rules into easily understandable, "low-threshold" concepts, eliminating fears of high risk and high capital investment. Key strategic points: Visualizing funds: Instead of displaying the total contract value, only the minimum entry amount (margin) is shown, lowering users' psychological barriers; Volatility Valuation: Clearly inform the profit or loss amount corresponding to the smallest unit of volatility, and establish specific profit expectations; Transparent Rules: The transaction code and core rules are presented in a minimalist manner, existing as a tool attribute; Phase Four: The Conversion Layer; Requirement objectives: [Path guidance] and [Final step]; Core task: To gather all attention and provide a unique, least-resistance execution path; Key strategic points: Strong interactive guidance: Button design should have clear action instructions (CTA) to transform "seeing" into "doing"; Tiered conversion: Provide two-way exit based on user maturity (aggressive users trade directly, conservative users experience the simulation). Please carefully read the card outline and hierarchy above, analyze which level the template below belongs to, and output it directly.

[0040] template; {template_content}; Output example; Phase Four: The Conversion.

[0041] All template tag text is organized together and used as an index for template retrieval. After the user enters the information they want to display, it is matched against this index one by one to find the most suitable template, and then the populated information is rendered in the template's HTML. Below is a complete tag example for a template: Card description; 1. This card uses a combination of "data visualization + incentive" layout. The upper part uses a bar chart to visually display the price increase trend over a time span, building user confidence in the product's added value through data comparison; the lower part highlights "Newcomer Benefits" and a prominent "Claim Now" button, using specific rewards (such as 30mg of gold) to stimulate user conversion. The overall color scheme uses a gold-red gradient to create a sense of wealth and urgency, resulting in a strong demand effect, making it suitable for inducing users to invest or register. Adapted text; 2. Suitable for texts with "historical comparison" and "instant reward" attributes: Trend comparison text: This type of text needs to show comparative data on price increases, revenue increases, user growth, etc., over time (e.g., 20xx vs 20xx). Promotional incentive text: Required text that includes specific amounts, gift names, and conditions for receiving the gift (e.g., give away XX yuan, or XX yuan worth of cash that can be withdrawn). Trust endorsement text: brief description of brand advantages or security guarantees (e.g., value preservation and appreciation, official certification); Adapt to the scenario; 3. This layout is suitable for the following scenarios: Financial product promotion: Showcasing the historical returns of financial products such as gold, funds, and stocks; Collectibles / Asset Appreciation: Displays market price trends; Membership / Benefits Upgrade: Showcases the year-on-year increase in the value of membership benefits, guiding members to upgrade; E-commerce promotional pre-sale: Compare historical price trends, highlight current discounts, and include coupons for claiming. Card title; 4. The appreciation trend of wealth management products and customer acquisition activities; To which level; 5. Action Conversion Layer.

[0042] S102, Generate a landing page outline for the requirements based on the requirements theme, including multiple sub-arguments, and determine the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base.

[0043] In this embodiment, the demand theme refers to the demand object or activity theme that the target page needs to express. For example, the demand theme can be a product promotion theme, an activity conversion theme, a service introduction theme, or a brand promotion theme. The demand landing page outline in this embodiment refers to the page content organization structure generated around the demand theme. Sub-arguments refer to the content units in the demand landing page outline used to express different demand intentions.

[0044] It should be further explained that, based on any of the above embodiments, the above page generates a landing page outline including multiple sub-arguments according to the demand theme, and determines the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base, which may include: S1021, Generate a requirement landing page outline based on the requirement theme, including a visual anchoring layer, a logical construction layer, a threshold resolution layer, and an action conversion layer; among which, the sub-arguments include the visual anchoring layer, the logical construction layer, the threshold resolution layer, and the action conversion layer; S1022, Generate the requirement objectives, requirement strategies and content descriptions corresponding to each sub-argument; S1023, match the demand objectives, demand strategies and content descriptions of each sub-argument with the template tags in the template knowledge base; S1024, determine the target card HTML template corresponding to each sub-argument based on the matching results. This embodiment can generate a landing page outline including a visual anchoring layer, a logical construction layer, a threshold resolution layer, and an action conversion layer based on the required theme. In this embodiment, the visual anchoring layer can be used to create visual appeal and theme positioning at the front of the page; the logical construction layer can be used to elaborate on product or activity value; the threshold resolution layer can be used to explain user concerns, usage thresholds, or participation thresholds; and the action conversion layer can be used to guide users to perform conversion actions such as clicking, consulting, registering, or purchasing. This embodiment does not limit the matching method. For example, sub-argument information can be matched with template tags based on keyword similarity; or sub-argument information can be matched with template tags based on semantic vector similarity; or the matching result can be determined by combining keyword similarity and semantic similarity. This embodiment can determine the target card HTML template corresponding to each sub-argument based on the matching results. Specifically, when the demand objective of a sub-argument matches the demand hierarchy of a card HTML template, and the content description of the sub-argument matches the appropriate text and context of the card HTML template, then the card HTML template can be identified as the target card HTML template corresponding to that sub-argument. This allows different sub-arguments to use card structures adapted to their respective expression objectives. This embodiment provides a specific method for determining the target card HTML template corresponding to each sub-argument, improving the accuracy of matching.

