Artificial intelligence-based data processing method and device, computer device and medium

Abstract

The application belongs to the technical field of artificial intelligence, and relates to a data processing method and device based on artificial intelligence, a computer device and a storage medium, which comprises the following steps: performing feature extraction and scene analysis processing on related data of a user; performing semantic analysis on obtained user portrait data and scene label data based on a large language model to obtain a semantic analysis result, and screening a matched target template from a template library; obtaining target node information related to the user portrait data and the scene label data from a knowledge graph; processing the target node information and input data of the user based on a content generation model to generate text content; filling the text content into the target template to obtain target text data; generating a target image data based on the target text data; and publishing the target text data and the target image data to a knowledge base. The application can be applied to a knowledge management scene in the field of financial technology, and effectively improves the generation efficiency of knowledge data.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology and can be applied to the financial technology field, particularly to data processing methods, devices, computer equipment and storage media based on artificial intelligence. Background Technology

[0002] In traditional automotive finance knowledge management scenarios, the generation and configuration of knowledge data primarily rely on manual page creation and field filling, lacking effective automated assistance, resulting in low efficiency in knowledge configuration. Specifically, companies need to manually configure unstructured knowledge such as product policies, risk control rules, and service processes to present them in a structured manner through H5 pages or reusing static templates. This process involves a large number of repetitive operations and is susceptible to human factors, making it difficult to guarantee the integrity and timeliness of knowledge data. This inefficient knowledge configuration method not only increases the company's operating costs but also restricts the dynamic updating capability of the knowledge graph.

[0003] For example, in the scenario of updating risk control rules in auto finance, when companies adjust loan approval conditions (such as adding an assessment dimension for the battery life of new energy vehicles), traditional methods require manual reconfiguration of relevant fields and page logic, resulting in a rule update cycle of up to several weeks, during which business risks may arise due to rule lag.

[0004] Therefore, there is an urgent need to provide an automated knowledge configuration method to improve the efficiency of knowledge data generation in the automotive finance field and support the real-time dynamic data updates of knowledge graphs. Summary of the Invention

[0005] The purpose of this application is to propose a data processing method, apparatus, computer equipment, and storage medium based on artificial intelligence, so as to solve the technical problem of low efficiency in the existing knowledge configuration methods in automotive finance knowledge management scenarios.

[0006] Firstly, an artificial intelligence-based data processing method is provided, including: Collect relevant data corresponding to users from multiple pre-set channels; User profile data corresponding to feature extraction processing of the relevant data, and scene tag data corresponding to scene analysis of the relevant data; Based on a preset large language model, semantic analysis is performed on the user profile data and the scene tag data to obtain corresponding semantic analysis results, and target templates that match the semantic analysis results are selected from a preset template library. Obtain target node information related to the user profile data and the scene tag data from a preset knowledge graph; The system receives input data from the user and processes the target node information and the input data based on a preset content generation model to generate corresponding text content. The text content is filled into the target template to obtain the corresponding target text data; Based on the target text data, image generation processing is performed to obtain the corresponding target image data; Based on the target text data and the target image data, corresponding target knowledge data is generated, and the target knowledge data is published to a preset knowledge base.

[0007] Secondly, an artificial intelligence-based data processing device is provided, comprising: The data collection module is used to collect relevant data corresponding to users from multiple preset channels; The first processing module is used to perform feature extraction processing on the relevant data to obtain corresponding user profile data, and to perform scene analysis on the relevant data to obtain corresponding scene tag data. The second processing module is used to perform semantic analysis on the user profile data and the scene tag data based on a preset large language model to obtain the corresponding semantic analysis results, and to filter out the target template that matches the semantic analysis results from a preset template library. The acquisition module is used to acquire target node information related to the user profile data and the scene tag data from a preset knowledge graph; The first generation module is used to receive the user's input data and process the target node information and the input data based on a preset content generation model to generate corresponding text content. The fill module is used to fill the text content into the target template to obtain the corresponding target text data; The second generation module is used to perform image generation processing based on the target text data to obtain the corresponding target image data; The third processing module is used to generate corresponding target knowledge data based on the target text data and the target image data, and to publish the target knowledge data to a preset knowledge base.

[0008] 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 artificial intelligence-based data processing method.

[0009] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the aforementioned artificial intelligence-based data processing method.

[0010] In the aforementioned scheme implemented by the data processing method, apparatus, computer equipment, and storage medium based on artificial intelligence, firstly, relevant data corresponding to the user is collected from multiple preset channels; then, feature extraction processing is performed on the relevant data to obtain corresponding user profile data, and scene analysis is performed on the relevant data to obtain corresponding scene tag data; then, semantic analysis is performed on the user profile data and the scene tag data based on a preset large language model to obtain corresponding semantic analysis results, and target templates matching the semantic analysis results are selected from a preset template library; next, target node information related to the user profile data and the scene tag data is obtained from a preset knowledge graph; subsequently, input data from the user is received, and the target node information and the input data are processed based on a preset content generation model to generate corresponding text content; then, the text content is filled into the target template to obtain corresponding target text data; and image generation processing is performed based on the target text data to obtain corresponding target image data; finally, corresponding target knowledge data is generated based on the target text data and the target image data, and the target knowledge data is published to a preset knowledge base. Based on the above automated processing flow, unlike existing knowledge configuration methods that rely on manual intervention, this application processes relevant user data to obtain user profile data and scene tag data. Then, by combining large language models, knowledge images, and content generation models, it processes and generates content based on the characteristics and needs of different modalities, satisfying diverse content generation requirements. Subsequently, based on the obtained target text data and target image data, it generates corresponding target knowledge data and publishes it to a preset knowledge base, thereby automatically and accurately realizing the generation and processing of knowledge data and effectively improving the efficiency of knowledge data generation. Attached Figure Description

[0011] To more clearly illustrate the solutions in this application, the accompanying drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is an exemplary system architecture diagram to which this application can be applied; Figure 2 This is a flowchart of an embodiment of the artificial intelligence-based data processing method according to this application; Figure 3 This is a schematic diagram of a structure of an embodiment of the artificial intelligence-based data processing apparatus according to this application; Figure 4This is a schematic diagram of the structure of one embodiment of the computer device according to this application. Detailed Implementation

[0013] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, in the specification, claims, and foregoing drawings of this application, are intended to cover non-exclusive inclusion. The terms "first," "second," etc., in the specification, claims, or foregoing drawings of this application are used to distinguish different objects, not to describe a particular order.

