Method, device and system for evaluating regulatory violations and reliability of functional ingredients based on artificial intelligence

Abstract

According to one embodiment, in a method for evaluating the legal violation status and reliability of an artificial intelligence-based functional ingredient performed by a device, the method comprises: receiving a request for analysis of a first product from a first user terminal; identifying each ingredient included in the first product based on an image of the first product or information of the first product; selecting a first ingredient, which is one of the ingredients included in the first product, as an analysis target; generating a first question requesting an analysis of whether the first ingredient violates regulations; inputting the first question into a first artificial intelligence model trained to analyze and answer whether the ingredient violates regulations; when the first artificial intelligence model generates a first answer as an answer to the first question, obtaining the first answer as an output result from the first artificial intelligence model; setting whether the first ingredient violates regulations based on the first answer; generating a second question requesting a reliability analysis of the first ingredient; and inputting the second question into a second artificial intelligence model trained to analyze and answer the reliability of the ingredient. A method for evaluating the legal violation status and reliability of an artificial intelligence-based functional ingredient is provided, comprising: a step of obtaining the second answer as an output result from the second artificial intelligence model when the second artificial intelligence model generates a second answer as an answer to the second question; a step of setting the reliability of the first ingredient based on the second answer; a step of generating an analysis report on the ingredients of the first product based on the legal violation status and reliability of each ingredient included in the first product when the legal violation status and reliability of each ingredient included in the first product are set; and a step of transmitting the analysis report on the ingredients of the first product to the first user terminal.

Description

Technology Field

[0001] The following examples relate to technology for evaluating the reliability and legal compliance of artificial intelligence-based functional ingredients. Background Technology

[0002] As various functional products, such as cosmetics, health supplements, and beauty products, become more widespread in the market, there is an increasing need for consumers and manufacturers to accurately determine whether the ingredients included in these products comply with national regulations.

[0003] Regulatory agencies in each country separately manage prohibited ingredients, restricted ingredients, permissible standards, and usage limits according to product categories, and since applicable standards differ from country to country even for the same raw material, a complex regulatory review process must be carried out to distribute products in multiple countries.

[0004] However, these regulations are continuously revised, and the volume of documents and academic materials regarding the safety of each raw material is vast, requiring a significant amount of time and specialized knowledge for a person to review and make a judgment on all the information.

[0005] Previously, it was common practice to manually check ingredients on product labels and individually consult legal documents; even when some automated systems existed, they provided only limited analysis capabilities at the level of simple keyword matching, making it difficult to adequately assess actual regulatory violations or the basis for safety.

[0006] In particular, although the reliability of raw material information is a crucial factor separate from regulatory compliance, existing systems lacked the capability to quantitatively evaluate reliability by integrating various evidence, such as safety studies, adverse effect reports, and market usage history for each raw material.

[0007] As a result, the problem of manufacturers, distributors, and consumers all facing difficulties in making accurate judgments has persisted. Accordingly, there is a need for an AI-based architecture capable of evaluating not only regulatory compliance by raw material but also reliability, and research is required on sophisticated automated analysis technologies that comprehensively reflect various criteria. Prior art literature

[0008] Korean Registered Patent No. 10-2607449 Korean Registered Patent No. 10-2494940 Korean Published Patent No. 10-2022-0143974 Korean Published Patent No. 10-2020-0003563 The problem to be solved

[0009] According to one embodiment, the purpose is to provide a method, apparatus, and system for evaluating whether an artificial intelligence-based functional ingredient violates regulations and for evaluating reliability.

[0010] The objectives of the present invention are not limited to those mentioned above, and other unmentioned objectives will be clearly understood from the description below. means of solving the problem

[0011] According to one embodiment, in a method for evaluating the legal violation status and reliability of an artificial intelligence-based functional ingredient performed by a device, the method comprises: receiving a request for analysis of a first product from a first user terminal; identifying each ingredient included in the first product based on an image of the first product or information of the first product; selecting a first ingredient, which is one of the ingredients included in the first product, as an analysis target; generating a first question requesting an analysis of whether the first ingredient violates regulations; inputting the first question into a first artificial intelligence model trained to analyze and answer whether the ingredient violates regulations; when the first artificial intelligence model generates a first answer as an answer to the first question, obtaining the first answer as an output result from the first artificial intelligence model; setting whether the first ingredient violates regulations based on the first answer; generating a second question requesting a reliability analysis of the first ingredient; and inputting the second question into a second artificial intelligence model trained to analyze and answer the reliability of the ingredient. A method for evaluating the legal violation status and reliability of an artificial intelligence-based functional ingredient is provided, comprising: a step of obtaining the second answer as an output result from the second artificial intelligence model when the second artificial intelligence model generates a second answer as an answer to the second question; a step of setting the reliability of the first ingredient based on the second answer; a step of generating an analysis report on the ingredients of the first product based on the legal violation status and reliability of each ingredient included in the first product when the legal violation status and reliability of each ingredient included in the first product are set; and a step of transmitting the analysis report on the ingredients of the first product to the first user terminal.

