Fragrance descriptor quantification method, system and equipment based on artificial intelligence large model multi-agent collaboration and medium

Abstract

The invention discloses a fragrance descriptor quantification method, system and device based on artificial intelligence large model multi-agent collaboration and a medium, and the method comprises the steps: converting a demand of a target customer, and obtaining a demand description corresponding to the demand; obtaining a first compound combination corresponding to the demand description based on a chemical mapping relationship in a preset knowledge graph; based on reinforcement learning, optimizing the component proportion of the first compound to obtain an optimized second compound combination; optimizing the second compound combination again to obtain a third compound combination; optimizing the third compound combination according to a taboo index and a non-additive limit index to obtain a fourth compound combination; and taking the fourth compound combination as a target compound combination of the target customer, and controlling the associated equipment to output the target compound combination. The invention belongs to the field of chemical odors. The present invention can generate a compound recipe based on the descriptors.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of chemical odor, in particular to a fragrance descriptor quantification method, system, device and medium based on artificial intelligence large model multi-agent collaboration. BACKGROUND

[0002] Fragrance blending refers to the creation of unique fragrances by mixing different fragrances (i.e. chemical substances with special odors). Perfumers carefully select and combine various fragrance ingredients according to specific goals, needs or creative concepts to achieve the desired fragrance effect. The fragrance blending process not only involves artistic creativity, but also requires deep professional knowledge and understanding of the characteristics of various fragrances and their interactions.

[0003] At the current industrial level, fragrance blending is increasingly focusing on intelligentization. How to achieve intelligent fragrance blending and how to link customer needs, chemical composition and equipment-generated fragrance blending are urgent problems to be solved. SUMMARY

[0004] The present application provides a fragrance descriptor quantification method, system, device and medium based on artificial intelligence large model multi-agent collaboration, which solves the technical problem of difficult intelligent fragrance blending in the prior art and achieves the technical effect of intelligent fragrance blending.

[0005] In a first aspect, the present application provides a fragrance descriptor quantification method based on artificial intelligence large model multi-agent collaboration, comprising: Converting the demand of the target customer to obtain a demand description corresponding to the demand; Based on the chemical mapping relationship in the preset knowledge graph, a first compound combination corresponding to the demand description is obtained, wherein the compound combination contains several compounds; Based on reinforcement learning, the ingredient ratio of the first compound is optimized to obtain a second compound combination after optimization; Based on molecular dynamics, the fragrance volatility of the second compound combination after optimization is predicted in time sequence, and a time decay index and a balance index are introduced to optimize the second compound combination again to obtain a third compound combination; According to the taboo index and the non-addition restriction index, the third compound combination is optimized to obtain a fourth compound combination; The fourth compound combination is taken as the target compound combination of the target customer, and the associated equipment is controlled to output the target compound combination.

[0006] Further, converting the demand of the target customer to obtain a demand description corresponding to the demand comprises: Based on a large language model, the demand of the target customer is converted into a descriptor; Based on the preset visual interaction model, the demand of the target customer is converted into a multi-dimensional user portrait. The descriptive words and the multi-dimensional user portrait are used as the demand description.

[0007] Further, based on reinforcement learning, the component ratio of the first compound is optimized to obtain an optimized second compound combination, including: Construct a black box function, wherein the variable of the black box function is the component ratio of the first compound, and the dependent variable is the user feedback; Based on the acquisition function UCB, the next component ratio of the first compound is determined; Based on the black box function, the dependent variables corresponding to the component ratios are iteratively compared, and when the preset iteration condition is met, the second compound combination is obtained.

[0008] Further, according to the taboo index and the non-addition restriction index, the third compound combination is optimized to obtain a fourth compound combination, including: Determine whether there is a compound in the third compound combination that is a taboo index or a non-addition restriction index; If so, after removing the compound, a compound similar in chemical properties to the removed compound is determined based on a preset knowledge graph, and the chemically similar compound is added to the third compound combination.

[0009] Further, it also includes: After using the target compound combination, collect user feedback; According to the user feedback, the compound ratio in the target compound combination is re-optimized.

[0010] Further, the multi-dimensional user portrait includes: Image selection, emotional coordinates, interest map, consumption capacity, and life cycle stage.

[0011] Further, the time decay function includes:

[0012] Wherein, is the fragrance intensity of the nth compound combination at the tth moment, is the fragrance intensity of the compound combination at the initial moment, is the decay coefficient of the nth compound combination.