[0045] S103, generate search terms corresponding to each sub-argument based on each sub-argument and the target card HTML template, and retrieve the content to be filled corresponding to each sub-argument based on the search terms.

[0046] In this embodiment, the search terms are information generated based on the sub-arguments and the target card HTML template to retrieve the content to be filled. The search terms in this embodiment may include at least one of the following: demand topic, sub-argument keywords, card title, applicable scenario, and content type. The content to be filled in this embodiment refers to the material content that needs to be filled into the target card HTML template, such as image descriptions, image links, text snippets, title content, or button text.

[0047] It should be further explained that, based on any of the above embodiments, the process of generating search terms corresponding to each sub-argument and the target card HTML template, and retrieving the content to be filled corresponding to each sub-argument based on the search terms, may include: calling a search tool to retrieve information based on the search terms; and using a multimodal large language model to assemble the retrieved information into content to be filled consisting of image descriptions and image links. This embodiment does not limit the specific type of search tool; for example, the search tool may be a web search interface; or it may be a material library search interface; or it may be an internal enterprise content library search interface.

[0048] This embodiment can generate search terms based on both sub-arguments and target card HTML templates, rather than just based on the required topic. This allows the retrieved content to be filled to simultaneously adapt to the expression requirements of the sub-arguments and the display structure of the target card HTML template, thereby reducing the probability of search results not matching the template structure.

[0049] Specifically, this embodiment involves several steps: determining the requirement topic, planning outline, template retrieval, information retrieval, and text planning. It completes the planning of the requirement text from the requirement topic to the adapted template. For easier understanding, please refer to... Figure 2 , Figure 2 This invention provides a framework diagram for determining content to be filled. Specifically, it may include: Step 1: Input the marketing theme and output the planning outline. The outline needs to conform to the basic methodology of marketing.

[0050] It should be noted that the marketing theme is only one application scenario of this invention, and it can be applied to content and media scenarios, enterprise services, etc.

[0051] Step 2: The outline is structured as sub-arguments, each with a title, justification, and brief description (see outline). A suitable template is retrieved based on each sub-argument. Specific search terms are planned for each sub-argument according to the content of this template. For example, the sub-argument about low-barrier participation in precious metals best matches audience conversion template A, which includes information on target price increases, conversion buttons, and trend introductions. The demand is then broken down into keywords based on the sub-arguments and template content for searching: futures trading qualifications, precious metal price increases, precious metal trends, and advantages of current deposits. The relevant keywords are retrieved using a search tool like Bocha, enriching the specific content of each sub-argument, and then filling the template with this content. Furthermore, all illustrations in the retrieved information are processed using a multimodal large language model, organizing them into a contextual information consisting of image descriptions and links. This embodiment involves two searches: the first search plans and designs the sub-arguments, then retrieves the template; the second search continues searching using elements from the template content. This ensures a strong fit between the retrieved content and the template, facilitating content filling.

[0052] S104. Based on the content to be filled and the target card HTML template, use rendering prompts to generate the card HTML with the content filled, and use page visual prompts to generate the target page based on the card HTML with the content filled.

[0053] In this embodiment, rendering prompts refer to the prompts used to guide the large language model in filling the target card's HTML template with the content to be filled. These prompts can constrain the position of the content to be filled, text length, image usage, and button content, etc. In this embodiment, page visual prompts refer to the prompts used to guide the model in designing the page visuals based on the card's HTML after the content has been filled. These page visual prompts can constrain the target page's header image, card borders, page color scheme, overall layout, and visual style, etc.

