A human-machine collaborative creation method, system and electronic device supporting combined creativity

Through the thinking divergent creative combination model of the large language model framework, the problem of combining creativity in human-computer collaborative creative tools is solved, and the efficiency, diversity and visual output of combination design is achieved, improving the design quality and user experience.

CN117932048BActive Publication Date: 2024-12-06ZHEJIANG UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410037031.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-01-10
Publication Date
2024-12-06
Estimated Expiration
2044-01-10

AI Technical Summary

Technical Problem

In the field of product design, human-machine collaborative creative tools lack effective combination creative tools, resulting in unsuccessful combinations, low assembly results, and inability to directly utilize outputs, especially for those who are not good at sketching or painting. The idea of ​​combination is difficult and time-consuming.

Method used

Provide a thinking divergent creative combination model based on the large language model framework, including keyword extraction module, thinking divergent module, creative combination module and visual output module. By obtaining concept design cases, the main body and auxiliary body database is constructed, and the pre-constructed The prompt word template extracts the main body and the secondary body, matches and combines, and generates a visual appearance map.

Benefits of technology

Effectively support the divergent thinking of users, generate more inspiring and novel combination references, reduce designers' cognitive burden, improve design quality and diversity, simplify the creative process, and reduce user pressure.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117932048B_ABST
    Figure CN117932048B_ABST
Patent Text Reader

Abstract

The present invention discloses a human-machine collaborative creation method supporting combination creativity, comprising the following steps: obtaining a collection of concept design cases, and dividing the concept design cases into a main body with a label and a sub-body used to describe the main body, so as to construct a corresponding database; constructing a divergent thinking creative combination model based on a large language model framework, wherein the divergent thinking creative combination model comprises a keyword extraction module, a divergent thinking module, a creative combination module, and a visualization output module: using the generated main body-sub-body concept combination and a visualization appearance diagram as a creative guide for the concept design. The present invention also provides a human-machine collaborative creation system and an electronic device. The method provided by the present invention can effectively diverge the user's needs and combine them between different concepts to provide a more inspiring and novel combination reference.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of computer-aided design, and in particular relates to a human-machine collaborative creation method, system and electronic equipment that supports combined creativity. Background Art

[0002] In creativity theory, combining two unrelated concepts into a novel idea is a common means to enhance creativity. People can integrate secondary concepts into the main concept to stimulate and promote creative tasks. However, trying to integrate different design elements in actual design scenarios is a very challenging task. Although some auxiliary creative design tools can significantly improve design output and designer satisfaction levels, it is worth noting that these tools often encounter problems such as unsuccessful combination, low quality of assembly results, and outputs that cannot be directly used. In recent years, with the continuous development of technological progress, generative artificial intelligence has shown extraordinary potential in the field of combinatorial creative design. However, the generation results of artificial intelligence are often uncontrollable. Therefore, combining human judgment with the powerful search ability of artificial intelligence often becomes a more optimized strategy. This human-computer collaborative approach provides a strong foundation for the application of combinatorial creativity in different professional fields. However, in the field of human-computer interaction, there is currently a lack of effective and powerful artificial intelligence combinatorial creative tools that serve product design. The research on integrating combinatorial creativity theory with artificial intelligence technology and applying it to the field of product design is still in the exploratory stage.

[0003] In creative thinking, a good concept representation often describes the core of the idea more accurately, which is also true in the context of combinatorial creativity. Combinatorial creative design ideas are composed of subjects and sub-subjects. In order to generate a creative new idea, we need to conceive their combination methods based on the characteristics of the subject and sub-subject nouns, and innovatively integrate different concepts and elements. Experienced designers like to express concepts with sketches and verbal ideas at the same time. This form of representation, which uses visual and verbal forms as conceptual stimuli, often helps designers transition to subsequent designs more effectively. However, it is very challenging to verbalize the combined ideas and then visualize the ideas. For those who are not good at sketching or painting, creating combined ideas is difficult and time-consuming. Even with the powerful computing power of computers, it is still difficult to simulate a complete combined idea.

