Artificial intelligence-based multi-platform content adaptation and output method, device, equipment and medium

Abstract

The application relates to the technical field of artificial intelligence, and discloses a multi-platform content adaptation and output method, device, equipment and medium based on artificial intelligence, the method comprising the following steps: obtaining input data of a user, and determining a task type based on the input data; determining a processing path according to the mode of the input data and the task type; determining an output platform and a user portrait of the user; determining a target template in a pre-constructed template library in combination with the input data, the task type, the processing path, the output platform and the user portrait; in the case that the target template is adopted, determining filling data in combination with the user portrait, the input data and popular content of the output platform within a target period; completing data filling of the target template based on the filling data, generating output data; and pushing the output data to the output platform. The application can be applied to the fields of financial technology and medical health, and can improve data propagation efficiency and effect.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, device and medium for multi-platform content adaptation and output based on artificial intelligence. Background Technology

[0002] With the diversification of social media and digital content platforms, users' demands for content format and platform adaptability are increasing. In existing technologies, content creators must manually adjust text, images, videos, and other materials to adapt to the format requirements of different platforms, such as WeChat, Douyin, and Xiaohongshu, leading to inefficiency and a high risk of errors. Furthermore, traditional methods lack the ability to intelligently analyze platform user preferences and content rules, making it difficult to achieve automated and personalized multi-platform content distribution. For example, in the fintech field, it is impossible to automatically generate personalized financial recommendation videos or articles that meet user preferences and platform requirements, or videos or articles on financial education and investor education that are easily understood and accepted by users. Similarly, in healthcare scenarios, it is impossible to automatically generate health education videos or articles, or videos or articles on health care or rehabilitation training, that meet user preferences and platform requirements. Summary of the Invention

[0003] This invention provides a method, apparatus, device, and medium for multi-platform content adaptation and output based on artificial intelligence, in order to solve the technical problems that content creators need to manually adjust text, images, videos, and other materials to adapt to different platforms, resulting in low efficiency and error-proneness, as well as the inability to intelligently analyze platform user preferences and content rules, making it difficult to achieve automated and personalized multi-platform content distribution.

[0004] Firstly, it provides a multi-platform content adaptation and output method based on artificial intelligence, including: Obtain user input data and determine the task type of the corresponding processing task based on the input data; The processing path is determined based on the modality and task type of the input data; Determine the output platform and the user profile of the user; The target template is determined from a pre-built template library by combining one or more of the input data, task type, processing path, output platform, and user profile. The template library contains multiple templates for filling data to form output data. If the target template is adopted, the fill data is determined by combining the user profile, input data, and popular content of the output platform during the target time period; Based on the filling data, the target template is filled with data to generate output data; The output data is pushed to the output platform.

[0005] Secondly, an artificial intelligence-based multi-platform content adaptation and output device is provided, including: The acquisition module is used to acquire user input data and determine the task type of the corresponding processing task based on the input data; The first determining module determines the processing path based on the modality and task type of the input data; The second determining module is used to determine the output platform and the user profile of the user. The third determining module is used to determine the target template in a pre-built template library by combining one or more of the input data, task type, processing path, output platform and user profile. The template library has multiple templates for filling data to form output data. The fourth determining module is used to determine the filling data by combining the user profile, input data, and popular content of the output platform during the target time period when the target template is adopted. The generation module is used to complete the data filling of the target template based on the filling data and generate output data; The output module is used to push the output data to the output platform.

[0006] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described multi-platform content adaptation and output method based on artificial intelligence.

[0007] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described multi-platform content adaptation and output method based on artificial intelligence.