[0012] When the first artificial intelligence model inputs the first question, it confirms through the first question that the first raw material is the subject of analysis, identifies the country of analysis as the first country based on the location of the first user terminal or the nationality of the first user, identifies the category of the first product containing the first raw material as the first category, compiles the raw materials prohibited from use in products of the first category in the first country, generates a first list, and if it is confirmed that the first raw material is included in the first list, it generates the first answer indicating that the first raw material has violated regulations, and if it is confirmed that the first raw material is not included in the first list, it identifies the content of the first raw material included in the first product as the first content, identifies the maximum content of the first raw material permitted for use in products of the first category in the first country as the second content, and if it is confirmed that the first content is higher than the second content, it generates the first answer indicating that the first raw material has violated regulations, and if it is confirmed that the first content is not higher than the second content, The above first answer can be generated as the above first raw material does not violate regulations.

[0013] When the second question is input, the second artificial intelligence model confirms that the first raw material is the subject of analysis through the second question, classifies papers mentioning the first raw material that were published during a preset reference period into a first group, classifies papers classified in the first group that mention efficacy regarding the first raw material into a first-1 group, and classifies papers classified in the first group that mention side effects regarding the first raw material into a first-2 group, identifies the number of papers classified in the first group as a first value, identifies the number of papers classified in the first-1 group as a second value, identifies the number of papers classified in the first-2 group as a third value, calculates a first ratio by dividing the second value by the first value, calculates a second ratio by dividing the third value by the first value, assigns a higher first score within a preset first score range the higher the first ratio, and assigns a higher second score within a preset second score range the lower the second ratio. Assigning a score, classifying patents disclosed during the reference period among the patents mentioning the first raw material into a second group, classifying patents among the patents classified into the second group that mention efficacy regarding the first raw material into a second-1 group, classifying patents among the patents classified into the second group that mention side effects regarding the first raw material into a second-2 group, confirming the number of patents classified into the second group as a fourth value, confirming the number of patents classified into the second-1 group as a fifth value, confirming the number of patents classified into the second-2 group as a sixth value, calculating a third ratio by dividing the fifth value by the fourth value, and calculating a fourth ratio by dividing the sixth value by the fourth value.The higher the above third ratio, the higher the third score is assigned within the preset third score range; the lower the above fourth ratio, the higher the fourth score is assigned within the preset fourth score range; among the clinical data from public institutions mentioning the above first raw material, the clinical data disclosed during the above reference period is classified into the third group; among the clinical data classified into the above third group, the clinical data mentioned as having efficacy for the above first raw material is classified into the third-1 group; among the clinical data classified into the above third group, the clinical data mentioned as having side effects for the above first raw material is classified into the third-2 group; the number of clinical data classified into the above third group is confirmed as the seventh value; the number of clinical data classified into the above third-1 group is confirmed as the eighth value; the number of clinical data classified into the above third-2 group is confirmed as the ninth value; the fifth ratio is calculated by dividing the above eighth value by the above seventh value; the sixth ratio is calculated by dividing the above ninth value by the above seventh value; and the higher the above fifth ratio, the higher the fifth score is assigned within the preset fifth score range. A score is assigned, and the lower the sixth ratio, the higher the sixth score is assigned within the preset range of sixth scores, and the first to sixth scores are summed to calculate a total sum, and the second answer can be generated as the reliability of the first raw material is the total sum. Effects of the invention

[0014] According to one embodiment, by providing a system capable of automatically evaluating compliance with national regulations and reliability for various raw materials included in functional products, the analysis process, which previously required a person to manually search through various laws, lists of prohibited and restricted ingredients, and safety-related documents to make judgments, can be significantly shortened, and the accuracy and consistency of product safety judgments can be significantly improved.

