Intelligent promotion planning method and system based on AI analysis and multi-model AIGC

Abstract

The invention relates to the technical field of artificial intelligence and computer application, and discloses an intelligent promotion planning method and system based on AI analysis and multi-model AIGC, which are used for solving the problems of low efficiency, strong subjectivity, shallow AIGC application and the like in the existing promotion planning. And then multiple models are dynamically called through an AIGC generation engine to generate high-quality promotion materials, the method comprises the core steps of demand disassembly, feature matching, planning instruction generation, multi-model routing, quality evaluation optimization and the like, end-to-end automation from demand analysis to material generation is achieved, and the efficiency and quality of promotion planning are remarkably improved.

Description

Technical Field

[0001] This invention relates to the fields of artificial intelligence and computer application technology, specifically to an intelligent promotion planning method and system based on AI analysis and multi-model AIGC. Background Technology

[0002] In current marketing practices, the planning and content creation stages heavily rely on manual labor. Planners must analyze promotional needs based on experience, determine the content style and structure, and then hand it over to designers and copywriters for execution. This model, under current technology, suffers from the following pain points:

[0003] Efficiency bottleneck: The process from understanding the needs to outputting the planning strategy takes too long, which cannot meet the high-frequency, fast-paced needs of digital marketing.

[0004] Highly subjective: The effectiveness of the planning relies heavily on personal experience, and there is a lack of objective and quantifiable standards for matching "target audience preferences", "brand tone" and "scene atmosphere", resulting in a disconnect between the final materials and actual market demand;

[0005] The application of AIGC is superficial: Existing AIGC solutions mostly call a single model in isolation (such as generating only text or only images), lacking systematic planning logic as input guidance in the early stage. This results in highly random content with low relevance to business goals, and still requires a lot of manpower for screening and secondary modification, failing to achieve end-to-end automation from requirements to finished product.

[0006] In existing technologies, there are some solutions that take into account these pain points. For example, some solutions generate copy by extracting keywords and calling language models, but they do not involve the collaborative planning and generation of multimodal content, nor do they have a deep analysis mechanism to accurately match content style with promotion scenarios. Another example is a personalized content recommendation method based on user profiles, which focuses on content distribution rather than the creative planning and generation of content. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent promotion planning method and system based on AI analysis and multi-model AIGC to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] A smart promotion planning method based on AI analysis and multi-model AIGC includes:

[0010] Based on the information interaction interface, the system receives promotional demand information and original materials input by the user. The promotional demand information includes the target audience profile, promotional scenario, core promotional selling points, and expected type.

[0011] Natural language processing is used to extract structured features from promotion demand information;

[0012] By using AI analysis models to process the structured features of promotion demand information in multiple levels, machine-readable style guidelines and content planning instructions are output.

[0013] The AIGC generation engine dynamically calls at least one AIGC model based on the material type requirements, and combines style guidelines, content planning instructions, and original materials to generate and optimize initial promotional materials.

[0014] As a further aspect of the present invention: the step of using an AI analysis model to perform multi-level processing on the structured features of promotion demand information includes:

[0015] Based on the industry demand knowledge base, the structured demand vector is orthogonally decomposed to obtain multiple analysis dimensions, including target dimension, style dimension and content dimension;

[0016] The target audience profile and promotion scenario are combined feature vectors, and similarity matching is performed based on a similarity algorithm and an industry style feature library to select tags with high matching degree to establish style guidance. The similarity algorithm adopts the cosine similarity algorithm.

[0017] A structured content planning instruction is generated based on the core promotional selling points and material requirements. The content planning instruction includes content structure, key information weights, and format specifications.

[0018] As a further embodiment of the present invention: the industry style feature library stores style tags and their corresponding visual element features and language features, all of which are used for similarity matching;

[0019] The content structure is used to characterize the distribution structure and layout of content elements in text and images, and the key information weights are used to guide the content focus of the AIGC model.

[0020] As a further aspect of the present invention: the AIGC generation engine dynamically calls at least one AIGC model according to the material type requirements, and generates and optimizes the initial promotional materials by combining style guidance, content planning instructions and original materials. The specific steps include:

[0021] Based on the material type requirements, the model is selected and called from the model library containing various types of AIGC models through the model routing controller;

[0022] Based on computer vision and NLP analysis, feature extraction and annotation of the original materials are performed and then integrated with content planning instructions;

[0023] Transform style guidelines and content planning instructions into enhanced prompts and generation parameters for a specific AIGC model;

[0024] The quality of the generated initial materials is evaluated using an assessment model. If the quality does not meet the threshold, the deduction items are analyzed and the parameters are adjusted to regenerate the materials. If the quality meets the threshold, the initial promotional materials are output.