[0013] In a second aspect, the present application provides a fragrance description word quantification device based on artificial intelligence large model multi-agent collaboration, and the method includes: ​​​The demand conversion module is configured to convert the demand of the target customer to obtain a demand description corresponding to the demand. The mapping module is configured to obtain a first compound combination corresponding to the demand description based on a chemical mapping relationship in a preset knowledge graph, wherein the compound combination comprises a plurality of compounds. The first optimization module is configured to optimize the proportion of the ingredients of the first compound based on reinforcement learning to obtain a second compound combination after optimization. The second optimization module is configured to optimize the fragrance volatility of the second compound combination after optimization in a time sequence based on molecular dynamics, and introduce a time decay index and a balance index to optimize the second compound combination again to obtain a third compound combination. The third optimization module is configured to optimize the third compound combination based on a taboo index and an unaddable restriction index to obtain a fourth compound combination. The compound output module is configured to take the fourth compound combination as a target compound combination of the target customer, and control an associated device to output the target compound combination.

[0014] In a third aspect, the present application provides an electronic device, comprising: a processor; a memory for storing processor-executable instructions; The processor is configured to execute to realize the sweet taste description word quantification method based on the artificial intelligence large model multi-agent collaboration provided in the first aspect.

[0015] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium, when the instructions in the non-transitory computer-readable storage medium are executed by the processor of the electronic device, the electronic device can execute the realization of the sweet taste description word quantification method based on the artificial intelligence large model multi-agent collaboration provided in the first aspect.

[0016] The one or more technical solutions provided in the present application have at least the following technical effects or advantages: The present application effectively solves the problem that the traditional fragrance modulation cannot quantify the customer's abstract description by the cooperation mechanism of the artificial intelligence large model and the multi-agent, and opens up the complete link from natural language understanding, chemical component mapping to device execution, realizes the precise modeling and automatic formula generation of the fragrance demand, and improves the individual customization ability and the system intelligence level. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort based on these drawings.

[0018] Figure 1 The flowchart of the fragrance descriptor quantification method based on the artificial intelligence large model multi-agent cooperation provided by the present application is shown in the figure. Figure 2 The structure diagram of the fragrance descriptor quantification device based on the artificial intelligence large model multi-agent cooperation provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0019] The embodiment of the present application provides a fragrance descriptor quantification method based on artificial intelligence large model multi-agent cooperation, which solves the technical problem of difficult intelligent fragrance blending in the prior art.

[0020] The technical solution of the present application is to solve the above technical problems, and the general idea is as follows: The fragrance descriptor quantification method based on artificial intelligence large model multi-agent cooperation comprises: converting the demand of a target customer to obtain a demand description corresponding to the demand; obtaining a first compound combination corresponding to the demand description based on a chemical mapping relationship in a preset knowledge graph, wherein the compound combination contains a plurality of compounds; optimizing the component ratio of the first compound based on reinforcement learning to obtain an optimized second compound combination; based on molecular dynamics, predicting the fragrance volatility of the optimized second compound combination in time sequence, and introducing a time decay index and a balance index to optimize the second compound combination again to obtain a third compound combination; optimizing the third compound combination according to a taboo index and an unaddable restriction index to obtain a fourth compound combination; taking the fourth compound combination as a target compound combination of the target customer, and controlling a related device to output the target compound combination.

[0021] In order to better understand the above technical solutions, the above technical solutions will be described in detail in the following in combination with the drawings in the specification and the specific embodiments.

[0022] Firstly, the term "and / or" appearing in this paper is only to describe the association relationship of the associated objects, which means that there can be three kinds of relationships, for example, A and / or B, which can represent the three cases of A alone, A and B together, and B alone. In addition, the character " / " in this paper generally represents an "or" relationship between the front and rear associated objects.

[0023] The present application provides a fragrance descriptor quantification method based on an artificial intelligence large model multi-agent cooperation, which comprises the following steps: Figure 1The illustrated artificial intelligence-based large model multi-agent collaborative fragrance descriptor quantification method includes steps S11-S16: Step S11, the demand of the target customer is converted to obtain the demand description corresponding to the demand.

[0024] Specifically, it includes: based on a large language model, converting the demand of the target customer into a description word; based on a preset visual interaction model, converting the demand of the target customer into a multi-dimensional user portrait; and taking the description word and the multi-dimensional user portrait as the demand description, wherein the multi-dimensional user portrait includes image selection, emotional coordinates, interest graph, consumption ability, and life cycle stage.