[0004] Patent document CN113221559A discloses a method and system for extracting Chinese key phrases in the field of scientific and technological innovation using semantic features. This method mines the features of Chinese scientific and technological innovation document corpus, constructs a library of Chinese stop words and stop patterns, and achieves high-performance filtering of invalid information. In addition, various key phrase extraction algorithms are quantitatively evaluated and analyzed with the help of domain expert annotations, so as to select an algorithm model that is more suitable for domain cognition, and use a variety of statistical rules for filtering to improve phrase extraction performance. Further, using the structural characteristics of the document, the subject semantics of the document is embedded in the vector space, and the semantic similarity between the extracted phrases and the document subject, and the semantic importance of the phrases are comprehensively used for calculation and ranking, so as to complete further screening of key phrases. This method can support a variety of downstream tasks and applications, including the construction of knowledge graphs in the field of scientific and technological innovation, semantic retrieval of scientific and technological innovation documents, and precise search of scientific and technological innovation entities.

[0005] Patent document CN115033772A discloses a creative stimulation method based on a semantic network, including: step 1, constructing a database for matching retrieval; step 2, obtaining keywords, obtaining corresponding node information in the semantic network through mapping, matching based on the node information in the database constructed in step 1, and obtaining a weight information set of all associated nodes; step 3, standardizing the statistical weights in the weight information set, and obtaining a standard statistical weight set corresponding to the associated nodes; step 4, based on the standard statistical weight set and the semantic weight set, calculating a path edge weight set; step 5, based on the path edge weight set, generating a force-directed graph for reflecting the implicit relationship between keywords and related words. Summary of the invention

[0006] The purpose of the present invention is to provide a human-computer collaborative creation method, system and electronic device that supports combined creativity. The method can effectively diverge the user's needs and combine different concepts to provide more inspiring and novel combination references.

[0007] In order to achieve the first object of the present invention, a human-machine collaborative creation method supporting combined creativity is provided, comprising the following steps:

[0008] Obtain a collection of conceptual design cases, and divide the conceptual design cases into a main body with a label and a sub-body used to describe the main body, so as to construct a corresponding database;

[0009] Construct a divergent thinking creative combination model based on a large language model framework, the divergent thinking creative combination model includes a keyword extraction module, a divergent thinking module, a creative combination module and a visual output module:

[0010] The keyword extraction module includes a pre-constructed prompt word template, and the prompt word template is used to extract the main body and the secondary body in the input text;

[0011] The divergent thinking module searches the database according to the tag to which the subject belongs to obtain to obtain a set of related sub-subjects, and matches the extracted sub-subjects with the set of related sub-subjects to obtain a set of matched sub-subjects;

[0012] The creative combination module combines the extracted subject with the set of matching sub-subjects to output a corresponding subject-sub-subject concept combination;

[0013] The visualization output module inputs the main body-subject concept combination into a pre-built text graph model to obtain a corresponding visualization appearance graph;

[0014] The generated main-sub-body concept combination and visual appearance diagram are used as creative guidance for concept design.

[0015] The present invention sets the relationship between the subject and the sub-sub ...

[0016] Specifically, the sub-body includes shape, material, color, image, structure and function.

[0017] Specifically, the expression of the divergent thinking module is as follows:

[0018] A Additive =M(Q A ), Q A =<Additive,Type>

[0019] Among them, A Additive represents the search result, M represents the large language model framework, Q A It indicates the first prompt word template, Type indicates the direction of divergent thinking, including shape, material, color, image, structure and function, and Additive indicates the secondary body.

[0020] Specifically, the Q A The template is: Please use your imagination to list <additive>of <type>feature.

[0021] Specifically, when combining the subject and the matching sub-subject set, the creative combination module also uses the A / B testing method to randomly re-sort the sub-subjects to generate multiple groups of subject-subject concept combinations with different sub-subject sortings.

[0022] Specifically, the creative combination module is expressed as follows:

[0023] A=M(Q B ), Q B = <Base,A Additive >

[0024] Among them, A represents the main body-subject concept combination, M represents the large language model framework, Base represents the main body in the input text, and A Additive Indicates the search result, Q B Represents the second prompt word template.