[0014] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0015] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0016] like Figure 1 As shown, system architecture 100 may include terminal device 101, network 102, and server 103. Terminal device 101 may be a laptop 1011, tablet 1012, or mobile phone 1013. Network 102 is used as a medium to provide a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0017] Users can use terminal device 101 to interact with server 103 via network 102 to receive or send messages, etc. Various communication client applications can be installed on terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social media platform software, etc.

[0018] Terminal device 101 can be various electronic devices with a display screen and support web browsing. In addition to laptops 1011, tablets 1012, or mobile phones 1013, terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 player (Moving Picture Experts Group Audio Layer IV), a laptop computer, and a desktop computer, etc.

[0019] Server 103 can be a server that provides various services, such as a backend server that provides support for the pages displayed on terminal device 101.

[0020] It should be noted that the artificial intelligence-based data processing method provided in the embodiments of this application is generally executed by a server / terminal device, and correspondingly, the artificial intelligence-based data processing device is generally set in the server / terminal device.

[0021] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0022] Continue to refer to Figure 2 The flowchart illustrates an embodiment of the AI-based data processing method according to this application. The order of steps in the flowchart can be changed, and some steps can be omitted, depending on different needs. The AI-based data processing method provided in this application can be applied to any scenario requiring knowledge management, and thus can be applied to products in these scenarios, such as knowledge management products in the financial and insurance fields. The AI-based data processing method includes the following steps: Step S201: Collect relevant data corresponding to the user from multiple preset channels.

[0023] In this embodiment, the data processing method based on artificial intelligence runs on an electronic device (e.g., Figure 1 The server / terminal device shown can acquire relevant data corresponding to the user through wired or wireless connections. 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-wideband) connections, and other currently known or future wireless connection methods.

[0024] This involves acquiring relevant user data (including user identity data, business line data, and historical behavior data) from multiple channels such as enterprise systems, business records, and user operation logs. User identity data is obtained through enterprise system interfaces, such as account information and department details; business line data is extracted from business systems based on business classification rules, specifying the user's business area, such as whether it is passenger car or commercial vehicle business in auto finance; historical behavior data is collected from user operation logs, covering operation records such as the page types configured by the user and the templates used.

[0025] Step S202: Perform feature extraction processing on the relevant data to obtain corresponding user profile data, and perform scene analysis on the relevant data to obtain corresponding scene tag data.

[0026] In this embodiment, a predictive model is constructed by using machine learning algorithms and a scene-aware engine to learn from and train on historical scene data. This predictive model can predict the user's current scene and generate corresponding scene labels based on the characteristics of real-time input data. For example, by training on a large amount of user behavior data in different scenarios, the model can learn the scene in which the user is most likely to be at a specific time, place, and device usage, thus accurately identifying scene labels when new data (such as the aforementioned related data) is input.

[0027] The specific implementation process of extracting features from the aforementioned data to obtain the corresponding user profile data will be described in more detail in subsequent embodiments of this application, and will not be elaborated upon here.

[0028] Step S203: Based on a preset large language model, perform semantic analysis on the user profile data and the scene tag data to obtain the corresponding semantic analysis results, and select target templates that match the semantic analysis results from a preset template library.

[0029] In this embodiment, the specific implementation process of performing semantic analysis on the user profile data and the scene tag data based on the preset large language model to obtain the corresponding semantic analysis results, and selecting the target template that matches the semantic analysis results from the preset template library will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0030] Step S204: Obtain target node information related to the user profile data and the scene tag data from the preset knowledge graph.

[0031] In this embodiment, the aforementioned knowledge graph is a business knowledge graph pre-constructed based on actual business needs, such as an automotive finance knowledge graph. Query conditions can be determined from the knowledge graph using a graph query algorithm, based on the current scenario (scenario tag data) and user needs (user profile data), and relevant node information can be extracted. For example, if the current scenario is "customer acquisition," the query can retrieve node information related to customer acquisition, such as product policies and service processes, and organize this information into a structured text format.

[0032] Step S205: Receive the user's input data, and process the target node information and the input data based on a preset content generation model to generate corresponding text content.

[0033] In this embodiment, the input data mentioned above is relevant information input by the user. This information may be the user's specific requirements for the generated content, such as whether they want the text style to be formal or lively, or it may be supplementary explanations, such as further emphasis on a certain product feature.

[0034] The specific implementation process of processing the target node information and the input data based on the preset content generation model to generate the corresponding text content will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0035] Step S206: Fill the target template with the text content to obtain the corresponding target text data.

[0036] In this embodiment, the target template predefines the content structure and some fixed elements. The generated text content simply needs to be filled into the corresponding positions in the target template. This improves the speed and efficiency of content generation. For example, when generating a monthly report on auto finance, the target template already includes the report's title page, table of contents, and fixed statistical chart framework. The content generation model only needs to generate the specific data and analysis content for the month and fill it in, significantly shortening the report generation time.

[0037] One approach is template-based population, where unique identifiers, such as specific tags or placeholders, are set for different content areas within the target template. For example, in an HTML template, this can be achieved using... As an identifier for the title area These identifiers serve as markers for the summary area. Furthermore, after generating text content, the content generation model uses these identifiers to populate the corresponding text content into the appropriate positions in the template. In programmatic implementation, techniques such as string replacement or DOM manipulation can be used to accomplish this population process. For example, Python's string replacement methods can be used to replace placeholders in the template string with the generated content.

[0038] Step S207: Perform image generation processing based on the target text data to obtain the corresponding target image data.

[0039] In this embodiment, the specific implementation process of generating image data based on the target text data to obtain the corresponding target image data will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0040] Step S208: Generate corresponding target knowledge data based on the target text data and the target image data, and publish the target knowledge data to a preset knowledge base.

[0041] In this embodiment, the specific implementation process of generating corresponding target knowledge data based on the target text data and the target image data, and publishing the target knowledge data to a preset knowledge base, will be described in further detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0042] This application first collects relevant user-related data from multiple preset channels; then performs feature extraction processing on the relevant data to obtain corresponding user profile data, and performs scene analysis on the relevant data to obtain corresponding scene tag data; next, based on a preset large language model, it performs semantic analysis on the user profile data and the scene tag data to obtain corresponding semantic analysis results, and selects target templates that match the semantic analysis results from a preset template library; then, it obtains target node information related to the user profile data and the scene tag data from a preset knowledge graph; subsequently, it receives input data from the user, and processes the target node information and the input data based on a preset content generation model to generate corresponding text content; then, it fills the text content into the target template to obtain corresponding target text data; and performs image generation processing based on the target text data to obtain corresponding target image data; finally, it generates corresponding target knowledge data based on the target text data and the target image data, and publishes the target knowledge data to a preset knowledge base. Based on the above automated processing flow, unlike existing knowledge configuration methods that rely on manual intervention, this application processes relevant user data to obtain user profile data and scene tag data. Then, by combining large language models, knowledge images, and content generation models, it processes and generates content based on the characteristics and needs of different modalities, satisfying diverse content generation requirements. Subsequently, based on the obtained target text data and target image data, it generates corresponding target knowledge data and publishes it to a preset knowledge base, thereby automatically and accurately realizing the generation and processing of knowledge data and effectively improving the efficiency of knowledge data generation.

