Credit card propaganda image-text generation method and device

Abstract

The invention discloses a credit card propaganda image-text generation method and device, and relates to the technical field of finance, and the method comprises the steps: screening a plurality of historical credit card propaganda prompt words according to the latest supervision requirements, and obtaining a plurality of compliant credit card propaganda prompt words; obtaining information of a plurality of credit card products, and determining a propaganda prompt word corresponding to each credit card product; inputting the basic information of the user into a credit card recommendation model, and outputting target credit card product information suitable for the user; inputting the character information of the target credit card product and the propaganda prompt word corresponding to the target credit card product into a financial document model, and outputting credit card propaganda document information; and inputting the basic information of the user, the target credit card product information and the credit card propaganda copywriting information into the text graph model to obtain a credit card propaganda image-text. According to the invention, high-quality and compliant credit card propaganda images and texts can be generated, and the efficiency of generating the credit card propaganda images and texts is improved.

Description

Technical Field

[0001] The present invention relates to the field of financial technology, and in particular to a method and device for generating credit card promotional graphics and texts. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the invention recited in the claims. No description herein is admitted to be prior art by inclusion in this section.

[0003] With the development of artificial intelligence technology, especially the advancement of large language models and artificial intelligence content generation technology, many fields have begun to adopt these technologies to improve work efficiency and quality. For example, in the financial field, traditional methods mainly rely on manual writing and operation, which is not only inefficient but also costly. Artificial intelligence technology can quickly generate pictures and texts to meet general picture and text generation needs. However, for specific financial industry transaction scenarios, such as the credit card field, it is difficult to generate graphic promotional manuscripts that meet actual needs. In the process of credit card graphic generation, it is easy to fail to meet regulatory requirements, and manual review is still required. Therefore, although artificial intelligence technology can automatically generate a large amount of high-quality content, the credibility is still low, and it cannot meet the requirements of credit card promotional manuscripts to be unique and compliant. Summary of the invention

[0004] The embodiment of the present invention provides a method for generating credit card promotional images and texts, which is used to generate high-quality and compliant credit card promotional images and texts, and improve the efficiency of generating credit card promotional images and texts. The method includes:

[0005] Collect multiple credit card promotional words from history;

[0006] According to the latest regulatory requirements, multiple historical credit card promotion prompts are screened to obtain multiple compliant credit card promotion prompts; the latest regulatory requirements include judgment information on whether the prompts are compliant;

[0007] Acquire multiple credit card product information; the credit card product information includes text information and pictures describing the credit card products;

[0008] Use multiple compliant credit card promotional prompts to match the text information of each credit card product to determine the promotional prompt corresponding to each credit card product;

[0009] Obtaining basic user information; the basic user information includes data describing user personal information and user consumption information;

[0010] Inputting the user's basic information into the credit card recommendation model, and outputting target credit card product information suitable for the user; the credit card recommendation model pre-uses the historical user's basic information and the credit card product information purchased by the user to train the multi-layer perceptron model;

[0011] Inputting text information of the target credit card product and the promotional prompt words corresponding to the target credit card product into a financial copywriting model, and outputting credit card promotional copywriting information; the financial copywriting model is obtained by fine-tuning the pre-trained large language model;

[0012] The user's basic information, target credit card product information and credit card promotional copy information are input into the text-generated graph model to obtain the credit card promotional image and text; the text-generated graph model includes an image and text generation module and a post-processing layer, the image and text generation module is obtained based on the diffusion model training, and the post-processing layer is used to perform compliance checks on the generated image after the image is generated.

[0013] The embodiment of the present invention also provides a credit card promotional image and text generation device, which is used to generate high-quality and compliant credit card promotional images and texts, and improve the efficiency of credit card promotional image and text generation. The device includes:

[0014] A historical prompt word collection module is used to collect multiple historical credit card promotion prompt words;

[0015] A compliance prompt word screening module is used to screen multiple historical credit card promotion prompt words according to the latest regulatory requirements to obtain multiple compliant credit card promotion prompt words; the latest regulatory requirements include judgment information on whether the prompt words are compliant;

[0016] A credit card product information processing module is used to obtain multiple credit card product information; the credit card product information includes text information and pictures describing the credit card product; using multiple compliant credit card promotion prompt words and the text information of each credit card product to match and determine the promotion prompt word corresponding to each credit card product;