[0025] Specifically, the Q B The template is: Design a <additive>of <type>, which can reflect <aadditive>Combination of creative <base> .

[0026] Specifically, the generation process of the visual appearance graph is as follows:

[0027] I=T(K), K=M(A)

[0028] Among them, I represents the visual appearance diagram of the main-subject concept combination visualization, T represents the text graph model, M represents the large language model framework, and A represents the main-subject concept combination.

[0029] Specifically, the generation process of the visual appearance image also includes image enhancement, and the expression of the image enhancement is as follows:

[0030] I Aug =T(E,I),E=M(Qs),Qs= <base>

[0031] Among them, I Aug represents the enhanced image result with added scene elements, T represents the large model of the text map, E represents the text scene element, I represents the visual appearance of the main-subject concept combination visualization, Q S Indicates the third prompt word template.

[0032] Specifically, Q S The template is: Please use your imagination to use some visual features to convey <base> The metaphorical core function.

[0033] In order to achieve the second invention objective of the present invention, a human-machine collaborative creation system is provided, which is implemented by the human-machine collaborative creation method supporting combined creativity provided above, including an input unit, a divergent thinking unit, a creative combination unit and a visual output unit;

[0034] The input unit is used to input the main body and related sub-bodies preset in the conceptual design;

[0035] The divergent thinking unit generates corresponding divergent thinking concepts of the sub-subject according to the input sub-subject and the prompt word template;

[0036] The creative combination unit generates concepts based on the input subject and related sub-subject thinking to generate corresponding subject-subject concept combinations;

[0037] The visualization output unit outputs a corresponding visualization appearance diagram according to the generated main body-subject concept combination.

[0038] In order to achieve the third purpose of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned human-computer collaborative creation method supporting combined creativity are implemented.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] The design elements of different concepts are structured into sub-entities, and the model is used to

[0041] A divergent search is performed on the input subject for related sub-sub ... BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flowchart of a human-machine collaborative creation method supporting combined creativity provided in this embodiment;

[0043] Figure 2 A divergent thinking flow chart of the divergent thinking module provided in this embodiment;

[0044] Figure 3 A workflow diagram of the creative combination module provided for this embodiment;

[0045] Figure 4 A workflow diagram of the visualization module provided in this embodiment;

[0046] Figure 5 A flowchart of image enhancement provided for this embodiment;

[0047] Figure 6 A schematic diagram of the expert evaluation results provided for this embodiment. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the technical scheme 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. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0049] like Figure 1 As shown, a human-machine collaborative creation method provided by this embodiment includes the following steps:

[0050] A collection of conceptual design cases is obtained, and the conceptual design cases are divided into a main body with a label and a sub-body used to describe the main body, so as to construct a corresponding database.

[0051] A divergent thinking creative combination model based on a large language model framework is constructed, wherein the divergent thinking creative combination model includes a keyword extraction module, a divergent thinking module, a creative combination module and a visualization output module.

[0052] The keyword extraction module includes a pre-constructed prompt word template, and the prompt word template is used to extract the main body and the secondary body in the input text.

[0053] The divergent thinking module searches the database according to the tag to which the subject belongs to obtain to obtain a set of related sub-subjects, and matches the extracted sub-subjects with the set of related sub-subjects to obtain a set of matched sub-subjects.

[0054] The creative combination module combines the extracted subject with the set of matching sub-subjects to output a corresponding subject-sub-subject concept combination.

[0055] The visualization output module inputs the subject-subject concept combination into a pre-built text graph model to obtain a corresponding visualization appearance graph.

[0056] The generated main-sub-body concept combination and visual appearance diagram are used as creative guidance for concept design.

[0057] More specifically, this embodiment proposes a human-machine collaborative creation method that integrates combination strategies and a decoder-only pre-trained large model to support combination creativity.

