Artificial intelligence-based copywriting generation method and device, computer equipment and medium

Abstract

The invention belongs to the technical field of artificial intelligence, and relates to an artificial intelligence-based copywriting generation method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining a theme keyword and a poster template input by a user; performing semantic analysis on the theme keyword to obtain semantic information; performing template adaptation processing on the poster template to obtain template typesetting constraint information; performing preliminary content generation processing on the semantic information based on a content generation model to obtain corresponding first content; based on the first content, performing reasoning processing on the semantic information through a content generation model to obtain corresponding second content; performing content filtering processing on the first content and the second content to obtain target content; and performing template constraint adaptation processing on the target content based on the template typesetting constraint information to generate a target copywriting. The method can be applied to business scenes of poster copywriting generation of related posters in the financial science and technology field and the medical health insurance field, and the efficiency and quality of poster copywriting generation are effectively improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology and can be applied to the fields of financial technology and medical health insurance, and in particular to artificial intelligence-based copywriting generation methods, devices, computer equipment and storage media. Background Art

[0002] In the poster copywriting scenario of the traditional insurance industry, agents currently rely heavily on general AI tools to complete copywriting. These tools usually use a thrust model with a fixed number of reasoning steps (such as a fixed execution of 20-50 steps of reasoning), resulting in low efficiency in poster copywriting generation. Specifically, because the model lacks the ability to dynamically adjust the reasoning path, agents need to wait a long time for the generation results of each copy, which severely limits the number of copies that can be processed daily, thereby affecting business development efficiency. This inefficient generation model not only increases the time cost of agents, but also makes it difficult to meet the insurance industry's demand for rapid response and efficient output.

[0003] For example, in insurance product marketing activities in the financial sector, agents need to customize differentiated poster copy for different customer groups. During the generation process, traditional AI tools need to complete a complete fixed-step reasoning regardless of the complexity of the copy, resulting in no significant difference in the generation time of simple copy (such as a single insurance introduction) and complex copy (such as a multi-insurance combination plan). If an agent needs to customize a comprehensive poster copy for a corporate client that includes a multi-insurance combination, insurance process, and claims service, traditional tools still need to generate it step by step according to a fixed number of steps, and some of the reasoning steps may only involve simple semantic repetition, resulting in a waste of computing resources. This inefficient generation method not only reduces the agent's work efficiency, but may also lead to missed marketing opportunities due to insufficient response speed.

[0004] Therefore, there is an urgent need to provide an efficient copy generation method to improve the efficiency of poster copy generation, meet the insurance industry's demand for efficient and accurate copy output, and improve the overall effectiveness of insurance marketing activities. Summary of the Invention

[0005] The purpose of the embodiments of the present application is to propose a method, device, computer equipment and storage medium for copywriting generation based on artificial intelligence to solve the technical problem of low generation efficiency of existing poster copywriting generation methods.

[0006] In a first aspect, a method for generating copywriting based on artificial intelligence is provided, comprising:

[0007] Get the theme keywords and poster templates entered by the user;

[0008] Perform semantic analysis on the topic keywords based on a preset analysis model to obtain corresponding semantic information;

[0009] Performing template adaptation processing on the poster template to obtain corresponding template typesetting constraint information;

[0010] Performing preliminary content generation processing on the semantic information based on a preset content generation model, terminating the initial content generation process and obtaining the corresponding first content when each layer of content is generated and detecting that the confidence level of the current layer reaches a preset exit threshold; wherein the content generation model is a model that integrates a dynamic early exit mechanism;

[0011] Based on the first content, the semantic information is inferred by the content generation model. When each layer of content is generated, if the confidence level of a specified number of layers triggers a preset absolute threshold exit condition, the inference process is terminated and the corresponding second content is obtained.

[0012] performing content filtering on the first content and the second content to obtain corresponding target content;

[0013] Based on the template typesetting constraint information, template constraint adaptation processing is performed on the target content to generate a corresponding target copy.

[0014] In a second aspect, a copywriting generation device based on artificial intelligence is provided, comprising:

[0015] The acquisition module is used to obtain the theme keywords and poster templates input by the user;

[0016] A parsing module, configured to perform semantic parsing on the subject keywords based on a preset parsing model to obtain corresponding semantic information;

[0017] A first processing module is used to perform template adaptation processing on the poster template to obtain corresponding template typesetting constraint information;

[0018] a second processing module configured to perform preliminary content generation processing on the semantic information based on a preset content generation model, terminating the initial content generation process and obtaining corresponding first content when each layer of content is generated and detecting that the confidence level of the current layer reaches a preset exit threshold; wherein the content generation model is a model that integrates a dynamic early exit mechanism;

[0019] a third processing module configured to perform inference processing on the semantic information based on the first content using the content generation model, and terminate the inference process when the confidence levels of a specified number of layers trigger a preset absolute threshold exit condition when each layer of content is generated, thereby obtaining corresponding second content;

[0020] a filtering module, configured to perform content filtering on the first content and the second content to obtain corresponding target content;

[0021] The first generating module is used to perform template constraint adaptation processing on the target content based on the template typesetting constraint information to generate a corresponding target copy.

[0022] In a third aspect, a computer device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned artificial intelligence-based copywriting generation method when executing the computer program.

[0023] In a fourth aspect, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned artificial intelligence-based copywriting generation method are implemented.