[0015] In addition, according to one embodiment, by automatically reflecting various conditions such as national regulatory frameworks, acceptance standards by product category, and usage restrictions, conformity can be quickly determined even for the same product when the country of sale or applicable laws differ. Through this, risk factors in the global distribution environment can be identified in advance, and the cost and time required to continuously manage product compliance with regulations can be reduced.

[0016] Meanwhile, the effects according to the embodiments are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0017] FIG. 1 is a schematic diagram showing the configuration of a system according to one embodiment. FIG. 2 is a flowchart illustrating the process of evaluating whether an artificial intelligence-based functional ingredient violates regulations and assesses reliability according to one embodiment. FIG. 3 is a flowchart illustrating the process of analyzing whether a raw material violates regulations according to one embodiment. FIGS. 4 and 5 are flowcharts for explaining the process of analyzing the reliability of raw materials according to one embodiment. FIG. 6 is a flowchart illustrating the process of recommending a substitute raw material for a raw material that violates regulations according to one embodiment. FIG. 7 is a flowchart illustrating the process of recommending a substitute raw material for a raw material with low reliability according to one embodiment. FIGS. 8 and 9 are flowcharts for explaining the process of displaying an analysis report according to one embodiment. FIG. 10 is an example diagram of the configuration of a device according to one embodiment. Specific details for implementing the invention

[0018] Hereinafter, embodiments are described in detail with reference to the attached drawings. However, various modifications may be made to the embodiments, and thus the scope of the patent application is not limited or restricted by these embodiments. It should be understood that all modifications, equivalents, and substitutions to the embodiments are included within the scope of the rights.

[0019] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed forms, and the scope of this specification includes modifications, equivalents, or substitutions that fall within the technical concept.

[0020] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0021] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0022] The terms used in the embodiments are for illustrative purposes only and should not be interpreted as intended to be limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "having" are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0023] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the embodiments pertain. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0024] In addition, when describing with reference to the attached drawings, identical components are assigned the same reference numeral regardless of drawing symbols, and redundant descriptions thereof are omitted. In describing the embodiments, if it is determined that a detailed description of related prior art could unnecessarily obscure the essence of the embodiments, such detailed description is omitted.

[0025] The embodiments can be implemented in various forms of products such as personal computers, laptop computers, tablet computers, smartphones, televisions, smart home appliances, intelligent automobiles, kiosks, and wearable devices.

[0026] In the embodiments, the Artificial Intelligence (AI) system is a computer system that implements human-level intelligence, and unlike existing rule-based smart systems, it is a system in which the machine learns and makes judgments on its own. As the AI ​​system improves in recognition rate and becomes capable of understanding sellers' preferences more accurately with use, existing rule-based smart systems are gradually being replaced by deep learning-based AI systems.

[0027] Artificial intelligence technology consists of machine learning and component technologies that utilize machine learning. Machine learning is an algorithmic technology that autonomously classifies and learns the characteristics of input data, while component technologies are technologies that mimic the functions of the human brain, such as cognition and judgment, by utilizing machine learning algorithms like deep learning; these technologies comprise fields such as linguistic understanding, visual understanding, reasoning / prediction, knowledge representation, and motion control.

[0028] The various fields where artificial intelligence technology is applied are as follows. Linguistic understanding refers to the technology of recognizing, applying, and processing human language and text, including natural language processing, machine translation, dialogue systems, question answering, and speech recognition / synthesis. Visual understanding refers to the technology of perceiving and processing objects like human vision, including object recognition, object tracking, image search, person recognition, scene understanding, spatial understanding, and image enhancement. Inference and prediction refers to the technology of logically reasoning and predicting by judging information, including knowledge / probability-based inference, optimization prediction, preference-based planning, and recommendation. Knowledge representation refers to the technology of automatically processing human experiential information into knowledge data, including knowledge construction (data generation / classification) and knowledge management (data utilization). Motion control refers to the technology of controlling autonomous driving of vehicles and the movement of robots, including motion control (navigation, collision, driving) and manipulation control (behavior control).

[0029] Generally, to apply machine learning algorithms to real-world situations, training is performed using a trial-and-error method due to the inherent characteristics of the fundamental methodologies. In particular, deep learning requires hundreds of thousands of iterations. Since it is impossible to execute this in a real physical external environment, training is instead performed through simulations that virtually recreate the actual physical environment on a computer.

[0030] FIG. 1 is a schematic diagram showing the configuration of a system according to one embodiment.

[0031] Referring to FIG. 1, a system according to one embodiment may include a plurality of user terminals (100) and devices (200) capable of communicating with each other through a communication network.