[0025] The large language model can be deepseek, chatgpt, etc. The powerful natural language processing capability of the large language model can be used to parse the natural language demand of the customer into specific description words.

[0026] For example, the customer proposes the demand for "fresh air after the rain", which can be converted into the image-based description word as wetness, green vitality, and air permeability.

[0027] The preset visual interaction model can include an emotional coordinate system, an image selection board, a sliding regulator, and virtual reality. The visual interaction model refers to a tool or framework that interacts with users through a graphical interface, collects user preferences and behavior data, and analyzes them.

[0028] For example, after the customer proposes "fresh air after the rain", the preset visual interaction model generates 4-5 images of different areas after the rain, and the customer can select the image with mineral ions brought by dust in the urban area image to participate in the formation of the smell.

[0029] Image selection: visual symbols corresponding to the user's preferred fragrance type (such as flowers, fruits, candies, etc.); Emotional coordinates: the emotions that the user wants the fragrance to convey (such as joy, relaxation, vitality, romance, etc.); Interest graph: the user's interest labels, such as "exercise", "reading", "party", etc. scenes; Consumption ability: the price range or brand preference that the user is willing to pay; Life cycle stage: age, gender, whether pregnant, whether children, and other physiological and life stage characteristics.

[0030] Through visual interaction (such as graphical selection, emotional labeling, etc.), the individual characteristics of the customer are quantified from multiple dimensions to form a three-dimensional user portrait.

[0031] Through the preset interaction interface (such as drag-and-drop sliders, icon selection), customer feedback is collected.

[0032] Map these feedback data into a predefined user profile model to generate quantified feature vectors. Combine the descriptive words generated by the large language model with the multi-dimensional user profile to form a complete demand expression.

[0033] For example: Customer demand: "I want a sweet fragrance suitable for summer, with a moderate price, and preferably reminiscent of the seaside." Conversion result: Descriptive words: "sweetness 3 / 5, freshness 4 / 5, oceanic tone 3 / 5, longevity 2 / 5."

[0034] User profile: "consumption ability: mid-end, interest graph: travel, photography, life cycle stage: young working class, emotional coordinates: light +2."

[0035] Integrate the above descriptive words and user profile as input for the subsequent matching of compound combinations by the system.

[0036] Step S12, based on the chemical mapping relationship in the preset knowledge graph, obtain the first compound combination corresponding to the demand description, wherein the compound combination contains several compounds.

[0037] The preset knowledge graph can be an odor molecule feature knowledge graph, which contains a physical and chemical property database of 25000+ fragrance ingredients, referencing the FEMA (Flavor Extract Manufacturers Association) standard, etc. It contains detailed chemical properties (such as molecular structure, volatility, odor intensity) and corresponding sensory labels (such as floral, fruity, woody aroma, etc.).

[0038] The odor molecule feature knowledge graph links the structure and chemical properties of odor molecules to the odor characteristics they produce. It is crucial for understanding how different compounds produce specific odors and designing new fragrances and flavor substances. Its mapping relationship includes: Relationship between molecular structure and odor: describes the association between specific chemical functional groups (such as aldehyde groups, ketone groups, alcohol hydroxyl groups, etc.) or molecular structures (such as linear, cyclic, aromatic, etc.) and odor types.

[0039] Volatility and odor intensity: describes the relationship between the volatility of a compound and its ability to emit odor in the air, as well as how these properties affect the intensity of human olfactory perception.

[0040] Chemical stability and longevity: analyzes the stability of compounds under different environmental conditions (such as temperature, humidity) and its impact on odor longevity.

[0041] Odor threshold: The minimum concentration at which a certain odor compound can be perceived by the human nose.

[0042] Interaction effects: Include possible interactions between different odor molecules, such as synergy, masking effect or contrast enhancement, which can affect the final sensory experience.

[0043] Biological activity and safety: Consider the potential impact of compounds on human health, including allergic reactions, toxicity and other safety-related information.

[0044] Synthesis path and origin: Provide information on how to synthesize specific odor compounds, as well as their natural sources, such as plant extracts, animal secretions or other natural resources.

[0045] Application areas: Indicate in which product categories various odor compounds are usually applied, such as perfumes, food flavorings, cleaning products, etc.

[0046] For example: Wetness mapping is provided using watermelon ketone to simulate the feeling of water molecules and geosmin to simulate the smell of mud after the rain.