[0054] It should be further explained that, based on any of the above embodiments, the process of generating a card HTML with filled content using rendering prompts based on the content to be filled and the target card HTML template, and generating a target page based on page visual prompts, may include: processing the content in the target card HTML template based on the content to be filled using rendering prompts to obtain the card HTML with filled content; and designing the page based on the card HTML with filled content using page visual prompts to obtain the target page; wherein the design content includes a header image, card borders, and page color scheme. This embodiment does not limit the specific form of the target page. For example, the target page may be an HTML page file; or it may be a web resource package that can be displayed in a browser; or it may be a landing page obtained by combining multiple card HTMLs with filled content.

[0055] Specifically, the rendering process described above may include: loading the HTML of the corresponding template, and replacing the original text in the template HTML with the text from the actual text requirements draft using design prompts. Attention should be paid to details in the prompt design, such as removing redundant card styles. For card templates lacking text, appropriate text should be added, and the content should be explained using a combination of text and images. For previously missing `image_placehold` values ​​in the template, the most suitable image should be found within the image information context and inserted into the corresponding position. The input is the text to be filled (i.e., the content to be filled) and the template HTML (i.e., the target card HTML template); the output is the card HTML after the required text has been filled.

[0056] Specifically, the above page design is used to unify and harmonize the overall page design by adjusting the colors of the various encapsulated card modules and main and subheadings. The design process may include: Input: HTML for each card after filling in the requirement text, landing page design prompts, main title, and subtitle. Step 1: Design the header image based on the main title. Establish header image templates for each requirement scenario—including color schemes, header text size, image element design, and constraints on the product's logo, etc. These constraints will be fixed. The user inputs the main title, subtitle, and central idea of ​​the current requirement page. This will generate a header image that meets online requirements.

[0057] Step Two: Next, design the card borders, which consist of border and header elements. Define a consistent border style, such as line thickness and corner radius. Make personalized border adjustments to suit the specific needs of the scenario. For example, a promotional scenario might require a more prominent card border, while a financial scenario might need simplified decorative elements. Different headers should be chosen for different scenarios. Verify the visual consistency of each module, such as ensuring a unified style between title cards and content cards, and output a draft of the adjusted layout. Maintain a consistent type of header border, card components, and decorative elements.

[0058] Step 3: Fill in the card's HTML content and, based on the header image's color, color-matching the entire page. Color hints will be provided by the designer. The overall scheme is based on the header image's color, referencing template styles, and scientifically planning the mapping relationship between the colors of various page elements and the header image's color. Once the color scheme is complete, you will have the final landing page HTML.

[0059] It should be further explained that, based on any of the above embodiments, after generating the filled card HTML using rendering prompts based on the content to be filled and the target card HTML template, and generating the target page based on the page visual prompts, the method may further include: obtaining the page evaluation level, custom text evaluation, and page modification instructions of the target page; wherein, the page evaluation level is determined by visual design, layout structure, interactive experience, requirement transformation, and scene adaptability; performing semantic parsing on the page modification instructions and converting the page modification instructions into structured page evaluations; and adjusting the models involved in the page generation process according to the page evaluation level, custom text evaluation, and structured page evaluations. This embodiment can perform semantic parsing on the page modification instructions and convert the page modification instructions into structured page evaluations. This embodiment does not limit the specific fields of the structured page evaluation. For example, the structured page evaluation may include the part to be adjusted, the object to be adjusted, the direction of adjustment, the intensity of adjustment, and the reason for adjustment; or it may include visual design evaluation, layout structure evaluation, interactive experience evaluation, requirement transformation evaluation, and scene adaptability evaluation. The models involved in this embodiment may include at least one of the following: a model for generating a landing page outline, a model for matching target card HTML templates, a model for generating search terms, a model for generating content-filled card HTML, and a model for generating the target page. By converting page modification instructions into structured page evaluations, subsequent model adjustments can have a parsable evaluation basis, thereby reducing the problem of difficulty in reusing evaluation information when adjustments are made solely based on natural language evaluations.

