Icon generation method, system and equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, and particularly provides an icon generation method, system and device and a storage medium, and the method comprises the steps: obtaining an icon description language, and extracting keywords from the icon description language, the keywords comprising a visual form, a function attribute, a state, an expression technique and a style; generating an icon instruction based on the keyword by utilizing a large model, wherein the icon instruction is used for limiting the attribute of an icon; calling an icon generation tool to generate a corresponding icon based on the icon instruction; an icon score fed back by the user is obtained, the icon and the icon score are fed back to a data set, and the data set is used for training the large model. According to the invention, automatic design and generation of icons are realized.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology, and specifically relates to an icon generation method, system, device and storage medium. Background Art

[0002] With the advent of the big data era and the rapid development of information visualization technology, visualization screens play a critical role in various fields. Whether used for real-time monitoring and decision-making in the command center or for information display and publicity in the exhibition center, visualization screens need to have clear, intuitive, and beautiful interfaces. As a core component of visualization screen interface design, the quality of UI icon design directly affects users' understanding of information and usage experience.

[0003] Traditional visual large-screen UI icon design mainly relies on manual design by professional designers. Designers need to rely on their own experience and expertise to go through a long process from the conception of design concepts, sketching, color matching to the perfection of the final icon. This method not only requires designers to have deep professional design skills, but also is often inefficient when faced with complex design requirements and tight time constraints. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention provides an icon generation method, system, device and storage medium to solve the above-mentioned technical problems.

[0005] In a first aspect, the present invention provides an icon generation method, comprising: Obtaining an icon description language, and extracting keywords from the icon description language, wherein the keywords include visual form, functional attributes, status, expression method and style; Generate icon instructions based on the keywords using the big model, wherein the icon instructions are used to define the attributes of the icon; Calling an icon generation tool to generate a corresponding icon based on the icon instruction; The icon ratings fed back by users are obtained, and the icons and icon ratings are fed back into a data set, where the data set is used to train the large model.

[0006] In an optional embodiment, the method further comprises: Construct icon generation instruction set, zero-shot recommendation instruction set and few-shot recommendation instruction set; Given a set of instructions , the large model As input, Assume that the length of the entire instruction is L, and maximize the conditional probability as the training objective function: ; in, is the large model parameter, Generate characters for large models The conditional probability of To generate characters The task input, To generate characters The task output; Use the icon generation instruction set to fine-tune the parameters of the large model and convert the icons in the dataset into icon description information; Use the zero-shot recommendation instruction set and the few-shot recommendation instruction set to fine-tune the parameters of the large model so that the large model can learn the mapping relationship between icon description information and attributes; Add a trainable prefix parameter to the Transformer layer of the large model to transform the input and output of the large model into , where P is the prefix parameter. When fine-tuning the parameters of the large model, only P is used for parameter update.

[0007] In an optional implementation, obtaining an icon description language and extracting keywords from the icon description language includes: Obtaining an icon generation request, and extracting encrypted data from the icon generation request; Decrypting the encrypted data using a local secret key to obtain an icon description language; A large model is used to extract keywords from the icon description language.

[0008] In an optional implementation, obtaining icon ratings from user feedback, and feeding the icons and icon ratings back to a data set, wherein the data set is used to train the large model, includes: A pop-up window is displayed while the icon is being generated, and feedback is obtained through the pop-up window; assigning tags to the icons based on keywords corresponding to the icons; Performing contextual semantic analysis on the feedback content to obtain sentiment classification, sentiment intensity and usage scenario, and labeling the icon with a feedback label based on the semantic analysis result; The labeled icons are saved to the data set, and the data set is cleaned and updated to delete invalid data in the data set; The sentiment classification and sentiment intensity of user feedback content within a specified period are counted, and the cumulative value of the product of the sentiment classification and the sentiment intensity is calculated. If the cumulative value is lower than a set threshold, the data set is used to train the large model.

[0009] In a second aspect, the present invention provides an icon generation system, comprising: A description acquisition module, used to acquire an icon description language and extract keywords from the icon description language, wherein the keywords include visual form, functional attributes, status, expression method and style; An instruction generation module, used to generate icon instructions based on the keywords using the large model, wherein the icon instructions are used to define the attributes of the icon; An icon generation module, used for calling an icon generation tool to generate a corresponding icon based on the icon instruction; The feedback processing module is used to obtain the icon ratings fed back by users and feed the icons and icon ratings back to a data set, where the data set is used to train the large model.