[0025] As a further aspect of the present invention: the evaluation indicators of the evaluation model include one or more of the semantic matching degree of the planning instructions and the visual / language fluency, and the evaluation process is based on a multimodal fusion strategy.

[0026] This invention aims to provide an intelligent promotion planning system based on AI analysis and multi-model AIGC, comprising:

[0027] The demand access module is used to receive promotion demand information and original materials input by users based on the information interaction interface. The promotion demand information includes target audience profile, promotion scenario, core promotion selling points and expected type.

[0028] The information preprocessing module is used to perform natural language processing on promotion demand information and extract structured features;

[0029] The intelligent planning module is used to process the structured features of promotion demand information in multiple levels using AI analysis models, and output machine-readable style guidelines and content planning instructions.

[0030] The AIGC generation module is used by the AIGC generation engine to dynamically call at least one AIGC model based on the material type requirements, and combine style guidelines, content planning instructions and original materials to generate and optimize initial promotional materials.

[0031] As a further aspect of the present invention: the intelligent planning module includes:

[0032] The requirement decomposition unit is used to orthogonally decompose the structured requirement vector based on the industry requirement knowledge base to obtain multiple analysis dimensions, including target dimension, style dimension and content dimension.

[0033] The feature matching unit is used to calculate the combined feature vector of the target audience profile and the promotion scenario, and to perform similarity matching based on the similarity algorithm and the industry style feature library to select tags with high matching degree to establish style guidance. The similarity algorithm adopts the cosine similarity algorithm.

[0034] The planning and generation unit is used to generate structured content planning instructions based on core promotional selling points and material requirements. The content planning instructions include content structure, key information weights, and format specifications.

[0035] As a further embodiment of the present invention: the industry style feature library stores style tags and their corresponding visual element features and language features, all of which are used for similarity matching;

[0036] The content structure is used to characterize the distribution structure and layout of content elements in text and images, and the key information weights are used to guide the content focus of the AIGC model.

[0037] As a further embodiment of the present invention: the AIGC generation module includes:

[0038] Based on the material type requirements, the model is selected and called from the model library containing various types of AIGC models through the model routing controller;

[0039] Based on computer vision and NLP analysis, feature extraction and annotation of the original materials are performed and then integrated with content planning instructions;

[0040] Transform style guidelines and content planning instructions into enhanced prompts and generation parameters for a specific AIGC model;

[0041] The quality of the generated initial materials is evaluated using an assessment model. If the quality does not meet the threshold, the deduction items are analyzed and the parameters are adjusted to regenerate the materials. If the quality meets the threshold, the initial promotional materials are output.

[0042] As a further aspect of the present invention: the evaluation indicators of the evaluation model include one or more of the semantic matching degree of the planning instructions and the visual / language fluency, and the evaluation process is based on a multimodal fusion strategy.

[0043] Compared with existing technologies, the beneficial effects of this invention are as follows: Through a two-level processing architecture, the promotion planning process is clearly divided into two independent but closely collaborative stages: intelligent planning and AIGC generation, achieving decoupling of planning logic and generation technology; a multi-level demand analysis model including demand decomposition, feature matching, and planning generation is constructed, realizing a gradual transformation from fuzzy demands to precise instructions; a dynamic model selection algorithm based on material type and performance indicators is proposed, enabling intelligent scheduling and collaborative work of multiple AIGC models; and a complete feedback loop including generation, evaluation, and optimization is established to ensure continuous improvement in the quality of output materials. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating the core steps of an intelligent promotion planning method based on AI analysis and multi-model AIGC.

[0045] Figure 2 This is a logical framework diagram of an intelligent promotion planning method based on AI analysis and multi-model AIGC.

[0046] Figure 3This is an optimized loop logic diagram in an intelligent promotion planning method based on AI analysis and multi-model AIGC.