[0017] The user credit card recommendation module is used to obtain basic user information; the basic user information includes data describing the user's personal information and the user's consumption information; the basic user information is input into the credit card recommendation model, and the target credit card product information suitable for the user is output; the credit card recommendation model is pre-trained using the historical basic user information and the credit card product information purchased by the user to obtain the multi-layer perceptron model;

[0018] A promotional copy generation module, used to input the text information of the target credit card product and the promotional prompt words corresponding to the target credit card product into a financial copy model, and output the credit card promotional copy information; the financial copy model is obtained by fine-tuning the pre-trained large language model;

[0019] The text-generated image processing module is used to input user basic information, target credit card product information and credit card promotional copy information into the text-generated image model to obtain credit card promotional images and texts; the text-generated image model includes an image-text generation module and a post-processing layer, the image-text generation module is obtained based on diffusion model training, and the post-processing layer is used to perform compliance checks on the generated images after the images are generated.

[0020] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned credit card promotional graphic and text generation method when executing the computer program.

[0021] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned credit card promotional image and text generation method is implemented.

[0022] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned credit card promotional image and text generation method is implemented.

[0023] In the embodiment of the present invention, a plurality of compliant credit card promotion prompt words are first collected, and then a suitable compliant credit card promotion prompt word is assigned to each credit card product. When promoting a credit card to a user, the credit card recommendation model is first used to determine the target credit card product information suitable for the user, and then the text information of the target credit card product and the promotion prompt word corresponding to the target credit card product are used to generate the credit card promotion copy information, and then the user basic information, the target credit card product information and the credit card promotion copy information are used to obtain the credit card promotion image and text, wherein the text-based graph model image and text generation module and the post-processing layer are used, and the image and text generation module is obtained based on the diffusion model training, and the post-processing layer is used to perform compliance checks on the generated image after the image is generated. Compared with the existing solution of simply using a large language model to generate images and texts, the embodiment of the present invention firstly uses multiple artificial intelligence models to generate copy information and promotional images and texts in a targeted manner, and the artificial intelligence model used is more suitable for actual needs and personalized needs, and is generated separately for a single credit card product and a specific user, and the generated credit card promotion image and text are of high quality, which can improve the user experience and satisfaction, and secondly, the compliance check is performed twice before and after, which ensures the review and filtering of non-compliant information, and generally improves the efficiency of credit card promotion image and text generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0025] Figure 1 It is a flow chart of a method for generating credit card promotional graphics and texts in an embodiment of the present invention;

[0026] Figure 2 A diagram showing a specific example of a method for generating credit card promotional graphics and texts in an embodiment of the present invention;

[0027] Figure 3 FIG. 1 is another specific example of the method for generating credit card promotional images and texts in an embodiment of the present invention;

[0028] Figure 4 Schematic diagram of a credit card promotional image and text generating device in an embodiment of the present invention;

[0029] Figure 5 Schematic diagram of a computer device in an embodiment of the present invention. DETAILED DESCRIPTION

[0030] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0031] First, the technical terms involved in this application are explained.

[0032] Diffusion model: The working principle is based on a step-by-step diffusion and de-diffusion process of the data distribution. In the generation process, it starts with a simple noise distribution, such as Gaussian noise, and then generates the target image through a series of denoising steps. The diffusion model includes two main processes: the forward diffusion process (converting the real image into noise) and the reverse diffusion process (generating the image from noise). The forward process is usually a Markov chain that gradually adds noise to the image until the image completely becomes a noise distribution. The reverse process uses a neural network (usually a U-Net architecture) to predict the noise that needs to be removed in each step to generate the image.

[0033] Multilayer Perceptron (MLP): A feedforward neural network consisting of multiple layers of neurons, each of which undergoes a nonlinear transformation through an activation function. MLP is one of the earliest deep learning models and is widely used in classification and regression tasks.

[0034] The acquisition, transmission, storage, use, and processing of data in the technical solution of this application comply with the relevant provisions of national laws and regulations.

[0035] It should be noted that in the embodiments of the present application, certain software, components, models and other existing solutions in the industry may be mentioned, and they should be regarded as exemplary. Their purpose is only to illustrate the feasibility of implementing the technical solution of the present application, but it does not mean that the applicant has or will necessarily use the solution.