[0010] In an optional embodiment, the system further includes: A collection construction module, used to construct an icon generation instruction set, a zero-shot recommendation instruction set, and a few-shot recommendation instruction set; Function determination module for a given set of instructions , the large model As input, Assume that the length of the entire instruction is L, and maximize the conditional probability as the training objective function: , in, is the large model parameter, Generate characters for large models The conditional probability of To generate characters The task input, To generate characters The task output; A first fine-tuning module, used to fine-tune the parameters of the large model using the icon generation instruction set, and convert the icons in the data set into icon description information; The second fine-tuning module is used to fine-tune the parameters of the large model using the zero-sample recommendation instruction set and the few-sample recommendation instruction set, so that the large model learns the mapping relationship between the icon description information and the attributes; The front-end setting module is used to add trainable prefix parameters to the Transformer layer of the large model, transforming the input and output of the large model into , where P is the prefix parameter. When fine-tuning the parameters of the large model, only P is used for parameter update.

[0011] In an optional implementation, the description acquisition module includes: A request acquisition unit, used for acquiring an icon generation request and extracting encrypted data from the icon generation request; A data decryption unit, used to decrypt the encrypted data using a local key to obtain an icon description language; The feature extraction unit is used to extract keywords from the icon description language using a large model.

[0012] In an optional implementation, the feedback processing module includes: A feedback collection unit, used to pop up a window while generating an icon, and obtain feedback content through the pop-up window; A label assigning unit, configured to assign a label to the icon based on a keyword corresponding to the icon; A sentiment labeling unit, used to perform contextual semantic analysis on the feedback content to obtain sentiment classification, sentiment intensity and usage scenario, and label the icon with a feedback label based on the semantic analysis result; A data cleaning unit, used to save the labeled icons to a data set, and to clean and update the data set to delete invalid data in the data set; The training trigger unit is used to count the sentiment classification and sentiment intensity of user feedback content within a specified period, calculate the cumulative value of the product of sentiment classification and sentiment intensity, and use the data set to train the large model if the cumulative value is lower than a set threshold.

[0013] In a third aspect, a device is provided, including: A memory, used for storing an icon generation program; A processor is used to implement the steps of the icon generation method provided in the first aspect when executing the icon generation program.

[0014] In a fourth aspect, a computer-readable storage medium is provided, on which an icon generation program is stored. When the icon generation program is executed by a processor, the steps of the icon generation method provided in the first aspect are implemented.

[0015] The beneficial effect of the present invention is that the icon generation method, system, device and storage medium provided by the present invention use a large model to learn keywords and icon elements by setting a keyword range, and then generate corresponding icon instructions based on the icon description language given by the user, and guide the icon generation tool to generate corresponding icons through the icon instructions. Automated design and generation of icons is achieved. In addition, a user feedback mechanism is added to timely correct the large model parameters to improve user experience.

[0016] In addition, the invention has a reliable design principle, a simple structure and a very broad application prospect. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention.

[0019] Figure 2 It is a schematic effect diagram of a method according to an embodiment of the present invention.

[0020] Figure 3 is a schematic block diagram of a system according to an embodiment of the present invention.

[0021] Figure 4 A schematic diagram of the structure of a device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0024] The key terms appearing in the present invention are explained below.

[0025] Large models refer to machine learning models with large-scale parameters and complex computational structures. These models are usually built with deep neural networks and have billions or even hundreds of billions of parameters. Large models are designed to improve the expressiveness and predictive performance of the model and to handle more complex tasks and data. Large models are widely used in various fields, including natural language processing, computer vision, speech recognition, and recommendation systems. Large models learn complex patterns and features by training massive amounts of data, have stronger generalization capabilities, and can make accurate predictions on unseen data.

[0026] The icon generation method provided in the embodiment of the present invention is executed by a computer device, and accordingly, the icon generation system runs in the computer device.

[0027] Figure 1 is a schematic flow chart of a method according to an embodiment of the present invention. Figure 1 The execution subject may be an icon generation system. According to different requirements, the order of the steps in the flow chart may be changed, and some may be omitted.