[0047] Green vitality mapping is provided using chlorophyll to simulate the bitter green feeling of broken grass stems or using violet leaf absolute to provide a metallic green leaf fragrance; Mineral coolness mapping is provided using aldehyde C12MNA (methyl nonyl acetaldehyde) to provide a cool metallic edge; Air permeability mapping is provided using Floralozone® (ozone floral molecule) to provide a light air feeling.

[0048] Step S13, based on reinforcement learning, optimizing the component ratio of the first compound to obtain the second compound combination after optimization; It includes: constructing a black box function, where the variable of the black box function is the component ratio of the first compound, and the dependent variable is the user feedback; determining the next component ratio of the first compound based on the acquisition function UCB; based on the black box function, iteratively compare the dependent variables corresponding to the component ratios, and when the preset iteration condition is met, obtain the second compound combination.

[0049] Specifically, a component ratio of the first compound can be randomly determined, and the user's feedback score is recorded. The entire optimization process is regarded as a black box function problem using a proxy model constructed by a Gaussian process, and the input variable is a vector representing the proportion of each compound.

[0050] Use the acquisition function UCB to find the next component ratio in the current model.

[0051] UCB is a commonly used acquisition function in Bayesian optimization, used to determine the next experimental point proportion combination.

[0052] Repeat the above until the optimal formula is found (i.e., the feedback score is the highest) or the upper limit of computing resources is reached.

[0053] Step S14, based on molecular dynamics, the optimized second compound combination is predicted in time sequence, and the time decay index and balance index are introduced to optimize the second compound combination again, and the third compound combination is obtained.

[0054] After obtaining the second compound combination optimized initially, the next step is to simulate the diffusion and volatilization behavior of the fragrance molecules in the air based on molecular dynamics, so as to predict the change trend of the fragrance over time. By introducing the time decay index (measuring the fragrance dissipation speed) and the balance index (evaluating the consistency of the release rhythm of different ingredients), the time dynamic characteristics of the fragrance are quantitatively evaluated.

[0055] Subsequently, under the premise of meeting the sensory and safety constraints, the ingredient proportion is adjusted by combining the optimization algorithm, so that the fragrance is not only pleasant at the initial moment, but also stable, coordinated and persistent during the entire release process. Finally, the third compound combination is obtained, which has the advantages of better time sequence control and overall balance of fragrance release, and is more in line with the needs of "fragrance life cycle" in actual use scenarios.

[0056] The time decay function includes:

[0057] Wherein, is the fragrance intensity of the first compound combination at the moment, is the fragrance intensity of the compound combination at the initial moment, is the decay coefficient of the first compound combination.

[0058] Balance function:

[0059] Wherein, is the balance index at the moment, is the number of moments, is the average fragrance intensity.

[0060] Step S15, according to the taboo index and the non-addition restriction index, the third compound combination is optimized to obtain the fourth compound combination.

[0061] Specifically comprising: judging whether there is a compound in the third compound combination that is a contraindicated index (a compound that cannot be added according to regulations) or an unaddable limited index (a compound that cannot be added artificially); if so, after removing the compound, determining a compound similar in chemical properties to the removed compound based on a preset knowledge graph, and adding the chemically similar compound to the third compound combination.

[0062] Step S16, taking the fourth compound combination as the target compound combination of the target customer, and controlling the associated device to output the target compound combination.

[0063] The associated device is a device for manufacturing compounds, which can be connected to mainstream fragrance blending devices through a device control protocol converter, convert the formula instructions into control commands recognizable by the device, and ensure the stable and accurate blending process of various high-precision fragrances with the help of a real-time viscosity monitoring and flow feedback system, so as to realize the automation and repeatable execution of the formula.

[0064] Further comprising: collecting user feedback after using the target compound combination; and re-optimizing the proportion of compounds in the target compound combination according to the user feedback.

[0065] Specifically comprising: Collecting user feedback: satisfaction, irritancy, persistence, etc. to build a feedback-formula mapping model Using a neural network to model the relationship between formula proportion and feedback, and outputting the influence factor of different component proportions on different dimension feedback.

[0066] Taking the maximization of user satisfaction as the objective function, and considering the regulatory constraints, cost, brand restrictions, etc. as boundary conditions. The optimization algorithm automatically searches for a new formula Each iteration generates a new combination, estimates the satisfaction score, and continuously optimizes.