[0036] Figure 1 FIG. 1 is a flow chart of a method for generating credit card promotional graphics and texts in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0037] Step 101, collecting multiple historical credit card promotion prompts;

[0038] Step 102: According to the latest regulatory requirements, multiple historical credit card promotion prompt words are screened to obtain multiple compliant credit card promotion prompt words; the latest regulatory requirements include judgment information on whether the prompt words are compliant;

[0039] Step 103: Acquire multiple credit card product information; the credit card product information includes text information and pictures describing the credit card products;

[0040] Step 104: using a plurality of compliant credit card promotional prompts and matching the text information of each credit card product, determine the promotional prompt corresponding to each credit card product;

[0041] Step 105: Obtain basic user information; the basic user information includes data describing user personal information and user consumption information;

[0042] Step 106: Input the basic information of the user into a credit card recommendation model, and output target credit card product information suitable for the user; the credit card recommendation model pre-uses historical basic information of the user and information of credit card products purchased by the user to train a multi-layer perceptron model;

[0043] Step 107: input the text information of the target credit card product and the promotional prompt words corresponding to the target credit card product into the financial copy model, and output the credit card promotional copy information; the financial copy model is fine-tuned based on the pre-trained large language model;

[0044] Step 108: Input the user's basic information, target credit card product information and credit card promotional copy information into the text-generated graph model to obtain the credit card promotional image and text; the text-generated graph model includes an image-text generation module and a post-processing layer, the image-text generation module is obtained based on diffusion model training, and the post-processing layer is used to perform compliance checks on the generated image after the image is generated.

[0045] from Figure 1 It can be seen from the process shown that in the embodiment of the present invention, according to the latest regulatory requirements, prompt words are first assigned and specified for each credit card product. Then, for any user, an artificial intelligence model is used to recommend a suitable target credit card, and a large language model is used to generate copy information. Then, the text-generated graph model is used to combine user information, copy information, pictures of the target credit card and other information to generate credit card promotional graphics and texts. The text-generated graph model used is trained based on the diffusion model to obtain a graphics and text generation module. The text-generated graph model also includes a post-processing layer for compliance inspection again. The embodiment of the present invention ensures the review and filtering of non-compliant information, and generally improves the efficiency and image quality of credit card promotional graphics and text generation.

[0046] The following is a detailed explanation of the credit card promotional image and text generation method in an embodiment of the present invention.

[0047] During implementation, first collect multiple historical credit card promotional words, such as low interest rates, interest-free periods, travel discounts, points rewards, membership discounts, partner merchants, etc. A collaborative office platform can be provided to receive credit card promotional words filled in and transmitted by multiple employees. Multiple credit card promotional words can also be obtained from historical credit card product data based on natural language processing.

[0048] Then, according to the latest regulatory requirements, multiple historical credit card promotion prompt words are screened to obtain multiple compliant credit card promotion prompt words, wherein the latest regulatory requirements include judgment information on whether the prompt words are compliant.

[0049] For example, sort out the words and phrases that are prohibited or restricted in the latest regulatory requirements, and then check one by one whether there are any non-compliance issues in multiple historical credit card promotional prompts.

[0050] Figure 2 FIG. 1 is a specific example of a method for generating credit card promotional graphics and texts in an embodiment of the present invention. Figure 2 As shown, the method includes:

[0051] Step 201: extracting lexical features and semantic features of each credit card promotional prompt word using natural language processing technology;

[0052] Step 202: Determine whether each credit card promotion word is compliant based on the latest regulatory requirements and the lexical features and semantic features of each credit card promotion word.

[0053] Preferably, natural language processing tools can be used to extract word features such as part of speech, stem, affix, etc. For example, for some words with negative affixes (such as "non" and "anti"), it may be necessary to focus on their meaning in the context to determine whether there is a potential non-compliance risk.

[0054] For semantic features, the semantic representation of words is extracted through the intermediate layer output of the word vector model or the deep learning classification model. For example, the BERT model can output the context-aware semantic vector of a word, and then calculate the similarity between the semantic vector and the semantic vector of non-compliant words in the latest regulatory requirements to determine the compliance of credit card promotional prompt words.