[0060] Specifically, the model involved in the aforementioned page generation process is actually an evaluation feedback and optimization process. It aims to create a complete closed loop encompassing user evaluation, behavioral feedback, AI (artificial intelligence) assessment, automatic iteration, and model evolution. Unlike traditional fixed generation models, this module relies on subjective user evaluations and implicit operational behaviors, combined with a self-developed large-scale evaluation model for demand pages and the ReAct inference mechanism. Through three rounds of standardized iterations to optimize page effects, the system achieves continuous autonomous upgrades, effectively improving the visual quality, scene adaptability, and conversion rate of demand pages. The system sets five page evaluation levels: A, B, C, D, and E, covering five dimensions: visual design, layout structure, interactive experience, demand conversion, and scene adaptability. It is open to all roles, with quantifiable and implementable level standards, allowing for rapid page quality assessment: Level A is excellent and can be directly launched; Level B is good and requires only minor adjustments; Level C is acceptable and requires targeted optimization; Level D needs optimization and requires comprehensive rectification; and Level E is unacceptable and requires regeneration. It supports both quick rating and custom text evaluation modes, balancing evaluation efficiency and accuracy. Users can complete AE level scoring with one click, and the system automatically archives data such as page tags and generation time. They can also manually input text feedback on page issues and optimization suggestions; all feedback data is entered into the database in real time, providing a basis for subsequent iterations. The system captures user page modification commands and, through semantic parsing and reverse reasoning, transforms user-initiated optimization behaviors into standardized page evaluations, uncovering implicit needs. For example, if a user commands "change the overall color to a warm tone," the system can automatically convert it into a structured evaluation of "the page's cool tone does not match the required scenario," filling the information gaps in purely text-based evaluations and maximizing the use of user operation data. Based on the finely tuned requirement page evaluation model, a ReAct reasoning + action mechanism is used to complete three rounds of closed-loop iteration and refinement. The first round addresses explicit page flaws based on user feedback, completing basic optimizations; the second round focuses on details, optimizing visual hierarchy and requirement conversion paths to improve scenario adaptability; the third round comprehensively reviews and verifies the effects, avoiding over-optimization and outputting the final, high-quality page. The system automatically archives all evaluation data, user modification commands, three rounds of iteration logs, and samples before and after page optimization, continuously accumulating a dataset of custom-designed requirement pages for incremental model fine-tuning. This continuously improves the model's ability to identify problems and make optimization decisions, achieving a self-evolving effect where the system becomes more accurate with use. The layers are: user interaction layer, data preprocessing layer, AI intelligent evaluation layer, iterative optimization execution layer, and data accumulation and evolution layer. Each layer works in tandem to complete the entire intelligent iteration process. The core advantages of this module are comprehensive evaluation dimensions, efficient utilization of user behavior data, standardized iteration process, and the system's ability to continuously evolve autonomously, significantly reducing the cost of manual intervention and steadily improving the quality of AI-generated requirement pages. For easier understanding, please refer to [link / reference needed]. Figure 3 , Figure 3 This is a feedback adjustment framework diagram provided for an embodiment of the present invention.

[0061] The page generation method provided in this invention involves determining a template knowledge base, generating a landing page outline including multiple sub-arguments based on the required theme, and determining the target card HTML template corresponding to each sub-argument based on the template tags in the template knowledge base. Then, it generates search terms based on each sub-argument and the target card HTML template, and retrieves the content to be filled based on the search terms. Finally, it generates the card HTML with the filled content using rendering prompts based on the content to be filled and the target card HTML template, and generates the target page based on the card HTML with the filled content using page visual prompts. As can be seen, compared with manually selecting templates and manually filling in materials, this invention can achieve matching between sub-arguments and card HTML templates through template tags, establish the association between sub-arguments, target card HTML templates, and content to be filled through search terms, and continuously generate the target page through rendering prompts and page visual prompts. This improves the adaptability of page content to template structure, reduces manual processing costs, and improves the accuracy and efficiency of page generation.