[0047] Figure 4 This is a diagram illustrating the main framework of an intelligent promotion planning system based on AI analysis and multi-model AIGC. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0049] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0050] like Figure 1 The above-described intelligent promotion planning method based on AI analysis and multi-model AIGC, as provided in one embodiment of the present invention, includes the following steps:

[0051] S1, receive promotional demand information and original materials input by the user based on the information interaction interface. The promotional demand information includes target audience profile, promotion scenario, core promotional selling points and expected type.

[0052] S2 performs natural language processing on the promotion demand information to extract structured features;

[0053] S3 utilizes AI analysis models to process the structured features of promotional demand information in multiple levels, outputting machine-readable style guidelines and content planning instructions.

[0054] S4, the AIGC generation engine, dynamically calls at least one AIGC model based on the material type requirements, and combines style guidelines, content planning instructions, and original materials to generate and optimize initial promotional materials.

[0055] This embodiment presents a technical solution that seamlessly integrates "intelligent demand analysis," "data-driven planning and decision-making," and "multi-model AIGC generation" to construct a closed-loop, automated intelligent promotion planning method. This aims to address issues in existing technologies such as low efficiency in promotion planning, poor style matching, isolated AIGC tool applications, and low usability of generated materials. Its core objective is to provide an automated and precise method that enables direct and rapid output from initial promotion needs to high-quality initial promotion materials, significantly reducing labor costs and skill dependence. Specifically, the core of this embodiment lies in constructing a two-level processing architecture comprising an intelligent planning module and an AIGC generation engine. This transforms vague user needs into precise, executable creative instructions, and under the comprehensive processing of the intelligent planning module, establishes initial promotion materials. Compared to existing technologies, GuangYuan offers several advantages: It achieves full-process automation and a leap in efficiency, compressing the traditional "day" planning and creation cycle to the "minute" level, enabling "instant planning and instant generation" of promotional materials; Data-driven decision-making and precise matching, using algorithms to replace human brains for style matching and content planning, eliminating subjective bias and increasing the alignment of output content with promotional goals to over 90%; Significantly improved material quality and usability, through a "planning-guided generation" mechanism, ensuring that AIGC-generated content has a clear commercial purpose and structure, increasing the direct usability of generated materials by over 40% and greatly reducing manual rework; Broad system compatibility and business adaptability, with modular design allowing the system to flexibly integrate new AIGC models and quickly adapt to different industries, offering strong scalability.

[0056] The execution of the method is based on the overall system architecture, and the system architecture on which the method depends mainly includes:

[0057] User interface layer: Provides a graphical interface and API interface, supporting various forms of user input;

[0058] Intelligent planning module: includes multi-level processing units of NLP preprocessing unit and AI analysis model;

[0059] AIGC generation engine: includes model library, model routing controller, material annotation unit and evaluation and optimization unit;

[0060] Knowledge base components: including an industry requirements knowledge base and an industry style feature base;

[0061] Output management module: responsible for the format conversion, storage and distribution of the final materials.

[0062] For step S1, the system receives promotional demand information input by the user through a graphical interface or API. This information can be structured or unstructured data, and includes at least: target audience profile (such as demographic attributes, interest tags, and spending power), promotional scenarios (such as e-commerce promotions, brand exposure, and new product launches), core promotional selling points (product features, price, and service advantages), and desired material types (such as copywriting, images, graphic combinations, and short videos). For step S2, the built-in Natural Language Processing (NLP) unit parses the unstructured text demand and converts it into a structured feature vector through technologies such as entity recognition and keyword extraction. The preprocessing process includes text cleaning, word segmentation, entity recognition, and sentiment analysis to ensure the quality and consistency of the input data.

[0063] like Figure 2 As shown, in another preferred embodiment of the present invention, the step of using an AI analysis model to perform multi-level processing on the structured features of promotion demand information includes:

[0064] Based on the industry demand knowledge base, the structured demand vector is orthogonally decomposed to obtain multiple analysis dimensions, including target dimension, style dimension and content dimension;

[0065] The target audience profile and promotion scenario are combined feature vectors, and similarity matching is performed based on a similarity algorithm and an industry style feature library to select tags with high matching degree to establish style guidance. The similarity algorithm adopts the cosine similarity algorithm.

[0066] A structured content planning instruction is generated based on the core promotional selling points and material requirements. The content planning instruction includes content structure, key information weights, and format specifications.

[0067] Furthermore, the industry style feature library stores style tags and their corresponding visual element features and language features, all of which are used for similarity matching.