[0067] To sum up, the present application provides a fragrance descriptor quantification method based on artificial intelligence large model multi-agent collaboration, comprising: converting the demand of a target customer to obtain a demand description corresponding to the demand; based on the chemical mapping relationship in the preset knowledge graph, obtaining a first compound combination corresponding to the demand description; based on reinforcement learning, optimizing the component ratio of the first compound to obtain an optimized second compound combination; based on molecular dynamics, predicting the fragrance volatility of the optimized second compound combination in time sequence, and introducing time decay index and balance index to optimize the second compound combination again to obtain a third compound combination; according to the taboo index and the non-addition restriction index, the third compound combination is optimized to obtain a fourth compound combination; the fourth compound combination is taken as the target compound combination of the target customer, and the associated equipment is controlled to output the target compound combination. The present application solves the problem that traditional fragrance modulation cannot quantify customer abstract description through the cooperation mechanism of artificial intelligence large model and multi-agent, and breaks through the complete link from natural language understanding, chemical component mapping to equipment execution, realizes the precise modeling and automatic formula generation of fragrance demand, and improves the personalized customization ability and system intelligence level.

[0068] Based on the same inventive concept, the present application provides a fragrance descriptor quantification device based on artificial intelligence large model multi-agent collaboration as shown in Figure 2 The method comprises: A demand conversion module 21 is configured to convert the demand of a target customer to obtain a demand description corresponding to the demand; A mapping module 22 is configured to obtain a first compound combination corresponding to the demand description based on the chemical mapping relationship in the preset knowledge graph, wherein the compound combination comprises a plurality of compounds; A first optimization module 23 is configured to optimize the component ratio of the first compound based on reinforcement learning to obtain an optimized second compound combination; A second optimization module 24 is configured to predict the fragrance volatility of the optimized second compound combination in time sequence based on molecular dynamics, and introduce time decay index and balance index to optimize the second compound combination again to obtain a third compound combination; A third optimization module 25 is configured to optimize the third compound combination according to the taboo index and the non-addition restriction index to obtain a fourth compound combination; A compound output module 26 is configured to take the fourth compound combination as the target compound combination of the target customer, and control the associated equipment to output the target compound combination.

[0069] Based on the same inventive concept, the present application further provides an electronic device, comprising: A processor; A memory for storing processor executable instructions; The processor is configured to implement the method for quantifying flavor descriptors based on artificial intelligence large model multi-agent collaboration as provided in the foregoing.

[0070] Based on the same inventive concept, the present application also provides a non-transitory computer-readable storage medium, when the instructions in the storage medium are executed by the processor of the electronic device, the electronic device can execute the method for quantifying flavor descriptors based on artificial intelligence large model multi-agent collaboration as provided in the foregoing.

[0071] Since the electronic device introduced in the embodiment is the electronic device used to implement the method for processing information in the embodiment of the present application, based on the method for processing information introduced in the embodiment of the present application, those skilled in the art can understand the specific implementation of the electronic device of the embodiment and its various forms, so the electronic device how to implement the method in the embodiment of the present application is not introduced in detail. As long as the electronic device used to implement the method for processing information in the embodiment of the present application is implemented by those skilled in the art, it belongs to the scope of the present application.

[0072] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer usable storage media containing computer usable program code (including but not limited to disk storage, CD-ROM, optical storage, etc.).

[0073] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices produce a device that implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks. Figure 1 The device that implements the functions specified in one flow or multiple flows and / or blocks.

[0074] These computer program instructions can also be stored in a computer readable storage medium that can guide the computer or other programmable data processing devices to work in a specific way, so that the instructions stored in the computer readable storage medium produce a product including instruction devices, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 ​one or more processes and / or blocks Figure 1 the function specified in the one or more blocks.

[0075] These computer program instructions can also be loaded into computer or other programmable data processing devices to cause a series of operational steps to be performed on the computer or other programmable devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable devices provide steps for implementing the flow Figure 1 one or more processes and / or blocks Figure 1 the function specified in the one or more blocks.

[0076] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they learn of the basic inventive concepts. Such additional variations and modifications should be considered as within the scope of the application as defined by the claims appended hereto. Accordingly, it is intended that all such additional variations and modifications be included within the scope of the application.

[0077] It will be apparent to those skilled in the art that various modifications and variations can be made to the present application without departing from the spirit or scope of the application. Thus, it is intended that the present application cover modifications and variations of this application provided they come within the scope of the appended claims and their equivalents.