[0008] The above-described solution, implemented by the AI-based multi-platform content adaptation and output method, apparatus, computer equipment, and storage medium, involves obtaining user input data and determining the task type based on the input data; determining the processing path based on the modality of the input data; determining the output platform and the user profile; determining the target template from a pre-built template library by combining the input data, task type, processing path, output platform, and user profile; if the target template is adopted, determining the fill data by combining the user profile, input data, and popular content of the output platform within the target time period; filling the target template with the fill data to generate output data; and pushing the output data to the output platform. This method allows the system to perform a series of processes on user input data using AI, including determining the task type and processing path, and determining the target template and fill content by combining the determined user profile, output platform, and current popular content of that platform. Finally, it generates output data that matches the output platform and meets the user's needs, all without user intervention. This significantly improves data processing efficiency and accuracy, and is applicable to various platforms with a wide range of adaptability. Attached Figure Description

[0009] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of an application environment for a multi-platform content adaptation and output method based on artificial intelligence, according to an embodiment of the present invention. Figure 2 This is a flowchart illustrating a multi-platform content adaptation and output method based on artificial intelligence in one embodiment of the present invention; Figure 3 yes Figure 1 A schematic diagram of a specific implementation method for step S10; Figure 4 This is a schematic diagram of the structure of an intelligent question-and-answer processing device according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 6 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0011] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] like Figure 1 and Figure 2 As shown, this embodiment of the invention provides a multi-platform content adaptation and output method based on artificial intelligence, including: S1: Obtain user input data and determine the task type of the corresponding processing task based on the input data; S2: Determine the processing path based on the modality and task type of the input data; S3: Determine the output platform and the user profile of the user; S4: Combine one or more of the input data, task type, processing path, output platform, and user profile to determine the target template in the pre-built template library, which has multiple templates for filling data to form output data; S5: If the target template is adopted, determine the fill data by combining the user profile, input data, and popular content of the output platform during the target time period; S6: Based on the filling data, complete the data filling of the target template and generate output data; S7: Push the output data to the output platform.

[0013] The solution described in this embodiment has a wide range of applications. For example, it can be applied to self-media operations to quickly generate content adapted to multiple platforms and improve operational efficiency; it can also be applied to corporate marketing to unify brand information and achieve precise cross-platform targeting; it can also be applied to content factories to mass-produce high-quality content adapted to different platforms, etc. In short, the solution in this embodiment can be applied in the media, finance, and medical fields, such as generating different types of output data for financial products, medical products, and medical knowledge, and distributing them to different platforms for promotion and explanation. It can also be applied to artificial intelligence platforms to generate corresponding financial product descriptions, medical product descriptions, or investor education content based on user needs, etc., and the specific applications are not limited to these.

[0014] In this embodiment, during execution, the system first obtains user input data. This data can be unimodal or multimodal, such as a mixture of text, images, and videos. The input data contains specific user instructions or information about how to process the data. Based on the input data, the system determines the task type, i.e., how the user wants the input data processed—whether it's generating videos, posters, or advertising text. Different processing tasks correspond to different task types. Simultaneously, the system determines the processing path based on the modality and task type of the input data. For example, if the input is plain text, the system enters the text-video script or text-multi-platform copywriting generation path; if the input is an image, it triggers the image-dynamic poster generation path. Next, the system determines the output platform. This output platform can be the default configuration, a platform specified by the user, a platform determined by the system based on platforms where the user has recently posted information, or a platform determined based on the current input data, task type, and the posting density of similar data on various platforms. The specific method is not unique. Next, user profiles are determined. For example, features can be extracted from user behavior logs, such as clicks and dwell time, as well as third-party data, such as user tag APIs provided by a platform, to build user profiles.

[0015] Specifically, in this embodiment, the system constructs a dynamic user profile for each creator (user). The data sources for constructing the user profile include the distribution of historical content types, completion rates, and interaction data; the creator's secondary editing behavior on the system-generated results, such as the extent of modification and the retention rate; and the style preferences actively set by the creator, such as the formality of tone and visual style tendencies. Based on the above data, the system analyzes and obtains the user's profile. The user profile is not fixed but is dynamically updated based on a time decay mechanism. Frequently occurring user interactions in the recent period are assigned high weight, which directly affects the generation of the current user profile.