[0032] First, the communication network can be configured regardless of the mode of communication, such as wired or wireless, and can be implemented in various forms to enable communication between servers and between servers and terminals.

[0033] Each of the multiple user terminals (100) can be implemented as a computing device equipped with a communication function, and, for example, can be implemented as a mobile phone, desktop PC, laptop PC, tablet PC, smartphone, etc., but is not limited thereto, and can be implemented as various types of communication devices that can be connected to an external server.

[0034] A plurality of user terminals (100) are terminals used by users, and may include a first user terminal (110) used by a first user, a second user terminal (120) used by a second user, etc.

[0035] Each of the plurality of user terminals (100) may be configured to perform all or part of the computational functions, storage / reference functions, input / output functions, and control functions that a conventional computer has. The plurality of user terminals (100) may be configured to communicate with the device (200) via wired or wireless means.

[0036] Each of the plurality of user terminals (100) may be connected to a web page operated by a person or organization providing a service using the device (200), or may have an application developed and distributed by a person or organization providing a service using the device (200) installed. Each of the plurality of user terminals (100) may be linked with the device (200) through a web page or an application.

[0037] Each of the multiple user terminals (100) can access the device (200) through a web page, application, etc. provided by the device (200).

[0038] The device (200) may be a private server owned by a person or organization providing a service using the device (200), a cloud server, or a peer-to-peer (P2P) set of distributed nodes. The device (200) may be configured to perform all or part of the computational functions, storage / reference functions, input / output functions, and control functions that a conventional computer possesses. The device (200) may be equipped with at least one artificial intelligence model that performs inference functions.

[0039] The device (200) may be configured to communicate with a plurality of user terminals (100) via wired or wireless means, control the operation of each of the plurality of user terminals (100), and control which information to display on the screen of each of the plurality of user terminals (100).

[0040] The device (200) is implemented as a server for evaluating whether functional ingredients violate regulations and for evaluating reliability, and can provide various services related thereto. Here, functional ingredients may refer to ingredients included in a product to implement or enhance the efficacy, performance, or physiological action of a specific product.

[0041] According to one embodiment, functional ingredients can be used in various functional product lines, such as cosmetics, health functional foods, medical and beauty treatment products, and personal care products, and can broadly include functions affecting the human body, ranging from external functions such as whitening, wrinkle improvement, UV protection, moisturizing, soothing, and elasticity improvement, to internal functions such as nutrient supply, metabolic regulation, and antioxidant action. In this case, functional ingredients can be composed in various forms, such as natural products, synthetic components, bio-based components, fermented products, peptides, vitamins, minerals, and plant extracts, and the intensity or stability of function expression may vary depending on the content, combination, and application method during the product manufacturing process. Furthermore, since the permissibility, usage limits, and certification standards for functional ingredients may vary depending on national regulations, they can be understood as core components that must be analyzed to determine whether a product complies with regulations.

[0042] Meanwhile, for convenience of explanation, only the first user terminal (110) and the second user terminal (120) among the plurality of user terminals (100) are shown in FIG. 1, but the number of terminals can vary depending on the embodiment. As long as the processing capacity of the device (200) allows, there is no particular limit to the number of terminals.

[0043] According to one embodiment, the device (200) can analyze whether the functional ingredient violates regulations based on artificial intelligence and analyze the reliability of the functional ingredient.

[0044] In the present invention, Artificial Intelligence (AI) refers to a technology that imitates human learning ability, reasoning ability, and perceptual ability, and implements them on a computer, and may include concepts such as machine learning and symbolic logic. Machine Learning (ML) is an algorithmic technology that classifies or learns the characteristics of input data on its own. AI technology can analyze input data as a machine learning algorithm, learn from the results of the analysis, and make judgments or predictions based on the results of the learning. Furthermore, technologies that mimic the functions of the human brain, such as cognition and judgment, by utilizing machine learning algorithms can also be understood as falling within the category of AI. For example, technological fields such as linguistic understanding, visual understanding, reasoning / prediction, knowledge representation, and motion control may be included.

[0045] Machine learning can refer to the process of training neural network models using experience in processing data. It implies that through machine learning, computer software improves its own data processing capabilities. A neural network model is constructed by modeling the correlations between data, and these correlations can be expressed by multiple parameters. A neural network model extracts and analyzes features from given data to derive correlations between them; machine learning can be defined as the process of optimizing the model's parameters by repeating this process. For example, a neural network model can learn the mapping (correlation) between inputs and outputs for data given as input-output pairs. Alternatively, even when only input data is provided, a neural network model can derive regularities between the given data and learn those relationships.