[0043] In some optional implementations, the user profile data corresponding to the feature extraction processing of the relevant data in step S202 includes the following steps: Obtain the preset preprocessing strategy.

[0044] In this embodiment, the preprocessing strategy includes the following: for user identity data, a data deduplication algorithm is used to remove duplicate account records by comparing key fields such as account number; for historical behavior data, a judgment rule for missing key information is set. If a record is missing key operation information, it is filled in according to business logic, such as filling in with the average value of similar operations, or directly removing the record. During preprocessing, date data is converted according to a unified standard format (such as "YYYY-MM-DD") to ensure that date data from different sources has a consistent format, facilitating subsequent processing.

[0045] The relevant data is processed based on the preprocessing strategy to obtain the corresponding processed data.

[0046] In this embodiment, preprocessing of the aforementioned data can be performed based on the strategy content of the preprocessing strategy to obtain the corresponding processed data.

[0047] Based on a preset lightweight model, feature extraction is performed on the processed data to obtain the corresponding feature data.

[0048] In this embodiment, the aforementioned lightweight model can specifically employ a lightweight embedding model. The preprocessed data is input into this lightweight embedding model, which analyzes historical user behavior data and uses deep learning algorithms to extract user operational preference features across different business scenarios. For example, the frequency with which users click on different types of pages in different business scenarios is used. These features are then converted into vector form. Combined with user identity data and business line data, a comprehensive "user-scenario" vector (feature data) is generated through vector concatenation or fusion algorithms. This vector contains various feature information about the user in different scenarios, such as basic information like age, gender, and occupation, as well as the user's behavioral habits and preferences in specific scenarios. For example, in a "shopping" scenario, the user profile might include features such as the categories of goods the user frequently purchases, the time period for purchases, and the price range of purchases.

[0049] Furthermore, the current configuration intent can be identified based on this vector. For example, by calculating the similarity between the vector and preset intent vectors such as "customer acquisition" and "pre-loan review", the vector with the highest similarity is the current configuration intent.

[0050] The feature data is used as the user profile data.

[0051] This application obtains a preset preprocessing strategy; then processes the relevant data based on the preprocessing strategy to obtain corresponding processed data; subsequently, it extracts features from the processed data based on a preset lightweight model to obtain corresponding feature data; and finally, it uses the feature data as the user profile data. Based on the above processing flow, this application ensures data quality by preprocessing the collected relevant data, and transforms the data into a quantifiable and analyzable vector form through feature extraction, thereby accurately identifying the user's intent in the current scenario and ensuring the accuracy of the generated user profile data. This is beneficial for providing a reliable data foundation for subsequent intelligent template recommendation and content generation.

[0052] In some optional implementations of this embodiment, step S203 includes the following steps: The user profile data and the scene tag data are associated based on a preset mapping algorithm to obtain the corresponding generated data.

[0053] In this embodiment, the identified scene tags are associated with user profiles using a mapping algorithm. The core idea of ​​the mapping algorithm is to find relevant feature dimensions in the user profile based on the scene tags. For example, when the "customer acquisition" scene tag is identified, the algorithm searches the user profile for relevant feature dimensions, such as the user's past customer acquisition habits, including the acquisition channels used and the types of acquisition information sent; and success stories, i.e., the methods the user has used to successfully acquire customers. These relevant features are then bound to the "customer acquisition" scene tag to form a complete "user-scene" association information, which is used as the corresponding generated data.

[0054] Based on the large language model, semantic analysis is performed on the generated data to obtain the corresponding semantic representation.

[0055] In this embodiment, the aforementioned large language model can be a finely tuned version of a large model (such as ChatGLM3 or Qwen). By inputting the generated data into this large language model, the model uses natural language processing techniques to perform semantic analysis. Through operations such as word segmentation, part-of-speech tagging, and syntactic analysis, the input text data is transformed into a semantic representation that a computer can understand. For example, for a user behavior description in the "customer acquisition" scenario, such as "frequently attracting customers by sending promotional information through social media," the large model will decompose it into individual words, analyze the relationships between the words, and understand the semantics of the entire sentence—that is, a user's customer acquisition behavior. In this way, the large model can understand the user's needs and the characteristics of the scenario.

[0056] Based on the large language model, semantic analysis is performed on all templates in the template library to obtain several corresponding key semantic features.

[0057] In this embodiment, semantic parsing is performed on all templates in the template library simultaneously. Natural language processing techniques are also used to extract key semantic features of the templates. For example, for an email template used for customer acquisition, the large model analyzes the email's subject and body content, extracting key information related to customer acquisition, such as descriptions of promotional activities and attractive features, and uses this information as the template's key semantic features.

[0058] This involves collecting template-related information from template management systems, user records, and other sources to build a corresponding template library. The applicable scenarios for templates are determined through manual annotation or automatic classification based on template functional characteristics; functional characteristics are extracted by analyzing the template's structure and content; historical usage data is obtained from user operation logs, including metrics such as click-through rate and usage frequency. This information is then organized into a structured data format, such as tables, for easier subsequent processing.

[0059] Calculate the similarity between the semantic representation and each of the key semantic features.

[0060] In this embodiment, the similarity between the semantic representation and each key semantic feature can be calculated using a cosine similarity algorithm. Cosine similarity measures the degree of similarity between two vectors by calculating the difference in their directions.

[0061] Select the specified template with the highest similarity from all the templates mentioned.

[0062] In this embodiment, the highest value of the specified similarity is determined by comparing all the calculated similarities, and then the specified template that matches the specified similarity is selected from all templates and used as the required target template.

[0063] Use the specified template as the target template.

[0064] This application uses a pre-defined mapping algorithm to associate user profile data and scene tag data to obtain corresponding generated data. Then, it performs semantic analysis on the generated data based on a large language model to obtain corresponding semantic representations. Next, it performs semantic analysis on all templates in the template library based on the large language model to obtain multiple key semantic features. Then, it calculates the similarity between the semantic representations and each key semantic feature. Subsequently, it selects the specified template with the highest similarity from all templates. Finally, it uses the specified template as the target template. Based on the above processing flow, this application can achieve accurate template recommendation by utilizing the semantic analysis capabilities of a large language model. The semantic matching and recommendation process fully considers user needs and template characteristics, effectively improving the accuracy and reliability of template recommendation.