[0016] After determining the user profile, the system combines the input data, task type, processing path, output platform, and user profile to determine the target template from a pre-built template library. In this embodiment, the template library contains multiple templates for filling data to form output data, i.e., the final document, video, poster, etc., published on the platform. These multiple templates involve multiple different template types, such as text, poster, and video, and can be more specific, such as emotional text, learning text, advertising posters, entertainment videos, learning videos, and science videos, etc., and are not unique. After determining the target template, if the user adopts the target template, the system combines the user profile, input data, and popular content on the output platform during the target time period to determine the filling data. Then, based on the filling data, the target template is filled to generate output data, and finally, the system automatically pushes the output data to the determined output platform.

[0017] The solution based on this embodiment can completely replace the user in the file publishing process and is applicable to various platforms. The system can automatically recognize the rules of different platforms, such as word count and formatting, enabling single-time synchronous creation and multi-platform output, saving repetitive work. Moreover, based on AI, it can adjust the language style according to user preferences, perform multilingual translation, and data-driven optimization to improve content matching and dissemination effectiveness. For example, the style of the output data can be adjusted comprehensively based on the type of input content and user preferences, making it humorous or tense to improve the viewing and dissemination effects. It also supports modal data processing, adapting to multiple types of content such as text, images, and videos to meet diverse dissemination needs. In summary, the solution of this embodiment can automate the production of output data, reduce labor costs, reduce error rates with AI verification, and significantly improve the utilization rate of materials through modular reuse.

[0018] In this embodiment, the system pre-builds a rule base to retrieve the rules for a given output platform, ensuring that the final processed output data meets the rules of that platform. This rule base can be built, for example, by crawling publicly available developer documentation from various platforms, such as WeChat's article guidelines and Douyin's video duration limits, or by obtaining real-time rules through API interfaces, such as Xiaohongshu's sensitive word database, thus collecting rule data from each platform. Then, the system uses a rule engine and a custom logic parser to perform architectural processing on the collected rules, such as converting them to JSON / YAML format before storage. To ensure the validity of the rules, the system sets up a dynamic update mechanism, monitoring for rule changes on each platform via scheduled tasks. If changes occur, the rule base is updated.

[0019] In one embodiment, obtaining user input data and determining the task type of the corresponding processing task based on the input data includes: S101: Determine the user's intent based on the input data; S102: The user intent is mapped to a task space through prompt word engineering, the task space including marketing copy generation, video script creation, graphic content rewriting, and dynamic poster production; S103: Determine the task type based on the mapping result.

[0020] For example, after receiving user input, the system autonomously determines "what type of content to generate" and "in what form" through the intent recognition and modality decision-making module. When user input (text, image, or mixed input) enters the system, the intent recognition model first performs semantic parsing and task classification on the input. This model is based on a large language model, such as, but not limited to, GPT-4. Next, through prompt word engineering, the user's input data is mapped to a predefined task space, which can be multiple, including but not limited to: marketing copy generation, video script creation, text and image content rewriting, and dynamic poster production. Finally, based on the actual mapping results, the task type for this processing task can be determined, such as generating marketing copy, preparing video scripts, modifying text and image content, or creating dynamic posters.

[0021] After determining the task type, the system further invokes the modal decision unit, which can autonomously select and generate links based on the following dimensions: Input modality and task type: If the input is plain text, the system will enter the text-video text-multi-platform copywriting generation path; if the input is an image, it will trigger the image-dynamic poster generation path. Alternatively, if the input is a combination of image and text, it will trigger the video script generation path.

[0022] In addition, the system can further improve the output data generation path based on the following two dimensions: Output Platform: The system maintains a platform feature knowledge base, recording content preferences for long articles on various platforms, such as Douyin, Xiaohongshu, and WeChat, including duration, copywriting style, and visual rhythm. The modal decision unit automatically matches the corresponding generation template and toolchain based on the target platform for subsequent applications.

[0023] Content complexity: If the input content contains structured information, such as product parameters and data indicators, the system will automatically call the structured extraction module to ensure the integrity of the information before stylizing and rewriting it.