[0062] For a clearer understanding of this invention, please refer to the following details. Figure 4 , Figure 4 This is a system architecture diagram of a page generation method provided by an embodiment of the present invention. The present invention is a requirement landing page HTML generation system based on a multimodal large language model. Specifically, it includes the following modules: a template processing module, a requirement copy planning module, a template copy rendering module, and a page visual module. First, the template processing module is a preprocessing module that supports users uploading visual card templates, continuously enriching the template library. In actual use, the first step is for the user to input the requirement theme into the requirement copy planning module. This module searches for real-time information and plans it into requirement text according to the requirement methodology, and retrieves the most suitable template based on the content. The template copy rendering module completes the precise assembly of the copy and template structure. It fills the loose text and images into the selected template frame HTML. Finally, the page visual module performs end-of-line optimization. After receiving the initially generated HTML page, it globally optimizes the header image, layout, and colors according to visual specifications and color scheme prompts, ensuring that the generated page not only matches the content but also possesses commercial-grade aesthetics. Through this connection of asset digitization, content structuring, automatic assembly, and visual processing, the system solves the problems of low efficiency in manual typesetting and the disconnect between copy and design in traditional requirement design. The specific process may include: The template processing module receives a landing page card component template image and uses it as a template to be processed. This image represents the card component style in the landing page for market information requirements. A multimodal large language model is used to recognize the landing page card component template image, and it is then reproduced as a card HTML template. This card HTML template has a page layout corresponding to the landing page card component template image. The container structure in the card HTML template is generated according to a preset container naming rule; image positions are labeled with `image_placehold`, and icon positions with `icon_placehold`. Image content, image size, and image style descriptions are recorded in the source code of the card HTML template, along with icon content, icon size, and icon style descriptions. A nested structure of the card HTML template is generated according to a preset hierarchical structure specification. The card HTML template is tagged to obtain template tags; these tags include card description, matching text, matching scenario, card title, and the required level. A template knowledge base is built based on the card HTML template and template tags; this knowledge base is used to recall matching card HTML templates based on the required content.

[0063] The requirement copywriting planning module receives the user's input requirement theme and generates a requirement landing page outline based on the theme. This outline includes multiple sub-arguments. The module determines the requirement level of each sub-argument based on a pre-defined market conversion page strategy outline. These requirement levels include visual anchoring, logical construction, threshold resolution, and action conversion. Based on the requirement title, requirement level, design rationale, and content description of each sub-argument, the module retrieves the corresponding card HTML template from the template knowledge base. Search terms are generated based on each sub-argument and its corresponding card HTML template; these terms are used to retrieve real-time information corresponding to the sub-argument. The module uses search tools to retrieve information based on the search terms and organizes the retrieved information according to the central idea of ​​each sub-argument into the data required for the template layout. Illustrations in the retrieved information are processed to obtain image context information, including image descriptions and links. Finally, text to be filled is generated based on the sub-arguments, the retrieved information, and the corresponding card HTML template; this text corresponds to the content area in the card HTML template. Load the corresponding card HTML template and replace the existing text in the card HTML template with the text to be filled. Based on the image context information, insert the image corresponding to the `image_placehold` in the card HTML template. Refine the card HTML template after filling in the text, removing redundant styles that do not match the current text, and adding explanatory text when the text content is insufficient, resulting in the card HTML template filled with the required text.

[0064] The page visual module receives the main title, subtitle, and central idea of ​​the page, and generates a header image based on these elements. The header image includes color scheme, title text, image elements, and a product identifier. Card borders and header materials are designed according to the current requirements, and the layout of the card HTML after filling in the requirement text is adjusted. Specifically, the page visual module standardizes the line thickness, corner radius, header materials, and decorative elements of the card borders. Based on the header image color and the template style of the card HTML, the page background color, card border color, button color, title color, and decorative element colors are uniformly adjusted.

[0065] Combine the header image, the revised card HTML layout, and the unified page color scheme to generate the final landing page HTML.

[0066] The online feedback and evolution module receives page evaluation data for the final landing page HTML. This data includes five levels of evaluation (A, B, C, D, E), custom text evaluations, and page modification instructions. The module performs semantic parsing on these instructions and converts them into structured page evaluations. For example, if the instruction is to change the overall color scheme to a warmer tone, the module converts this into a structured evaluation indicating a mismatch between the page's color scheme and the desired scenario. Based on these structured evaluations, the final landing page HTML undergoes three rounds of iterative optimization. The first round addresses visible page flaws, the second round optimizes visual hierarchy and the requirement conversion path, and the third round re-examines the optimization results. The optimized landing page HTML is output, and the page evaluation data, page modification instructions, iteration logs, and pre- and post-optimization samples are archived. Based on these archived data, the required page dataset is created, and the large-scale required page evaluation model is incrementally fine-tuned using this dataset.