Claims

1. A method for quantifying flavor descriptors based on the cooperation of multiple artificial intelligence large models and multiple agents, characterized in that, The method comprises the following steps: Converting the demand of the target customer to obtain a demand description corresponding to the demand; Based on the chemical mapping relationship in the preset knowledge graph, a first compound combination corresponding to the demand description is obtained, wherein the compound combination contains a plurality of compounds; Based on reinforcement learning, the proportion of the first compound is optimized to obtain an optimized second compound combination; Based on molecular dynamics, the fragrance volatility of the optimized second compound combination is predicted in time sequence, and time decay index and balance index are introduced to optimize the second compound combination again to obtain a third compound combination; According to the taboo index and the non-addition restriction index, the third compound combination is optimized to obtain a fourth compound combination; The fourth compound combination is taken as the target compound combination of the target customer, and an associated device is controlled to output the target compound combination.

2. The method of claim 1, wherein the method is based on a plurality of artificial intelligence large models. The demand of the target customer is converted to obtain a demand description corresponding to the demand, including: Based on a large language model, the demand of the target customer is converted into a description word; Based on a preset visual interaction model, the demand of the target customer is converted into a multi-dimensional user portrait; The description word and the multi-dimensional user portrait are taken as the demand description. 3.The method of claim 1, wherein the method comprises: Based on reinforcement learning, the proportion of the first compound is optimized to obtain an optimized second compound combination, including: A black box function is constructed, wherein the variable of the black box function is the proportion of the first compound, and the dependent variable is the user feedback; Based on the acquisition function UCB, the next component proportion of the first compound is determined; Based on the black box function, the dependent variables corresponding to the component proportions are iteratively compared, and when the preset iteration condition is met, the second compound combination is obtained. 4.The method of claim 1, wherein the method comprises: According to the taboo index and the non-addition restriction index, the third compound combination is optimized to obtain a fourth compound combination, including: Determine whether there is a compound in the third compound combination that meets the taboo index or the non-addition restriction index; If so, after removing the compound, determine a compound similar in chemical properties to the removed compound based on the preset knowledge graph, and add the chemically similar compound to the third compound combination. 5.The method of claim 1, wherein the method further comprises: determining a plurality of AI model parameters based on the plurality of AI model parameters and the plurality of AI model parameters of the AI model; and determining a plurality of AI model parameters based on the plurality of AI model parameters and the plurality of AI model parameters of the AI model. Further comprising: After using the target compound combination, collect user feedback; According to the user feedback, the proportion of the compound in the target compound combination is optimized again. 6.The method of claim 1, wherein the method comprises: The multi-dimensional user portrait includes: Image selection, emotional coordinates, interest map, consumption ability and life cycle stage.

7. The method of claim 1, wherein the method is based on a plurality of artificial intelligence large models, and the plurality of artificial intelligence large models are trained by a plurality of agents. The time decay function includes: wherein is the compound combination in question is the compound combination in question is the intensity of the fragrance at the moment in question is the intensity of the fragrance of the compound combination at the initial moment is the compound combination in question is the decay coefficient of the compound combination in question 8. The device for quantifying the scent descriptors based on the multi-agent collaboration of the artificial intelligence large model, characterized in that, The method comprises: A demand conversion module for converting the demand of the target customer to obtain a demand description corresponding to the demand; A mapping module for obtaining a first compound combination corresponding to the demand description based on the chemical mapping relationship in the preset knowledge graph, wherein the compound combination contains a plurality of compounds; A first optimization module for optimizing the proportion of the first compound based on reinforcement learning to obtain an optimized second compound combination; A second optimization module for predicting the fragrance volatility of the optimized second compound combination in time sequence based on molecular dynamics, and introducing time decay index and balance index to optimize the second compound combination again to obtain a third compound combination; The third optimization module is configured to optimize the third compound combination according to the taboo index and the non-addition limitation index, and obtain a fourth compound combination. The compound output module is configured to output the fourth compound combination as a target compound combination of the target customer, and control an associated device to output the target compound combination.

9. An electronic device, comprising: The method comprises: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to execute to implement the sweet taste descriptor quantification method based on the artificial intelligence large model multi-agent collaboration according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, comprising: When the instructions in the non-transitory computer-readable storage medium are executed by the processor of the electronic device, the electronic device can execute the sweet taste descriptor quantification method based on the artificial intelligence large model multi-agent collaboration according to any one of claims 1 to 7.

Images