[0065] In some alternative implementations, step S205 includes the following steps: Call the preset content generation model.

[0066] In this embodiment, the selection of the above content generation model is not specifically limited, and a general LLM model, such as GPT4, Qwen, etc., can be used.

[0067] Based on the target node information, the content generation model is used to process the input data to generate content, resulting in corresponding title data, abstract data, and a draft of the main text.

[0068] In this embodiment, target node information and user input data are fed into the content generation model. The model then uses deep learning algorithms to perform semantic understanding and analysis on this data. Combining the logical relationships and semantic information within the knowledge graph, and following preset text generation rules and business requirements (such as requiring concise and clear titles and summaries that summarize the main content), it generates corresponding titles, summaries, and initial drafts of the main text. For example, based on the logical relationships of product policies and service processes within the knowledge graph, it generates main text content that conforms to business logic.

[0069] The title data, abstract data, and initial draft of the main text are integrated to obtain the corresponding integrated data.

[0070] In this embodiment, the generated title data, abstract data, and initial draft of the main text are integrated and processed, and the resulting integrated data is used as the corresponding text content.

[0071] The integrated data is used as the text content.

[0072] This application invokes a pre-defined content generation model; then, based on target node information, it uses the content generation model to process the input data, obtaining corresponding title data, abstract data, and a draft of the main text; subsequently, it integrates the title data, abstract data, and draft of the main text to obtain integrated data; finally, it uses the integrated data as the text content. Based on the above processing flow, this application provides users with accurate and professional knowledge support by integrating knowledge graph information, while processing user input to meet users' personalized needs, and by leveraging the powerful capabilities of the content generation model to generate high-quality text content, ensuring the accuracy of the text content.

[0073] In some alternative implementations, step S207 includes the following steps: The target text data is semantically parsed using natural language processing technology to obtain the corresponding semantic parsing results.

[0074] In this embodiment, the target text data can be semantically analyzed using natural language processing (NLP) technology. First, lexical analysis is performed, breaking down sentences into words and determining the part of speech of each word. For example, "customer acquisition guide" is broken down into "customer," "acquisition," and "guide," which are identified as nouns, verbs, and nouns, respectively. Then, syntactic analysis is performed to construct the grammatical structure of the sentences and understand the grammatical relationships between words. Through these analyses, key information is extracted, such as the theme, which can be determined by statistically analyzing the core words that appear frequently in the text. For example, in "customer acquisition guide," "customer acquisition" is the theme. Sentiment can be judged by analyzing the sentiment words used in the text, such as positive, negative, or neutral. Important elements can be found using keyword extraction algorithms, such as extracting key product features and service advantages from the main text. These algorithms filter words based on factors such as their importance and frequency in the text.

[0075] Call the preset image generation model.

[0076] In this embodiment, the selection of the image generation model is not specifically limited and can be determined according to actual business needs, such as using a Stable Diffusion model.

[0077] Based on the semantic parsing results, image generation instructions are generated that match the input requirements of the image generation model.

[0078] In this embodiment, based on the semantic parsing results, suitable image generation instructions are generated according to the input requirements of the image generation model. The instructions need to clearly define the image's theme, representing the topic information extracted from the target text data, such as "product display in a customer acquisition scenario." Style requirements are specified according to business needs; for example, "Xiaohongshu style" is characterized by vibrant colors and lively visuals, while "Enterprise WeChat blue" emphasizes a blue tone and a simple, elegant style. Key elements are the important information extracted from the text, such as product features and service advantages. A rule engine transforms the semantic information into an instruction format that the model can understand. The rule engine is used to convert human-understood semantic information according to the syntax and format specified by the model; for example, it transforms "displaying product features in a customer acquisition scenario in a Xiaohongshu style" into a specific code or instruction format that the image generation model can recognize.

[0079] The image generation instruction is processed based on the image generation model to generate corresponding image data.

[0080] In this embodiment, the specific implementation process of processing the image generation instruction based on the image generation model to generate the corresponding image data will be further described in detail in subsequent specific embodiments of this application, and will not be elaborated on here.

[0081] The image data is used as the target image data.

[0082] This application performs semantic parsing on target text data using natural language processing technology to obtain corresponding semantic parsing results. Then, it calls a pre-defined image generation model and generates image generation instructions that match the input requirements of the image generation model based on the semantic parsing results. Subsequently, the image generation model processes the image generation instructions to generate corresponding image data, which is then used as the target image data. Based on this processing flow, this application generates image generation instructions based on semantic parsing results, and then processes these instructions using an image generation model, ensuring that the generated image data accurately expresses the core information of the target text data, thus guaranteeing the accuracy of the target image data.

[0083] In some optional implementations of this embodiment, the step of processing the image generation instruction based on the image generation model to generate corresponding image data includes the following steps: The image generation instruction is processed based on the image generation model to generate the corresponding initial image data.

[0084] In this embodiment, image generation instructions are input into an image generation model, which then uses algorithms such as generative adversarial networks (GANs) to generate cover images, carousel images, and other similar images. The GAN consists of a generator and a discriminator. The generator is responsible for generating images according to the instructions, while the discriminator determines whether the generated images meet the requirements. The two work against each other, continuously optimizing the process, ultimately generating high-quality images, i.e., initial image data. This initial image data should be related to the brand's business but does not yet possess the brand's style; examples include generated ordinary product cover images and event carousel images.

[0085] Get the preset style requirements.

[0086] In this embodiment, a sufficient number of corporate brand style images representing the company's brand style are collected in advance. These images should cover various visual elements commonly used by the company, such as product promotional images, posters, and official website screenshots. The images must be of high quality, with appropriate resolution, and clearly present the key characteristics of the brand style, such as color, texture, and font. Then, style feature analysis is performed on the corporate brand style images to obtain the corresponding style requirements. The specific implementation process includes: 1) Color Distribution Analysis. Extracting Color Histograms: For each corporate brand style image, image processing techniques are used to extract its color histogram. Color histograms statistically analyze the distribution of different color pixels in an image, reflecting key information such as the brand image's dominant color tone and color scheme proportions. For example, analysis might reveal that blue pixels account for the highest proportion in a certain brand image, followed by white and gray, with blue exhibiting a certain distribution across different brightness levels. This indicates that the brand uses blue as its primary color tone, paired with white and gray to create a simple and professional style. Determining Color Mapping Relationships: Based on the color histograms of the generated image and the brand style image, a color mapping relationship is established. This determines which colors in the generated image should be mapped to which colors in the brand style image, ensuring consistency in color between the generated image and the brand style. For example, light blue in the generated image might be mapped to dark blue in the brand style image to enhance color contrast and saturation.