[0024] The above-mentioned judgment process is completed autonomously through a hybrid reasoning of system rules and models, and the judgment results can be passed to downstream components, such as dynamic content generation and format conversion engines, in the form of "generated task descriptors".

[0025] After determining the generation path, the system will further determine the user profile and output platform. Then, based on the input data, task type, processing path, output platform, and user profile, the system will determine the target template from a pre-built template library, including: S401: Based on the input data, task type and processing path, perform semantic similarity matching with the feature labels of each template in the template library to determine the first candidate template; S402: Based on the user profile and the user's historical output data, a second candidate template is obtained by filtering from the first candidate template; S403: Based on the platform rules of the output platform and the structural characteristics of popular content in the target time period, the target template is obtained by filtering from the third candidate template.

[0026] For example, before executing this operation, the system pre-builds a template library. This library includes multiple preset templates, specifically tailored to different content types, such as product recommendations, narrative performances, and knowledge dissemination, and for different platforms, providing various structured templates. For instance, the Douyin "product promotion video" template predefines a narrative structure of "golden 3-second opening - pain point demonstration - product introduction - price anchor - order guidance"; the Xiaohongshu "image and text notes" template predefines a layout format of "clickbait title - contextualized images - segmented key points - hashtags". When executing the above process, the system first performs semantic similarity matching with the feature tags of each template in the template library based on the input data, task type, and processing path to determine the first candidate template. For example, it calculates the semantic similarity between the user's input data and the applicable scenario tags of the template, prioritizing templates with consistent content types. Then, it filters the first candidate templates based on the user profile and the user's historical output data to obtain the second candidate template. For example, it calls the creator profile module to analyze the template types that the creator has used most frequently and with the best interaction results in the past, and gives higher weight to the corresponding templates. Then, based on the platform rules of the output platform and the structural characteristics of popular content within the target time period, the target template is obtained by filtering from the third candidate templates. For example, the structural characteristics of popular content on the current target platform can be obtained through the platform trend awareness thread, and templates that match the characteristics of recent hit content can be prioritized.

[0027] Furthermore, determining the target template also includes: S404: Using a large language model, perform small-sample learning on the template architecture of the output samples that meet the threshold in the target time period and match the input data, task type, and user profile of the output platform; S405: Generate the target template based on the learning results.

[0028] In other words, if the method based on the aforementioned embodiments fails to directly retrieve a suitable template as the target template from the template library, the system can call a large language model to perform few-sample learning. Specifically, the sample can be extracted in real time from the popular content on the platform, and then the content structure can be abstracted into a new template by learning the content structure. Based on this, the target template is determined, and then the template is dynamically stored in the template library, so that the template system has the ability to self-evolve.

[0029] In another embodiment, the fill data is determined by combining the user profile, input data, and popular content from the output platform during the target time period, including: S501: Determine the user's style preference for the output data based on the user profile; S502: Determine the multi-dimensional features of popular content on the output platform within the target time period, wherein the multi-dimensional features include title features, text features, video features, and tag features; S503: Perform feature statistics on the multi-dimensional features and determine the injection weight of each dimension feature based on the statistical results; S504: Combine the style preference, multi-dimensional features and corresponding injection weights to process the input data and generate the filling data.

[0030] For example, collaborative filtering methods can be combined to recommend suitable data style formats based on the historical preferences of similar users, or based on the user's historical preferences. Alternatively, a decision tree model can be used to predict the user's style preferences based on user profiles, thus determining the suitable data style format. In this embodiment, the system will run a platform trend awareness thread to collect features of popular content on the platform in real-time or near real-time, including: title sentence structure (suspenseful, numerical, question-based), text length range, the rhythm features of the first 3 seconds of the video, and popular topic tags. Then, the platform's popular features are extracted into injectable "hot topic weight vectors" through a feature statistical model. Finally, the system combines the user's style preferences, multi-dimensional features, and corresponding weight vectors to comprehensively process the input data and generate filler data that matches various needs.

