AI-enhanced link table recording method and system

Abstract

The invention discloses an AI-enhanced link table recording method and system. The method comprises the steps that a blank link graph is initialized; the user inputs a design idea according to the design task; a user inputs a design idea according to a given design task and converts the design idea into a design action through a large language model; adding the design action as a node into the link graph, and completing a link relationship; extracting a key design action from the link graph, and converting the key design action into visual stimulation by using an image generator; if a user generates a new design idea due to visual stimulation, inputting the design idea and repeating the generating step and the linking step; evaluating and analyzing; and repeating the inputting step to the analyzing step until the link graph meeting the requirement is generated. According to the method, a large language model and an image generation technology are integrated, the tedious process of a traditional link table recording method is innovated, and the creative excitation and dynamic analysis capability in the design process is remarkably enhanced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an AI-enhanced link notation method and system. Background Art

[0002] Getting inspiration or ideas through brainstorming is a key part of many design tasks. How to effectively promote creativity in the ideation process has become a research hotspot in recent years. Link notation, as a design thinking analysis method, originated in the mid-1990s and was first proposed by designer Gabriela Goldschmidt (Gabriela Goldschmidt. 1995. The designer as a team of one. Design studies 16, 2 (1995), 189–209.), which aims to track, visualize and analyze thinking patterns in the design process. In this method, designers are asked to verbally express their ideas while thinking, and these ideas are then recorded and formed into protocols. Researchers break down these protocols into constructed or inferred analysis units and develop a classification scheme to encode these paragraphs to study the connections between them. With the development of technology, link notation has been extended to analyze the design process of the team. For example, (Tellervo 2023.Mobile gaze tracking and an extended linkography for collaborative sketching and designing.International Journal of Technology and Design Education 33,2(2023),379–413.) used eye tracking technology to solve design problems in the collaborative process and demonstrated an effective method to use gaze data and link notation to explore solutions to design problems. Blom (Nicolaas Blom and Alfred Bogaers.2020.Using Linkography to investigate students'thinking and information use during a STEM task.International Journal of Technology and Design Education 30(2020),1–20.) combined link notation and color archives to show students' thinking process in the design process. In addition, some studies have tried to use complex mathematical tools to improve link table analysis, such as Kan and Gero (Jeff WT Kan and John S Gero. 2008. Acquiring information from linkography in protocol studies of designing. Design studies 29, 4 (2008), 315–337.) using clustering and Shannon entropy, two different strategies, to abstract information from link tables. Since 1990, the field of human-computer interaction has continued to pay more attention to creativity. Shneiderman pointed out (Shneiderman, B. 2007. Creativity support tools: accelerating discovery and innovation. Commun. ACM 50, 12 (2007), 20-32) that the development of creativity support systems is a major challenge in the field of human-computer interaction.A large number of research is devoted to providing various collaborative creation tools such as ideation (Yuyu Lin, Jiahao Guo, Yang Chen, Cheng Yao, and Fangtian Ying. 2020. It Is Your Turn: Collaborative Ideation with a Co-Creative Robot through Sketch. In Proc of SIGCHI. ​​1-14), emotional communication (Cecilia R. Aragon and Alison Williams. 2011. Collaborative Creativity: A Complex Systems Model with Distributed Afect. In Proc of SIGCHI. ​​1875-1884) and content editing (Zhao Z., Badam SK, Chandrasegaran S., Park D., Elmqvist N., Kisselburgh LG and Ramani K. 2014. skWiki: A Multimedia Sketching System for Collaborative Creativity. In Proc of SIGCHI, 1235-1244.), and there is also some work (Chaudhuri S. and Koltun V. 2010. Data-Driven Suggestions for Creativity Support in 3D Modeling. ACM Trans. Graph. 29, 6 (2010), Article No. 183) is committed to providing various suggestions in the design process to give users more inspiration.

[0003] Existing link notation methods mainly rely on manual recording and analysis, which is cumbersome, time-consuming, and prone to errors, making it difficult to meet the efficiency requirements of the actual design process. Especially as the design process deepens, the number of ideas surges, and manual management becomes more difficult, and may even result in omissions or misjudgments. In addition, existing design tools lack effective automated tools to support the classification and association of ideas, which makes it difficult to effectively sort out and integrate ideas during the design process, hindering the generation and development of creativity. In the long-term design process, it is difficult to maintain the coherence and integrity of the thinking path, which makes it easy for design ideas to be scattered and difficult to effectively guide and evaluate the design process. Summary of the invention

[0004] The purpose of the present invention is to solve the problems in the prior art.