[0055] During implementation, multiple compliant credit card promotion prompt words obtained through screening are used as the underlying prompt words of the final text image, which can be stored in a vector knowledge base, and the vector knowledge base can be updated regularly afterwards.

[0056] When the regulatory requirements are updated, the search is directly based on the updated vector knowledge base, searching for prompt word related information, and submitting the feedback results and prompt words to a large language model in the financial field for prompt word judgment. If it does not meet the latest regulatory requirements, it will be deleted; otherwise, the prompt words are sorted to obtain the optimal financial prompt word sorting.

[0057] Get multiple credit card product information. The credit card product information includes text information and pictures describing the credit card product. For example, if the obtained credit card product information exists in the form of pictures or posters, then the text and picture are decomposed to obtain the text information and pictures describing the credit card product.

[0058] By matching multiple compliant credit card promotional words with the text information of each credit card product, the promotional words corresponding to each credit card product are determined.

[0059] During implementation, since different credit card products may have different characteristics, credit card promotion prompt words are assigned to each credit card product in advance.

[0060] In one embodiment, multiple compliant credit card promotional words are matched with the text information of each credit card product to determine the promotional word corresponding to each credit card product, which may include: for each credit card product: calculating the cosine similarity of each compliant credit card promotional word and the text information of the credit card product in turn, and using the credit card promotional word with a similarity exceeding a specified threshold as the promotional word corresponding to the credit card product.

[0061] For example, the word text information is converted into a vector in advance to obtain the selected credit card promotion prompt word vector E and the credit card product text information vector F, and the cosine similarity is calculated according to the following formula:

[0062] Cosine similarity C = (w × E × F) / (||E|| × ||F||);

[0063] Among them, w is the weight of vector E, which can be determined according to the ranking of the credit card promotion prompt words in the vector knowledge base. The higher the ranking, the greater the weight.

[0064] In step 105, basic user information is obtained, and the basic user information includes data describing the user's personal information and user consumption information, such as the user's gender, age, hobbies, historical consumption records, and other information.

[0065] In step 106, the user's basic information is input into a credit card recommendation model, and target credit card product information suitable for the user is output; the credit card recommendation model pre-uses historical user basic information and information on credit card products purchased by the user to train a multi-layer perceptron model.

[0066] In one embodiment, before inputting the user's basic information into the credit card recommendation model, the method may further include:

[0067] The user's basic information and each credit card product information are preprocessed to obtain user characteristics and credit card characteristics; the preprocessing includes missing value processing such as filling or deleting missing values, standardization processing and feature selection.

[0068] Standardization is to standardize or normalize numerical features so that the model can learn better. Feature selection is to select features related to the recommendation task and reduce redundant features.

[0069] Step 106 inputs the user's basic information into a credit card recommendation model and outputs target credit card product information suitable for the user, which may include: inputting user characteristics and credit card characteristics into a credit card recommendation model and outputting target credit card product information suitable for the user.

[0070] In one embodiment, the credit card recommendation model includes an input layer, multiple hidden layers and an output layer, the number of neurons in the input layer is equal to the total number of input user features and credit card features, the activation function of the hidden layer is the ReLU function, and the activation function of the output layer is the Softmax function.

[0071] The credit card recommendation model in the embodiment of the present invention is obtained by training based on a multi-layer perceptron model.

[0072] Input layer: The number of neurons in the input layer is equal to the total number of user features and credit card features.

[0073] Hidden layer: number of neurons = (number of neurons in the input layer + number of neurons in the output layer) / 2.

[0074] Output layer: The number of neurons in the output layer is equal to the number of credit cards to be recommended.

[0075] The ReLU activation function is selected for the hidden layer because it can effectively alleviate the gradient disappearance problem.

[0076] Select Adam optimization algorithm for optimization algorithm.

[0077] During training, hyperparameters such as learning rate, batch size, and number of hidden layers are adjusted to optimize model performance.

[0078] When the performance stops improving, the training is terminated early to prevent overfitting.

[0079] After obtaining the target credit card product information suitable for the user, in step 107, the text information of the target credit card product and the promotional prompt words corresponding to the target credit card product are input into the financial copy model, and the credit card promotional copy information is output; the financial copy model is obtained by fine-tuning the pre-trained large language model.

[0080] During implementation, the user's basic information can also be input into the financial copy model to output credit card promotional copy information with user preferences.