[0046] An artificial intelligence learning model or neural network model can be designed to implement the structure of the human brain on a computer and may include multiple network nodes that have weights and simulate neurons of a human neural network. The multiple network nodes may have interconnected relationships by simulating the synaptic activity of neurons, where neurons exchange signals through synapses. In an artificial intelligence learning model, multiple network nodes may be located in layers of different depths and exchange data according to convolutional connections. The artificial intelligence learning model may be, for example, an Artificial Neural Network (ANN) or a Convolutional Neural Network (CNN). As an embodiment, the artificial intelligence learning model may be machine learned according to methods such as supervised learning, unsupervised learning, and reinforcement learning. Machine learning algorithms for performing machine learning may include Decision Tree, Bayesian Network, Support Vector Machine, Artificial Neural Network, Ada-boost, Perceptron, Genetic Programming, and Clustering.

[0047] Among these, CNNs are a type of multilayer perceptron designed to use minimal preprocessing. CNNs consist of one or more convolutional layers and standard artificial neural network layers stacked on top, additionally utilizing weights and pooling layers. Thanks to this structure, CNNs can fully utilize two-dimensional input data. Compared to other deep learning architectures, CNNs demonstrate good performance in both image and audio fields. CNNs can also be trained using standard backpropagation. CNNs have the advantage of being easier to train than other feedforward artificial neural network techniques and using a small number of parameters.

[0048] Convolutional networks are neural networks comprising sets of nodes with bounded parameters. Many computer vision tasks have been significantly improved, driven by the increased size of available training data and the availability of computational power, combined with algorithmic advancements such as discriminative linear units and dropout training. In the case of massive datasets, such as those available for many tasks today, outfitting is not critical, and increasing the network size improves test accuracy. Optimal utilization of computing resources becomes a limiting factor. To address this, distributed, scalable implementations of deep neural networks can be employed.

[0049] FIG. 2 is a flowchart illustrating the process of evaluating whether an artificial intelligence-based functional ingredient violates regulations and assesses reliability according to one embodiment.

[0050] Referring to FIG. 2, first, in step S201, the device (200) may receive an analysis request for a first product from a first user terminal (110). At this time, the analysis request may include product images, product label information, product name, product category information, etc., for identifying the first product, and the format of the analysis request may include various forms such as image uploads, OCR-recognized character information, and product data provided through API integration. To this end, the device (200) may identify that the product for analysis request is the first product based on the information included in the analysis request.

[0051] According to one embodiment, the first user terminal (110) can take a picture of the first product through a camera module. At this time, the part where the raw material of the first product is displayed can be taken. Subsequently, the first user terminal (110) can transmit an analysis request for the first product, including the image of the first product taken, to the device (200), and the device (200) can verify the image of the first product taken through the analysis request.

[0052] According to another embodiment, when the product name of the first product is entered by the first user terminal (110) through a text input window or voice recognition, the first user terminal (110) can transmit an analysis request for the first product including the product name of the first product to the device (200), and when the device (200) confirms the product name of the first product through the analysis request, the device (200) can obtain the first product information. Here, the first product information may include various information related to the product, such as the product name, category, manufacturer, type of raw material, and content of raw material, and may be obtained through the device (200)'s database, an external server, information posted on the web, etc.

[0053] In step S202, the device (200) can identify each of the raw materials included in the first product based on an image of the first product or information about the first product. At this time, the first product may be a functional product designed to exhibit specific efficacy or function, such as whitening, wrinkle improvement, UV protection, moisturizing, or nutrient supply, and each of the raw materials included in the first product may correspond to a functional raw material as a component that contributes to the manifestation of efficacy of the functional product.

[0054] Specifically, when the device (200) receives an image-based analysis request for a first product from a first user terminal (110), it recognizes the ingredient label area where ingredients are indicated within the image based on the image of the first product, and extracts the text contained in that area through an OCR engine to identify each ingredient included in the first product. At this time, the device (200) can be configured to apply a dictionary-based normalization algorithm, a typographical correction module, and international INCI name mapping rules to correct character recognition errors that may occur due to the shape of the label, lighting, distortion, language differences, etc., so that the final list of ingredients included in the first product is stably derived. In addition, if additional information such as ingredient descriptions, complex ingredients, or concentration indications are included in the ingredient list in parentheses, the device (200) can structure this information to additionally verify the efficacy, side effects, content, etc., for each ingredient.