[0005] The technical solution adopted by the present invention to solve the technical problem is: to provide an AI enhanced link notation method, comprising the following steps:

[0006] Initialization step, initializing a blank link graph;

[0007] Input step, the user enters the design idea according to the given design task;

[0008] The generation step converts design ideas into design action expressions through a large language model to generate design actions;

[0009] In the linking step, the design action is added as a node in the linking graph and the linking relationship is completed. The linking process refers to the design action linking suggestions automatically provided by the large language model according to the preset standards.

[0010] Stimulation step: extract key design actions from the link graph and convert them into visual stimuli using an image generator; if the user generates a new design idea due to the visual stimulation, the user enters the design idea and repeats the generation step and the linking step;

[0011] Analyze steps, evaluate the completion of design tasks and analyze user characteristics;

[0012] Repeat the steps, from the input step to the analysis step until a link diagram that meets the requirements is generated.

[0013] Preferably, the generating step includes the following steps:

[0014] For design ideas inputted via text, preprocessing is directly performed; for design ideas inputted via voice, voice recognition software is first called to convert the voice into text, and then preprocessing is performed; the preprocessing includes cleaning the text through regular expressions and text processing libraries, filtering stop words, and repeating and completing incomplete sentences to obtain preprocessing information;

[0015] Define a prompt word template and use a large language model to automatically extract design actions from preprocessed information, expressed as:

[0016] GC=M pt1 ((PT(x)));

[0017] Among them, GC represents the generated design action content, x represents the preprocessing information, PT represents the prompt template used to package the input content, and M pt1 Represents a large language model used to reason about design examples.

[0018] Preferably, the generating step further includes: a large language model generating new design actions according to existing design actions.

[0019] Preferably, in the linking step, the design action is presented in the form of visual nodes, and the user links the nodes in two ways; one way is that the user manually draws the link relationship between the nodes by mouse drawing; the other way is that the user selects the link relationship between the nodes based on the automatic link suggestion view provided by the large language model.

[0020] Preferably, the process of generating an automatic link suggestion view using a large language model comprises the following steps:

[0021] Identify relationships between design actions;

[0022] Determine precisely the specific types and possibilities of these relationships;

[0023] Finally, several fan graphs are designed for each node. A fan graph represents a link relationship, and a fan in the fan graph represents the possibility that a certain node has a corresponding link relationship with that node.

[0024] Preferably, in the stimulation step, the process of generating visual stimulation comprises the following steps:

[0025] Determine whether the design action is a critical design action based on the number of links;

[0026] Design actions that convert text to images using the Image Generator;

[0027] Organize all the images into a collage as a visual stimulus.

[0028] Preferably, in the analysis step, the completion of the design task is evaluated by calculating the forward chain entropy. A larger entropy value indicates a more active design process and more diverse ideas involved in the design task. The forward chain entropy H forward It is expressed as:

[0029]

[0030] Among them, p(x i ) represents the design action x i is the probability of a forward link accounting for the total links during the brainstorming process, and n is the number of design actions.

[0031] Preferably, the analysis step includes identifying the thinking patterns of different users, which specifically includes the following steps:

[0032] Nodes with more than three forward links are classified as divergent thinking actions, and nodes with more than three backward links are classified as convergent thinking actions;

[0033] Analyze and count the divergence and convergence of all design actions of different users to identify user thinking types.

[0034] Preferably, the analysis step includes identifying the degree of interaction between users, specifically calculating the link frequency between different users, expressed as:

[0035] C uv =freq(u,v);

[0036] Among them, C uv represents the link frequency of the design actions of user u and user v; C uv The larger the value, the more interactions there are between user i and user j, and the more suitable they are for collaborative design tasks.

[0037] The present invention also provides an AI-enhanced link notation system, comprising:

[0038] Initialization module, initializes a blank link graph;

[0039] Input module, where users input design ideas according to given design tasks;

[0040] The generation module converts design ideas into design action expressions through a large language model and generates design actions;

[0041] The linking module adds the design action as a node to the linking graph and completes the linking relationship. The linking process refers to the design action linking suggestions automatically provided by the large language model according to the preset standards.