[0067] The beneficial effects of this embodiment are as follows: It supports all stages of landing page creation, significantly shortens the delivery cycle, and reduces cross-departmental collaboration costs and design manpower investment. It enhances landing page design quality by ensuring visual consistency across all channels through unified card-based style specifications and a high-quality landing page component template library, thereby enhancing brand awareness. It flexibly adjusts styles to adapt to different demand scenarios, meeting user needs and significantly increasing user dwell time and conversion rates. Furthermore, it improves demand conversion rates by expanding into segmented scenarios through the copywriting planning module combined with the use of the component template library, supporting refined operations.

[0068] The following describes a page generation apparatus provided by an embodiment of the present invention. The page generation apparatus described below can be referred to in correspondence with the page generation method described above.

[0069] Please refer to the details. Figure 5 , Figure 5 A schematic diagram of a page generation device provided in an embodiment of the present invention may include: The template knowledge base determination module 100 is used to determine the template knowledge base; wherein, the template knowledge base construction process includes processing the landing page card component template image to obtain the card HTML template, performing tagging processing on the card HTML template to obtain template tags, and constructing the template knowledge base based on the card HTML template and the template tags; the template tags include card description, adaptation text, adaptation scenario, card title, and requirement level; The target card HTML template building module 200 is used to generate a landing page outline including multiple sub-arguments based on the demand theme, and to determine the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base. The content to be filled module 300 is used to generate search terms corresponding to each sub-argument based on each sub-argument and the target card HTML template, and to retrieve the content to be filled corresponding to each sub-argument based on the search terms; The target page generation module 400 is used to generate a card HTML with filled content based on the content to be filled and the target card HTML template using rendering prompts, and to generate a target page based on the card HTML with filled content using page visual prompts.

[0070] Furthermore, based on any of the above embodiments, the template knowledge base determination module 100 may include: The image recognition unit is used to input the landing page card component template image into the multimodal large language model, so that the multimodal large language model can recognize the text area, image area, icon area, button area, border style and hierarchical structure in the landing page card component template image; The card HTML template generation unit is used to generate the card HTML template based on the recognition results output by the multimodal large language model; wherein, the container structure in the card HTML template is generated according to the preset container naming rules, and the icon content, icon size and icon style description are recorded; the nested structure of the card HTML template is generated according to the preset hierarchical structure specification.

[0071] Furthermore, based on any of the above embodiments, the template knowledge base determination module may further include: The card description generation module is used to generate the card description based on the layout and intended purpose of the card HTML template. The adaptive text generation module is used to generate the adaptive text based on the text content of the card HTML template; The adaptation scenario generation module is used to generate the adaptation scenario based on the required scenario of the card HTML template; The card title and demand level determination module is used to generate the card title according to the adaptation scenario and determine the demand level of the card HTML template. The template knowledge base construction module is used to associate and store the card HTML template, the card description, the adaptation text, the adaptation scenario, the card title and the requirement hierarchy to obtain the template knowledge base for template retrieval.

[0072] Furthermore, based on any of the above embodiments, the target card HTML template construction module 200 may include: The requirement landing page outline generation unit is used to generate the requirement landing page outline, including a visual anchoring layer, a logical construction layer, a threshold resolution layer, and an action conversion layer, based on the requirement theme; wherein, the sub-arguments include the visual anchoring layer, the logical construction layer, the threshold resolution layer, and the action conversion layer; The sub-argument information generation unit is used to generate the demand objectives, demand strategies and content descriptions corresponding to each sub-argument. The template tag matching unit is used to match the requirement objective, requirement strategy and content description of each sub-argument with the template tags in the template knowledge base; The target card HTML template determination unit is used to determine the target card HTML template corresponding to each sub-argument based on the matching results.

[0073] Furthermore, based on any of the above embodiments, the above-mentioned content-to-be-filled determination module 300 may include: The information retrieval unit is used to invoke search tools to retrieve information based on the search terms; The content to be filled generation unit is used to use a multimodal large language model to assemble the retrieved information into content to be filled, consisting of image descriptions and image links.

[0074] Furthermore, based on any of the above embodiments, the target page generation module 400 may include: The content filling unit is used to process the content in the target card HTML template based on the rendering prompt words and the content to be filled, so as to obtain the card HTML after the content is filled. The page design unit is used to design the page based on the visual cues of the page and the HTML of the card after the filled content to obtain the target page; wherein, the design content includes header image, card border and page color scheme.