[0024] In the solution implemented by the above-mentioned artificial intelligence-based copywriting generation method, device, computer equipment and storage medium, the following steps are firstly used: first, the user-inputted theme keywords and poster templates are obtained; then, the theme keywords are semantically parsed based on a preset parsing model to obtain corresponding semantic information; and the poster template is template-adapted to obtain corresponding template typesetting constraint information; then, preliminary content generation processing is performed on the semantic information based on a preset content generation model. When each layer of content is generated, if it is detected that the confidence of the current layer reaches a preset exit threshold, the initial content generation process is terminated, and the corresponding first content is obtained; wherein, the content generation model is a model that integrates a dynamic early exit mechanism; subsequently, based on the first content, the semantic information is inferred through the content generation model. When each layer of content is generated, if it is detected that the confidence of a specified number of layers triggers a preset absolute threshold exit condition, the inference process is terminated, and the corresponding second content is obtained; the first content and the second content are further content-filtered to obtain the corresponding target content; finally, the target content is template-constrained based on the template typesetting constraint information to generate the corresponding target copywriting. This application obtains the subject keywords and poster templates input by the user, then performs semantic analysis on the subject keywords based on the use of a parsing model to obtain semantic information, and performs template adaptation processing on the poster template to obtain template typesetting constraint information. Subsequently, based on the use of a content generation model that integrates a dynamic early exit mechanism, it achieves efficient construction of high-quality copy content through two stages: preliminary content generation and reasoning processing, breaking through the generation limit of a fixed number of steps and effectively improving the efficiency of copy generation. Subsequently, the copy generated by the content generation model will be further optimized and adapted intelligently through the two steps of content filtering and template constraint adaptation, thereby ensuring that the final output copy meets high-quality standards in both content and form, thereby improving the quality of the generated target copy. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the solutions in this application, a brief introduction will be given below to the drawings required for use in the description of the embodiments of this application. Obviously, the drawings described below are some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0026] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0027] Figure 2 is a flow chart of an embodiment of the artificial intelligence-based copywriting generation method according to the present application;

[0028] Figure 3 This is a structural diagram of an embodiment of an artificial intelligence-based copywriting generation device according to the present application;

[0029] Figure 4 It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art to which this application belongs. The terms used in the specification of the application are for the purpose of describing specific embodiments only and are not intended to limit this application. The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0031] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0032] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0033] like Figure 1As shown, system architecture 100 may include a terminal device 101, a network 102, and a server 103. Terminal device 101 may be a laptop computer 1011, a tablet computer 1012, or a mobile phone 1013. Network 102 is a medium for providing a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0034] The user can use the terminal device 101 to interact with the server 103 via the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0035] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer and a desktop computer, etc.

[0036] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .

[0037] It should be noted that the artificial intelligence-based copy generation method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based copy generation device is generally set in the server / terminal device.

[0038] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0039] Continue to refer Figure 2, showing a flowchart of an embodiment of the method for generating text based on artificial intelligence according to the present application. According to different needs, the order of the steps in the flowchart can be changed, and some steps can be omitted. The method for generating text based on artificial intelligence provided in the embodiment of the present application can be applied to any scenario where product recommendations are required, and the method for generating text based on artificial intelligence can be applied to products in these scenarios, for example, poster copy generation scenarios for related posters in the fields of financial technology and medical health insurance. The method for generating text based on artificial intelligence comprises the following steps:

[0040] Step S201: Acquire the subject keywords and poster template input by the user.

[0041] In this embodiment, the method for generating text based on artificial intelligence is executed on the electronic device (e.g. Figure 1 The server / terminal device shown in the figure can receive user-entered theme keywords and poster templates via a wired or wireless connection. It should be noted that the aforementioned wireless connection methods may include, but are not limited to, 3G / 4G / 5G connections, WiFi connections, Bluetooth connections, WiMAX connections, Zigbee connections, UWB (Ultra Wide Band) connections, and other currently known or future developed wireless connection methods. The execution subject of this application is specifically a copywriting generation system, which may be referred to as the system. This application can be applied to business scenarios for poster copywriting generation in the fields of financial technology and healthcare insurance. The aforementioned users may refer to insurance agents. Based on their actual poster copywriting generation requirements, users can enter corresponding theme keywords (such as "pension insurance products," "critical illness insurance products," "accident insurance products," etc.) and relevant information corresponding to the poster template through the system page, including the template size (such as A4) and color scheme (such as warm tones). The system will receive and store this input information as basic data for subsequent steps. The theme keywords and poster templates entered by the user provide raw data support for semantic parsing and template adaptation, ensuring that subsequent steps are carried out according to the agent's needs.

[0042] Step S202: semantically analyze the topic keywords based on a preset analysis model to obtain corresponding semantic information.

[0043] In this embodiment, the above parsing model is a model with semantic parsing function. The selection of this parsing model is not specifically limited and can be determined according to actual business requirements. For example, a pre-trained BERT-Base model can be selected. Among them, the input topic keywords can be semantically parsed according to the selected parsing model. Specifically, the parsing model extracts the core entities in the keywords (such as "annuity receipt", "retirement security"), filters out meaningless stop words (such as "of", "already"), and returns the corresponding output results. The output result is structured semantic information, focusing on key entities and core concepts. In addition, the result of semantic parsing (semantic information) provides the core semantic elements for the content generation process of the content generation model, which helps to ensure that the generated copy content closely follows the theme and avoids deviating from the user's needs.

[0044] Step S203: Perform template adaptation processing on the poster template to obtain corresponding template layout constraint information.

[0045] In this embodiment, the specific implementation process of performing template adaptation processing on the poster template to obtain corresponding template layout constraint information will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated here too much.

[0046] Step S204: Based on a preset content generation model, perform preliminary content generation processing on the semantic information. When generating each layer of content, if it is detected that the confidence level of the current layer reaches the preset exit threshold, the initial content generation process is terminated, and the corresponding first content is obtained; wherein, the content generation model is a model integrated with a dynamic early exit mechanism.