[0042] The stimulation module extracts key design actions from the link graph and converts them into visual stimuli using an image generator; if the user generates a new design idea due to the visual stimulation, the user enters the design idea and repeats the generation and linking steps;

[0043] Analysis module, to evaluate the completion of design tasks and analyze user characteristics;

[0044] Repeat the module and repeat the entry step to the analysis step until a link diagram that meets the requirements is generated.

[0045] The present invention has the following beneficial effects:

[0046] (1) Automated design analysis and idea generation: By integrating a large language model, the system can automatically analyze and process design ideas, quickly extract and classify design actions, and generate new design ideas to stimulate creativity, greatly improving the efficiency and creativity of the design process.

[0047] (2) Visual link notation construction: Use visualization technologies such as D3.js to dynamically construct and display the logical relationship between design ideas, making the design process intuitive and visual, making it easier for users to understand and track the design thinking path, and promoting effective communication and collaboration among teams;

[0048] (3) Multi-level interactive exploration and dynamic analysis: It provides multiple interactive modes such as overview view and keyword view to support users' in-depth exploration from macro to micro. It monitors the innovation dynamics of the design process in real time through methods such as entropy analysis, providing a scientific basis for team optimization and creativity improvement.

[0049] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments, but the present invention is not limited to the embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 A method step diagram of an embodiment of the present invention;

[0051] Figure 2 It is a schematic diagram of a flow chart of an embodiment of the present invention;

[0052] Figure 3 This is a display interface diagram of an embodiment of the present invention;

[0053] Figure 4 A schematic diagram for generating link suggestions according to an embodiment of the present invention;

[0054] Figure 5 4 is a system structure diagram of an embodiment of the present invention. DETAILED DESCRIPTION

[0055] See also Figure 1 and Figure 2 The method step diagram and flow chart of an embodiment of the present invention are shown, which include the following steps:

[0056] S101, initialization step, initializing a blank link graph;

[0057] S102, input step, the user inputs the design idea according to the given design task;

[0058] S103, generation step, converting the design idea into design action expression through a large language model to generate the design action;

[0059] S104, linking step, adding the design action as a node to the linking graph and completing the linking relationship; the linking process refers to the design action linking suggestions automatically provided by the large language model according to the preset standards;

[0060] S105, stimulation step, extracting key design actions from the link graph and converting them into visual stimulation using an image generator; if the user generates a new design idea due to the visual stimulation, the user enters the design idea and repeats the generation step and the linking step;

[0061] S106, analysis step, evaluating the completion of the design task and analyzing user characteristics;

[0062] S107, repeat the steps, repeat the input step to the analysis step until a link diagram that meets the requirements is generated.

[0063] Specifically, the design tasks include advertising design tasks, film and television design tasks, game design tasks and thinking course design tasks.

[0064] Specifically, the link graph that meets the requirements means that all participating users reach a unified and clear idea, and also generate a corresponding link graph.

[0065] The following introduces the implementation details of the key steps by taking the design of a competitive game as an example. Figure 3 shown.

[0066] (1) Input and extraction of design ideas.

[0067] Initialize design task information ( Figure 3 1), the user inputs ideas via voice or text ( Figure 3 2) in the sentence: "Design a survival game, for example, players are on a virtual island and survive to the end by finding resources, building shelters, and competing with other players"; if the idea is input in the form of voice, the Alibaba Cloud speech recognition API is called to transcribe the user's speech into text in real time. For the text data input by the user, it is cleaned through regular expressions and text processing libraries to filter stop words and repeat and complete incomplete sentences to obtain preprocessing information. Define a prompt word template and use a large language model to automatically extract design actions from the preprocessing information. The prompt word template can be regarded as a sentence wrapper used to organize ideas. Its expression is as follows:

[0068] GC=M pt1 ((PT(x)));

[0069] Among them, GC represents the generated design action content, x represents the preprocessing information of the input idea, PT represents the prompt template used to package the input content; M pt1 It represents a large language model for reasoning about design examples. The embodiment of the present invention adopts gpt3.5-turbo. Specifically, the large language model classifies these idea information into different design actions, such as "inspiration", "question", "knowledge" and "comment". These design actions are then used to build the basic structure of the link notation.

[0070] In addition, the embodiment of the present invention also has the function of automatically generating design actions, see Figure 2In the AI ​​design action, "AI design action" is realized through "design action generation instruction". When the user clicks the "inspiration" button on the user interface, all existing design actions will be added to the preset prompt word template by default, and new design actions will be created through the large language model; the user can also manually select several design actions and let the large language model generate new design actions based on these design actions.