[0058] It consists of two phases: Phase 1 includes the divergent thinking phase and the convergent thinking phase, delivered by AI and humans respectively;

[0059] The second stage is the visualization stage, which effectively visualizes the verbal combination ideas into a combined image;

[0060] Ultimately, users can use the generated verbal concepts as well as images as stimuli for subsequent designs. It is designed to help amateur designers and creative enthusiasts accelerate concept design and support creativity.

[0061] In the first stage, the user inputs two concepts - the main body and the secondary body. In the combination of "main body + secondary body", since the main body is the main feature of the creativity, the creativity of the combination creativity depends largely on how to introduce novel secondary bodies. Generally speaking, the combination design process has a certain level of abstraction, and there is a lack of clear connection and specific expression between different concepts. Due to the cognitive limitations of individuals, it is difficult to manually combine secondary bodies with the main body directly. An effective way to alleviate this problem is combination fixing, which means that before the combination, the system automatically comes up with relevant associations.

[0062] like Figure 1 As shown in the figure, the secondary body is positioned as the input of the model to help users jump out of the combination of the main body and the secondary body and realize cross-context thinking. LLM's rich knowledge and reasoning ability are used to diverge the secondary body from six aspects: shape, material, color, image, structure and function, and realize the associated search of the secondary body. The divergence process of the divergent thinking module can be expressed as:

[0063] A Additive =M(Q A ), Q A =<Additive,Type>

[0064] Among them A Additive represents the result of divergent thinking, M represents the large language model used in the system, Q A Indicates the standard prompt for inputting the large model during the divergent thinking phase of the secondary body. A It is constructed using three concepts: formatted prompt template, Additive and Type. Additive represents the secondary concept, and Type represents the direction of divergent thinking, ranging from one of the six aspects of shape, material, color, image, structure and function.

[0065] The Q A The template is: Please use your imagination to list <additive>of <type>feature.

[0066] For the convergence phase, a human-centered interaction strategy was adopted. Given that convergent thinking requires complex constraints and selection considerations, and that designs often reflect individual personality and preferences, we aim to enhance the user experience by empowering users to take charge of the selection process and granting them decision-making power over implementation options. To reduce the burden of decision-making, we placed no limit on the number of implementation options a user could choose from and allowed for individual preference selection. To reduce potential biases introduced by sorting and ranking, an A / B testing approach was employed to ensure different category orders for different users. This added complexity refined the design task and provided a clearer design trajectory for the large language model.

[0067] The process of its A / B testing method is as follows:

[0068] The model displays the embodiments in the form of rectangular selection boxes and presents them according to six major characteristics to reduce confusion and facilitate user comparison and selection. However, the embodiments in each category are randomly sorted.

[0069] In order to avoid the influence of arrangement and order on the selection, we applied A / B testing to make different users see different classification order to reduce the possible deviation caused by classification order. For example, one sorting order is: shape, material, color, image, structure and function; another sorting order is: material, structure, color, shape, function and image. (There are a total of sorting methods The system randomly assigns users to different classification sortings to ensure that each user sees only one sorting method, avoiding certain features from receiving disproportionate attention because they are ranked higher, thereby averaging out the bias caused by the classification order.

[0070] like Figure 3 As shown, the filtered sub-subject divergence results and the subject are input into the creative combination module as new inputs and combined into text concepts. For example, "Now given the subject is a chair and the sub-subject is love, please use your imagination to combine chairs and love to design a chair that involves the 'structural' features of love and can 'reflect emotional connection'." By explicitly specifying the subject and sub-subject, we provide the LLM with a well-defined request that can effectively reduce ambiguity and directional confusion in the design. In this prompt, an inspiring key prompt is also included: "Please use your imagination to the fullest." The purpose of this is to allow the large language model to simulate the imaginative thinking process, thereby triggering creative responses that go beyond traditional design. The concept combination process can be expressed as:

[0071] A=M(Q B ), Q B = <Base,A Additive >

[0072] A represents the result of concept combination, M represents the large language model used in the system, and Q B Indicates the standard prompt for input into the large model at the stage of combining the main and secondary concepts. B Is to use the formatted prompt template, Base and A Additive These three concepts are constructed. Base represents the main concept, A Additive Indicates the result of divergent thinking.