[0068] The content structure is used to characterize the distribution structure and layout of content elements in text and images, and the key information weights are used to guide the content focus of the AIGC model.

[0069] In this embodiment, the multi-layered structure includes: a demand decomposition layer: connected to an industry demand knowledge base of a prediction system. This knowledge base stores typical promotion templates and logical rules for various industries (such as e-commerce, education, finance, and cultural tourism). Based on this knowledge base, this layer decomposes the structured demand vector into three orthogonal analysis dimensions: target dimension (who it is targeting), style dimension (what attributes), and content dimension (what it says). The decomposition process uses a neural network model based on an attention mechanism, combined with knowledge graph technology for multi-dimensional demand analysis; and a feature matching layer: connected to a dynamic industry style feature library. This library stores style tags (such as "youthful and lively style," "high-end business style," and "warm family style") and their corresponding... The corresponding visual elements (color, font, composition) and linguistic features (tone, vocabulary) are used in this layer. An improved cosine similarity algorithm is employed to calculate the similarity between the combined feature vector of the "target audience tag" and the "promotion scenario" and the vectors of each style tag. The tag with the highest similarity is selected as the output style guide. This guide is specific, for example: "The color scheme is mainly based on low-saturation Meilandi colors, the font is a sans-serif typeface, the language style is friendly and encouraging, and the emphasis is on the product's safety and professionalism." The planning generation layer generates structured content planning instructions based on the "core promotional selling points" and "material type requirements." These instructions are machine-readable JSON or XML objects that clearly define:

[0070] Content structure: such as the three-part structure of copywriting: "pain point introduction - selling point explanation - call to action", and the "product main body position - background element - selling point text label position" of images;

[0071] Key information weighting: quantifies the importance of each selling point, such as "core function selling point weight 0.5, price advantage weight 0.3", which is used to guide the content focus of the AIGC model;

[0072] Formatting specifications: such as hard constraints on word count, image resolution, and video length; the planning generation layer adopts template-based dynamic generation technology, combined with genetic algorithms to optimize the weight allocation of key information, ensuring the scientific nature and effectiveness of planning instructions.

[0073] like Figure 3 As shown, in another preferred embodiment of the present invention, the AIGC generation engine dynamically calls at least one AIGC model according to the material type requirements, and generates and optimizes the initial promotional materials by combining style guidance, content planning instructions and original materials. The specific steps include:

[0074] Based on the material type requirements, the model is selected and called from the model library containing various types of AIGC models through the model routing controller;

[0075] Based on computer vision and NLP analysis, feature extraction and annotation of the original materials are performed and then integrated with content planning instructions;

[0076] Transform style guidelines and content planning instructions into enhanced prompts and generation parameters for a specific AIGC model;

[0077] The quality of the generated initial materials is evaluated using an assessment model. If the quality does not meet the threshold, the deduction items are analyzed and the parameters are adjusted to regenerate the materials. If the quality meets the threshold, the initial promotional materials are output.

[0078] Furthermore, the evaluation metrics of the evaluation model include one or more of the semantic matching degree of the planning instructions and visual / language fluency, and the evaluation process is based on a multimodal fusion strategy.

[0079] In this embodiment, the AIGC model library integrates at least three types of models: large language models (such as DeepseekR1), text-to-image models (such as Stable Diffusion), and text-to-video / audio models (such as Wan). Each model has a model routing controller that automatically selects and calls one or more models based on the "material type requirement." The routing decision is based on a multi-index evaluation system, including model performance, response time, cost efficiency, and generation quality. For example, when the requirement is "text and image combination," the large language model and the text-to-image model are called in parallel; when the requirement is "short video," the large language model is called sequentially to generate the script, then the video generation model is called, and the audio model may be called to generate the dubbing.