[0028] like Figure 1 As shown, the method includes: S1. Obtain an icon description language, and extract keywords from the icon description language, wherein the keywords include visual form, functional attributes, status, expression method and style.

[0029] Use natural language processing technology (such as word segmentation, part-of-speech tagging, named entity recognition, etc.) to extract keywords from the icon description language. Establish a keyword library and classify and organize the extracted keywords.

[0030] From the icon description language, we need to extract keywords that can represent the core features of the icon. These keywords can be divided into several categories: Visual form: describes the appearance characteristics of an icon, such as shape, color, line thickness, etc.

[0031] Functional attributes: Indicates the function or purpose represented by the icon, such as "file transfer", "deletion", "fault prompt", etc.

[0032] Status: static icon or dynamic icon. Static icons, that is, the visual form does not change over time, are the more common icon types. Dynamic icons are usually composed of continuous graphics and presented in the form of rhythmic animation.

[0033] Expression technique: refers to the design technique used in the icon, such as "flat", "three-dimensional", "hand-painted", etc.

[0034] Style: Describes the overall style of the icon, such as "simple", "cartoon", "retro", etc.

[0035] S2. Generate icon instructions based on the keywords using the large model, wherein the icon instructions are used to define the attributes of the icon.

[0036] Use deep learning techniques (such as Transformer, BERT, etc.) to build a large model that can understand and generate icon instructions. Take the extracted keywords as input and use the large model to generate icon instructions with clear attribute definitions. These instructions should be able to clearly describe the icon features required by users.

[0037] When training a large model, a large amount of data pairs containing icon descriptions and corresponding instructions are used for training. When generating icon instructions, keywords are used as input, and instructions that meet the requirements are generated through the reasoning ability of the large model.

[0038] S3. Calling an icon generation tool to generate a corresponding icon based on the icon instruction.

[0039] Select appropriate icon generation tools according to actual needs. These tools can be rule-based generators, machine learning models, or deep learning models, etc. Input the generated icon instructions into the icon generation tool, and the tool generates the corresponding icon according to the instructions.

[0040] S4. Obtain icon ratings from user feedback, and feed the icons and icon ratings back to a data set, where the data set is used to train the large model.

[0041] The generated icons are displayed to users, and their ratings and feedback are collected. The user-rated icons and corresponding rating data are fed back into the dataset used to train the large model.

[0042] In an embodiment of the present invention, based on step S1, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0043] In the processing flow of the icon generation system, first, the system needs to obtain an icon generation request from the user or application. This request usually contains rich information to guide the system on how to generate the icon required by the user. Among them, the encrypted data contained in the request is particularly critical. This data is encrypted by the user or application through a certain encryption algorithm to ensure information security. They may contain key descriptive information such as the icon style, color, shape, etc., and are the core basis for generating icons.

[0044] Once the system receives the icon generation request, it will immediately start processing the encrypted data. To this end, one or more local keys are pre-stored in the system. These keys are the only keys to decrypt the encrypted data and have a high degree of confidentiality and security. The system will use these local keys to decrypt the encrypted data through the corresponding decryption algorithm. After the decryption is successful, the originally encrypted icon description language can be restored and become plain text information that the system can understand and process.

[0045] Next, the system needs to use a large model to process these icon description languages. Large models usually have powerful natural language processing and knowledge reasoning capabilities, and can accurately understand and analyze text information. In this step, the large model will deeply parse the icon description language and extract keywords from it.

[0046] Keywords include visual form, functional attributes, status, expression method and style. These keywords can be divided into several categories: Visual form: describes the appearance characteristics of an icon, such as shape, color, line thickness, etc.

[0047] Functional attributes: Indicates the function or purpose represented by the icon, such as "file transfer", "deletion", "fault prompt", etc.

[0048] Status: static icon or dynamic icon. Static icons, that is, the visual form does not change over time, are the more common icon types. Dynamic icons are usually composed of continuous graphics and presented in the form of rhythmic animation.

[0049] Expression technique: refers to the design technique used in the icon, such as "flat", "three-dimensional", "hand-painted", etc.

[0050] Style: Describes the overall style of the icon, such as "simple", "cartoon", "retro", etc.