[0031] Specifically, the filling data includes text data and / or visual data, such as images or image streams; The process of determining the fill data includes generating text data and / or visual data, and determining the fill data based on the text data and / or visual data. Generating the text data includes: S506: Determine the current hot phrases of the output platform based on the multi-dimensional features and the corresponding injection weights; S507: Using the aforementioned style preferences and popular sentence patterns as guidance, generate the text data based on the input data; Generating the visual data includes: S508: Determine the current popular visual features of the output platform based on the multi-dimensional features and the corresponding injection weights; S509: Using the style preferences and popular visual features as guidance, generate the visual data based on the input data; The method further includes: S8: When the output data includes both text data and visual data, perform modal alignment and semantic alignment on the text data and visual data.

[0032] For example, the system invokes a dynamic content generation and format conversion engine to generate content in stages based on the type of template placeholders. In the text generation stage, it can utilize tools like GPT-4 to generate scripts or copy. The system dynamically combines "style constraints" from the user profile with "hot topic phrases" from popular platform content in the prompts, achieving dual control and generating copy content that conforms to the template framework. In the visual generation stage, it can use tools like Stable Diffusion to generate images and Pika to add animations. The system determines the generation parameters based on the user's historical visual style preferences, such as color scheme and composition, and popular platform visual features, such as dynamic text appearance, generating visual materials that match the template framework. Then, in the modal alignment stage, it can use CLIP to verify video footage and voiceover text. The system prioritizes modal consistency, ensuring that the generated text and visual content are semantically matched, avoiding semantic deviations caused by style intervention.

[0033] When generating cross-modal content from data, such as text-to-video scripts, large models like GPT-4 can be used to generate video scripts. These scripts can be combined with script structures (e.g., "beginning-conflict-climax-ending") and platform styles (e.g., the fast pace of TikTok, the tutorial style of Xiaohongshu) to generate video scripts. After script generation, video generation tools, such as Runway ML, can be used to convert the script into a storyboard, including scene descriptions and voice-over text. Another example is image-to-animated posters. Stable Diffusion can be used to generate static images, which can then be animated using tools like Pika to add dynamic effects, including but not limited to floating text and background gradients. Furthermore, OpenCV can be used to detect key points in images, such as faces and objects, to generate dynamic effects like blinking and rotation.

[0034] Afterwards, the system can perform format conversions according to actual needs. For example, text compression / expansion; assuming a scenario of long articles on WeChat or short articles on Xiaohongshu, the BART (Briefing and Abstraction) model can be used to extract core information, and the tone can be adjusted using a StyleGAN (Style Transfer AN) model, such as from formal to colloquial. Next, multimodal alignment is performed, using a CLIP (Content-Based Interaction) model to ensure modal consistency of the generated content, such as semantic matching between video footage and voice-over text. In practical applications, the system can utilize multiple pre-built automated toolchains, such as a microservice architecture (Spring Cloud), to call different generation tools via APIs, such as text generation, image generation, and video compositing, achieving an end-to-end process and further improving data processing and generation efficiency.

[0035] In one embodiment, the tags and titles in the populated data can be prepared using the following methods. For tags, reinforcement learning, such as the PPO algorithm, can be used to simulate the platform's recommendation mechanism, training the model to generate high-weight tags, such as the "Challenge" tag on Douyin, and then dynamically adjusting the tag combination by combining platform APIs, such as the keyword recommendation interface of Xiaohongshu. For titles, a sequence-to-sequence model can be trained, taking the original title and a platform style description, such as "Douyin: Suspenseful Opening," as input, to obtain a suitable title.

[0036] After the preparation of the filler data is completed, such as Figure 3 As shown, the system will complete the data filling of the target template based on the filling data and generate output data, including: S601: Identify fillable placeholders in the target template; S602: Map the fillable placeholders to the key data in the input data; S603: Based on the mapping relationship between the fillable placeholders and the key data, the fillable data is filled into the target template accordingly to generate the output data.