[0080] For structured material labeling: User-uploaded raw materials (product images, logos, white papers, etc.) undergo computer vision (CV) or NLP analysis to extract features and label them (e.g., "product image - 45-degree angle - white background", "logo - rectangle - red"), transforming them into digital assets that match content planning instructions. The labeling process employs a multimodal AI model to ensure accuracy and completeness. For model parameter configuration and cue word engineering: "Style guidelines" and "content planning instructions" are translated into reinforcement cue words and generation parameters understandable by each AIGC model. Cue word engineering combines template-based and dynamic parameter filling methods to ensure a high degree of consistency between generated content and planning requirements. For example, setting negative values ​​for the Stable Dif-fusion model... The prompt function excludes unwanted elements and provides a system role to constrain the language style of the language model. For the multi-round generation and optimization mechanism: after the initial materials are generated, a built-in evaluation model evaluates their quality. This evaluation model scores based on predefined indicators (such as semantic matching degree with the planning instructions, image aesthetic quality, and text fluency). The evaluation process adopts a multimodal fusion strategy, combining traditional quality indicators and semantic matching degree evaluation based on deep learning. If the score is lower than a preset threshold (such as 80 points), the system will automatically analyze the deduction items, fine-tune the model parameters or reconstruct the prompt words, and initiate a new round of generation until the initial promotional materials that meet the quality requirements are output. The optimization process adopts an adaptive optimization algorithm, which dynamically adjusts the generation strategy according to the evaluation results.

[0081] The following example illustrates this with a typical application scenario: promoting a new beauty product on an e-commerce platform.

[0082] First, the user inputs their needs, including promotional requirements and original materials:

[0083] Promotional requirements: "Targeting women aged 25-30 with sensitive skin, promote a new foundation with an 8-hour wear time and skin-nourishing ingredients during a major sales event. A set of graphic and text materials is needed, including one main image and a text description of no more than 200 words." Original materials: Upload a real photo of the product's front (transparent background) and a vector file of the brand logo.

[0084] The user input data is preprocessed, and the NLP unit parses the requirement text to extract key feature words such as "T25-30 years old", "female", "sensitive skin", "big promotion", "8-hour makeup lasting", "skin care ingredients", "foundation", and "images and text".

[0085] Then, the AI ​​analysis model performs multi-level processing, including: a demand decomposition layer, which, based on the "beauty e-commerce" knowledge base, decomposes the demand into: target dimension (young women with sensitive skin), style dimension (requiring professionalism, safety, and natural beauty), and content dimension (highlighting the dual selling points of long-lasting makeup and skincare); a feature matching layer, which calculates the similarity between combined features and the style library, matching "natural, light luxury, and professional style," and outputs style guidelines as follows: "Main color: off-white, light pink; secondary color: light gold; font: thin sans-serif; image style: clean, bright, and textured macro photography feel; language style: professional, gentle, and trustworthy, emphasizing 'zero burden on skin' and 'makeup and skincare in one'"; and a planning generation layer, which outputs content planning instructions (JSON format).

[0086] Then, the AIGC generation engine model routing controller identifies it as "image and text combination," and calls the text-to-image model (StableDif-fusion) and the large language model (such as DeepseekR1) in parallel. For image generation, the CV module analyzes the original product image as "cosmetic bottle - isolated subject," and transforms the image part in the style guidance and content planning instructions into reinforcement prompts: "Professional product photography of a foundation bottle, centered on a clean light pink to white gradient background, with delicate gold accents and petal textures, minimalist, luxury style, high detail, 8K, logo on top-right, taglines '8-Hour Wear' and 'Skin-Caring Formula' elegantly displayed below," generating the initial main image. For copywriting generation, the system role instruction is passed to the large language model: "You are a professional beauty copywriter with a gentle, professional, and trustworthy writing style," and user prompts: "Write an article for young women with sensitive skin for a major promotion, highlighting 8-hour makeup wear and skin-nourishing ingredients, structured as…, word count within 200," generating the initial copy.

[0087] During the iterative optimization process before output, if the main image scores 75 points due to "unclear selling point text," which is below the 80-point threshold, the system automatically analyzes the deduction items and identifies the problem as "insufficient recognizability of selling point text." The optimization engine adjusts the prompt word parameters, adds the description "clear, legible text for taglines," and increases the text contrast parameter. After regeneration, the main image score is raised to 85 points, and together with the copy (initially scoring 88 points), it is output to the user as the initial promotional material.

[0088] In the technical implementation of the AI ​​analysis model, the requirement decomposition layer adopts a neural network model based on the attention mechanism, combined with knowledge graph technology to perform multi-dimensional requirement analysis; the feature matching layer uses an improved cosine similarity algorithm to support the similarity calculation of multi-modal feature vectors; and the planning generation layer adopts a template-based dynamic generation technology to support multiple output formats such as JSON and XML.