[0081] In step 108, the user's basic information, target credit card product information and credit card promotional copy information are input into the text-generated graph model to obtain the credit card promotional image and text; the text-generated graph model includes an image-text generation module and a post-processing layer, the image-text generation module is obtained based on the diffusion model training, and the post-processing layer is used to perform compliance checks on the generated image after the image is generated.

[0082] During implementation, the text-based graph model also includes a text encoder, which is used to encode the text information in the user's basic information, target credit card product information, and credit card promotional copy information into a text vector. The text encoder can extract the semantic information in the text and convert it into a numerical vector that the model can process for subsequent fusion and interaction with image features.

[0083] The image information in the user's basic information, target credit card product information, and credit card promotional copy information is first mapped to a low-dimensional latent space to obtain the latent representation of the image.

[0084] The text vector and the potential representation of the image are fused and input into the image-text generation module.

[0085] The image-text generation module can be built based on the UNet architecture, including an encoder and a decoder; the encoder is used to reduce the spatial dimension of the image and increase the number of channels to extract the features of the image; the decoder is used to reduce the number of channels and restore the spatial dimension of the image to generate the image.

[0086] The generated image is then passed through the post-processing layer for compliance check. If the compliance check fails, the generated image and the reason for failure are fed back to the input layer, and the text vector and image low-dimensional representation are regenerated to obtain the regenerated image. It is then passed through the post-processing layer for compliance check until the compliance check passes and the credit card promotional image and text are generated.

[0087] Afterwards, the credit card promotional images and texts are pushed to the user terminal.

[0088] Figure 3 FIG. 1 is another specific example of the method for generating credit card promotional images and texts in an embodiment of the present invention. Figure 3 As shown, the method includes: first, inputting a preliminary prompt word, searching a credit card scenario financial vocabulary, generating a financial prompt word set, filtering the prompt words according to regulatory requirements, and generating the optimal financial prompt word A after filtering; synchronously, obtaining basic user information, obtaining a credit card product library, selecting a credit card product suitable for the user, performing text-image decomposition on the credit card product, and obtaining a credit card text description information B and a credit card picture set; wherein the credit card text description information B and the financial prompt word A are merged and input into a financial large language model, and the prompt word is filtered according to regulatory requirements to generate text information FT; then the text information FT and the credit card picture set are input into a text-image model, a credit card promotion picture set is output, and picture verification is performed to generate a picture PS, and finally the text information FT and the picture PS are merged to generate a picture-text promotion manuscript for the credit card scenario. Figure 3 The aforementioned artificial intelligence method in the embodiment of the present invention is used in the process of determining the credit card products, copy information, text images, etc. that are suitable for the user, and will not be described in detail here.

[0089] The present invention also provides a device for generating credit card promotional images and texts, as described in the following embodiments. Since the principle of solving the problem by the device is similar to that of the method for generating credit card promotional images and texts, the implementation of the device can refer to the implementation of the method for generating credit card promotional images and texts, and the repeated parts will not be repeated.

[0090] Figure 4 FIG. 1 is a schematic diagram of a device for generating credit card promotional graphics and texts in an embodiment of the present invention. Figure 4 As shown, the device 400 includes:

[0091] A historical prompt word collection module 401 is used to collect multiple historical credit card promotion prompt words;

[0092] The compliance prompt word screening module 402 is used to screen a plurality of historical credit card promotion prompt words according to the latest regulatory requirements to obtain a plurality of compliant credit card promotion prompt words; the latest regulatory requirements include judgment information on whether the prompt words are compliant;

[0093] The credit card product information processing module 403 is used to obtain a plurality of credit card product information; the credit card product information includes text information and pictures describing the credit card product; and to determine the promotional prompt word corresponding to each credit card product by matching a plurality of compliant credit card promotional prompt words with the text information of each credit card product;

[0094] The user credit card recommendation module 404 is used to obtain basic user information; the basic user information includes data describing the user's personal information and the user's consumption information; the basic user information is input into the credit card recommendation model, and the target credit card product information suitable for the user is output; the credit card recommendation model is pre-trained using the historical basic user information and the credit card product information purchased by the user to obtain the multi-layer perceptron model;

[0095] The promotional copy generation module 405 is used to input the text information of the target credit card product and the promotional prompt words corresponding to the target credit card product into the financial copy model, and output the credit card promotional copy information; the financial copy model is fine-tuned based on the pre-trained large language model;

[0096] The text-generated image processing module 406 is used to input the user's basic information, the target credit card product information and the credit card promotional copy information into the text-generated image model to obtain the credit card promotional image and text; the text-generated image model includes an image and text generation module and a post-processing layer, the image and text generation module is obtained based on the diffusion model training, and the post-processing layer is used to perform compliance checks on the generated image after the image is generated.