[0055] According to another embodiment, when the device (200) receives an analysis request based on the product name of the first product from the first user terminal (110), it can obtain information about the first product through a database, an external server, information posted on the web, etc., and then verify each of the ingredients included in the first product based on the obtained information about the first product. To this end, the first product information may include information regarding the ingredient name, efficacy, side effects, content, etc., for each ingredient included in the first product.

[0056] Through such a processing method, the device (200) can use either one or both of the image-based raw material verification method and the product information-based raw material verification method, and can finally accurately verify each raw material included in the first product.

[0057] In step S203, the device (200) may select a first raw material, which is one of the raw materials included in the first product, as the subject of analysis.

[0058] That is, the device (200) can select one of the raw materials confirmed to be included in the first product and determine it as the first raw material, and perform an analysis to evaluate whether the first raw material violates regulations and its reliability. At this time, the selection of raw materials may use a sequential iterative structure or a priority-based structure. For example, if a user requests an analysis of a specific raw material first, that raw material may be set as the first raw material, otherwise, the first raw material in the extracted raw material list may be automatically set as the first raw material.

[0059] In step S204, the device (200) may generate a first question requesting an analysis of whether the first raw material violates regulations.

[0060] That is, the device (200) can automatically construct a first question in a form that an artificial intelligence model can understand to analyze whether the first raw material violates regulations. For example, the first question can be constructed in the form of a clear query, such as “Does the first raw material violate regulations?”

[0061] In step S205, the device (200) may input a first question to a first artificial intelligence model trained to analyze whether the raw material violates regulations and provide an answer. Here, the first artificial intelligence model may be in a state of being pre-trained to analyze whether a specific raw material violates regulations and provide an answer.

[0062] According to one embodiment, the first artificial intelligence model may be a generative artificial intelligence model. Here, generative artificial intelligence is a type of artificial intelligence specialized in generating new data, and is an artificial intelligence technology that creates various content such as text, audio, and images.

[0063] That is, the first artificial intelligence model is a conversational artificial intelligence model that analyzes text to identify a question and generates an answer to the question. It may be a model that identifies which raw material is the subject of analysis through the question, analyzes whether the raw material violates regulations, and outputs the analyzed result as an answer. In this case, the first artificial intelligence model can analyze whether the raw material violates regulations by determining whether it is a prohibited ingredient or whether there is a content limit when the raw material is used in a specific category of products in a specific country. A detailed explanation regarding this will be described later with reference to Fig. 3.

[0064] According to one embodiment, the first artificial intelligence model is a regulatory judgment model designed to automatically determine whether a functional ingredient violates regulations, and may be configured to include a natural language processing module for interpreting an input question, a rule-based module for matching country and category information, an information processing module for normalizing ingredient information and content, and a classification module that finally performs a regulatory judgment. In this case, the natural language processing module can contextually recognize ingredient names, product information, country information, etc. included in the question using a Transformer structure or an LLM-based language model, and the classification module may be configured with an output node to determine whether a specific ingredient corresponds to a prohibited ingredient or exceeds the permitted content.

[0065] Additionally, the first artificial intelligence model internally maintains a mapping table that integrates raw material names, INCIs, synonyms, code systems, etc., for linkage with the database of the device (200), and may include a multi-condition branching structure to apply different regulatory conditions depending on the country and product category. This structure can be designed to perform all interpretation processes necessary for regulatory judgment using only the information included in the question.

[0066] In step S206, if the device (200) generates a first answer as an answer to the first question in the first artificial intelligence model, it can obtain the first answer as an output result from the first artificial intelligence model.

[0067] For example, if the device (200) inputs a first question, “Does the first raw material violate regulations?” into the first artificial intelligence model, and a first answer, “The first raw material does not violate regulations,” is generated through the first question, the device can obtain the first answer from the first artificial intelligence model.

[0068] In other words, the first AI model can identify the raw material to be analyzed through a question, analyze whether the raw material violates regulations, and output the analyzed result as an answer. To this end, the first AI model may be pre-trained to analyze whether the raw material violates regulations by considering factors such as whether it is a prohibited ingredient or if content limits exist when the raw material is used in a specific category of products in a specific country.