[0051] In an embodiment of the present invention, based on step S2, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0052] The present invention uses the ChatGPT-4 model [1][2] to generate icon instructions based on keywords. The ChatGPT-4 model is used to parse the user's natural language input and extract key information, such as the specific requirements, visual style, and functional requirements of the icon. ChatGPT-4 analyzes the description: First, in the industrial production monitoring scenario, equipment failure is usually related to elements such as shutdown and fault warning. The simple and modern style means that the icon elements should be simple and avoid overly complex designs. Based on the red and white color tones, red can be used to indicate danger and warning, and white can be used to highlight key points or as a background color. Based on these analyses, ChatGPT-4 outputs structured design instructions. For example, the icon shape can be a simple circle with a red cross inside to indicate a fault. The lines of the cross should be simple and smooth, the color should be pure red, and the background of the circle should be white.

[0053] During the image feature evaluation phase, if the icon is not recognizable enough, such as the red cross is not prominent enough against the white circular background, the outline of the fault element (red cross) is enhanced, such as by thickening the line or increasing the contrast between red and white, so that the icon is more eye-catching on the visualization screen. Throughout the design process, from user input description to final icon output, each data transmission link is encrypted using an encryption algorithm.

[0054] On the basis of the above embodiments, in order to further improve the reliability of the data classification results provided by the above embodiments, as an implementable method, the following training method is adopted to reduce the requirement on the number of samples in the training data set and improve the training efficiency.

[0055] (1) Constructing icon generation instruction sets, zero-shot recommendation instruction sets, and few-shot recommendation instruction sets; Icon generation instruction set: Purpose: Convert an icon to icon description information.

[0056] Example directive: Directive 1: Convert the provided icon image into concise, descriptive text.

[0057] Instruction 2: Please generate description information based on the icon, including color, shape and main elements.

[0058] Instruction 3: Icon description generation task: Please provide a text description for the given icon.

[0059] Input: Icon image. Output: Icon description information (e.g. "A blue circular icon with a white arrow pointing to the right.").

[0060] Zero-shot recommended instruction set: Purpose: To recommend related attributes or categories based on icon description information without any additional samples.

[0061] Example directive: Instruction 1: Recommend the category to which the icon belongs based on its description information.

[0062] Instruction 2: Based on the description of the following icon, please predict its possible function or use.

[0063] Instruction 3: Icon attribute recommendation: Assign the most appropriate attribute label to the icon based on the description information.

[0064] Input: Icon description information. Output: Attributes or categories (eg: "This is a file transfer icon, so its category is 'File Operations'.").

[0065] Few sample recommended instruction sets: Purpose: Given a small number of samples, learn the mapping relationship between icon description information and attributes and make recommendations.

[0066] Example directive: Instruction 1: Please refer to the following examples to recommend related attributes based on the icon description information.

[0067] Instruction 2: Given the correspondence between several icon descriptions and attributes, assign appropriate attributes to the new icon description.

[0068] Instruction 3: Few-shot attribute mapping: Predict the attributes of a new icon description based on a few training samples.

[0069] Input: The correspondence between a small amount of icon description information and attributes + new icon description information. Output: Attributes (e.g., "Based on the training samples, the most likely attribute of this icon description is 'notification'.").

[0070] (2) Given a set of instructions , the large model As input, Assume that the length of the entire instruction is L, and maximize the conditional probability as the training objective function: ; in, is the large model parameter, Generate characters for large models The conditional probability of To generate characters The task input, To generate characters The task output; (3) Use the icon generation instruction set to fine-tune the parameters of the large model and convert the icons in the dataset into icon description information; Task: Convert icon images into icon description information.

[0071] Methods: We train a large model to generate descriptive text from icon images by fine-tuning it using an icon generation instruction set.

[0072] Objective: Maximize the conditional probability, where is the large model parameter, is the icon image, and is the generated icon description information.

[0073] (4) Use the zero-shot recommendation instruction set and the few-shot recommendation instruction set to fine-tune the parameters of the large model so that the large model can learn the mapping relationship between icon description information and attributes; Task: Learn the mapping relationship between icon description information and attributes.

[0074] method: Zero-shot recommendation: directly use icon description information as input to train large models to predict attributes or categories.

[0075] Few-sample recommendation: Given a small number of training samples, a large model is trained to learn the mapping relationship and applied to new icon description information.

[0076] Goal: To enable large models to accurately map icon description information to related attributes or categories.

