E-commerce product image and text generation optimization system based on generative AI

Abstract

The present invention discloses an e-commerce product image and text generation optimization system based on generative AI, which relates to the field of image and text optimization technology and is used to solve the problem that it is difficult to generate differentiated image and text content for different user groups, resulting in insufficient image and text attractiveness and low conversion rate. By collecting basic information of products and user behavior data, the products are classified and marked and the attention conditions are judged. After screening, the core features of the products and the user preference features are extracted to determine the user interest weight value of each marked product; feedback information of the products under the corresponding classification mark is collected, and image and text information is generated in combination with the user interest weight. The update time is set to collect historical optimization effect data, and the data is used to update the user behavior data, thereby realizing the optimization of images and texts of different products, improving the matching degree between content and user preferences and display attractiveness.

Description

Technical Field

[0001] The present invention relates to the field of image and text optimization technology, and more specifically, to an e-commerce product image and text generation optimization system based on generative AI. Background Art

[0002] With the rapid development of e-commerce platforms, graphic and text information, as an important form of product display, directly affects users' browsing experience and decision-making. Currently, mainstream e-commerce platforms usually generate and display product graphic and text content based on product titles, descriptions, specifications and some static images with the help of fixed templates or rule engines, so that the product content has certain structural and visual features on the front-end page, which helps to improve the neatness of the overall interface and the consistency of user browsing.

[0003] The existing technology has the following deficiencies:

[0004] At present, the image and text display method based on static templates lacks a dynamic matching mechanism with users' personalized interests, making it difficult to generate differentiated image and text content for different user groups, which in turn leads to insufficient image and text attractiveness and low conversion rate. Therefore, an e-commerce product image and text generation optimization system based on generative AI is proposed.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an e-commerce product image and text generation optimization system based on generative AI, which solves the problems raised in the above-mentioned background technology by applying product feature modeling, user behavior analysis, feedback data closed-loop update and generative content generation fusion mechanism.

[0007] To achieve the above objectives, the present invention provides the following technical solutions: an e-commerce product image and text generation optimization system based on generative AI, comprising a data integration module, a feature extraction module, an image and text generation module, and an image and text output module, wherein each module is electrically connected;

[0008] The data integration module is used to collect basic product information and user behavior data, classify products according to the basic product information, calculate the attention coefficient of each category-tagged product using the user behavior data, filter products according to the attention coefficient of each category tag, retain the user behavior data of the filtered products, and pass it together with the basic product information and category tags of the corresponding products to the feature extraction module;

[0009] The feature extraction module extracts the core features of each category-tagged product and the user preference features from the basic product information of the filtered products and the user behavior data, determines the user interest weight value of each category-tagged product based on the core features of each category-tagged product and the user preference features, and transmits the user interest weight value to the image and text generation module;

[0010] After receiving the user interest weight values ​​of each category tag, the image and text generation module collects feedback information of the products under the corresponding category tag, combines the user interest weight values ​​of each category tag and the feedback information to generate image and text information, and transmits the image and text information to the image and text output module;

[0011] The image and text output module is used to receive image and text information and output it to the user end.

[0012] In a preferred embodiment, the basic product information in the data integration module includes product tags, product shipments, and product clicks;

[0013] The data integration module classifies and marks the products according to their product labels. If the product labels of the products are the same, they will be marked as the same category.

[0014] A period of time is selected as the collection time, and the product shipment volume and product click volume are obtained by counting the product shipment volume and the product click volume during the collection time.

[0015] In a preferred embodiment, user behavior data includes the number of product tag searches, product returns, and product return visits;

[0016] The product tag search volume is the total number of product searches for the same product tag during the collection period;

[0017] The product return volume is the total number of product returns with the same product label during the collection period;

[0018] When collecting product return visits, the visit value of the product after deducting one from the number of product visits during the collection time is recorded, and then the visit values ​​of products with the same product label are summed up to obtain the product return visit value of the category tag corresponding to the product label.