[0087] 2) Font Style Analysis. Font Recognition: Using font recognition tools or algorithms to identify the fonts in the company's brand style images. This involves determining commonly used font types, such as Song, Hei, and Kai, as well as special font styles, such as bold, italics, and underlines. Font Feature Extraction: Further analysis of font features, such as stroke thickness, character spacing, and line spacing. These features are crucial for maintaining brand style consistency. For example, a brand's font might have thicker strokes, moderate character spacing, and larger line spacing, giving a sense of grandeur and stability.

[0088] 3) Texture and Pattern Analysis. Texture Extraction: If the corporate brand style image contains specific textures, such as wood grain or metal grain, texture extraction algorithms are used to separate these textures from the image. Textures can be accurately described by analyzing features such as frequency, direction, and roughness. Pattern Recognition: Identify common patterns in the brand image, such as the company logo and specific decorative patterns. Determine the position, size, and proportion of these patterns in the image so that they can be accurately applied to the generated image during style transfer.

[0089] Based on the style requirements, the initial image data is adjusted using a preset style transfer algorithm to obtain the adjusted first image data.

[0090] In this embodiment, the above adjustment process is a style transfer application, and the implementation process includes: 1. Color Adjustment. Color Map-Based Adjustment: Based on the established color mapping relationship, the color of each pixel in the generated image is adjusted. The colors in the generated image are replaced with the corresponding colors in the brand style image to achieve color uniformity. For example, image editing software or specialized image processing algorithms can be used to modify the pixel colors of the generated image one by one according to the color map table. Color Balance and Contrast Adjustment: After color replacement, the color balance and contrast of the generated image are further adjusted to better match the brand style requirements. For example, if the brand style image has high contrast, the contrast of the generated image can be appropriately increased to enhance its visual impact.

[0091] 2. Font Replacement. Text Location: Locate the text region in the generated image where the font needs to be replaced. Text detection algorithms can be used to automatically identify the text location and content in the image. Font Replacement and Layout Adjustment: Replace the located text with the brand-specified font and adjust the text layout according to the brand font characteristics, such as character spacing and line spacing. Ensure that the replaced text is visually consistent with the brand style and does not affect the overall layout and readability of the image.

[0092] 3. Texture and Pattern Addition. Texture Compositing: Apply the extracted brand texture to the corresponding areas of the generated image. Texture compositing algorithms can be used to naturally blend the texture into the image based on its content and structure. For example, if the generated image is a product display, the brand's wood texture can be applied to the product background to create an atmosphere consistent with the brand style. Pattern Insertion: Accurately insert the identified brand pattern into the specified position in the generated image. Adjust its display effect in the image according to its size and proportion to ensure consistency between the pattern and the image content. For example, inserting the company logo in a corner or center of the generated image enhances brand recognition.

[0093] The first image data is processed using a preset optimization strategy to obtain the corresponding second image data.

[0094] In this embodiment, the image optimization processing includes: 1) Sharpness optimization. Sharpening: Image sharpening algorithms are used to enhance the edges and details of the image, improving image sharpness. Sharpening can make objects in the image more clearly distinguishable, enhancing the visual effect. For example, Gaussian sharpening or Laplacian sharpening algorithms are used, and the sharpening parameters are adjusted appropriately to avoid over-sharpening that leads to noise and artifacts in the image. Super-resolution reconstruction: If the resolution of the style-transferred image is low, super-resolution reconstruction techniques can be used to improve the image resolution. Super-resolution reconstruction algorithms can learn the intrinsic structure and features of the image to recover a high-resolution image from a low-resolution image, making the image clearer and more detailed. For example, deep learning-based super-resolution reconstruction methods, such as SRCNN and ESRGAN, can achieve good reconstruction results.

[0095] 2) Color Optimization. Color Correction: Further correct the colors of the style-transferred image to ensure accuracy and naturalness. Color correction tools or algorithms can be used to adjust parameters such as color temperature, hue, and saturation to eliminate color deviations and make the image colors more consistent with human visual perception. For example, if an image is yellowish, the color temperature can be appropriately lowered to restore normal colors. Color Enhancement: While ensuring color accuracy, enhance the colors of the image to improve vibrancy and contrast. This can be achieved by adjusting color channel curves or using color enhancement algorithms to make the colors in the image more vivid and saturated, enhancing the image's visual appeal. For example, increasing the contrast of the red and green channels can make plants and flowers in an image more vibrant and eye-catching.

[0096] 3) Detail Optimization. Noise Reduction: During style transfer, noise may be introduced, affecting image quality. Noise reduction algorithms are used to remove noise and improve image clarity. Common algorithms include median filtering, Gaussian filtering, and nonlocal mean denoising. Appropriate methods and parameters can be selected based on the image's characteristics. Detail Enhancement: Important details in the image are enhanced to make them more prominent and clear. Detail enhancement algorithms, such as wavelet transform-based methods, can be used to extract high-frequency detail information and appropriately amplify and enhance it, improving the image's detail representation. For example, enhancing facial expression details makes the person more lifelike.

[0097] The second image data is used as the image data.

[0098] This application obtains preset style requirements; then, based on these requirements, it adjusts the initial image data using a preset style transfer algorithm to obtain adjusted first image data; subsequently, it optimizes the first image data using a preset optimization strategy to obtain corresponding second image data; finally, it uses the second image data as the final image data. Based on this processing flow, this application adjusts the initial image data using a style transfer algorithm based on style requirements to obtain adjusted first image data, ensuring that the image conforms to the corporate brand style. Furthermore, it optimizes the first image data using an optimization strategy to further improve image quality.

[0099] In some optional implementations of this embodiment, step S208 includes the following steps: Select specified text content from the target text data; wherein the specified text content is content that meets the preset conditions for generating speech.

[0100] In this embodiment, the portion of the generated target text data from which speech needs to be generated can be selected as the specified text content based on actual business requirements. For example, in training materials, only the main text may need to be generated as speech. The selected specified text content undergoes format checking and preprocessing. Format checking includes checking for special characters, garbled text, etc., to ensure the text is standardized. Preprocessing may include word segmentation and pause marking, such as adding pause markers to places requiring emphasis, making the generated speech more in line with human expression habits. The text must also meet the requirements of the speech synthesis model. Different speech synthesis models may have different requirements for text format and encoding, requiring corresponding adjustments.

[0101] Obtain preset speech synthesis parameters.