[0077] (5) Add a trainable prefix parameter to the Transformer layer of the large model to transform the input and output of the large model into , where P is the prefix parameter. When fine-tuning the parameters of the large model, only P is used for parameter update.

[0078] Add trainable prefix parameters to the Transformer layer of the large model so that only these prefix parameters are updated during fine-tuning, thereby improving training efficiency and model generalization ability.

[0079] Add a prefix parameter P to the input and output of the Transformer layer.

[0080] During fine-tuning, only the prefix parameter P is updated while keeping other parameters of the large model unchanged.

[0081] Initialize the prefix parameter P to a random value or a specific value.

[0082] During the training process, the value of the prefix parameter P is updated through the back-propagation algorithm.

[0083] By continuously optimizing this training objective function, the large model can achieve better performance in icon generation and attribute recommendation tasks.

[0084] In an embodiment of the present invention, based on step S3, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0085] For example, the icon instruction is to generate an icon with a circular shape, a blue background, a white border, and a simple style. In this case, this instruction is expressed in a structured way, such as: { "shape": "circle", "backgroundColor": "blue", "borderColor": "white", "style": "simple"}.

[0086] Input it into the icon generation tool. The tool will use its internal deep learning model to generate the corresponding icon based on the information in the instruction. In this process, it is necessary to ensure that the icon generation tool can understand and correctly parse the icon instruction we provide. In this embodiment, the icon generation tool uses Iconify.

[0087] The final generated icon is as follows Figure 3 shown.

[0088] In an embodiment of the present invention, based on step S4, a possible embodiment is given below to illustrate its specific implementation scheme in a non-limiting manner.

[0089] S401. A window pops up while the icon is generated, and feedback content is obtained through the pop-up window.

[0090] Implementation steps: Icon generation and window triggering: When the icon generation task is completed, a front-end event is triggered to display a modal dialog box (Modal) or a pop-up window (Popup) to collect user feedback.

[0091] Feedback collection: The pop-up window contains a text input box, a star rating system (to quantify the intensity of sentiment), and an optional drop-down menu or checkbox (to collect usage scenario information). Users can use these elements to provide specific feedback on the icon.

[0092] Submission mechanism: After the user completes the feedback, click the "Submit" button, and the front-end will send the feedback data (including text, ratings, and usage scenarios) to the back-end server for storage and processing.

[0093] S402. Assign a label to the icon based on the keyword corresponding to the icon.

[0094] Implementation steps: Keyword extraction: During the icon generation process, the system will record the keywords or themes used to generate the icon.

[0095] Tag assignment: The backend server selects the most appropriate tag from the predefined tag library and assigns it to the icon based on these keywords. The tag library can be a fixed set based on business logic or dynamically updated.

[0096] Tag storage: The assigned tags are stored in the database in association with the icon ID.

[0097] S403. Perform contextual semantic analysis on the feedback content to obtain sentiment classification, sentiment intensity and usage scenario, and annotate the icon with a feedback label based on the semantic analysis result.

[0098] Implementation steps: Text preprocessing: Preprocess the collected feedback text, including removing stop words, punctuation marks, stemming, etc.

[0099] Sentiment analysis: Use sentiment analysis models (such as BERT, RoBERTa and other pre-trained models) to classify the sentiment (positive, negative, neutral) and score the sentiment intensity (such as 1-5 stars) of the pre-processed text.

[0100] Usage scenario identification: Identify the usage scenarios mentioned in the feedback through keyword matching or semantic understanding models.

[0101] Feedback label annotation: According to the sentiment classification, sentiment intensity and usage scenario, the icons are labeled with corresponding feedback labels, such as "positive review-high satisfaction-business use".

[0102] S404. Save the labeled icons to the data set, and clean and update the data set to delete invalid data in the data set.

[0103] Implementation steps: Data storage: Save the annotated icon information (including the icon itself, labels, feedback labels, etc.) to a dataset. The dataset can be a relational database, NoSQL database, or file storage system.

[0104] Data cleaning: Run data cleaning scripts regularly to check whether there are duplicates, missing values, or outliers in the data set. Delete or correct invalid data, such as empty feedback, data in non-standard formats, etc.

[0105] Data Updates: As new icons are generated and user feedback is collected, the dataset is continuously updated to ensure that it contains the latest and most comprehensive information.