[0097] In one embodiment, the compliance prompt word screening module 402 is specifically used to:

[0098] Using natural language processing technology, extract the lexical and semantic features of each credit card promotional word;

[0099] Determine whether each credit card promotion word is compliant based on the latest regulatory requirements and the lexical and semantic features of each credit card promotion word.

[0100] In one embodiment, the credit card product information processing module 403 is specifically used for: for each credit card product: calculating the cosine similarity between each compliant credit card promotional word and the text information of the credit card product in turn, and using the credit card promotional word with a similarity exceeding a specified threshold as the promotional word corresponding to the credit card product.

[0101] In one embodiment, the apparatus 400 further includes a preprocessing module. The preprocessing module is used to:

[0102] Before the user credit card recommendation module 404 inputs the user basic information into the credit card recommendation model, the user basic information and each credit card product information are preprocessed to obtain user features and credit card features; the preprocessing includes missing value processing, standardization processing and feature selection;

[0103] Input the user's basic information into the credit card recommendation model, and output the target credit card product information suitable for the user, including:

[0104] Input user characteristics and credit card characteristics into the credit card recommendation model and output target credit card product information suitable for the user.

[0105] In one embodiment, the credit card recommendation model includes an input layer, multiple hidden layers and an output layer, the number of neurons in the input layer is equal to the total number of input user features and credit card features, the activation function of the hidden layer is the ReLU function, and the activation function of the output layer is the Softmax function.

[0106] In one embodiment, the image-text generation module includes an encoder and a decoder; the encoder is used to reduce the spatial dimension of the image and increase the number of channels to extract the features of the image; the decoder is used to reduce the number of channels and restore the spatial dimension of the image to generate the image.

[0107] Figure 5 Schematic diagram of a computer device in an embodiment of the present invention. Figure 5 As shown, an embodiment of the present invention further provides a computer device 500, including a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501, wherein the processor 501 implements the above-mentioned credit card promotional image and text generation method when executing the computer program 503.

[0108] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned credit card promotional image and text generation method is implemented.

[0109] An embodiment of the present invention further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the above-mentioned credit card promotional image and text generation method is implemented.

[0110] In the embodiment of the present invention, a plurality of compliant credit card promotion prompt words are first obtained, and then appropriate compliant credit card promotion prompt words are assigned to each credit card product. When promoting a credit card to a user, the credit card recommendation model is first used to determine the target credit card product information suitable for the user, and then the text information of the target credit card product and the promotion prompt words corresponding to the target credit card product are used to generate the credit card promotion copy information, and then the user basic information, the target credit card product information and the credit card promotion copy information are used to obtain the credit card promotion image and text, wherein the text-based graph model image and text generation module and the post-processing layer are used, and the image and text generation module is obtained based on the diffusion model training, and the post-processing layer is used to perform compliance checks on the generated image after the image is generated. Compared with the existing solution of simply using a large language model to generate images and texts, the embodiment of the present invention firstly uses multiple artificial intelligence models to generate copy information and promotional images and texts in a targeted manner, and the artificial intelligence model used is more suitable for actual needs and personalized needs, and is generated separately for a single credit card product and a specific user, and the generated credit card promotion image and text are of high quality, which can improve the user experience and satisfaction, and secondly, the compliance check is performed twice before and after, which ensures the review and filtering of non-compliant information, and generally improves the efficiency of credit card promotion image and text generation.

[0111] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0112] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0115] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of 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.