[0102] In this embodiment, speech synthesis parameters can be set as needed. Regarding language, if the training materials are geared towards international clients, multilingual playback parameters can be set through the speech synthesis parameter configuration interface. Different languages ​​have different speech libraries and pronunciation rules. For example, English and Chinese pronunciation methods and intonations differ significantly, requiring the selection of the appropriate speech library and configuration of correct pronunciation rules based on the specific language. The speech rate parameter can be adjusted according to the importance of the content and the audience's comprehension ability. For example, the speech rate can be slowed down appropriately for important content to allow for better understanding. The intonation parameter can be set with different intonation modes, such as formal or lively, to adapt to different scenarios and content styles.

[0103] Based on a preset speech synthesis model, the specified text content and the speech synthesis parameters are processed to generate speech data, thereby obtaining the corresponding target speech data.

[0104] In this embodiment, by inputting the specified text content and speech synthesis parameters into the speech synthesis model, the model employs a speech synthesis algorithm, such as deep learning-based speech synthesis technology. This technology first extracts features from the text, converting it into digital features that a computer can process. Then, based on the set speech synthesis parameters, such as speech rate and intonation, these digital features are adjusted to meet the requirements. Finally, these digital features are converted into digital human narration audio to generate fluent and natural speech audio.

[0105] The target text data, target image data, and target speech data are integrated and processed to obtain the corresponding target knowledge data.

[0106] In this embodiment, the target text data, target image data and target speech data are integrated and processed, and the resulting integrated target data is used as the corresponding target knowledge data.

[0107] The target knowledge data is published to the knowledge base using a preset data import tool.

[0108] In this embodiment, various types of target knowledge data generated can be integrated into a knowledge base using a data import tool. The knowledge content is categorized based on business scenarios and knowledge types, such as categorizing it according to business processes in auto finance (customer acquisition, pre-loan review, etc.) and knowledge types (product policies, service processes, etc.). The knowledge content is also annotated with keywords, tags, and other information to facilitate subsequent retrieval and management. For example, text, images, and audio related to the "Customer Acquisition Guide" can be categorized and stored according to business scenarios and knowledge types, and tags such as "Customer Acquisition" and "Product Policy" can be added.

[0109] In addition, the system also has the following data processing functions: Graph matching data processing: Based on the relationship between knowledge content and the "Automotive Finance Knowledge Graph," graph matching algorithms are used for graph matching. Entities and relationships in the knowledge content are analyzed and associated with relevant nodes in the knowledge graph, enriching the knowledge graph's content. For example, the new product policy in the "Customer Acquisition Guide" is associated with product nodes in the knowledge graph; by establishing links between entities, the structure and information of the knowledge graph are improved.

[0110] Review Flow Data Processing: This process collects user-submitted review requests and related information, extracting information such as the review target and requirements from the requests. Following pre-defined review rules and procedures, review tasks are assigned to appropriate reviewers. Reviewers then review the content according to the rules, recording their comments and results, and providing feedback to the user via message notifications or system feedback. For example, after a user submits a review request for a "Customer Acquisition Guide," the system assigns the task to reviewers in the relevant field based on the review rules. The reviewers then record their comments and provide feedback to the user.

[0111] Access statistics processing: User access data for knowledge content is collected using front-end tracking technology, including dwell time, conversion rate to favorites, and sharing paths. The collected data is statistically analyzed, and data analysis algorithms are used to generate access statistics reports. For example, data such as the number of visits to the "Customer Acquisition Guide" within a week, average dwell time per user, and conversion rate to favorites are compiled and presented in charts and other formats to generate reports, providing a basis for optimizing and recommending knowledge content.

[0112] Data feedback processing: Access statistics and other relevant data are fed back to the large language model and content generation model. This data is cleaned and preprocessed to remove noise and outliers, and transformed into a format suitable for model training. This data is then used as training data or optimization parameters to adjust the model's recommendation and generation strategies, enabling iterative optimization. For example, based on user access data for the "Customer Acquisition Guide," the large language model's strategy for recommending related templates can be adjusted to improve recommendation accuracy.

[0113] This application selects specified text content from target text data, where the specified text content meets preset conditions for speech generation. Then, preset speech synthesis parameters are obtained. Based on a preset speech synthesis model, the specified text content and speech synthesis parameters are processed to generate speech, resulting in corresponding target speech data. Subsequently, the target text data, target image data, and target speech data are integrated to obtain corresponding target knowledge data. Finally, the target knowledge data is published to a knowledge base using a preset data import tool. Based on this processing flow, this application further enriches the presentation of content through speech synthesis. By selecting appropriate text and setting parameters, natural and fluent speech is generated. Furthermore, by publishing the target knowledge data to a knowledge base, the quality and utilization efficiency of knowledge content can be improved, providing users with better knowledge services.

[0114] In some optional implementations, this application also provides data processing functions for the dynamic master control system, the specific implementation process of which includes: Master page definition data collection: Collect relevant data about master page definitions from system configuration files, business rule documents, and other sources. The whitelist attribute set is determined through business personnel definition and system analysis, clarifying which attributes are configurable, such as whether certain fields on a page are allowed to be modified by users; the default value strategy sets default values ​​for configurable attributes based on business needs and historical data, such as the default display content of a field when no user configuration is provided; permission rules obtain the different roles' operation permissions for the master page through the role management system, such as which roles can edit and view the master page.

[0115] Role-Scene Data Association: User role information is obtained from the user management system, and current scene data is obtained from the scene perception engine. An association algorithm is then used to link the user role information with the current scene data. For example, a role-scene mapping table is created. Based on the user role and the current scene, the corresponding role characteristics are looked up in the table to clarify the role characteristics of a passenger vehicle operations user in the "customer acquisition" scenario.

[0116] Dynamic field enabling / disabling: Based on the association data between roles and scenarios, as well as the permission rules and whitelist attribute sets in the master template definition, a decision algorithm dynamically determines which fields need to be enabled and which need to be disabled. For example, for commercial vehicle users, in specific scenarios, based on permission rules and whitelist attribute sets, only user-related fields in the video template are allowed to be operated, while other irrelevant fields are disabled through front-end technology, achieving a simplified interface that is "personalized for each user".

[0117] The dynamic master control system's data processing, through the collection of master definition data and the association of role and scenario data, enables the dynamic enabling and disabling of fields. It flexibly adjusts the interface display content according to different user roles and scenario needs, improving system applicability and user experience, and meeting diverse business requirements.

[0118] In some alternative implementations, the user information obtained is subject to user consent and complies with relevant laws and policies.

[0119] Furthermore, any software tools or components not belonging to our company that appear in the embodiments of this application are merely illustrative examples and do not represent actual use.

[0120] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order 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.

[0121] It should be emphasized that, in order to further ensure the privacy and security of the aforementioned target knowledge data, the target knowledge data can also be stored in a node of a blockchain.