[0089] In AIGC model integration and management, the model routing controller adopts a rule-based decision tree algorithm and performs dynamic routing in conjunction with real-time model performance monitoring; it supports standardized interfaces for various AIGC models, including multiple communication protocols such as RESTAPI and gRPC; the model performance monitoring system tracks key indicators such as generation quality and response time of each model in real time.

[0090] The quality assessment model employs a multimodal fusion strategy, combining traditional image / text quality assessment metrics with deep learning-based semantic matching evaluation. The assessment model supports online learning and incremental updates, continuously optimizing assessment criteria based on user feedback. It provides fine-grained assessment reports to help the system accurately identify shortcomings in generated materials.

[0091] like Figure 4 As shown, the present invention also provides an intelligent promotion planning system based on AI analysis and multi-model AIGC, which includes:

[0092] The demand access module 100 is used to receive promotion demand information and original materials input by users based on the information interaction interface. The promotion demand information includes target audience profile, promotion scenario, core promotion selling points and expected type.

[0093] The information preprocessing module 200 is used to perform natural language processing on the promotion demand information and extract structured features;

[0094] The intelligent planning module 300 is used to process the structured features of promotion demand information in multiple levels using AI analysis models, and output machine-readable style guidelines and content planning instructions.

[0095] AIGC Generation Module 400 is used by the AIGC generation engine to dynamically call at least one AIGC model based on the material type requirements, and combine style guidelines, content planning instructions and original materials to generate and optimize initial promotional materials.

[0096] In another preferred embodiment of the present invention, the intelligent planning module includes:

[0097] The requirement decomposition unit is used to orthogonally decompose the structured requirement vector based on the industry requirement knowledge base to obtain multiple analysis dimensions, including target dimension, style dimension and content dimension.

[0098] The feature matching unit is used to calculate the combined feature vector of the target audience profile and the promotion scenario, and to perform similarity matching based on the similarity algorithm and the industry style feature library to select tags with high matching degree to establish style guidance. The similarity algorithm adopts the cosine similarity algorithm.

[0099] The planning and generation unit is used to generate structured content planning instructions based on core promotional selling points and material requirements. The content planning instructions include content structure, key information weights, and format specifications.

[0100] As another preferred embodiment of the present invention, the industry style feature library stores style tags and their corresponding visual element features and language features, all of which are used for similarity matching.

[0101] The content structure is used to characterize the distribution structure and layout of content elements in text and images, and the key information weights are used to guide the content focus of the AIGC model.

[0102] In another preferred embodiment of the present invention, the AIGC generation module includes:

[0103] Based on the material type requirements, the model is selected and called from the model library containing various types of AIGC models through the model routing controller;

[0104] Based on computer vision and NLP analysis, feature extraction and annotation of the original materials are performed and then integrated with content planning instructions;

[0105] Transform style guidelines and content planning instructions into enhanced prompts and generation parameters for a specific AIGC model;

[0106] The quality of the generated initial materials is evaluated using an assessment model. If the quality does not meet the threshold, the deduction items are analyzed and the parameters are adjusted to regenerate the materials. If the quality meets the threshold, the initial promotional materials are output.

[0107] As another preferred embodiment of the present invention, the evaluation indicators of the evaluation model include one or more of the semantic matching degree of the planning instructions and the visual / language fluency, and the evaluation process is based on a multimodal fusion strategy.

[0108] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0109] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the disclosure in the specification and embodiments. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

[0110] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A smart promotion planning method based on AI analysis and multi-model AIGC, characterized in that, Include: Based on the information interaction interface, the system receives promotional demand information and original materials input by the user. The promotional demand information includes the target audience profile, promotional scenario, core promotional selling points, and expected type. Natural language processing is used to extract structured features from promotion demand information; By using AI analysis models to process the structured features of promotion demand information in multiple levels, machine-readable style guidelines and content planning instructions are output. The AIGC generation engine dynamically calls at least one AIGC model based on the material type requirements, and combines style guidelines, content planning instructions, and original materials to generate and optimize initial promotional materials.

2. The intelligent promotion planning method based on AI analysis and multi-model AIGC according to claim 1, characterized in that, The steps of using AI analysis models to process the structured features of promotion demand information at multiple levels include: Based on the industry demand knowledge base, the structured demand vector is orthogonally decomposed to obtain multiple analysis dimensions, including target dimension, style dimension and content dimension; The target audience profile and promotion scenario are combined feature vectors, and similarity matching is performed based on a similarity algorithm and an industry style feature library to select tags with high matching degree to establish style guidance. The similarity algorithm adopts the cosine similarity algorithm. A structured content planning instruction is generated based on the core promotional selling points and material requirements. The content planning instruction includes content structure, key information weights, and format specifications.