[0069] Specifically, the first AI model can be trained based on a large-scale regulatory dataset that includes regulatory documents from various countries and industries, lists of prohibited ingredients, standards for permitted ingredients and their content, and product ingredient table data. During the training process, the first AI model can be trained using a supervised learning method to understand the relationship between ingredient names and country-specific regulatory information. Furthermore, it can enhance its regulatory judgment capabilities by repeatedly receiving training samples that include labels indicating whether each ingredient is a prohibited substance, the maximum content allowed in a specific category, and usage conditions.

[0070] Furthermore, the first AI model may undergo pre-training or fine-tuning using a text corpus consisting of regulatory documents, legal manuals, and product information to secure language model-based natural language processing capabilities. In addition, by utilizing actual product ingredient lists and content data to learn the determination of regulatory violations based on content comparison, the first AI model can be processed to normalize and make judgments regarding various notation methods and unit systems. During the training process, evaluation indicators such as precision and recall are used to assess whether the output of the first AI model aligns with actual regulatory standards, and cases of misjudgment can be supplemented with additional training data to improve judgment accuracy. Through such training procedures, the first AI model can be trained to determine legal violations by consistently reflecting regulatory standards, even when new raw materials, products, or countries are input.

[0071] In addition to the learning method described above, the first artificial intelligence model may also be trained through various learning methods of the known artificial intelligence model.

[0072] In step S207, the device (200) can determine whether the first raw material violates regulations based on the first answer.

[0073] For example, if the device (200) confirms that the first answer is “The first raw material does not violate regulations,” it can set the violation of regulations of the first raw material to “no violation of regulations.”

[0074] The device (200) can determine whether each raw material included in the first product violates regulations through the same method as determining whether the first raw material violates regulations, and to this end, steps S204 through S207 can be performed repeatedly.

[0075] In step S208, the device (200) can generate a second question requesting a reliability analysis for the first raw material.

[0076] That is, the device (200) can automatically construct a second question in a form that an artificial intelligence model can understand to analyze the reliability of the first raw material. For example, the second question can be constructed in the form of a clear query, such as “What is the reliability of the first raw material?”

[0077] In step S209, the device (200) may input a second question into a second artificial intelligence model trained to analyze the reliability of raw materials and provide an answer. Here, the second artificial intelligence model may be in a state of being pre-trained to analyze the degree of reliability of a specific raw material and provide an answer.

[0078] According to one embodiment, the second artificial intelligence model may be a generative artificial intelligence model. Here, generative artificial intelligence is a type of artificial intelligence specialized in generating new data, and is an artificial intelligence technology that creates various content such as text, audio, and images.

[0079] That is, the second artificial intelligence model is a conversational artificial intelligence model that analyzes text to identify questions and generates answers to questions. It may be a model that identifies which raw material is the subject of analysis through the question, analyzes the reliability of the raw material, and outputs the analyzed results as answers. In this case, the second artificial intelligence model can analyze the reliability of the raw material based on papers mentioning efficacy of the raw material, papers mentioning side effects of the raw material, patents mentioning efficacy of the raw material, patents mentioning side effects of the raw material, public data mentioning efficacy of the raw material, public data mentioning side effects of the raw material, etc. A detailed explanation related to this will be described later with reference to FIGS. 4 and 5.

[0080] According to one embodiment, the second artificial intelligence model is configured to analyze various data sources, such as papers, patents, and clinical data from public institutions, and convert them into quantified scores in order to calculate the reliability of a functional ingredient. It may be configured to include a natural language processing module, a literature classification module, a ratio calculation module, a score conversion module, and a reliability calculation module. In this case, the natural language processing module may perform the role of extracting the component to be analyzed from the input question and understanding text data such as the title, abstract, and body of a literature, or patent specifications and clinical reports; the literature classification module may filter each of the papers, patents, and clinical data according to a reference period and group literature mentioning efficacy and literature mentioning side effects; the ratio calculation module may calculate a ratio based on the number of literatures in the efficacy group and the side effects group; the score conversion module may be configured to assign scores within a score range corresponding to the ratio; and the reliability calculation module may derive a final reliability value by summing multiple scores calculated based on the literature, patents, and clinical data.

[0081] In other words, the second artificial intelligence model can integrate and analyze public evidence data of different natures through this structure and generate quantitative results regarding raw material reliability.

[0082] In step S210, if the device (200) generates a second answer as an answer to the second question in the second artificial intelligence model, it can obtain the second answer as an output result from the second artificial intelligence model.

[0083] For example, if the device (200) inputs a second question, “What is the reliability of the first raw material?” into the second artificial intelligence model, and a second answer, “The reliability of the first raw material is 300,” is generated through the second question, the device can obtain the second answer from the second artificial intelligence model.