[0019] In a preferred embodiment, when calculating the attention coefficient of each type of tagged product, the data integration module counts the number of product tag searches for different categories of tags, and calculates the attention coefficient of the corresponding category tag based on the number of product tag searches: , where s is the number of product tag retrievals for the current category, v is the sum of the number of product tag retrievals for all categories, and a is the attention coefficient for the current category tag;

[0020] The attention coefficient of each category mark is compared with the preset screening threshold. When the attention coefficient of the category mark corresponding to the product exceeds the screening threshold, the corresponding product is screened out.

[0021] In a preferred embodiment, the feature extraction module extracts the core features of the product based on the shipment volume and click volume of the product marked in each category. The specific steps are as follows:

[0022] After normalizing the logarithms of the shipment volume and click volume of each category-labeled product, the core features of the corresponding category-labeled product are calculated using the geometric mean method: , where b is the normalized result of the shipment volume of the product marked in the corresponding category, c is the normalized result of the click volume of the product marked in the corresponding category, and h is the core feature of the product marked in the corresponding category;

[0023] After taking the logarithm of the product return volume and product return volume of each category mark and normalizing them, the user preference characteristics of the corresponding category mark are obtained by subtracting the normalized result of the product return volume from the normalized result of the product return volume.

[0024] In a preferred embodiment, the feature extraction module sums the core features of the product with the corresponding category mark and the user preference features and multiplies the sum by a preset weight adjustment ratio to generate a user interest weight value for the product with the corresponding category mark.

[0025] In a preferred embodiment, the feedback information of the products under the corresponding category mark is the average frequency of image and text replacement;

[0026] By setting a time window, we obtain a set of products with the same category tag. For each product, we count the actual number of times the image and text content is changed within the time window. The number of times the image and text are changed for all products is summarized and averaged to obtain the average image and text change frequency, which is then normalized.

[0027] In a preferred embodiment, the average frequency of image and text replacement and the user interest weight value are substituted into the logistic regression formula to calculate the ranking coefficient. The specific formula is expressed as follows: ;

[0028] Where L is the result of logistic regression, i.e., the ranking coefficient, e is the natural base, and y is the linear combination term of the logistic regression model. The specific setting of y is: ;

[0029] Where, is the bias term, For the The average frequency of image and text replacement for each category label, For the The user interest weight value of the category label, as well as are the regression coefficients of the mean frequency of picture and text replacement and the user interest weight value respectively.

[0030] In a preferred embodiment, the ranking coefficients are collected and sorted in descending order based on the numerical values ​​of the ranking coefficients to obtain the ranking results of each category mark, and product images are provided through the product database;

[0031] The sorting results of each category tag, the product category tag and the product image are passed to the image and text output module as image and text information.

[0032] In a preferred embodiment, the graphic output module obtains graphic information and outputs it, and transmits the sorting results of each category tag, the product category tag and the product picture to the user end respectively, and arranges and displays the products corresponding to each category tag according to the sorting results of each category tag.

[0033] Technical effects and advantages of the present invention:

[0034] The present invention collects basic information of commodities and user behavior data, classifies and marks commodities according to the basic information of commodities, uses user behavior data to judge the attention conditions of marked commodities, filters the basic information of commodities and user behavior data according to the attention conditions of marked commodities, extracts the core features of commodities and the user preference features from the filtered basic information of commodities and user behavior data, determines the user interest weight values ​​of various marked commodities, collects feedback information of commodities under corresponding classification marks, generates graphic and text information by integrating the user interest weight values ​​and feedback information of various classification marks, sets update time to collect historical optimization effect data, and transmits the historical optimization effect data to the data integration module to update the user behavior data, thereby improving the graphic and text optimization of different types of commodities, effectively improving the adaptability of graphic and text content to the actual preferences of users, and making subsequent display content more attractive and guiding. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a flowchart of the implementation of the e-commerce product image and text generation optimization system based on generative AI in the present invention.