[0075] Furthermore, based on any of the above embodiments, the page generation apparatus may further include: The page rating acquisition module is used to acquire the page rating level, custom text rating, and page modification instructions of the target page; wherein, the page rating level is determined by visual design, layout structure, interactive experience, demand conversion, and scene adaptability. The page modification instruction parsing module is used to perform semantic parsing on the page modification instruction and convert the page modification instruction into a structured page evaluation. The model adjustment module is used to adjust the model involved in the page generation process based on the page evaluation level, custom text evaluation, and structured page evaluation.

[0076] It should be noted that the order of the modules and units in the above-mentioned page generation device can be changed without affecting the logic.

[0077] The page generation device provided in this embodiment of the invention searches for real-time information based on the required topic and plans it into text to be filled. It also determines the template corresponding to each sub-argument, thereby filling the text to be filled into the selected target card HTML template. Finally, it generates the target page based on the page visual cues. Since the target page can be generated directly based on the required topic, the generation efficiency can be improved. Furthermore, since the target card HTML template and the text to be filled are determined based on each sub-argument, the accuracy of the target page generation is improved.

[0078] The following describes a page generation device provided by an embodiment of the present invention. The page generation device described below and the page generation method described above can be referred to in correspondence.

[0079] Please refer to Figure 6 , Figure 6 A schematic diagram of a page generation device provided in an embodiment of the present invention may include: Memory 10 is used to store computer programs; Processor 20 is used to execute computer programs to implement the page generation method described above.

[0080] The memory 10, processor 20, and communication interface 30 all communicate with each other through the communication bus 40.

[0081] In this embodiment of the invention, the memory 10 is used to store one or more programs. The programs may include program code, which includes computer operation instructions. In this embodiment of the invention, the memory 10 may store programs for implementing the following functions: The template knowledge base is determined; the construction process of the template knowledge base includes processing the landing page card component template images to obtain card HTML templates, tagging the card HTML templates to obtain template tags, and constructing the template knowledge base based on the card HTML templates and template tags; the template tags include card description, adaptation text, adaptation scenario, card title, and requirement level; Generate a landing page outline with multiple sub-arguments based on the demand theme, and determine the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base; Generate search terms corresponding to each sub-argument based on each sub-argument and the target card HTML template, and retrieve the content to be filled for each sub-argument based on the search terms; Based on the content to be filled and the target card HTML template, render prompts to generate the card HTML with the content filled, and then generate the target page based on the card HTML with the content filled and page visual prompts.

[0082] In one possible implementation, the memory 10 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function; and the data storage area may store data created during use.

[0083] Furthermore, memory 10 may include read-only memory and random access memory, providing instructions and data to the processor. A portion of the memory may also include NVRAM. The memory stores operating systems and operating instructions, executable modules, or data structures, or subsets thereof, or extended sets thereof, wherein the operating instructions may include various operating instructions for implementing various operations. The operating system may include various system programs for implementing various basic tasks and handling hardware-based tasks.

[0084] Processor 20 can be a central processing unit (CPU), an application-specific integrated circuit, a digital signal processor, a field-programmable gate array, or other programmable logic device. Processor 20 can be a microprocessor or any conventional processor. Processor 20 can call programs stored in memory 10.

[0085] The communication interface 30 can be an interface for the communication module, used to connect with other devices or systems.

[0086] Of course, it should be noted that, Figure 6 The structure shown does not constitute a limitation on the page generation device in the embodiments of the present invention. In practical applications, the page generation device may include devices such as... Figure 6 More or fewer components as shown, or combinations of certain components.

[0087] The following describes the computer-readable storage medium provided in the embodiments of the present invention. The computer-readable storage medium described below can be referred to in correspondence with the page generation method described above.

[0088] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the page generation method described above.

[0089] The computer-readable storage medium may include various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0090] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0091] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0092] Finally, it should be noted that in this document, relationships such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0093] The foregoing has provided a detailed description of a page generation method, apparatus, device, and computer-readable storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A page generation method, characterized in that, include: A template knowledge base is determined; the construction process of the template knowledge base includes processing the landing page card component template image to obtain a card HTML template, tagging the card HTML template to obtain template tags, and constructing the template knowledge base based on the card HTML template and the template tags; the template tags include card description, adaptation text, adaptation scenario, card title, and requirement level; Generate a landing page outline for the requirements based on the requirements theme, including multiple sub-arguments, and determine the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base. Based on each sub-argument and the target card HTML template, generate search terms corresponding to each sub-argument, and retrieve the content to be filled corresponding to each sub-argument based on the search terms; Based on the content to be filled and the target card HTML template, rendering prompts are used to generate the card HTML with the filled content, and the target page is generated based on the card HTML with the filled content using page visual prompts.