[0122] The blockchain referred to in this application is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0123] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. Foundational artificial intelligence technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0124] 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 instructing related hardware through computer-readable instructions. These computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, optical disk, or read-only memory (ROM), or random access memory (RAM).

[0125] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated 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 steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0126] Further reference Figure 3 As a response to the above Figure 2 To implement the method shown, this application provides an embodiment of a data processing device based on artificial intelligence, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0127] like Figure 3 As shown, the artificial intelligence-based data processing device 300 described in this embodiment includes: a collection module 301, a first processing module 302, a second processing module 303, an acquisition module 304, a first generation module 305, a filling module 306, a second generation module 307, and a third processing module 308. Wherein: The collection module 301 is used to collect relevant data corresponding to the user from multiple preset channels; The first processing module 302 is used to perform feature extraction processing on the relevant data to obtain corresponding user profile data, and to perform scene analysis on the relevant data to obtain corresponding scene tag data. The second processing module 303 is used to perform semantic analysis on the user profile data and the scene tag data based on a preset large language model to obtain the corresponding semantic analysis results, and to filter out the target template that matches the semantic analysis results from the preset template library. The acquisition module 304 is used to acquire target node information related to the user profile data and the scene tag data from a preset knowledge graph; The first generation module 305 is used to receive the user's input data and process the target node information and the input data based on a preset content generation model to generate corresponding text content. The filling module 306 is used to fill the text content into the target template to obtain the corresponding target text data; The second generation module 307 is used to perform image generation processing based on the target text data to obtain the corresponding target image data; The third processing module 308 is used to generate corresponding target knowledge data based on the target text data and the target image data, and to publish the target knowledge data to a preset knowledge base.

[0128] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data processing method in the aforementioned embodiments, and will not be repeated here. In some optional implementations of this embodiment, the first processing module 302 includes: The first acquisition submodule is used to acquire the preset preprocessing strategy; The first processing submodule is used to process the relevant data based on the preprocessing strategy to obtain the corresponding processed data; The extraction submodule is used to extract features from the processed data based on a preset lightweight model to obtain the corresponding feature data. The first determining submodule is used to use the feature data as the user profile data.

[0129] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data processing method in the aforementioned embodiments, and will not be repeated here. In some optional implementations of this embodiment, the second processing module 303 includes: The association submodule is used to associate the user profile data and the scene tag data based on a preset mapping algorithm to obtain the corresponding generated data; The first analysis submodule is used to perform semantic analysis on the generated data based on the large language model to obtain the corresponding semantic representation; The second analysis submodule is used to perform semantic analysis on all templates in the template library based on the large language model to obtain multiple corresponding key semantic features. A computational submodule is used to calculate the similarity between the semantic representation and each of the key semantic features; The first filtering submodule is used to filter out the specified template with the highest similarity from all the templates; The second determining submodule is used to use the specified template as the target template.

[0130] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data processing method in the aforementioned embodiments, and will not be repeated here. In some optional implementations of this embodiment, the first generation module 305 includes: The first calling submodule is used to call the preset content generation model; The first generation submodule is used to process the input data based on the target node information using the content generation model to generate the corresponding title data, summary data, and initial draft of the main text. The first integration submodule is used to integrate the title data, abstract data, and initial draft of the main text to obtain the corresponding integrated data; The third determining submodule is used to use the integrated data as the text content.

[0131] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data processing method in the aforementioned embodiments, and will not be repeated here. In some optional implementations of this embodiment, the second generation module 307 includes: The parsing submodule is used to perform semantic parsing on the target text data based on natural language processing technology to obtain the corresponding semantic parsing results; The second submodule is used to invoke a preset image generation model; The second generation submodule is used to generate image generation instructions that match the input requirements of the image generation model based on the semantic parsing results; The third generation submodule is used to process the image generation instruction based on the image generation model to generate corresponding image data; The fourth determining submodule is used to use the image data as the target image data.

[0132] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data processing method in the aforementioned embodiments, and will not be repeated here. In some optional implementations of this embodiment, the third generation submodule includes: The generation unit is used to process the image generation instruction based on the image generation model to generate corresponding initial image data; The acquisition unit is used to acquire preset style requirements; An adjustment unit is used to adjust the initial image data based on the style requirements using a preset style transfer algorithm to obtain the adjusted first image data. An optimization unit is used to perform image optimization processing on the first image data based on a preset optimization strategy to obtain the corresponding second image data; A determining unit is used to use the second image data as the image data.

[0133] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data processing method in the aforementioned embodiments, and will not be repeated here. In some optional implementations of this embodiment, the third processing module 308 includes: The second filtering submodule is used to filter out specified text content from the target text data; wherein, the specified text content is content that meets preset conditions for generating speech; The second acquisition submodule is used to acquire preset speech synthesis parameters; The third generation submodule is used to perform speech generation processing on the specified text content and the speech synthesis parameters based on a preset speech synthesis model to obtain the corresponding target speech data. The second integration submodule is used to integrate the target text data, the target image data, and the target speech data to obtain the corresponding target knowledge data. The publishing submodule is used to publish the target knowledge data to the knowledge base based on a preset data import tool.

[0134] In this embodiment, the operations performed by the above modules or units correspond one-to-one with the steps of the artificial intelligence-based data processing method in the aforementioned embodiments, and will not be repeated here. To address the aforementioned technical problems, embodiments of this application also provide a computer device. Please refer to [link / reference needed]. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0135] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected via a system bus. It should be noted that only the computer device 4 with components 41-43 is shown in the figure; however, it should be understood that it is not required to implement all the shown components, and more or fewer components can be implemented alternatively. Those skilled in the art will understand that the computer device described here is a device capable of automatically performing 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.

[0136] The computer device can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device can interact with the user via a keyboard, mouse, remote control, touchpad, or voice control.

[0137] 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 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the computer device 4. Of course, the memory 41 may also include both the internal storage unit and its external storage device of the computer device 4. In this embodiment, the memory 41 is typically used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for data processing methods based on artificial intelligence. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or will be output.

[0138] 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 typically used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or to process data, for example, to execute computer-readable instructions of the artificial intelligence-based data processing method.

[0139] The network interface 43 may include a wireless network interface or a wired network interface, which is typically used to establish communication connections between the computer device 4 and other electronic devices.

[0140] Compared with the prior art, the embodiments of this application have the following beneficial effects: In this embodiment, user profile data and scene tag data are obtained by processing relevant user data. Then, by combining large language models, knowledge images, and content generation models, the user profile data and scene tag data are processed and content is generated according to the characteristics and needs of different modalities, thus meeting diverse content generation needs. Furthermore, based on the obtained target text data and target image data, corresponding target knowledge data is generated and published to a preset knowledge base, thereby automatically and accurately realizing the generation and processing of knowledge data and effectively improving the generation efficiency of knowledge data.