[0036] Figure 2 This is a schematic diagram of the steps of the e-commerce product image and text generation optimization system based on generative AI of the present invention. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0038] The present invention collects basic information of commodities and user behavior data, classifies and marks commodities according to the basic information of commodities, uses user behavior data to judge the attention conditions of marked commodities, filters the basic information of commodities and user behavior data according to the attention conditions of marked commodities, extracts the core features of commodities and the user preference features from the filtered basic information of commodities and user behavior data, determines the user interest weight values ​​of various marked commodities, collects feedback information of commodities under corresponding classification marks, generates graphic information by integrating the user interest weight values ​​and feedback information of various classification marks, sets update time to collect historical optimization effect data, and transmits the historical optimization effect data to the data integration module to update the user behavior data, thereby improving the graphic optimization of different types of commodities and improving the accurate arrangement of different commodities.

[0039] Example 1: E-commerce product image and text generation optimization system based on generative AI, such as Figures 1 to 2 As shown, it includes a data integration module, a feature extraction module, a graphic and text generation module, and a graphic and text output module, and each module is connected by electrical signals;

[0040] The functions of each module are as follows:

[0041] The data integration module is used to collect basic product information and user behavior data, classify products according to the basic product information, calculate the attention coefficient of each category-tagged product using the user behavior data, filter products according to the attention coefficient of each category tag, retain the user behavior data of the filtered products, and pass it together with the basic product information and category tags of the corresponding products to the feature extraction module;

[0042] The feature extraction module extracts the core features of each category-tagged product and the user preference features from the basic product information of the filtered products and the user behavior data, determines the user interest weight value of each category-tagged product based on the core features of each category-tagged product and the user preference features, and transmits the user interest weight value to the image and text generation module;

[0043] After receiving the user interest weight values ​​of each category tag, the image and text generation module collects feedback information of the products under the corresponding category tag, combines the user interest weight values ​​of each category tag and the feedback information to generate image and text information, and transmits the image and text information to the image and text output module;

[0044] The image and text output module is used to receive image and text information and output it to the user end.

[0045] It should be noted that the user behavior data is updated to re-determine the attention conditions of different types of tagged products, thereby achieving dynamic self-optimization of product images and texts.

[0046] The basic product information in the data integration module includes product tags, product shipments, and product clicks;

[0047] Product tags are the initial tags assigned to products when they are put on the shelves. They are used to categorize the products. For example, when a keypad phone and a touchscreen phone are put on the shelves, the initial tags assigned to them are both "Mobile Phones."

[0048] The data integration module classifies and marks the products according to their product labels, and marks the products in the same category when their product labels are consistent.

[0049] Select a period of time as the collection time. The product shipment volume and product click volume are obtained by counting the number of shipments and clicks on the product during the collection time. The product shipment volume and product click volume reflect the activity level of the corresponding product. The more product shipment volume or product click volume, the higher the activity level of the product.

[0050] User behavior data includes product tag search volume, product return volume, and product return visit volume;

[0051] The number of product tag searches is the total number of product searches for the same product tag during the collection period. The more product searches for the same product tag, the higher the attention of the corresponding product tag, and the higher the attention of the category tag corresponding to the product tag;

[0052] The product return volume is the total number of product returns with the same product label during the collection period. When collecting product return visits, the visit value of the product is recorded after the number of product visits is reduced by one during the collection period. Then, the visit values ​​of products with the same product label are summed to obtain the product return volume of the category tag corresponding to the product label.

[0053] For products with the same category tag, the lower the product return volume or the higher the product return visit volume, the higher the user interest in the category tag;

[0054] When the data integration module calculates the attention coefficient of each type of tagged product based on the product tag search volume, it counts the product tag search volume of different categories of tags respectively, and calculates the attention coefficient of the corresponding category tag based on the product tag search volume: , where s is the number of product label retrievals for the current category, v is the sum of the number of product label retrievals for all categories, and a is the attention coefficient for the current category.