[0047] In this embodiment, the above content generation model (which can be simply referred to as the model) is a model integrated with a dynamic early exit mechanism, that is, the DEER (Dynamic Early Exit in Reasoning) dynamic early exit mechanism is applied to the AI model for content generation. The selection of this AI model is not specifically limited. For example, a Transformer model can be used. The DEER mechanism can be integrated into the model as a component or module for content generation. And the performance of the DEER mechanism in the content generation of the AI model can be optimized by adjusting parameters such as the routing matrix, confidence threshold, and resource allocation strategy.

[0048] Specifically, the Dynamic Early Exit (DEER) mechanism achieves a balance between generation efficiency and quality through multi-level confidence assessment and dynamic exit decisions. Its technical process is divided into three phases: 1. Phased Inference and Confidence Calculation: The generation process is split into 20 independent levels (Level 1 to Level 20), each of which outputs a hidden state Hi; each level corresponds to a Transformer block in the model. For example, Level 1 handles basic semantic parsing, Level 5 generates core selling points, and Level 15 completes compliance verification. Dynamic Routing: Computing resources are dynamically allocated through a routing matrix. For simple scenarios (such as short copy), only the first five levels are activated (accounting for 15% of the computational effort), while for complex scenarios (such as long copy with multiple selling points), all 20 levels are activated (accounting for 100% of the computational effort). The confidence level Ci = σ(Wc·Hi+bc) is calculated using the Sigmod function, reflecting the consistency between the current content and the target semantics (the closer Ci is to 1, the more reliable the content). 2. Dynamic Exit Decision: Absolute Threshold (0.92): If the confidence level of three consecutive layers exceeds 0.92, inference is terminated immediately (for example, after generating the core selling point "100 Diseases Covered," subsequent modifiers become redundant); Relative Threshold (15%): If the confidence level of the current layer increases by >15% compared to the previous layer, an accelerated exit is triggered (for example, the confidence level jumps from "Basic Coverage" to "Payment Upon Diagnosis"). 3. Adaptive Resource Allocation: For simple scenarios (short copy), the first 1 / 3 layers of the model (3B parameters) are activated, with a computational latency of <50ms; for complex scenarios (long copy), the entire model (175B parameters) is enabled for parallel computation, maintaining a response time of <200ms.

[0049] The advantages of using the DEER mechanism in content generation of AI models include: 1. Improving generation efficiency: Through dynamic exit decisions, the DEER mechanism avoids unnecessary reasoning steps and reduces the waste of computing resources. In simple scenarios, it can quickly generate copy that meets the needs, improving overall generation efficiency. 2. Ensuring generation quality: The DEER mechanism ensures that the content generated at each layer is consistent with the target semantics through confidence assessment. In complex scenarios, the accuracy and completeness of the generated copy are guaranteed by enabling parallel computing of the complete model. 3. Adapting to different needs: The DEER mechanism can adaptively allocate computing resources according to the complexity and needs of copy generation. Whether it is a short copy or a long copy, efficient generation can be achieved by adjusting the number of reasoning layers and computing resources.

[0050] In addition, the initial content generation process includes the following: 1) Input integration: Receive core entities from semantic parsing (e.g., "annuity collection" and "retirement security") and layout constraint information from template adaptation (e.g., key area location and visual complexity). This information is integrated as the initial input for copy generation. 2) Content generation: Based on this integrated input, the model begins generating preliminary copy content layer by layer. For example, the first layer might generate "[Pension Insurance] covers post-retirement living security," while the second layer adds "Designed specifically for middle-aged and elderly individuals." Each layer of content generation builds on the results of the previous layer, gradually enriching the copy's detail and completeness. 3) Confidence calculation: After each layer of content is generated, the confidence of the current layer is calculated using the Sigma ID function. The Sigma ID function maps content quality to a confidence interval between 0 and 1 (e.g., C1 = 0.78, C2 = 0.82). The confidence level reflects the reliability and relevance of the currently generated content. 4) Exit condition determination: Check whether the current confidence level reaches the preset exit threshold (e.g., confidence levels > 0.92 for three consecutive layers). If the threshold is not reached, the next layer of content will be generated; if the threshold is reached, the initial content generation phase will be terminated. In addition, the results of the initial content generation provide a basic framework and content direction for the subsequent generation of core selling points, ensuring the overall coherence of the copy.

[0051] Step S205: Based on the first content, the semantic information is inferred through the content generation model. When each layer of content is generated, if the confidence of a specified number of layers is detected to trigger a preset absolute threshold exit condition, the inference process is terminated and the corresponding second content is obtained.

[0052] In this embodiment, reasoning processing refers to the process of generating core selling points, which specifically includes the following implementation steps: 1) In-depth reasoning: Based on the preliminary content, further reasoning is performed to generate core selling point content. For example, the sixth layer generates "monthly annuity payment, stable income for life", and the seventh layer adds "flexible payment methods to meet different needs". The generation of core selling points pays more attention to highlighting the unique value and competitive advantage of the product. 2) Dynamic change of confidence: As the reasoning deepens, the confidence may change due to the improvement of content quality (such as C6 = 0.93, C7 = 0.94). The improvement in confidence indicates that the generated content is more focused, accurate and highly compatible with the template. 3) Absolute threshold exit condition: When the confidence of a specified number of consecutive layers (for example, 3 layers) exceeds a threshold (such as 0.92), the absolute threshold exit condition is triggered. Terminate the reasoning process to avoid generating redundant or low-quality content.

[0053] The core selling point generation process is the essence of the copywriting, directly relying on the foundation of initial content generation and the constraints of template adaptation. A dynamic early exit mechanism ensures the efficiency of the content generation process and the optimization of content quality.

[0054] Step S206: Perform content filtering on the first content and the second content to obtain corresponding target content.

[0055] In this embodiment, the specific implementation process of performing content filtering on the first content and the second content to obtain the corresponding target content will be further described in detail in subsequent specific embodiments of this application and will not be elaborated on here.