[0106] S405. Count the sentiment classification and sentiment intensity of user feedback content within a specified period, calculate the cumulative value of the product of sentiment classification and sentiment intensity, and use the data set to train the large model if the cumulative value is lower than a set threshold.

[0107] Implementation steps: Data aggregation: Aggregate user feedback data from the dataset based on a specified period (e.g., weekly, monthly), and calculate the product of the sentiment classification and sentiment intensity of each icon.

[0108] Cumulative value calculation: add up the products of all icons to get the cumulative value.

[0109] Threshold comparison: Compare the accumulated value with the preset threshold. The threshold can be set based on business needs, historical data or expert advice.

[0110] Model training: If the accumulated value is lower than the threshold, it indicates that the current model may not meet user needs well. At this time, the model is retrained or fine-tuned using the latest dataset to improve the ability of icon generation and feedback processing.

[0111] Model evaluation and iteration: After training is completed, the new model is evaluated and necessary iterations and optimizations are performed based on the evaluation results.

[0112] Through the above detailed implementation plan, it can be ensured that the icon generation system can continuously collect user feedback, optimize the icon generation model based on the feedback, and improve user experience and satisfaction.

[0113] In some embodiments, the icon generation system may include a plurality of functional modules composed of computer program segments. The computer programs of the various program segments in the icon generation system may be stored in a memory of a computer device and executed by at least one processor to perform (see Figure 1 Description) icon generation function.

[0114] In this embodiment, the icon generation system can be divided into multiple functional modules according to the functions it performs, such as Figure 3 As shown. The functional modules of the system 300 may include: a description acquisition module 310, an instruction generation module 320, an icon generation module 330 and a feedback processing module 340. The module referred to in the present invention refers to a series of computer program segments that can be executed by at least one processor and can complete fixed functions, which are stored in a memory. In this embodiment, the functions of each module will be described in detail in subsequent embodiments.

[0115] A description acquisition module, used to acquire an icon description language and extract keywords from the icon description language, wherein the keywords include visual form, functional attributes, status, expression method and style; An instruction generation module, used to generate icon instructions based on the keywords using the large model, wherein the icon instructions are used to define the attributes of the icon; An icon generation module, used for calling an icon generation tool to generate a corresponding icon based on the icon instruction; The feedback processing module is used to obtain the icon ratings fed back by users and feed the icons and icon ratings back to a data set, where the data set is used to train the large model.

[0116] Optionally, as an embodiment of the present invention, the system further includes: A collection construction module, used to construct an icon generation instruction set, a zero-shot recommendation instruction set, and a few-shot recommendation instruction set; Function determination module for a given set of instructions , the large model As input, Assume that the length of the entire instruction is L, and maximize the conditional probability as the training objective function: , in, is the large model parameter, Generate characters for large models The conditional probability of To generate characters The task input, To generate characters The task output; A first fine-tuning module, used to fine-tune the parameters of the large model using the icon generation instruction set, and convert the icons in the data set into icon description information; The second fine-tuning module is used to fine-tune the parameters of the large model using the zero-sample recommendation instruction set and the few-sample recommendation instruction set, so that the large model learns the mapping relationship between the icon description information and the attributes; The front-end setting module is used to add trainable prefix parameters to the Transformer layer of the large model, transforming the input and output of the large model into , where P is the prefix parameter. When fine-tuning the parameters of the large model, only P is used for parameter update.

[0117] Optionally, as an embodiment of the present invention, the description acquisition module includes: A request acquisition unit, used for acquiring an icon generation request and extracting encrypted data from the icon generation request; A data decryption unit, used to decrypt the encrypted data using a local key to obtain an icon description language; The feature extraction unit is used to extract keywords from the icon description language using a large model.

[0118] Optionally, as an embodiment of the present invention, the feedback processing module includes: A feedback collection unit, used to pop up a window while generating an icon, and obtain feedback content through the pop-up window; A label assigning unit, configured to assign a label to the icon based on a keyword corresponding to the icon; A sentiment labeling unit, used to perform contextual semantic analysis on the feedback content to obtain sentiment classification, sentiment intensity and usage scenario, and label the icon with a feedback label based on the semantic analysis result; A data cleaning unit, used to save the labeled icons to a data set, and to clean and update the data set to delete invalid data in the data set; The training trigger unit is used to count the sentiment classification and sentiment intensity of user feedback content within a specified period, calculate the cumulative value of the product of sentiment classification and sentiment intensity, and use the data set to train the large model if the cumulative value is lower than a set threshold.