Claims

1. A method for generating credit card promotional images and texts, characterized in that: include: Collect multiple credit card promotional words from history; According to the latest regulatory requirements, multiple historical credit card promotional reminders are screened to obtain multiple compliant credit card promotional reminders; The latest regulatory requirements include information on whether the prompt words are compliant; Acquire multiple credit card product information; the credit card product information includes text information and pictures describing the credit card products; Use multiple compliant credit card promotional reminder words to match the text information of each credit card product to determine the promotional reminder words corresponding to each credit card product; Obtaining basic user information; the basic user information includes data describing user personal information and user consumption information; Inputting the user's basic information into the credit card recommendation model, and outputting target credit card product information suitable for the user; the credit card recommendation model pre-uses the historical user's basic information and the credit card product information purchased by the user to train the multi-layer perceptron model; Inputting text information of the target credit card product and the promotional prompt words corresponding to the target credit card product into a financial copywriting model, and outputting credit card promotional copywriting information; the financial copywriting model is obtained by fine-tuning the pre-trained large language model; The user's basic information, target credit card product information and credit card promotional copy information are input into the text-generated graph model to obtain the credit card promotional image and text; the text-generated graph model includes an image and text generation module and a post-processing layer, the image and text generation module is obtained based on the diffusion model training, and the post-processing layer is used to perform compliance checks on the generated image after the image is generated.

2. The method according to claim 1, characterized in that According to the latest regulatory requirements, multiple historical credit card promotional reminders were screened to obtain multiple compliant credit card promotional reminders, including: Using natural language processing technology, extract the lexical and semantic features of each credit card promotional word; Based on the latest regulatory requirements and the lexical and semantic features of each credit card promotional word, determine whether each credit card promotional word is compliant.

3. The method according to claim 1, characterized in that Use multiple compliant credit card promotional words to match the text information of each credit card product to determine the promotional words corresponding to each credit card product, including: For each credit card product: calculate the cosine similarity between each compliant credit card promotion word and the text information of the credit card product in turn, and use the credit card promotion word with a similarity exceeding a specified threshold as the promotion word corresponding to the credit card product.

4. The method according to claim 1, characterized in that Before inputting the user's basic information into the credit card recommendation model, it also includes: Preprocessing the basic information of the user and the product information of each credit card to obtain user characteristics and credit card characteristics; the preprocessing includes missing value processing, standardization processing and feature selection; Input the user's basic information into the credit card recommendation model, and output the target credit card product information suitable for the user, including: Input user characteristics and credit card characteristics into the credit card recommendation model and output target credit card product information suitable for the user.

5. The method according to claim 4, characterized in that The credit card recommendation model includes an input layer, multiple hidden layers and an output layer. The number of neurons in the input layer is equal to the total number of input user features and credit card features. The activation function of the hidden layer is the ReLU function, and the activation function of the output layer is the Softmax function.

6. The method according to claim 1, characterized in that The image-text generation module includes an encoder and a decoder; the encoder is used to reduce the spatial dimension of the image and increase the number of channels to extract the features of the image; The decoder is used to reduce the number of channels while restoring the spatial dimensions of the image to generate an image.

7. A credit card promotional image and text generating device, characterized in that: include: A historical prompt word collection module is used to collect multiple historical credit card promotion prompt words; The compliance prompt word screening module is used to screen multiple historical credit card promotion prompt words according to the latest regulatory requirements to obtain multiple compliant credit card promotion prompt words; The latest regulatory requirements include information on whether the prompt words are compliant; A credit card product information processing module is used to obtain multiple credit card product information; the credit card product information includes text information and pictures describing the credit card product; using multiple compliant credit card promotion prompt words and the text information of each credit card product to match and determine the promotion prompt word corresponding to each credit card product; The user credit card recommendation module is used to obtain basic user information; the basic user information includes data describing the user's personal information and the user's consumption information; the basic user information is input into the credit card recommendation model, and the target credit card product information suitable for the user is output; the credit card recommendation model is pre-trained using the historical basic user information and the credit card product information purchased by the user to obtain the multi-layer perceptron model; A promotional copy generation module, used to input the text information of the target credit card product and the promotional prompt words corresponding to the target credit card product into a financial copy model, and output the credit card promotional copy information; the financial copy model is obtained by fine-tuning the pre-trained large language model; The text-generated image processing module is used to input user basic information, target credit card product information and credit card promotional copy information into the text-generated image model to obtain credit card promotional images and texts; the text-generated image model includes an image-text generation module and a post-processing layer, the image-text generation module is obtained based on diffusion model training, and the post-processing layer is used to perform compliance checks on the generated images after the images are generated.

8. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images