[0056] Step S207 : performing template constraint adaptation processing on the target content based on the template typesetting constraint information to generate a corresponding target copy.

[0057] In this embodiment, the above-mentioned specific implementation process of performing template constraint adaptation processing on the target content based on the template typesetting constraint information to generate the corresponding target copy will be further described in detail in subsequent specific embodiments of this application and will not be elaborated on here.

[0058] This application first obtains the subject keywords and poster templates input by the user; then performs semantic analysis on the subject keywords based on a preset parsing model to obtain corresponding semantic information; and performs template adaptation processing on the poster template to obtain corresponding template layout constraint information; then performs preliminary content generation processing on the semantic information based on a preset content generation model, and when each layer of content is generated, if it is detected that the confidence of the current layer reaches a preset exit threshold, the initial content generation process is terminated, and the corresponding first content is obtained; wherein, the content generation model is a model that integrates a dynamic early exit mechanism; subsequently, based on the first content, the semantic information is inferred through the content generation model, and when each layer of content is generated, if it is detected that the confidence of a specified number of layers triggers a preset absolute threshold exit condition, the inference process is terminated, and the corresponding second content is obtained; further content filtering processing is performed on the first content and the second content to obtain the corresponding target content; finally, template constraint adaptation processing is performed on the target content based on the template layout constraint information to generate the corresponding target copy. This application obtains the subject keywords and poster templates input by the user, then performs semantic analysis on the subject keywords based on the use of a parsing model to obtain semantic information, and performs template adaptation processing on the poster template to obtain template typesetting constraint information. Subsequently, based on the use of a content generation model that integrates a dynamic early exit mechanism, it achieves efficient construction of high-quality copy content through two stages: preliminary content generation and reasoning processing, breaking through the generation limit of a fixed number of steps and effectively improving the efficiency of copy generation. Subsequently, the copy generated by the content generation model will be further optimized and adapted intelligently through the two steps of content filtering and template constraint adaptation, thereby ensuring that the final output copy meets high-quality standards in both content and form, thereby improving the quality of the generated target copy.

[0059] In some optional implementations, step S203 includes the following steps:

[0060] The poster template is segmented into key areas based on a preset segmentation model to obtain corresponding area segmentation results.

[0061] In this embodiment, the segmentation model is a model with image segmentation capabilities. There is no specific limitation on the selection of the segmentation model. For example, a pre-trained SAM (Segment Anything Model) model can be used. The poster template can be input into the selected segmentation model, and it is ensured that the template image format of the poster is compatible with the segmentation model. Then, the segmentation model automatically segments the poster template and identifies different areas in the image, such as the price block, the product highlight display area, the data chart area, the QR code location, etc. Among them, the segmentation model accurately divides the boundaries of each key area through pixel-level classification and boundary detection. Subsequently, a regional segmentation result containing key area labels and boundary coordinates is generated, such as the coordinate range of the price block, the location of the product highlight display area, etc. This information is output as structured data to facilitate subsequent steps. In addition, the generated key area segmentation result provides specific visual element position information for template adaptation, so that subsequent typesetting and visual adaptation can be performed based on the actual template layout.

[0062] The visual complexity of the poster template is calculated based on a preset calculation strategy.

[0063] In this embodiment, the specific implementation process of calculating the visual complexity of the poster template based on the preset calculation strategy will be further described in detail in subsequent specific embodiments of this application and will not be elaborated on here.

[0064] The region segmentation result and the visual complexity are integrated to generate a corresponding template visual characteristic description.

[0065] In this embodiment, the generated region segmentation result and the visual complexity may be integrated to form a complete template visual characteristic description.

[0066] Based on the template visual characteristic description, template layout constraint information corresponding to the poster template is generated.

[0067] In this embodiment, matching template layout constraint information can be automatically generated based on the location and visual complexity of the key areas included in the generated template visual feature description. For example, for a simple template, the generated template layout constraint information may include: recommending the use of basic alignment (e.g., center alignment). For a complex template, the generated template layout constraint information may include: recommending the use of more sophisticated alignment and spacing constraints (e.g., maintaining a specific spacing from a data chart).

[0068] This application performs key area segmentation on the poster template based on a preset segmentation model to obtain corresponding area segmentation results; then calculates the visual complexity of the poster template based on a preset calculation strategy; then integrates the area segmentation results and the visual complexity to generate a corresponding template visual feature description; and subsequently generates template typesetting constraint information corresponding to the poster template based on the template visual feature description. This application performs key area segmentation on the poster template based on the use of a segmentation model to obtain area segmentation results, and calculates the visual complexity of the poster template based on the use of a calculation strategy, then integrates the area segmentation results and the visual complexity to generate a corresponding template visual feature description, and then based on the use of the template visual feature description, it can achieve efficient and accurate generation of template typesetting constraint information corresponding to the poster template, thereby improving the generation efficiency of the template typesetting constraint information and ensuring the accuracy of the obtained template typesetting constraint information. Moreover, the generated template typesetting constraint information can be used to provide important constraint conditions for the content generation processing of the content generation model, thereby effectively ensuring that the generated target copy can be coordinated with the visual elements of the poster template to avoid typesetting conflicts or information overload.

[0069] In some optional implementations of this embodiment, calculating the visual complexity of the poster template based on a preset calculation strategy includes the following steps:

[0070] The poster template is preprocessed to obtain a corresponding target image.

[0071] In this embodiment, the format of the poster template is an image format. The poster template can be preprocessed, including grayscale conversion, edge detection, etc., to obtain a corresponding target image, so as to facilitate the subsequent calculation of visual complexity.

[0072] The edge of the target image is extracted based on a preset edge detection algorithm to obtain corresponding edge information.