[0037] For example, after the target template is selected, the system parses the template structure and identifies fillable placeholders, such as "product name," "core selling points," "opening remarks," and "introduction." Then, it performs an initial mapping between these placeholders and key information in the user's input data. Once both the target template and the fillable data are determined, the system, based on the initial mapping and the correspondence between the fillable data and the input data, fills in the specific content at each placeholder, ultimately generating the output data.

[0038] The system-generated output data will also be reviewed by the user. During this stage, the user can modify and adjust the output content. The system will record the user's modification behavior and update the user profile accordingly.

[0039] Once the final output data is determined, the system will push it to the designated output platform for dissemination. To achieve a real-time feedback loop, the system can use event tracking technology and platform APIs, such as Douyin's Creator Center API, to obtain metrics like click-through rate and completion rate in real time. Then, an online learning framework is used to update the generation strategy in real time, such as "shortening video length when the completion rate is low." Furthermore, the system can compare the effects of different strategies through traffic diversion to select the optimal solution. Additionally, user feedback data, such as click-through rate, can be used as new samples to incrementally train and update the generation model, forming a closed-loop feedback system.

[0040] In one application example, the user inputs raw content through the system, such as a long article and the target platform, such as WeChat Official Accounts, Douyin, or Xiaohongshu. The platform rule parsing module automatically calls the API interfaces of each platform to obtain content specifications and production strategies, such as the maximum number of characters in WeChat articles (20,000 characters) and the maximum video length on Douyin (3 minutes). Then, the AI ​​model generates core content, i.e., fill-in content, based on the user's input and breaks it down into multiple versions, for example: - WeChat version: Generate text and image layouts, insert relevant images and paragraph titles; - TikTok version: Generates a short video script and calls a video generation tool to synthesize a 15-second vertical video; - Xiaohongshu version: Generate image and text notes with keywords and add popular platform tags.

[0041] The system then automatically pushes the adapted content to various platforms and collects user feedback through data interfaces, such as the completion rate on Douyin and the number of likes on Xiaohongshu. Based on the feedback data, the generation strategy is adjusted, such as increasing the short video length to 30 seconds and optimizing keyword density.

[0042] like Figure 4 As shown, another embodiment of the present invention also provides a multi-platform content adaptation and output device based on artificial intelligence, including: The acquisition module is used to acquire user input data and determine the task type of the corresponding processing task based on the input data; The first determining module is used to determine the processing path based on the modality and task type of the input data; The second determining module is used to determine the output platform and the user profile of the user. The third determining module is used to determine the target template in a pre-built template library by combining the input data, task type, processing path, output platform and user profile. The template library has multiple templates for filling data to form output data. The fourth determining module is used to determine the filling data by combining the user profile, input data, and popular content of the output platform during the target time period when the target template is adopted. The generation module is used to complete the data filling of the target template based on the filling data and generate output data; The output module is used to push the output data to the output platform.

[0043] In one embodiment, determining the target template from a pre-built template library by combining the input data, task type, processing path, output platform, and user profile includes: Based on the input data, task type, and processing path, semantic similarity matching is performed with the feature labels of each template in the template library to determine the first candidate template; A second candidate template is obtained by filtering the first candidate template based on the user profile and the user's historical output data. The target template is obtained by filtering from the third candidate templates based on the platform rules of the output platform and the structural characteristics of popular content in the target time period.

[0044] In one embodiment, the device further includes: The learning module is used to perform small-sample learning on the template architecture of the output samples of the output platform that meet the threshold in the target time period and match the input data, task type, and user profile using the large language model. The template generation module is used to generate the target template based on the learning results.

[0045] In one embodiment, the data to be filled is determined by combining the user profile, input data, and popular content from the output platform during the target time period, including: Based on the user profile, determine the user's style preference for the output data; Determine the multi-dimensional features of popular content on the output platform within the target time period. The multi-dimensional features include title features, text features, video features, and tag features. Feature statistics are performed on the multi-dimensional features, and the injection weights of each dimension feature are determined based on the statistical results; The input data is processed by combining the style preference, multi-dimensional features and corresponding injection weights to generate the filling data.