3. The intelligent promotion planning method based on AI analysis and multi-model AIGC according to claim 2, characterized in that, The industry style feature library stores style tags and their corresponding visual element features and language features, all of which are used for similarity matching. The content structure is used to characterize the distribution structure and layout of content elements in text and images, and the key information weights are used to guide the content focus of the AIGC model.

4. The intelligent promotion planning method based on AI analysis and multi-model AIGC according to claim 3, characterized in that, The AIGC generation engine dynamically calls at least one AIGC model based on the material type requirements, and combines style guidelines, content planning instructions, and original materials to generate and optimize the initial promotional materials. The specific steps include: Based on the material type requirements, the model is selected and called from the model library containing various types of AIGC models through the model routing controller; Based on computer vision and NLP analysis, feature extraction and annotation of the original materials are performed and then integrated with content planning instructions; Transform style guidelines and content planning instructions into enhanced prompts and generation parameters for a specific AIGC model; The quality of the generated initial materials is evaluated using an assessment model. If the quality does not meet the threshold, the deduction items are analyzed and the parameters are adjusted to regenerate the materials. If the quality meets the threshold, the initial promotional materials are output.

5. The intelligent promotion planning method based on AI analysis and multi-model AIGC according to claim 4, characterized in that, The evaluation metrics of the evaluation model include one or more of the semantic matching degree of the planning instructions and the visual / language fluency. The evaluation process is based on a multimodal fusion strategy.

6. An intelligent promotion planning system based on AI analysis and multi-model AIGC, characterized in that, Include: The demand access module is used to receive promotion demand information and original materials input by users based on the information interaction interface. The promotion demand information includes target audience profile, promotion scenario, core promotion selling points and expected type. The information preprocessing module is used to perform natural language processing on promotion demand information and extract structured features; The intelligent planning module is used to process the structured features of promotion demand information in multiple levels using AI analysis models, and output machine-readable style guidelines and content planning instructions. The AIGC generation module is used by the AIGC generation engine to dynamically call at least one AIGC model based on the material type requirements, and combine style guidelines, content planning instructions and original materials to generate and optimize initial promotional materials.

7. The intelligent promotion planning system based on AI analysis and multi-model AIGC according to claim 6, characterized in that, The intelligent planning module includes: The requirement decomposition unit is used to orthogonally decompose the structured requirement vector based on the industry requirement knowledge base to obtain multiple analysis dimensions, including target dimension, style dimension and content dimension. The feature matching unit is used to calculate the combined feature vector of the target audience profile and the promotion scenario, and to perform similarity matching based on the similarity algorithm and the industry style feature library to select tags with high matching degree to establish style guidance. The similarity algorithm adopts the cosine similarity algorithm. The planning and generation unit is used to generate structured content planning instructions based on core promotional selling points and material requirements. The content planning instructions include content structure, key information weights, and format specifications.

8. The intelligent promotion planning system based on AI analysis and multi-model AIGC according to claim 7, characterized in that, The industry style feature library stores style tags and their corresponding visual element features and language features, all of which are used for similarity matching. The content structure is used to characterize the distribution structure and layout of content elements in text and images, and the key information weights are used to guide the content focus of the AIGC model.

9. The intelligent promotion planning system based on AI analysis and multi-model AIGC according to claim 8, characterized in that, The AIGC generation module includes: Based on the material type requirements, the model is selected and called from the model library containing various types of AIGC models through the model routing controller; Based on computer vision and NLP analysis, feature extraction and annotation of the original materials are performed and then integrated with content planning instructions; Transform style guidelines and content planning instructions into enhanced prompts and generation parameters for a specific AIGC model; The quality of the generated initial materials is evaluated using an assessment model. If the quality does not meet the threshold, the deduction items are analyzed and the parameters are adjusted to regenerate the materials. If the quality meets the threshold, the initial promotional materials are output.

10. The intelligent promotion planning system based on AI analysis and multi-model AIGC according to claim 9, characterized in that, The evaluation metrics of the evaluation model include one or more of the semantic matching degree of the planning instructions and the visual / language fluency. The evaluation process is based on a multimodal fusion strategy.

Images