[0073] In this embodiment, the selection of the edge detection algorithm is not specifically limited and can be determined based on actual business needs. For example, the Canny edge detection algorithm can be selected. Specifically, the edge of the target image can be extracted based on the selected edge detection algorithm to extract edge information from the target image.

[0074] Based on the edge information, the total edge length and the total image area of ​​the target image are calculated.

[0075] In this embodiment, the total edge length (Edge) and the total image area (Area) of the target image may be calculated based on the use of the generated edge information.

[0076] The ratio between the total edge length and the total area of ​​the image is calculated.

[0077] In this embodiment, the ratio of the total edge length to the total image area, ie, Edge / area, may be calculated according to a ratio calculation formula.

[0078] A visual complexity corresponding to the poster template is generated based on the ratio.

[0079] In this embodiment, the layout complexity of the above-mentioned poster template can be evaluated based on the obtained ratio value: a low ratio (such as <0.2): a simple template with concise layout and fewer visual elements. A medium ratio (such as 0.2-0.5): a medium-complexity template containing a certain number of visual elements. A high ratio (such as >0.5): a complex template with dense layout and rich visual elements. Then, the visual complexity of the above-mentioned poster template is generated based on the obtained layout complexity, such as "Complexity: Medium".

[0080] This application obtains the corresponding target image by preprocessing the poster template; then extracts the edge of the target image based on a preset edge detection algorithm to obtain the corresponding edge information; then calculates the total edge length and the total image area of ​​the target image based on the edge information; subsequently calculates the ratio between the total edge length and the total image area; and finally generates the visual complexity corresponding to the poster template based on the ratio. This application obtains the target image by preprocessing the poster template; then extracts the edge of the target image based on the edge detection algorithm to obtain edge information; and calculates the total edge length and the total image area of ​​the target image based on the edge information, and then calculates the ratio between the total edge length and the total image area, and automatically and accurately generates the visual complexity corresponding to the poster template based on the obtained ratio, thereby ensuring the accuracy of the obtained visual complexity. In addition, the generated visual complexity can be used to provide a reference for the complexity of the layout for template adaptation, thereby helping subsequent steps to formulate appropriate constraints to adapt to templates of different complexities.

[0081] In some optional implementations, step S206 includes the following steps:

[0082] Confidence data of each layer of content recorded in the content generation process corresponding to the content generation model is extracted.

[0083] In this embodiment, a comprehensive review of all the content generated by the above-mentioned content generation model, including the preliminary content (first content) and the core selling point (second content), can be conducted to ensure that every level and every expression in the content generation process is covered. Among them, the system will automatically identify and mark redundant modifiers or phrases with low confidence, such as "absolute safety" and "ultimate guarantee". These redundant contents are usually repeated, exaggerated or not strongly related to the core selling point. The marking process can be assisted by natural language processing technology (such as keyword extraction and semantic analysis). In addition, the marked redundant content can also be classified for subsequent processing. For example, repeated expressions are classified as "redundant repetitions" and exaggerated expressions are classified as "redundant exaggerations". In addition, during the content generation process of the above-mentioned content generation model, the confidence data of each layer of content will be automatically recorded. Specifically, the confidence data is calculated and recorded by the Sigma ID function during the content generation process.

[0084] Based on the confidence data, third content having a confidence lower than a preset confidence threshold is filtered out from the first content and the second content.

[0085] In this embodiment, there is no specific limitation on the setting of the above-mentioned confidence threshold, which can be determined according to actual business needs. For example, it can be set to 0.85. For example, if the confidence of a certain part of the content is 0.78, it will be marked as low-confidence content. After filtering out the third content with a confidence lower than the confidence threshold according to the above-mentioned confidence threshold, the third content can be further marked, that is, a low-confidence label is added to these contents. In addition, the results of the confidence threshold screening provide a high-quality content foundation for copywriting simplification, ensuring that only the most relevant and reliable information is processed in subsequent steps.

[0086] The third content is filtered from the first content and the second content to obtain corresponding fourth content.

[0087] In this embodiment, all redundant content marked as low confidence (third content) can be removed from the first content and the second content to ensure that only core information and key selling points with high confidence are retained, thereby obtaining the corresponding fourth content.

[0088] The fourth content is subjected to content simplification processing to obtain the corresponding target content.

[0089] In this embodiment, the above-mentioned specific implementation process of streamlining the fourth content to obtain the corresponding target content will be further described in detail in subsequent specific embodiments of this application and will not be elaborated on here.

[0090] This application extracts the confidence data of each layer of content recorded in the content generation process corresponding to the content generation model; then, based on the confidence data, filters out the third content whose confidence is lower than the preset confidence threshold from the first content and the second content; then filters the third content from the first content and the second content to obtain the corresponding fourth content; and then performs content simplification on the fourth content to obtain the corresponding target content. This application extracts the confidence data of each layer of content recorded in the content generation process corresponding to the content generation model, and based on the use of the confidence data, filters out the third content whose confidence is lower than the preset confidence threshold from the first content and the second content; then filters the third content from the first content and the second content to obtain the fourth content, and then performs content simplification on the fourth content, thereby completing the corresponding content filtering process and generating the required target content. This application optimizes and simplifies the content generated by the content generation model by combining the use of confidence threshold screening and copy simplification processing methods. Each step of processing is based on the results of the previous step and is strictly verified and adjusted, so as to effectively ensure that the final target copy can meet high-quality standards in terms of content, effectively improving the quality of the generated target content.

[0091] In some optional implementations, performing content streamlining on the fourth content to obtain the corresponding target content includes the following steps:

[0092] Content integration is performed on the fourth content to obtain corresponding first processed content.