[0046] In one embodiment, the fill data includes text data and / or visual data; Determining the filling data includes: The current hot phrases of the output platform are determined based on the multi-dimensional features and the corresponding injection weights. Guided by the aforementioned style preferences and popular sentence patterns, the text data is generated based on the input data; Determining the filling data includes: Based on the multi-dimensional features and corresponding injection weights, the current popular visual features of the output platform are determined. Guided by the aforementioned style preferences and popular visual features, the visual data is generated based on the input data; The device further includes: The alignment module is used to perform modal alignment and semantic alignment on the text data and visual data when the output data includes both text data and visual data.

[0047] In one embodiment, the step of filling the target template with the filled data and generating output data includes: Identify fillable placeholders in the target template; Map the fillable placeholders to the key data in the input data; Based on the mapping relationship between fillable placeholders and key data, the fillable data is filled into the target template to generate the output data.

[0048] In one embodiment, obtaining user input data and determining the task type of the corresponding processing task based on the input data includes: Determine the user's intent based on the input data; The user intent is mapped to a task space through prompt word engineering. The task space includes marketing copy generation, video script creation, graphic content rewriting, and dynamic poster production. The task type is determined based on the mapping results.

[0049] The AI-based multi-platform content adaptation and output device of this invention can obtain user input data and determine the task type based on the input data; determine the processing path according to the modality of the input data; determine the output platform and the user profile; determine the target template in a pre-built template library by combining the input data, task type, processing path, output platform, and user profile; if the target template is adopted, determine the fill data by combining the user profile, input data, and popular content of the output platform in the target time period; complete the data filling of the target template based on the fill data to generate output data; and push the output data to the output platform. Through this device, the system can combine AI to perform a series of processes on user input data, ultimately generating output data that matches the output platform and meets the user's needs. The entire process requires no user intervention, significantly improving data processing efficiency and accuracy, and is applicable to various platforms with a wide range of adaptability.

[0050] For specific limitations regarding AI-based multi-platform content adaptation and output devices, please refer to the limitations of the intelligent question-answering method described above, which will not be repeated here. Each module in the aforementioned intelligent question-answering processing device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0051] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the computer program is executed by the processor, it implements the functions or steps of a server-side intelligent question-answering method based on artificial intelligence.

[0052] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 6 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When the computer program is executed by the processor, it implements client-side functions or steps of an artificial intelligence-based intelligent question-answering processing method.

[0053] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain user input data and determine the task type of the corresponding processing task based on the input data; The processing path is determined based on the modality and task type of the input data; Determine the output platform and the user profile of the user; The target template is determined from a pre-built template library by combining one or more of the input data, task type, processing path, output platform, and user profile. The template library contains multiple templates for filling data to form output data. If the target template is adopted, the fill data is determined by combining the user profile, input data, and popular content of the output platform during the target time period; Based on the filling data, the target template is filled with data to generate output data; The output data is pushed to the output platform.

[0054] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor: Obtain user input data and determine the task type based on the input data, wherein the task type represents the type of task for processing the input data; The processing path is determined based on the modality and task type of the input data; Determine the output platform and the user profile of the user; The target template is determined from a pre-built template library by combining the input data, task type, processing path, output platform, and user profile. The template library contains multiple templates for filling data to form output data, and the multiple templates involve multiple different template types. If the target template is adopted, the fill data is determined by combining the user profile, input data, and popular content of the output platform during the target time period; Based on the filling data, the target template is filled with data to generate output data; The output data is pushed to the output platform.

[0055] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0056] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0057] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0058] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

[0059] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for multi-platform content adaptation and output based on artificial intelligence, characterized in that, include: Obtain user input data and determine the task type of the corresponding processing task based on the input data; The processing path is determined based on the modality and task type of the input data; Determine the output platform and the user profile of the user; The target template is determined from a pre-built template library by combining one or more of the input data, task type, processing path, output platform, and user profile. The template library contains multiple templates for filling data to form output data. If the target template is adopted, the fill data is determined by combining the user profile, input data, and popular content of the output platform during the target time period; Based on the filling data, the target template is filled with data to generate output data; The output data is pushed to the output platform.