[0071] (2) Construction and visualization of link notation. The designed actions are rendered by D3.js and presented in the form of visual nodes on the link diagram of the user interface ( Figure 3 3), each node represents a design action, and the user can manually link two nodes by drawing with the mouse. In addition to manually building link notations, an automatic link suggestion view is also provided ( Figure 3 4) in the figure helps users reduce the burden of links. Among them, relationship types such as parallel, incremental, alternative and tangent are represented by circles, and each relationship type is assigned a unique color. Each circle supports expansion operation. When the mouse clicks on the circle, its outer circle will expand to form 10 sector-shaped areas. These areas are arranged in a clockwise direction, representing the potential relationship possibilities between the current design action and the past 10 design actions, respectively. Different shades of color are used to represent different strengths of possibility. When the mouse clicks on these sectors, the corresponding relationship is immediately displayed intuitively in the form of nodes and connecting lines. The link relationship includes: parallel, indicating that the two nodes are repeated ideas in two different contexts; incremental, indicating that the new node is based on the previous node. Addition, indicating that the new node replaces part of the content of the previous node; tangent, indicating that the two nodes belong to the same topic and form a cognitive association.

[0072] The link suggestion strategy will automatically use a large language model to analyze the classified design actions to identify the logical connections between them, see Figure 4 As shown in Figure 1, it is mainly divided into two stages: first, identifying the relationship between design actions, and second, accurately determining the specific categories and possibilities of these relationships. As a preferred method for identifying relationships, constructing prompt sentences to guide the large language model to output idea relationships, the prompt sentences are as follows: Figure 4 As shown in 41; as a preferred method, the method for determining the specific category of the relationship is: introducing the result of idea relationship recognition, constructing a prompt sentence to guide the large language model to determine the type of relationship, the prompt sentence is as follows Figure 4 As shown in Figure 42, users can intuitively view the structure and dynamics of the entire design process through the interface.

[0073] (3) Coarse-to-fine exploration interaction. The embodiments of the present invention provide node views at different levels, allowing users to grasp the structure of the entire design discussion from a macro perspective, or to delve into the details of a specific design idea. The nodes marked with links are divided into two views: an overview view and a detail view, which the user can switch using the mouse wheel. The overview view helps users fully understand the overall situation of the link mark, while the detail view displays the keywords of each design action and the proportional distribution of the types of relationships they link to. In view of the large space required for the display of the detail view, it is impossible to fully display the keywords of all nodes on a single canvas, so a sliding timeline is introduced so that users can browse the detail views of different design actions. This interactive design allows users to flexibly switch between levels according to their needs, thereby obtaining a comprehensive perspective or local details of the design process.

[0074] Keyword extraction in the detail view uses the precise mode in the Java third-party package jieba to segment the design action content, and then uses the stop word dictionary to remove stop words. Finally, each design action can get a series of candidate words. Furthermore, for the series of candidate words obtained after segmentation and stop word removal, consider their distribution in the entire design action set, calculate the TF-IDF value respectively, and then sort each word in descending order according to the TF-IDF value, and select and extract multiple words with larger TF-IDF values ​​as keywords for each design action.

[0075] (4) Creative support and inspiration. The present invention also provides a visual stimulation view ( Figure 35), automatically generate visual stimuli and new design concepts based on current design ideas and discussion content to stimulate users' innovative thinking. In the context of the link graph, based on the number of links, select key design actions from the ideation process (for example, select the first few nodes with the largest number of links), and use an image generator to convert text design actions into visual stimuli, and then organize all visual stimuli (images) into a collage. The specific method is as follows: first, the shape-aware slicing algorithm is used to divide the triangular collage area into visually pleasing sub-areas; the stimulus image (visual stimulus) is retrieved for keyword association and detected with salient areas; then the layout of the segmented collage area is optimized by analyzing the input image, with the goal of maximizing the total salient area of ​​the image; these visual stimuli are presented in a compact layout to maximize the use of limited screen space (Dong-Yi Wu, Sheng-Yi Yao, Yun-Chen Lin, Tong-Yee Lee, et al. 2023. Image Collage on Arbitrary Shape via Shape-Aware Slicing and Optimization. IEEE Transactions on Visualization and Computer Graphics (2023).).

[0076] (5) Status monitoring and dynamic analysis: Entropy analysis, thinking pattern recognition, and user connection view are used to monitor the innovation dynamics during the design process.