[0073] Where Q B The template is: "Combinatorial creativity is an innovative design approach in which new ideas, designs or solutions are generated by combining existing, often seemingly unrelated elements or concepts. In the common noun-noun combination, the main or basic concept related to product design is called 'Base', and the other additional concept that forms the combination is called 'Additive'. Each combinatorial creative product design idea has a given set of 'Base' and 'Additive'.

[0074] Now for Base: <base> , and Additive: <additive>Please use your imagination to combine Base and Additive to design a <additive>of <type>, which can reflect <aadditive>Combination of creative <base> . "

[0075] After obtaining the integrated design concepts, they must be further screened or optimized based on their quality. First of all, a successful combination design should follow its core principle, that is, the final designed product should clearly reflect the main body while seamlessly integrating the functions of the subsidiary body. If the synthesis does not meet expectations, potential manifestations may include that the final product is inconsistent with the main body or fails to fully reflect the properties of the additive. For these suboptimal design results, the model provides users with the option of regenerating or fine-tuning the details. It is worth noting that although in theory the main body can be combined with multiple subsidiaries, the method provided in this embodiment mainly involves the scenario of a single subsidiary.

[0076] In the second stage, the text combination description is visualized through T2I to synthesize a concept map that contains both the main and secondary features. This stage consists of an appearance generator and a scene element generator. The former, as a T2I-Prompt synthesis system, is used to efficiently synthesize visual concepts that contain the main features of the main and secondary features, while the latter uses visual metaphor theory to generate metaphorical visual elements based on the functions and usage scenarios of the products. The product is used to supplement the contextual elements to further enrich the combination method.

[0077] In the image generator, in order to obtain the most clear and accurate image output, the input text should be concise, specific, and avoid being vague or overly abstract. Keyword inputs are considered an excellent choice mainly because they are clear and concise, directly capturing the core elements of the generated image without processing extraneous or redundant information. This not only speeds up the computational efficiency of the model, but also reduces the risk of ambiguity and inaccurate generation. In addition, the concise structure of keywords also allows users to easily combine them to create more complex images with multiple attributes or features, thereby improving system performance and user experience while ensuring the accuracy and clarity of image generation. The keyword can be any noun phrase, such as a word "chair", or a noun phrase modified by a prepositional phrase, such as "the cat stands on the table". Therefore, for example Figure 4 As shown in Figure 2, we extract the keyword list from the creative combination module in the first stage to make it more suitable as the input of the image generator. The visualization process can be expressed as:

[0078] I=T(K), K=M(A)

[0079] I represents the result image of combined concept visualization, T represents the large model of the text map used, K represents the keyword list input into the large model in the visualization stage, M represents the large language model used, and A represents the result of concept combination in the previous stage.

[0080] In the model, users will see a list of keywords in a multiple-choice format. Adjacent to each keyword is a checkbox, which not only allows users to choose whether to check each keyword, but also allows them to directly edit the keyword content or add personalized keywords to more accurately describe the characteristics of the desired image. It is particularly important to emphasize that in the combined concept, the main body and the phrases related to it should be listed first in the list; the secondary body, as an additional feature of the main body, should be placed immediately after it.

[0081] In the visualization output module, the large language model is used to assist in exploring the visual representation elements of the subject based on the functional metaphor. First, the large language model conducts an in-depth and detailed analysis of the core functional meaning of the subject metaphor and its related keywords. For example, the main function of a hair dryer is to quickly dry hair or other objects through air flow, and the keyword is air flow. With keywords as a guide, the large language model can generate visual features that can intuitively convey this metaphorical function, such as using dynamic lines to simulate the effect of air flow. These initially selected visual elements also need to be interpreted from multiple angles and carefully optimized through user feedback to ensure that they have a close and creative connection with the metaphorical function.