[0093] In this embodiment, the above-mentioned content integration includes: integrating the filtered fourth content to ensure the coherence and completeness of the information. For example, the initial content "[Pension Insurance] Covers Post-Retirement Life Security" and the core selling point "Monthly Annuity, Lifelong Stable Income" are integrated into a concise statement.

[0094] The first processing content is optimized and expressed to obtain corresponding second processing content.

[0095] In this embodiment, the optimized expression processing includes optimizing the expression of the first processed content to make it more concise, direct, and easy to understand. For example, "[Pension Insurance] covers post-retirement living security and is absolutely safe and reliable" is optimized to "[Pension Insurance] covers post-retirement living security."

[0096] The second processing content is verified based on a preset verification strategy.

[0097] In this embodiment, the verification strategy includes verifying the streamlined second processing content to ensure that its information is complete and meets the expected simplicity. The verification of the second processing content can be performed using an automated tool, and a corresponding verification result can be generated. The verification result can include whether the second processing content has passed verification or whether the second processing content has failed verification.

[0098] If the second processed content passes the verification, the second processed content is used as the target content.

[0099] In this embodiment, only when it is detected that the second processed content passes the verification, the second processed content will be used as the required target content.

[0100] This application integrates the fourth content to obtain the corresponding first processed content; then optimizes the expression of the first processed content to obtain the corresponding second processed content; then verifies the second processed content based on a preset verification strategy; if the second processed content passes the verification, the second processed content is used as the target content. This application automatically and accurately completes the content streamlining of the fourth content by integrating the content, optimizing the expression, and verifying the fourth content, so that the generated target content retains the core information and key selling points with high confidence, while ensuring the simplicity and directness of the expression, effectively improving the quality of the generated target content.

[0101] In some optional implementations of this embodiment, step S207 includes the following steps:

[0102] Based on the template layout constraint information, corresponding layout instructions are added to the target content to obtain a corresponding first copy.

[0103] In this embodiment, based on the core content (i.e., target content) obtained after redundant content filtering, specific layout instructions can be added to the target content based on the template layout constraint information (such as key area location and visual complexity) provided during the template adaptation phase. For example, adding the "@right-align" instruction ensures that the copy is right-aligned with the price block; or adding the "@10px from chart" instruction maintains an appropriate distance from the data chart.

[0104] Get the preset multimodal alignment strategy.

[0105] In this embodiment, the multimodal alignment strategy includes verifying the alignment of the copy with the visual elements of the poster template (such as the price block, QR code, and data charts) through visual preview or simulated layout. Based on the alignment results, the layout constraints are adjusted accordingly to ensure that the copy is visually consistent with the overall template design and avoid occlusion or information overlap.

[0106] The first text is aligned based on the multimodal alignment strategy to obtain a corresponding second text.

[0107] In this embodiment, the first text may be aligned according to the policy content of the multimodal alignment policy, thereby generating a corresponding second text.

[0108] The second document is used as the target document.

[0109] This application obtains the corresponding first copy by adding corresponding typesetting instructions to the target content based on the template typesetting constraint information; then obtains a preset multimodal alignment strategy; and aligns the first copy based on the multimodal alignment strategy to obtain the corresponding second copy; and subsequently uses the second copy as the target copy. This application obtains the corresponding first copy by adding corresponding typesetting instructions to the target content based on the use of template typesetting constraint information; and then aligns the first copy based on the use of multimodal alignment strategy, thereby completing further adaptation processing of the target content efficiently and intelligently, effectively ensuring that the target copy ultimately generated can meet high-quality standards in form, and improving the quality of the generated target copy.

[0110] In some optional implementations of this embodiment, after step S207, the electronic device may further perform the following steps:

[0111] Call up preset graphic design tools.

[0112] In this embodiment, there is no specific limitation on the selection of the above-mentioned graphic design tools, which can be determined according to actual business needs. For example, automatic synthesis software can be used.

[0113] The target text and the poster template are synthesized based on the graphic design tool to generate a corresponding visual poster.

[0114] In this embodiment, based on the use of the above-mentioned graphic design tools, the target copy after typeset adaptation can be synthesized with the above-mentioned poster template to ensure that the position, size, color and other visual attributes of the target copy in the poster template meet the design requirements, thereby generating a corresponding visual poster.

[0115] The format of the visualization poster is converted to obtain a corresponding target poster.

[0116] In this embodiment, after generating a visual poster, a quality check can be performed on the visual poster to ensure that the content is accurate, the layout is beautiful, and it is highly consistent with the visual elements of the template. Specifically, an automated tool can be used to perform a quality check to ensure that the final output poster meets the expected standards. The final visual poster can be converted into an appropriate format (such as PNG, JPEG, PDF, etc.) according to actual business needs to generate the corresponding target poster.

[0117] Output processing is performed on the target poster.

[0118] In this embodiment, the generated target poster can be delivered to an agent or a relevant user to ensure that the target poster can be successfully applied to the required scenario, thereby completing the output processing of the target poster.

[0119] This application generates a visual poster by synthesizing the target copy and the poster template based on the graphic design tool; then converting the format of the visual poster to obtain the corresponding target poster; and subsequently outputting the target poster. This application generates a visual poster by synthesizing the target copy and the poster template based on the use of a graphic design tool, and then converting the format of the visual poster to obtain the target poster and output it, thereby automatically and accurately generating a visual target poster that meets the content requirements and has a high degree of visual fit, thereby improving the generation efficiency and accuracy of the target poster. In addition, the generation process of the target poster makes full use of the information in the template adaptation stage, effectively ensuring the perfect integration of the target copy and the poster template in terms of vision and content.

[0120] In some optional implementations, the user information obtained is obtained with the user's consent and complies with relevant laws and policies.

[0121] In addition, any software tools or components not provided by our company that appear in the embodiments of this application are merely examples and do not represent actual use.