[0055] The attention coefficient of each category mark is compared with the preset screening threshold. When the attention coefficient of the category mark corresponding to the product exceeds the screening threshold, the corresponding product is screened out.

[0056] It should be noted that the screening threshold is set by professionals in this field according to the requirements of the graphic information list. For example, there are currently 5 category tags, and the attention coefficients are 0.62, 0.43, 0.51, 0.74, and 0.45 respectively. If the graphic information list requires 3 category tags, that is, only products with 3 category tags are provided in the graphic information list, the screening threshold can be set to 0.5, and products with a category tag attention coefficient exceeding 0.5 will be excluded, and so on. No further analysis will be done here.

[0057] The data integration module transfers the shipment volume, click volume, return volume, and return visit volume of the filtered products to the feature extraction module;

[0058] The feature extraction module extracts the core features of products based on the shipment volume and click volume of products marked in each category. The specific steps are as follows:

[0059] After normalizing the logarithms of the shipment volume and click volume of each category-labeled product, the core features of the corresponding category-labeled product are calculated using the geometric mean method: , where b is the normalized result of the shipment volume of the product marked in the corresponding category, c is the normalized result of the click volume of the product marked in the corresponding category, and h is the core feature of the product marked in the corresponding category;

[0060] Similarly, the feature extraction module normalizes the logarithms of the product return volume and product return volume for each category label, and then subtracts the normalized result of product return volume from the normalized result of product return volume to obtain the user preference feature for the corresponding category label.

[0061] When determining the user interest weight value of each category-tagged product based on the core features of the products tagged in each category and the user preference features, the user interest weight value of the corresponding category-tagged product is generated by summing the core features of the products tagged in the corresponding category and the user preference features and then multiplying it by a preset weight adjustment ratio.

[0062] It should be noted that the weight adjustment ratio is used to adjust the core features of the product and the user preference features to the range of the category tag weight. For example, the core feature of a product marked in a certain category is 0.6, the user preference feature is 0.8, and the weight adjustment ratio is 1.2. Then the user interest weight of the corresponding category-tagged product is 0.576. The specific value of the weight adjustment ratio is not unique and can be limited according to actual conditions. It will not be elaborated here.

[0063] The image and text generation module is used to receive the user interest weight values ​​generated by the feature extraction module for different category marked products;

[0064] Among them, the user interest weight value represents the user's attention and preference for various products, which will not be elaborated here;

[0065] Specifically, the feedback information for products under the corresponding category tag refers to a set of statistical indicators of the actual update performance data of the image and text content of each product in the product set with the same category tag on the platform front end within a preset time window, that is, the average image and text replacement frequency. The acquisition logic is to obtain the product set with the same category tag by setting a time window, and for each product, count the actual number of image and text content changes within the time window. The number of image and text changes for all products is summarized and averaged to obtain the average image and text replacement frequency;

[0066] It should be noted that the actual number of times the graphic content is changed within the time window can be determined by any event, such as updating the title or main text content, replacing the main or detail image of the product, switching the product display template, and regenerating the content generation model and uploading new graphic content, etc., which will not be detailed here.

[0067] Furthermore, the triggering method for judging events can be obtained based on image recognition methods and similarity calculations. When a certain image area or image similarity is lower than a preset similarity threshold, it is considered that the main image or detail image of the product has been replaced, etc. The specific triggering method for judging events is set by our experimenters based on the specific e-commerce operation platform and will not be elaborated here.

[0068] Furthermore, the time window is set to , set the set of products with the same category tag as ,in, Indicates the number of products under this category. The calculation formula for the mean frequency of image and text replacement is: ;

[0069] Where, is the mean frequency of picture and text replacement, For products in the time window Number of times the image and text are changed, The total number of products marked in the current category, is the length of the preset time window, The index number marked for the product category, indicating the Category product collection, , The total number of categories set in the system. Category tags are pre-set by the data integration module based on the basic information of the product.

[0070] Normalize the mean frequency of picture and text change so that it is in the same dimension as the user interest weight value.