[0119] Figure 4The icon generation method provided for the embodiment of the present application can be applied to the device. It will be appreciated by those skilled in the art that the device structure involved in the embodiment of the present invention does not constitute a limitation on the device, and the device may include more or less components than shown, or combine certain components, or arrange different components. In an embodiment of the present invention, the device includes but is not limited to a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or required herein.

[0120] The device 400 may include: a processor 410, a memory 420 and a communication unit 430. These components communicate via one or more buses. Those skilled in the art will appreciate that the server structure shown in the figure does not limit the present invention, and it may be a bus structure or a star structure, and may include more or fewer components than shown in the figure, or combine certain components, or arrange components differently.

[0121] The memory 420 may be used to store the execution instructions of the processor 410, and the memory 420 may be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. When the execution instructions in the memory 420 are executed by the processor 410, the device 400 is enabled to perform some or all of the steps in the following method embodiments.

[0122] The processor 410 is the control center of the storage device, and uses various interfaces and lines to connect various parts of the entire electronic device. It runs or executes software programs and / or modules stored in the memory 420, and calls data stored in the memory to perform various functions of the electronic device and / or process data. The processor can be composed of an integrated circuit (IC), for example, it can be composed of a single packaged IC, or it can be composed of a plurality of packaged ICs with the same or different functions. For example, the processor 410 can include only a central processing unit (CPU). In an embodiment of the present invention, the CPU can be a single computing core or multiple computing cores.

[0123] The communication unit 430 is used to establish a communication channel so that the storage device can communicate with other devices, receive user data sent by other devices or send user data to other devices.

[0124] The present invention also provides a computer storage medium, wherein the computer storage medium may store a program, and when the program is executed, the program may include some or all of the steps in each embodiment provided by the present invention. The storage medium may be a magnetic disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM).

[0125] Those skilled in the art can clearly understand that the technology in the embodiments of the present invention can be implemented by means of software plus a necessary general hardware platform. Based on this understanding, the technical solution in the embodiments of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes, including several instructions for enabling a computer device (which can be a personal computer, a server, or a second device, a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention.

[0126] In this specification, the same or similar parts between the various embodiments can be referred to each other. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.

[0127] In the several embodiments provided by the present invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of systems or modules, which can be electrical, mechanical or other forms.

[0128] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0129] In addition, each functional module in each embodiment of the present invention may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0130] Although the present invention has been described in detail with reference to the accompanying drawings and in combination with the preferred embodiments, the present invention is not limited thereto. Without departing from the spirit and essence of the present invention, a person of ordinary skill in the art may make various equivalent modifications or substitutions to the embodiments of the present invention, and these modifications or substitutions shall be within the scope of the present invention. Any person of ordinary skill in the art may easily think of changes or substitutions within the technical scope disclosed by the present invention, and these shall be within the scope of protection of the present invention.

Claims

1. A method for generating an icon, characterized in that: include: Obtaining an icon description language, and extracting keywords from the icon description language, wherein the keywords include visual form, functional attributes, status, expression method and style; Generate icon instructions based on the keywords using the big model, wherein the icon instructions are used to define the attributes of the icon; Calling an icon generation tool to generate a corresponding icon based on the icon instruction; The icon ratings fed back by users are obtained, and the icons and icon ratings are fed back into a data set, where the data set is used to train the large model.

2. The method according to claim 1, characterized in that: The method further comprises: Construct icon generation instruction set, zero-shot recommendation instruction set and few-shot recommendation instruction set; Given a set of instructions , the large model As input, Assume that the length of the entire instruction is L, and maximize the conditional probability as the training objective function: ； in, is the large model parameter, Generate characters for large models The conditional probability of To generate characters The task input, To generate characters The task output; Use the icon generation instruction set to fine-tune the parameters of the large model and convert the icons in the dataset into icon description information; Use the zero-shot recommendation instruction set and the few-shot recommendation instruction set to fine-tune the parameters of the large model so that the large model can learn the mapping relationship between icon description information and attributes; Add a trainable prefix parameter to the Transformer layer of the large model to transform the input and output of the large model into , where P is the prefix parameter. When fine-tuning the parameters of the large model, only P is used for parameter update.