[0141] This application also provides another embodiment, namely, providing a computer-readable storage medium storing computer-readable instructions that can be executed by at least one processor to cause the at least one processor to perform the steps of the artificial intelligence-based data processing method described above.

[0142] Compared with the prior art, the embodiments of this application have the following main advantages: In this embodiment, user profile data and scene tag data are obtained by processing relevant user data. Then, by combining large language models, knowledge images, and content generation models, the user profile data and scene tag data are processed and content is generated according to the characteristics and needs of different modalities, thus meeting diverse content generation needs. Furthermore, based on the obtained target text data and target image data, corresponding target knowledge data is generated and published to a preset knowledge base, thereby automatically and accurately realizing the generation and processing of knowledge data and effectively improving the generation efficiency of knowledge data.

[0143] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they 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 this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0144] Obviously, the embodiments described above are only some embodiments of this application, not all embodiments. The accompanying drawings show preferred embodiments of this application, but do not limit the patent scope of this application. This application can be implemented in many different forms; rather, the purpose of providing these embodiments is to provide a more thorough and comprehensive understanding of the disclosure of this application. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this application's specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the scope of patent protection of this application.

[0145] 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.

Claims

1. A data processing method based on artificial intelligence, characterized in that, Includes the following steps: Collect relevant data corresponding to users from multiple pre-set channels; User profile data corresponding to feature extraction processing of the relevant data, and scene tag data corresponding to scene analysis of the relevant data; Based on a preset large language model, semantic analysis is performed on the user profile data and the scene tag data to obtain corresponding semantic analysis results, and target templates that match the semantic analysis results are selected from a preset template library. Obtain target node information related to the user profile data and the scene tag data from a preset knowledge graph; The system receives input data from the user and processes the target node information and the input data based on a preset content generation model to generate corresponding text content. The text content is filled into the target template to obtain the corresponding target text data; Based on the target text data, image generation processing is performed to obtain the corresponding target image data; Based on the target text data and the target image data, corresponding target knowledge data is generated, and the target knowledge data is published to a preset knowledge base.

2. The data processing method based on artificial intelligence according to claim 1, characterized in that, The step of performing feature extraction processing on the relevant data to obtain the corresponding user profile data specifically includes: Obtain the preset preprocessing strategy; The relevant data is processed based on the preprocessing strategy to obtain the corresponding processed data; Based on a preset lightweight model, feature extraction is performed on the processed data to obtain the corresponding feature data; The feature data is used as the user profile data.

3. The data processing method based on artificial intelligence according to claim 1, characterized in that, The step of performing semantic analysis on the user profile data and the scene tag data based on a preset large language model to obtain corresponding semantic analysis results, and then selecting target templates that match the semantic analysis results from a preset template library, specifically includes: The user profile data and the scene tag data are associated based on a preset mapping algorithm to obtain the corresponding generated data; Based on the large language model, semantic analysis is performed on the generated data to obtain the corresponding semantic representation; Based on the large language model, semantic analysis is performed on all templates in the template library to obtain multiple corresponding key semantic features; Calculate the similarity between the semantic representation and each of the key semantic features; Select the specified template with the highest similarity from all the templates mentioned; Use the specified template as the target template.

4. The data processing method based on artificial intelligence according to claim 1, characterized in that, The step of processing the target node information and the input data based on the preset content generation model to generate corresponding text content specifically includes: Call the preset content generation model; Based on the target node information, the content generation model is used to process the input data to generate content, resulting in corresponding title data, summary data, and a first draft of the main text. The title data, abstract data, and initial draft of the main text are integrated to obtain the corresponding integrated data. The integrated data is used as the text content.

5. The data processing method based on artificial intelligence according to claim 1, characterized in that, The step of performing image generation processing based on the target text data to obtain the corresponding target image data specifically includes: The target text data is semantically parsed using natural language processing technology to obtain the corresponding semantic parsing results. Call the preset image generation model; Based on the semantic parsing results, an image generation instruction matching the input requirements of the image generation model is generated; The image generation instruction is processed based on the image generation model to generate corresponding image data; The image data is used as the target image data.

6. The data processing method based on artificial intelligence according to claim 5, characterized in that, The step of processing the image generation instruction based on the image generation model to generate corresponding image data specifically includes: The image generation instruction is processed based on the image generation model to generate the corresponding initial image data; Obtain the preset style requirements; Based on the style requirements, the initial image data is adjusted using a preset style transfer algorithm to obtain the adjusted first image data; The first image data is processed by a preset optimization strategy to obtain the corresponding second image data. The second image data is used as the image data.

7. The data processing method based on artificial intelligence according to claim 1, characterized in that, The step of generating corresponding target knowledge data based on the target text data and the target image data, and publishing the target knowledge data to a preset knowledge base, specifically includes: Select specified text content from the target text data; wherein, the specified text content is content that meets preset conditions for generating speech; Obtain preset speech synthesis parameters; Based on a preset speech synthesis model, the specified text content and the speech synthesis parameters are processed to generate speech data, thereby obtaining the corresponding target speech data. The target text data, the target image data, and the target speech data are integrated and processed to obtain the corresponding target knowledge data; The target knowledge data is published to the knowledge base using a preset data import tool.

8. A data processing device based on artificial intelligence, characterized in that, include: The data collection module is used to collect relevant data corresponding to users from multiple preset channels; The first processing module is used to perform feature extraction processing on the relevant data to obtain corresponding user profile data, and to perform scene analysis on the relevant data to obtain corresponding scene tag data. The second processing module is used to perform semantic analysis on the user profile data and the scene tag data based on a preset large language model to obtain the corresponding semantic analysis results, and to filter out the target template that matches the semantic analysis results from a preset template library. The acquisition module is used to acquire target node information related to the user profile data and the scene tag data from a preset knowledge graph; The first generation module is used to receive the user's input data and process the target node information and the input data based on a preset content generation model to generate corresponding text content. The fill module is used to fill the text content into the target template to obtain the corresponding target text data; The second generation module is used to perform image generation processing based on the target text data to obtain the corresponding target image data; The third processing module is used to generate corresponding target knowledge data based on the target text data and the target image data, and to publish the target knowledge data to a preset knowledge base.

9. A computer device, characterized in that, It includes a memory and a processor, wherein the memory stores computer-readable instructions, and the processor executes the computer-readable instructions to implement the steps of the data processing method based on artificial intelligence as described in any one of claims 1 to 7.

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 data processing method based on artificial intelligence as described in any one of claims 1 to 7.

Images