[0071] It should be noted that the standardization methods include but are not limited to standard linear transformation based on interval scaling, statistical Standardization method or normalization method based on nonlinear mapping function. The application method of standardization processing will not be described in detail here;

[0072] Substitute the mean frequency of image and text replacement and the user interest weight into the logistic regression formula to calculate the ranking coefficient. The specific formula is as follows: ;

[0073] Where L is the result of logistic regression, i.e., the ranking coefficient, e is the natural base, and y is the linear combination term of the logistic regression model. Specifically, y is set as: ;

[0074] Where, is the bias term, For the The average frequency of image and text replacement for each category label, For the The user interest weight value of the category label, as well as are the regression coefficients of the mean frequency of picture and text replacement and the user interest weight value;

[0075] It should be noted that when the mean value of the frequency of image and text replacement and the user interest weight value are larger, the ranking coefficient is larger, which means that the product corresponding to the current category label is ranked higher, indicating that the content update demand of the product corresponding to the current category label is stronger and the user attention is higher. At this time, the ranking priority of the category product is higher. Conversely, when the mean value of the frequency of image and text replacement and the user interest weight value are smaller, the ranking coefficient is smaller, which means that the user attention of the current category product is lower and the image and text stability is higher.

[0076] Collect all the ranking coefficients and sort them in descending order according to the numerical values ​​of the ranking coefficients to obtain the ranking results of each category mark;

[0077] Furthermore, product images are provided through a product database. Specifically, a distributed object storage system is used to manage product images. The product database stores image path information or access links of corresponding products. The image path information is bound to the product unique identifier, allowing the image and text generation module to call the product image and participate in the generation process.

[0078] It should be noted that the product database is a data management system built on a relational database structure, usually implemented using a database engine such as MySQL or PostgreSQL. It is mainly used to store structured product images, and in this example, is used to provide product images corresponding to corresponding category tags;

[0079] The sorting results of each category mark, the product category mark and the product image are transmitted to the image and text output module as image and text information;

[0080] The image and text output module obtains the image and text information and outputs it, and transmits the sorting results of each category tag, the product category tag and the product image to the user end respectively. According to the sorting results of each category tag, the products corresponding to each category tag are arranged and displayed.

[0081] The graphic and text output module transmits the sorting results of each category tag, product category tag and product image to the user end, and arranges and displays the products according to the sorting results. It can effectively improve the matching degree between graphic and text content and user interests, enhance the attractiveness of product display and click-through conversion rate, and optimize user browsing efficiency, reduce search time, and help realize personalized content distribution and efficient scheduling of graphic and text resources, further support the system's real-time response and dynamic optimization of user behavior, and enhance the overall interactive experience.

[0082] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0083] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0084] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application of the technical solution and the invention constraints. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0085] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article, or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article, or device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or device that includes the element.

[0086] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referenced to each other.

[0087] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An e-commerce product image and text generation optimization system based on generative AI, characterized by: It includes a data integration module, a feature extraction module, a graphic and text generation module, and a graphic and text output module, and each module is connected by electrical signals; The data integration module is used to collect basic product information and user behavior data, classify products according to the basic product information, calculate the attention coefficient of each category-tagged product using the user behavior data, filter products according to the attention coefficient of each category tag, retain the user behavior data of the filtered products, and pass it together with the basic product information and category tags of the corresponding products to the feature extraction module; The feature extraction module extracts the core features of each category-tagged product and the user preference features from the basic product information of the filtered products and the user behavior data, determines the user interest weight value of each category-tagged product based on the core features of each category-tagged product and the user preference features, and transmits the user interest weight value to the image and text generation module; After receiving the user interest weight values ​​of each category tag, the image and text generation module collects feedback information of the products under the corresponding category tag, combines the user interest weight values ​​of each category tag and the feedback information to generate image and text information, and transmits the image and text information to the image and text output module; The feedback information for products under the corresponding category tag is the average frequency of image and text changes; By setting a time window, we obtain a set of products with the same category tag. For each product, we count the actual number of times the image and text content is changed within the time window. We then aggregate and average the number of image and text changes for all products to obtain the mean image and text change frequency, which is then normalized. Substitute the mean frequency of image and text replacement and the user interest weight into the logistic regression formula to calculate the ranking coefficient. The specific formula is as follows: ； Where L is the result of logistic regression, i.e., the ranking coefficient, e is the natural base, and y is the linear combination term of the logistic regression model. The specific setting of y is: ； Where, is the bias term, is the mean frequency of image and text replacement for the j-th category label, is the user interest weight value of the j-th category label, as well as are the regression coefficients of the mean frequency of picture and text replacement and the user interest weight value; The image and text output module is used to receive image and text information and output it to the user end.