[0082] In order to preserve the details in the high-quality images output by the image generator to the greatest extent possible while avoiding excessive interference from scene elements, such as Figure 5 As shown in Figure 1, the visualization output module combines these carefully selected and optimized text visual elements with the high-quality images output by the image generator to form multimodal input data, which are integrated into T2I to generate product images full of rich scene elements and profound metaphors. The image enhancement process can be expressed as:

[0083] I Aug =T(E,I),E=M(Qs),Qs= <base>

[0084] Among them I Aug represents the enhanced image result with scene elements added, T represents the text image model used in the system, E represents the text scene element, and I represents the result image of the combined concept visualization. M represents the large language model used in the system. Qs represents the standard prompt input to the large model in the combined image enhancement stage. Qs is constructed using the two concepts of formatted prompt template and Base.

[0085] Base represents the main concept. The Qs template is: "Please use your imagination to use some visual features to convey <base> metaphorical core function.”

[0086] This embodiment also provides a human-machine collaborative creation system, which is implemented by the human-machine collaborative creation method provided by the above embodiment, including an input unit, a divergent thinking unit, a creative combination unit and a visual output unit;

[0087] The input unit is used to input the main body and related sub-bodies preset in the conceptual design;

[0088] The divergent thinking unit generates corresponding divergent thinking concepts of the sub-subject according to the input sub-subject and the prompt word template;

[0089] The creative combination unit generates concepts based on the input subject and related sub-subject thinking to generate corresponding subject-subject concept combinations;

[0090] The visualization output unit outputs a corresponding visualization appearance diagram according to the generated main body-subject concept combination.

[0091] This embodiment also provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the human-computer collaborative creation method provided in the above embodiment are implemented.

[0092] In order to better illustrate the technical effect of the method provided in this embodiment, 30 novice designers with similar design capabilities were recruited to participate in this design experiment. All participants had an undergraduate education background in design, spanning 3 to 4 years. The participants were divided into 3 groups, 10 people in each group, of which 5 people participated in Experiment 1 and Experiment 2 respectively. The first group adopted a strategy-only approach, and participants generated design solutions based entirely on pre-provided combination templates in the first stage. Once they had verbal ideas, they were asked to use non-artificial intelligence tools such as Adobe Photoshop or hand-drawing in the second stage to visualize the ideas by using visual templates. The second group, using only AI methods, required designers to autonomously generate verbal design ideas without using combination templates in the first stage. Then, in the second stage, they were asked to freely use T2I to visualize the idea without the guidance of visual templates. The third group, called "Strategy + AI", used the method provided in this embodiment to combine the combination templates that guide the large language model with the visual templates that guide T2I for creative product design.

[0093] In order to accurately evaluate the design concepts, three experts with more than 5 years of design experience were hired as evaluators. All ideas were evaluated under a uniform set of guidelines and inter-rater agreement on a 7-point scale. Four criteria were used for the evaluation: quantity, novelty, quality, and diversity. Quantity was measured by counting the total number of design ideas generated by each person, while diversity was quantified by the types of design concepts.

[0094] like Figure 6 As shown in the figure, all three different design methods can produce expected and high-quality results. Among them, our method has achieved significant advantages in quantity, novelty, quality and diversity indicators, which shows that our system can excellently complete the combination of creative design tasks, greatly improve the quality and diversity of design ideas, and reduce the complexity of search and combination of implementation examples. It also proves that the method of divergent and convergent thinking proposed in this embodiment is stronger than traditional design methods in generating high-quality creative thinking. This method also helps designers more easily generate visualization effects consistent with their expectations in the visualization stage.

[0095] In summary, the method provided by this embodiment provides diverse and innovative results in its two stages, automates the process of creative generation and association, and users can focus on exploring and refining innovative combinations, thereby driving the combined creativity forward. In human-computer interaction, it represents a user-centered design system that effectively generates tailored information, which not only simplifies the creative process, but also reduces cognitive and physical stress, and improves overall user satisfaction and interaction fluidity. It embodies the synergy between artificial intelligence and humans, enabling humans to freely navigate complex design fields.

[0096] Of course, the above description is only a specific embodiment of the present invention and is not intended to limit the scope of implementation of the present invention. All equivalent changes or modifications made according to the structure, characteristics and principles described in the patent application scope of the present invention should be included in the patent application scope of the present invention.