3. The method according to claim 1, characterized in that Obtaining an icon description language and extracting keywords from the icon description language, including: Obtaining an icon generation request, and extracting encrypted data from the icon generation request; Decrypting the encrypted data using a local secret key to obtain an icon description language; A large model is used to extract keywords from the icon description language.

4. The method according to claim 1, characterized in that: Obtaining icon ratings from user feedback, and feeding the icons and icon ratings back to a data set, wherein the data set is used to train the large model, including: A pop-up window is displayed while the icon is being generated, and feedback is obtained through the pop-up window; assigning tags to the icons based on keywords corresponding to the icons; Performing contextual semantic analysis on the feedback content to obtain sentiment classification, sentiment intensity and usage scenario, and labeling the icon with a feedback label based on the semantic analysis result; The labeled icons are saved to the data set, and the data set is cleaned and updated to delete invalid data in the data set; The sentiment classification and sentiment intensity of user feedback content within a specified period are counted, and the cumulative value of the product of the sentiment classification and the sentiment intensity is calculated. If the cumulative value is lower than a set threshold, the data set is used to train the large model.

5. An icon generation system, characterized in that: include: A description acquisition module, used to acquire an icon description language and extract keywords from the icon description language, wherein the keywords include visual form, functional attributes, status, expression method and style; An instruction generation module, used to generate icon instructions based on the keywords using the large model, wherein the icon instructions are used to define the attributes of the icon; An icon generation module, used for calling an icon generation tool to generate a corresponding icon based on the icon instruction; The feedback processing module is used to obtain the icon ratings fed back by users and feed the icons and icon ratings back to a data set, where the data set is used to train the large model.

6. The system according to claim 5, characterized in that The system further comprises: A collection construction module, used to construct an icon generation instruction set, a zero-shot recommendation instruction set, and a few-shot recommendation instruction set; Function determination module for a given set of instructions , the large model As input, Assume that the length of the entire instruction is L, and maximize the conditional probability as the training objective function: ， in, is the large model parameter, Generate characters for large models The conditional probability of To generate characters The task input, To generate characters The task output; A first fine-tuning module, used to fine-tune the parameters of the large model using the icon generation instruction set, and convert the icons in the data set into icon description information; The second fine-tuning module is used to fine-tune the parameters of the large model using the zero-sample recommendation instruction set and the few-sample recommendation instruction set, so that the large model learns the mapping relationship between the icon description information and the attributes; The front-end setting module is used to add trainable prefix parameters to the Transformer layer of the large model, transforming the input and output of the large model into , where P is the prefix parameter. When fine-tuning the parameters of the large model, only P is used for parameter update.

7. The system according to claim 5, characterized in that The description acquisition module includes: A request acquisition unit, used for acquiring an icon generation request and extracting encrypted data from the icon generation request; A data decryption unit, used to decrypt the encrypted data using a local key to obtain an icon description language; The feature extraction unit is used to extract keywords from the icon description language using a large model.

8. The system according to claim 5, characterized in that The feedback processing module comprises: A feedback collection unit, used to pop up a window while generating an icon, and obtain feedback content through the pop-up window; A label assigning unit, configured to assign a label to the icon based on a keyword corresponding to the icon; A sentiment labeling unit, used to perform contextual semantic analysis on the feedback content to obtain sentiment classification, sentiment intensity and usage scenario, and label the icon with a feedback label based on the semantic analysis result; A data cleaning unit, used to save the labeled icons to a data set, and to clean and update the data set to delete invalid data in the data set; The training trigger unit is used to count the sentiment classification and sentiment intensity of user feedback content within a specified period, calculate the cumulative value of the product of sentiment classification and sentiment intensity, and use the data set to train the large model if the cumulative value is lower than a set threshold.

9. A device, characterized in that: include: A memory, used for storing an icon generation program; A processor, used to implement the steps of the icon generation method as described in any one of claims 1 to 4 when executing the icon generation program.

10. A computer-readable storage medium storing a computer program, characterized in that: The readable storage medium stores an icon generation program, and when the icon generation program is executed by a processor, the steps of the icon generation method according to any one of claims 1 to 4 are implemented.

Images