[0077] Entropy analysis is a method of measuring uncertainty or information content, which is used here to measure the progress of the entire brainstorming process. Specifically, two types of entropy analysis views are defined: Figure 3 6): Forward link entropy and reverse link entropy. Taking forward link entropy as an example, its calculation formula is:

[0078]

[0079] Among them, H forward (x) represents the forward link entropy, p(x i ) represents a specific design action x i The probability that the forward link of accounts for the total links in the brainstorming process, n is the number of design ideas. If a design idea x i is cited multiple times, then p(x i ) is larger; on the contrary, if a design idea is rarely cited, then p(x i ) is small. Entropy H forwardThe larger the value of (x), the more uncertain the transfer of ideas during brainstorming, which means that the ideas are more diverse; the smaller the value of entropy, the higher the certainty of idea transfer, which may mean that the ideas are more single.

[0080] Thinking Pattern Recognition View ( Figure 3 7) in the figure uses a stacked bar chart to represent the proportional distribution of thinking patterns, with each column representing a different user and each row representing the proportional distribution of thinking patterns of different users in the current conception stage. If the number of forward links of a node exceeds three, it is classified as a divergent thinking action; similarly, if the number of reverse links exceeds three, it is classified as a convergent thinking action. By analyzing and counting the divergence and convergence of all user design actions, the thinking type can be identified: users who think that divergent thinking actions are more than convergent thinking actions are divergent users; users who think that convergent thinking actions are more than divergent thinking actions are convergent users. Identifying user thinking types is conducive to optimizing team composition and improving creative efficiency: divergent thinkers are good at creating new ideas, while convergent thinkers are good at selecting the most feasible one from multiple solutions; therefore, the ideal team should contain a certain proportion of people of the two thinking types; in addition, the team composition can be dynamically adjusted according to the different stages of the project, such as increasing divergent thinkers in the concept generation stage and increasing the proportion of convergent thinkers in the execution stage.

[0081] User Contact View( Figure 3 8) in the paper uses a heat map to show the degree of interaction between different users in design actions, where the depth of color represents the intensity of interaction. By visualizing the correlation between user design actions, it helps to find out which users have close cooperative relationships, thereby promoting the generation of more valuable design ideas. This analysis is based on the correlation matrix of link notation, which can be obtained by calculating the link frequency of design actions between users:

[0082] C uv =freq(u,v);

[0083] Among them, C uv represents the link frequency of the design actions of user u and user v; C uv The larger the value, the more interactions there are between user i and user j, and the more suitable they are for collaborative design tasks. In this way, the user combination that is most likely to spark innovation can be effectively identified.

[0084] See also Figure 5 FIG. 1 is a system structure diagram of an embodiment of the present invention, including:

[0085] Initialization module 501, initializing a blank link graph;

[0086] Input module 502, the user inputs the design idea according to the given design task;

[0087] A generation module 503 converts the design idea into a design action expression through a large language model to generate a design action;

[0088] Linking module 504, adds the design action as a node to the linking graph and completes the linking relationship; the linking process refers to the design action linking suggestions automatically provided by the large language model according to the preset standards;

[0089] The stimulation module 505 extracts key design actions from the link diagram and converts them into visual stimulation using an image generator; if the user generates a new design idea due to the visual stimulation, the user enters the design idea and repeats the generation step and the linking step;

[0090] Analysis module 506, evaluating the completion of the design task and analyzing user characteristics;

[0091] Repeat module 507, repeat the input step to the analysis step until a link diagram that meets the requirements is generated.

[0092] This application not only innovates the cumbersome process of the traditional link notation method by seamlessly integrating large-scale language models and image generation technology, but also significantly enhances the creative stimulation and dynamic analysis capabilities in the design process. In the field of product design, AI-enhanced link notation technology can greatly accelerate the creative output efficiency of the design team, and can help advertising planners, film and television producers, and game developers to quickly break the limitations of thinking and stimulate a steady stream of creative inspiration when facing complex and changing design requirements, thereby optimizing the design plan and creating more innovative and attractive works. At the same time, this technology can play an important role in the fields of education and training; in design thinking courses, AI-enhanced link notation tools can not only guide students to conduct efficient brainstorming and creative design, but also analyze and feedback students' ideas in real time, greatly improving teaching effectiveness and learning experience. In addition, this technology can be used as an important part of the online learning platform, providing remote learners with convenient creative support tools and promoting the unbounded dissemination and sharing of knowledge.