2. The e-commerce product image and text generation optimization system based on generative AI according to claim 1 is characterized by: The basic product information in the data integration module includes product tags, product shipments, and product clicks; The data integration module classifies and marks the products according to their product labels. If the product labels of the products are the same, they will be marked as the same category. A period of time is selected as the collection time, and the product shipment volume and product click volume are obtained by counting the product shipment volume and the product click volume during the collection time.

3. The e-commerce product image and text generation optimization system based on generative AI according to claim 1 is characterized by: User behavior data includes product tag search volume, product return volume, and product return visit volume; The product tag search volume is the total number of product searches for the same product tag during the collection period; The product return volume is the total number of product returns with the same product label during the collection period; When collecting product return visits, the visit value of the product after deducting one from the number of product visits during the collection time is recorded, and then the visit values ​​of products with the same product label are summed up to obtain the product return visit value of the category tag corresponding to the product label.

4. The generative AI-based e-commerce product image and text generation optimization system according to claim 3 is characterized by: When the data integration module calculates the attention coefficient of each tagged product, it counts the number of product tag searches for different categories and calculates the attention coefficient of the corresponding category tag based on the number of product tag searches: , where s is the number of product tag retrievals for the current category, v is the sum of the number of product tag retrievals for all categories, and a is the attention coefficient for the current category tag; The attention coefficient of each category mark is compared with the preset screening threshold. When the attention coefficient of the category mark corresponding to the product exceeds the screening threshold, the corresponding product is screened out.

5. The generative AI-based e-commerce product image and text generation optimization system according to claim 3 is characterized by: The feature extraction module extracts the core features of products based on the shipment volume and click volume of products marked in each category. The specific steps are as follows: After normalizing the logarithms of the shipment volume and click volume of each category-labeled product, the core features of the corresponding category-labeled product are calculated using the geometric mean method: , where b is the normalized result of the shipment volume of the product marked in the corresponding category, c is the normalized result of the click volume of the product marked in the corresponding category, and h is the core feature of the product marked in the corresponding category; After taking the logarithm of the product return volume and product return volume of each category mark and normalizing them, the user preference characteristics of the corresponding category mark are obtained by subtracting the normalized result of the product return volume from the normalized result of the product return volume.

6. The generative AI-based e-commerce product image and text generation optimization system according to claim 5, characterized in that: The feature extraction module sums the core features of the products marked in the corresponding category and the user preference features and multiplies the sum by a preset weight adjustment ratio to generate the user interest weight value of the products marked in the corresponding category.

7. The generative AI-based e-commerce product image and text generation optimization system according to claim 1, characterized in that: Collect all sorting coefficients and sort them in descending order according to the numerical value of the sorting coefficient to obtain the sorting results of each category mark, and provide product images through the product database; The sorting results of each category tag, the product category tag and the product image are passed to the image and text output module as image and text information.

8. The generative AI-based e-commerce product image and text generation optimization system according to claim 7, characterized in that: The image and text output module obtains the image and text information and outputs it, and transmits the sorting results of each category tag, the product category tag and the product image to the user end respectively. According to the sorting results of each category tag, the products corresponding to each category tag are arranged and displayed.

Images