[0097] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention is described in detail with reference to the above-described embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-described embodiments within the technical scope disclosed by the present invention, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.< / aadditive> < / type> < / additive> < / additive> < / type> < / additive> < / aadditive> < / type> < / additive> < / type> < / additive>

Claims

1. A human-machine collaborative creation method supporting combinatorial creativity, characterized in that: The following steps are involved: Obtain a collection of conceptual design cases, and divide the conceptual design cases into a main body with a label and a sub-body used to describe the main body, so as to construct a corresponding database; Construct a divergent thinking creative combination model based on a large language model framework, the divergent thinking creative combination model includes a keyword extraction module, a divergent thinking module, a creative combination module and a visual output module: The keyword extraction module includes a pre-constructed prompt word template, and the prompt word template is used to extract the main body and the secondary body in the input text; The divergent thinking module searches the database according to the tag to which the subject belongs to obtain to obtain a set of related sub-subjects, and matches the extracted sub-subjects with the set of related sub-subjects to obtain a set of matched sub-subjects; The creative combination module combines the extracted subject with the set of matching sub-subjects to output a corresponding subject-sub-subject concept combination; The visualization output module inputs the main body-subject concept combination into a pre-built text graph model to obtain a corresponding visualization appearance graph; The generated main-sub-body concept combination and visual appearance diagram are used as creative guidance for concept design.

2. The human-machine collaborative creation method supporting combined creativity according to claim 1, characterized in that: The sub-body includes shape, material, color, image, structure and function.

3. The human-machine collaborative creation method supporting combined creativity according to claim 1, characterized in that: The expression of the divergent thinking module is as follows: A Additive =M(Q A ),Q A =<Additive,Type> Among them, A Additive represents the search result, M represents the large language model framework, Q A It indicates the first prompt word template, Type indicates the direction of divergent thinking, including shape, material, color, image, structure and function, and Additive indicates the secondary body.

4. The human-machine collaborative creation method supporting combined creativity according to claim 1, characterized in that: When combining the subject and the matching sub-subject set, the creative combination module also uses the A / B testing method to re-randomly sort the sub-subjects to generate multiple groups of subject-sub-subject concept combinations with different sub-subject sortings.

5. The human-machine collaborative creation method supporting combined creativity according to claim 1, characterized in that: The creative combination module is expressed as follows: A=M(Q B ),Q B =<Base,A Additive > Among them, A represents the main body-subject concept combination, M represents the large language model framework, Base represents the main body in the input text, and A Additive Indicates the search result, Q B Represents the second prompt word template.

6. The human-machine collaborative creation method supporting combined creativity according to claim 1, characterized in that: The generation process of the visual appearance graph is as follows: I=T(K), K=M(A) Among them, I represents the visual appearance diagram of the main-subject concept combination visualization, T represents the text graph model, M represents the large language model framework, and A represents the main-subject concept combination.

7. The human-machine collaborative creation method supporting combined creativity according to claim 1, characterized in that: The generation process of the visual appearance graph also includes image enhancement, and the expression of the image enhancement is as follows: I Aug =T(E,I),E=M(Qs),Qs= <base> Among them, I Aug represents the enhanced image result with added scene elements, T represents the large model of the text map, E represents the text scene element, I represents the visual appearance of the main-subject concept combination visualization, Q s Indicates the third prompt word template.

8. A human-computer collaborative creation system, characterized in that: The method is implemented by a human-machine collaborative creation method supporting combined creativity as claimed in any one of claims 1 to 7, comprising an input unit, a divergent thinking unit, a creative combination unit and a visual output unit; The input unit is used to input the main body and related sub-bodies preset in the conceptual design; The divergent thinking unit generates corresponding divergent thinking concepts of the sub-subject according to the input sub-subject and the prompt word template; The creative combination unit generates concepts based on the input subject and related sub-subject thinking to generate corresponding subject-subject concept combinations; The visualization output unit outputs a corresponding visualization appearance diagram according to the generated main body-subject concept combination.

9. An electronic device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the human-computer collaborative creation method supporting combined creativity described in any one of claims 1 to 7 are implemented.