[0093] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. An AI-enhanced link notation method, characterized in that: The following steps are involved: Initialization step, initializing a blank link graph; Input step, the user enters the design idea according to the given design task; The generation step converts design ideas into design action expressions through a large language model to generate design actions; Linking step: adding the design action as a node to the linking graph and completing the linking relationship; The linking process refers to the large language model to automatically provide design action linking suggestions based on preset criteria; Stimulation step: extract key design actions from the link graph and convert them into visual stimuli using an image generator; if the user generates a new design idea due to the visual stimulation, the user enters the design idea and repeats the generation step and the linking step; Analyze steps, evaluate the completion of design tasks and analyze user characteristics; Repeat the steps, from the input step to the analysis step until a link diagram that meets the requirements is generated.

2. The AI ​​enhanced link notation method according to claim 1, characterized in that: The generating step includes the following steps: For design ideas inputted via text, preprocessing is directly performed; for design ideas inputted via voice, voice recognition software is first called to convert the voice into text, and then preprocessing is performed; the preprocessing includes cleaning the text through regular expressions and text processing libraries, filtering stop words, and repeating and completing incomplete sentences to obtain preprocessing information; Define a prompt word template and use a large language model to automatically extract design actions from preprocessed information, expressed as: GC=M pt1 ((PT(x))); Among them, GC represents the generated design action content, x represents the preprocessing information, PT represents the prompt template used to package the input content, and M pt1 Represents a large language model used to reason about design examples.

3. The AI ​​enhanced link notation method according to claim 1, characterized in that: The generation step also includes: a large language model generates new design actions based on existing design actions.

4. The AI ​​enhanced link notation method according to claim 1, characterized in that: In the linking step, the design actions are presented in the form of visual nodes, and the user links the nodes in two ways; One way is that the user manually draws the link relationship between nodes by mouse drawing; Another way is that users select link relationships between nodes based on the automatic link suggestion view provided by a large language model.

5. The AI ​​enhanced link notation method according to claim 4, characterized in that: The process of generating automatic link suggestion views from a large language model includes the following steps: Identify relationships between design actions; Determine precisely the specific types and possibilities of these relationships; Finally, several fan graphs are designed for each node. A fan graph represents a link relationship, and a fan in the fan graph represents the possibility that a certain node has a corresponding link relationship with that node.

6. The AI ​​enhanced link notation method according to claim 1, characterized in that: In the stimulation step, the process of generating visual stimulation includes the following steps: Determine whether the design action is a critical design action based on the number of links; Design actions that convert text to images using the Image Generator; Organize all the images into a collage as a visual stimulus.

7. The AI ​​enhanced link notation method according to claim 1, characterized in that: In the analysis step, the forward chain entropy is calculated to evaluate the completion of the design task. The larger the entropy value, the more active the design process is and the more diverse the ideas involved in the design task are. The forward chain entropy H forward It is expressed as: Among them, p(x i ) represents the design action x i is the probability of a forward link accounting for the total links during the brainstorming process, and n is the number of design actions.

8. The AI ​​enhanced link notation method according to claim 1, characterized in that: The analysis step includes identifying the thinking patterns of different users, which specifically includes the following steps: Nodes with more than three forward links are classified as divergent thinking actions, and nodes with more than three backward links are classified as convergent thinking actions; Analyze and count the divergence and convergence of all design actions of different users to identify user thinking types.

9. The AI ​​enhanced link notation method according to claim 1, characterized in that: The analysis step includes identifying the degree of interaction between users, specifically calculating the link frequency between different users, which is expressed as: C uv =freq(u,v); Among them, C uv represents the link frequency of the design actions of user u and user v; C uv The larger the value, the more interactions there are between user i and user j, and the more suitable they are for collaborative design tasks.

10. An AI-enhanced link notation system, characterized in that: include: Initialization module, initializes a blank link graph; Input module, where users input design ideas according to given design tasks; The generation module converts design ideas into design action expressions through a large language model and generates design actions; Link module, adds the design action as a node to the link graph and completes the link relationship; The linking process refers to the large language model to automatically provide design action linking suggestions based on preset criteria; The stimulation module extracts key design actions from the link graph and converts them into visual stimuli using an image generator; if the user generates a new design idea due to the visual stimulation, the user enters the design idea and repeats the generation and linking steps; Analysis module, to evaluate the completion of design tasks and analyze user characteristics; Repeat the module and repeat the entry step to the analysis step until a link diagram that meets the requirements is generated.

Images