[0122] It should be understood that the order of execution of the steps in the above embodiments does not necessarily mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0123] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned target document, the above-mentioned target document can also be stored in a node of a blockchain.

[0124] The blockchain referred to in this application refers to a new application model for computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a series of data blocks generated using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity of the information (to prevent counterfeiting) and generate the next block. Blockchain can include the blockchain underlying platform, the platform product service layer, and the application service layer.

[0125] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0126] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0127] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware via computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes in the above-described method embodiments. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0128] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0129] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a copywriting generation device based on artificial intelligence, which is similar to the embodiment of the present invention. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0130] like Figure 3 As shown, the artificial intelligence-based copywriting generation device 300 described in this embodiment includes: an acquisition module 301, a parsing module 302, a first processing module 303, a second processing module 304, a third processing module 305, a filtering module 306 and a first generation module 307.

[0131] in:

[0132] Acquisition module 301, used to acquire the theme keywords and poster templates input by the user;

[0133] The parsing module 302 is used to perform semantic parsing on the topic keywords based on a preset parsing model to obtain corresponding semantic information;

[0134] The first processing module 303 is used to perform template adaptation processing on the poster template to obtain corresponding template layout constraint information;

[0135] The second processing module 304 is configured to perform preliminary content generation processing on the semantic information based on a preset content generation model. When generating each layer of content, if the confidence level of the current layer reaches a preset exit threshold, the initial content generation process is terminated and the corresponding first content is obtained. The content generation model is a model that integrates a dynamic early exit mechanism.

[0136] A third processing module 305 is configured to perform inference processing on the semantic information based on the first content using the content generation model. When generating each layer of content, if the confidence level of a specified number of layers triggers a preset absolute threshold exit condition, the inference process is terminated and the corresponding second content is obtained.

[0137] A filtering module 306 is configured to perform content filtering on the first content and the second content to obtain corresponding target content;

[0138] The first generating module 307 is configured to perform template constraint adaptation processing on the target content based on the template typesetting constraint information to generate a corresponding target copy.

[0139] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based copywriting generation method in the aforementioned embodiment, and will not be repeated here.

[0140] In some optional implementations of this embodiment, the first processing module 303 includes:

[0141] A segmentation submodule, configured to segment the poster template into key areas based on a preset segmentation model to obtain corresponding area segmentation results;

[0142] A calculation submodule, configured to calculate the visual complexity of the poster template based on a preset calculation strategy;

[0143] An integration submodule, configured to integrate the region segmentation result with the visual complexity to generate a corresponding template visual feature description;

[0144] The generating submodule is used to generate template layout constraint information corresponding to the poster template based on the template visual characteristic description.

[0145] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based copywriting generation method in the aforementioned embodiment, and will not be repeated here.

[0146] In some optional implementations of this embodiment, the calculation submodule includes:

[0147] A preprocessing unit, configured to preprocess the poster template to obtain a corresponding target image;

[0148] An extraction unit, configured to extract edges of the target image based on a preset edge detection algorithm to obtain corresponding edge information;

[0149] a first calculating unit, configured to calculate a total edge length and a total image area of ​​the target image based on the edge information;

[0150] a second calculating unit, configured to calculate a ratio between the total edge length and the total area of ​​the image;

[0151] A generating unit is configured to generate a visual complexity corresponding to the poster template based on the ratio.

[0152] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based copywriting generation method in the aforementioned embodiment, and will not be repeated here.

[0153] In some optional implementations of this embodiment, the filtering module 306 includes:

[0154] An extraction submodule, configured to extract confidence data of each layer of content recorded in the content generation process corresponding to the content generation model;

[0155] a screening submodule, configured to screen out third content having a confidence level lower than a preset confidence threshold from the first content and the second content based on the confidence data;

[0156] a filtering submodule, configured to filter the third content from the first content and the second content to obtain corresponding fourth content;

[0157] The processing submodule is configured to perform content simplification processing on the fourth content to obtain the corresponding target content.

[0158] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based copywriting generation method in the aforementioned embodiment, and will not be repeated here.

[0159] In some optional implementations of this embodiment, the processing submodule includes:

[0160] an integration unit, configured to perform content integration on the fourth content to obtain corresponding first processed content;

[0161] an optimization unit, configured to perform optimized expression processing on the first processing content to obtain corresponding second processing content;

[0162] a verification unit, configured to verify the second processing content based on a preset verification strategy;

[0163] A determining unit is configured to use the second processed content as the target content if the second processed content passes the verification.

[0164] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based copywriting generation method in the aforementioned embodiment, and will not be repeated here.

[0165] In some optional implementations of this embodiment, the first generating module 307 includes:

[0166] An adding submodule, configured to add corresponding typesetting instructions to the target content based on the template typesetting constraint information to obtain a corresponding first copy;

[0167] Get submodule, used to obtain the preset multimodal alignment strategy;

[0168] an alignment submodule, configured to align the first text based on the multimodal alignment strategy to obtain a corresponding second text;

[0169] A submodule is determined, configured to use the second text as the target text.

[0170] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based copywriting generation method in the aforementioned embodiment, and will not be repeated here.

[0171] In some optional implementations of this embodiment, the artificial intelligence-based copywriting generation device further includes:

[0172] Calling module, used to call preset graphic design tools;

[0173] A second generating module is configured to synthesize the target text and the poster template based on the graphic design tool to generate a corresponding visual poster;

[0174] A conversion module, configured to convert the format of the visual poster to obtain a corresponding target poster;

[0175] An output module is used to perform output processing on the target poster.

[0176] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based copywriting generation method in the aforementioned embodiment, and will not be repeated here.

[0177] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0178] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 having components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital signal processors (DSPs), embedded devices, etc.