2. The multi-platform content adaptation and output method based on artificial intelligence according to claim 1, characterized in that, Based on the input data, task type, processing path, output platform, and user profile, the target template is determined from the pre-built template library, including: Based on the input data, task type, and processing path, semantic similarity matching is performed with the feature labels of each template in the template library to determine the first candidate template; A second candidate template is obtained by filtering the first candidate template based on the user profile and the user's historical output data. The target template is obtained by filtering from the third candidate templates based on the platform rules of the output platform and the structural characteristics of popular content in the target time period.

3. The multi-platform content adaptation and output method based on artificial intelligence according to claim 1, characterized in that, Determining the target template further includes: The template architecture of the output samples that meet the threshold in the target time period and match the input data, task type, and user profile of the output platform is learned by using a large language model with few samples. The target template is generated based on the learning results.

4. The multi-platform content adaptation and output method based on artificial intelligence according to claim 1, characterized in that, Based on the user profile, input data, and popular content from the output platform during the target time period, the population data is determined, including: Based on the user profile, determine the user's style preference for the output data; Determine the multi-dimensional features of popular content on the output platform within the target time period. The multi-dimensional features include title features, text features, video features, and tag features. Feature statistics are performed on the multi-dimensional features, and the injection weights of each dimension feature are determined based on the statistical results; The input data is processed by combining the style preference, multi-dimensional features and corresponding injection weights to generate the filling data.

5. The multi-platform content adaptation and output method based on artificial intelligence according to claim 4, characterized in that, The filling data includes text data and / or visual data; Determining the filling data includes: The current hot phrases of the output platform are determined based on the multi-dimensional features and the corresponding injection weights. Guided by the aforementioned style preferences and popular sentence patterns, the text data is generated based on the input data; Determining the filling data includes: Based on the multi-dimensional features and corresponding injection weights, the current popular visual features of the output platform are determined. Guided by the aforementioned style preferences and popular visual features, the visual data is generated based on the input data; The method further includes: When the output data includes both text data and visual data, modal alignment and semantic alignment are performed on the text data and visual data.

6. The multi-platform content adaptation and output method based on artificial intelligence according to claim 1, characterized in that, The process of filling the target template with the filled data and generating output data includes: Identify fillable placeholders in the target template; Map the fillable placeholders to the key data in the input data; Based on the mapping relationship between fillable placeholders and key data, the fillable data is filled into the target template to generate the output data.

7. The multi-platform content adaptation and output method based on artificial intelligence according to claim 1, characterized in that, The step of obtaining user input data and determining the task type of the corresponding processing task based on the input data includes: Determine the user's intent based on the input data; The user intent is mapped to a task space through prompt word engineering. The task space includes marketing copy generation, video script creation, graphic content rewriting, and dynamic poster production. The task type is determined based on the mapping results.

8. A multi-platform content adaptation and output device based on artificial intelligence, characterized in that, include: The acquisition module is used to acquire user input data and determine the task type of the corresponding processing task based on the input data, wherein the task type represents the task type of the processing task for the input data; The first determining module is used to determine the processing path based on the modality and task type of the input data; The second determining module is used to determine the output platform and the user profile of the user. The third determining module is used to determine the target template in a pre-built template library by combining the input data, task type, processing path, output platform and user profile. The template library has multiple templates for filling data to form output data, and the multiple templates involve multiple different template types. The fourth determining module is used to determine the filling data by combining the user profile, input data, and one or more popular contents of the output platform during the target time period when the target template is adopted. The generation module is used to complete the data filling of the target template based on the filling data and generate output data; The output module is used to push the output data to the output platform.

9. 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, it implements the steps of the multi-platform content adaptation and output method based on artificial intelligence as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-platform content adaptation and output method based on artificial intelligence as described in any one of claims 1 to 7.

Images