2. The page generation method according to claim 1, characterized in that, The landing page card component template image is processed to obtain the card HTML template, including: The landing page card component template image is input into a multimodal large language model, enabling the multimodal large language model to recognize the text area, image area, icon area, button area, border style, and hierarchical structure in the landing page card component template image; The card HTML template is generated based on the recognition results output by the multimodal large language model; wherein, the container structure in the card HTML template is generated according to the preset container naming rules, and the icon content, icon size and icon style description are recorded; the nested structure of the card HTML template is generated according to the preset hierarchical structure specification.

3. The page generation method according to claim 1, characterized in that, Before obtaining the template knowledge base, it also includes: The card description is generated based on the layout of the card HTML template and the intended purpose; The adapted text is generated based on the text content of the card's HTML template; The adaptation scenario is generated based on the required scenario of the card HTML template; The card title is generated based on the adaptation scenario, and the requirement level of the card HTML template is determined. The card HTML template, the card description, the adapted text, the adapted scenario, the card title, and the requirement hierarchy are associated and stored to obtain the template knowledge base for template retrieval.

4. The page generation method according to claim 1, characterized in that, Generate a landing page outline with multiple sub-arguments based on the demand theme, and determine the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base, including: Based on the stated requirements, a landing page outline is generated, comprising a visual anchoring layer, a logical construction layer, a threshold resolution layer, and an action conversion layer; wherein, the sub-arguments include the visual anchoring layer, the logical construction layer, the threshold resolution layer, and the action conversion layer; Generate the corresponding requirement objectives, requirement strategies, and content descriptions for each of the sub-arguments; The requirement objectives, requirement strategies, and content descriptions of each sub-argument are matched with the template tags in the template knowledge base; The target card HTML template corresponding to each sub-argument is determined based on the matching results.

5. The page generation method according to any one of claims 1 to 4, characterized in that, After generating the filled card HTML based on the content to be filled and the target card HTML template using rendering prompts, and generating the target page based on page visual prompts, the process further includes: Obtain the page rating, custom text evaluation, and page modification instructions for the target page; wherein, the page rating is determined by visual design, layout structure, interactive experience, demand conversion, and scene adaptability; The page modification instructions are semantically parsed and converted into structured page evaluations; The model involved in the page generation process is adjusted based on the page rating, custom text rating, and structured page rating.

6. The page generation method according to claim 1, characterized in that, Based on each sub-argument and the target card HTML template, generate search terms corresponding to each sub-argument, and retrieve the content to be filled corresponding to each sub-argument based on the search terms, including: Use the search terms to retrieve information using the search tool. The retrieved information is used to form the content to be filled, consisting of image descriptions and image links.

7. The page generation method according to claim 1, characterized in that, Based on the content to be filled and the target card HTML template, using rendering prompts, generate the filled card HTML, and generate the target page based on the page visual prompts, including: Based on the rendering prompt, the content in the target card HTML template is processed according to the content to be filled, to obtain the card HTML after the content is filled; Based on the visual cues on the page, the HTML of the card after the content is filled is designed to obtain the target page; wherein, the design content includes header image, card border and page color scheme.

8. A page generation apparatus, characterized in that, include: The template knowledge base determination module is used to determine the template knowledge base. The construction process of the template knowledge base includes processing the landing page card component template image to obtain the card HTML template, performing tagging processing on the card HTML template to obtain template tags, and constructing the template knowledge base based on the card HTML template and the template tags. The template tags include card description, adaptation text, adaptation scenario, card title, and requirement level. The target card HTML template building module is used to generate a landing page outline including multiple sub-arguments based on the demand theme, and to determine the target card HTML template corresponding to each sub-argument based on each sub-argument and the template tags in the template knowledge base. The module for determining the content to be filled is used to generate search terms corresponding to each sub-argument based on each sub-argument and the target card HTML template, and to retrieve the content to be filled corresponding to each sub-argument based on the search terms; The target page generation module is used to generate a card HTML with filled content based on the content to be filled and the target card HTML template using rendering prompts, and to generate a target page based on the card HTML with filled content using page visual prompts.

9. A page generation device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the page generation method 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 that, when executed by a processor, implements the steps of the page generation method as described in any one of claims 1 to 7.

Images