[0179] The computer device may be a desktop computer, notebook computer, PDA, cloud server, etc. The computer device may interact with the user via a keyboard, mouse, remote control, touchpad, or voice control device.

[0180] The memory 41 includes at least one type of readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (SMC), a secure digital (SD) card, a flash card, etc. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of the artificial intelligence-based copywriting generation method. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0181] In some embodiments, the processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions or process data stored in the memory 41, such as computer-readable instructions for executing the artificial intelligence-based copywriting generation method.

[0182] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0183] The present application also provides another embodiment, namely, providing a computer-readable storage medium, which stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the above-mentioned artificial intelligence-based copywriting generation method.

[0184] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0185] Obviously, the embodiments described above are only some of the embodiments of the present application, rather than all of the embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present application specification and the accompanying drawings, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present application.

Claims

1. A copywriting generation method based on artificial intelligence, characterized in that: The steps include: Get the theme keywords and poster templates entered by the user; Perform semantic analysis on the topic keywords based on a preset analysis model to obtain corresponding semantic information; Performing template adaptation processing on the poster template to obtain corresponding template typesetting constraint information; Performing preliminary content generation processing on the semantic information based on a preset content generation model, terminating the initial content generation process and obtaining the corresponding first content when each layer of content is generated and detecting that the confidence level of the current layer reaches a preset exit threshold; wherein the content generation model is a model that integrates a dynamic early exit mechanism; Based on the first content, the semantic information is inferred by the content generation model. When each layer of content is generated, if the confidence level of a specified number of layers triggers a preset absolute threshold exit condition, the inference process is terminated and the corresponding second content is obtained. performing content filtering on the first content and the second content to obtain corresponding target content; Based on the template typesetting constraint information, template constraint adaptation processing is performed on the target content to generate a corresponding target copy.

2. The method for generating text based on artificial intelligence according to claim 1, characterized in that: The step of performing template adaptation processing on the poster template to obtain corresponding template typesetting constraint information specifically includes: Segmenting the poster template into key areas based on a preset segmentation model to obtain corresponding area segmentation results; Calculating the visual complexity of the poster template based on a preset calculation strategy; Integrating the region segmentation result and the visual complexity to generate a corresponding template visual feature description; Based on the template visual characteristic description, template layout constraint information corresponding to the poster template is generated.

3. The method for generating text based on artificial intelligence according to claim 2, characterized in that: The step of calculating the visual complexity of the poster template based on a preset calculation strategy specifically includes: Preprocessing the poster template to obtain a corresponding target image; Extracting edges of the target image based on a preset edge detection algorithm to obtain corresponding edge information; Calculating the total edge length and total image area of ​​the target image based on the edge information; Calculating a ratio between the total edge length and the total area of ​​the image; A visual complexity corresponding to the poster template is generated based on the ratio.

4. The method for generating text based on artificial intelligence according to claim 1, characterized in that: The step of performing content filtering on the first content and the second content to obtain corresponding target content specifically includes: Extracting confidence data of each layer of content recorded in the content generation process corresponding to the content generation model; Based on the confidence data, screening out third content having a confidence lower than a preset confidence threshold from the first content and the second content; filtering the third content from the first content and the second content to obtain corresponding fourth content; The fourth content is subjected to content simplification processing to obtain the corresponding target content.

5. The method for generating text based on artificial intelligence according to claim 4, characterized in that: The step of performing content simplification processing on the fourth content to obtain the corresponding target content specifically includes: performing content integration on the fourth content to obtain corresponding first processed content; Performing optimized expression processing on the first processing content to obtain corresponding second processing content; Verifying the second processing content based on a preset verification strategy; If the second processed content passes the verification, the second processed content is used as the target content.

6. The method for generating text based on artificial intelligence according to claim 1, characterized in that: The step of performing template constraint adaptation processing on the target content based on the template typesetting constraint information to generate a corresponding target copy specifically includes: Adding corresponding typesetting instructions to the target content based on the template typesetting constraint information to obtain a corresponding first copy; Get the preset multimodal alignment strategy; Performing alignment processing on the first copy based on the multimodal alignment strategy to obtain a corresponding second copy; The second document is used as the target document.

7. The method for generating text based on artificial intelligence according to claim 1, characterized in that: After the step of performing template constraint adaptation processing on the target content based on the template typesetting constraint information to generate a corresponding target copy, the method further includes: Call preset graphic design tools; Combining the target text and the poster template based on the graphic design tool to generate a corresponding visual poster; Converting the format of the visualization poster to obtain a corresponding target poster; Output processing is performed on the target poster.

8. A copywriting generation device based on artificial intelligence, characterized in that: include: The acquisition module is used to obtain the theme keywords and poster templates input by the user; A parsing module, configured to perform semantic parsing on the subject keywords based on a preset parsing model to obtain corresponding semantic information; A first processing module is used to perform template adaptation processing on the poster template to obtain corresponding template typesetting constraint information; a second processing module configured to perform preliminary content generation processing on the semantic information based on a preset content generation model, terminating the initial content generation process and obtaining corresponding first content when each layer of content is generated and detecting that the confidence level of the current layer reaches a preset exit threshold; wherein the content generation model is a model that integrates a dynamic early exit mechanism; a third processing module configured to perform inference processing on the semantic information based on the first content using the content generation model, and terminate the inference process when the confidence levels of a specified number of layers trigger a preset absolute threshold exit condition when each layer of content is generated, thereby obtaining corresponding second content; a filtering module, configured to perform content filtering on the first content and the second content to obtain corresponding target content; The first generating module is used to perform template constraint adaptation processing on the target content based on the template typesetting constraint information to generate a corresponding target copy.

9. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the artificial intelligence-based copywriting generation method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the artificial intelligence-based copywriting generation method according to any one of claims 1 to 7.

Images