[0084] In other words, the second AI model can identify the raw material to be analyzed through a question, analyze the reliability of the raw material, and output the analyzed result as an answer. To this end, the second AI model may be pre-trained to analyze the reliability of the raw material based on papers mentioning efficacy, papers mentioning side effects, patents mentioning efficacy, patents mentioning side effects, public data mentioning efficacy, and public data mentioning side effects.

[0085] Specifically, the second AI model can be trained using various types of literature data and public databases to perform information processing necessary for calculating the reliability of functional ingredients. During the training process, the second AI model can be trained using a supervised learning method to identify the relationship between ingredient names and expressions related to efficacy and side effects for each of the papers, patents, and clinical data. To this end, training samples may be provided in which literature mentioning the ingredient is labeled as evidence of efficacy, evidence of side effects, or neutral literature; the second AI model can then learn text patterns, keywords, and sentence structures within the literature to recognize semantic associations between the ingredient and its effects or side effects.

[0086] In addition, the second artificial intelligence model learns how to process literature metadata (publication date, release date, etc.) for reference period filtering, and can learn parameters for rule-based or learning-based algorithms that calculate the efficacy ratio and side effect literature ratio relative to the number of literature and convert them into scores.

[0087] In the case of patent data, the learning process may include analyzing technical content regarding ingredient mentions and effects across various domains, such as patent specifications, abstracts, and claims, and fine-tuning of the language model may be performed to understand the technical sentences unique to patent literature. For clinical data as well, the second AI model may be trained by extracting meaning from the structure of clinical reports, trial summaries, and clinical results. During the learning process, a loss function may be used to minimize the difference between the calculated score and the reference reliability score evaluated by actual experts, thereby continuously improving the reliability calculation performance of the second AI model. Finally, through learning, the second AI model acquires the ability to comprehensively reflect the strength of evidence from various data sources, enabling it to calculate consistent reliability scores for new ingredients based on literature trends within the reference period.

[0088] In addition to the learning method described above, the second artificial intelligence model may also be trained through various learning methods of the known artificial intelligence model.

[0089] In step S211, the device (200) can set the reliability of the first raw material based on the second answer.

[0090] For example, the device (200) can set the reliability of the first raw material to “300” if the second answer is confirmed as “the reliability of the first raw material is 300.”

[0091] The device (200) can set the reliability for each raw material included in the first product through the same method as the method of setting the reliability of the first raw material, and to this end, steps S208 through S211 can be performed repeatedly.

[0092] In step S212, the device (200) may generate an analysis report on the raw materials of the first product based on whether there is a violation of regulations and the reliability of each raw material included in the first product, if the violation of regulations and the reliability of each raw material included in the first product are set. At this time, the analysis report may include whether there is a violation of regulations and the reliability level for each raw material, and visualization to enhance user understanding may be performed, such as separately indicating only the raw materials that have violated regulations or separately indicating only the raw materials whose reliability is below the standard.

[0093] When the device (200) generates an analysis report on the raw material of the first product, it may recommend alternative raw materials and add the recommendation information to the analysis report, and a detailed explanation related thereto will be described later with reference to FIGS. 6 and 7.

[0094] In step S213, the device (200) can transmit an analysis report on the raw material of the first product to the first user terminal (110). At this time, the analysis report on the raw material of the first product can be displayed on the screen of the first user terminal (110).

[0095] When the device (200) transmits an analysis report on the raw material of the first product to the first user terminal (110), it may generate and transmit a page for displaying the analysis report, and a detailed explanation related thereto will be described later with reference to FIGS. 8 and 9.

[0096] As described above, by automatically analyzing the legal compliance and reliability of each ingredient included in the product based on artificial intelligence, it enables faster and more consistent judgments compared to the existing method where humans manually verify laws and ingredient information. Furthermore, by providing an analysis report that quantitatively presents the legal compliance and reliability levels for each ingredient, it can help intuitively identify product safety and problematic ingredients.

[0097] FIG. 3 is a flowchart illustrating the process of analyzing whether a raw material violates regulations according to one embodiment.

[0098] Referring to Fig. 3, first, in step S301, when the first question is input, the first artificial intelligence model can confirm that the first raw material is the subject of analysis through the first question.

[0099] Specifically, when a first question is input from the device (200), the first artificial intelligence model can extract the name of the ingredient subject to regulatory judgment within the first question using a natural language processing-based analysis module. To this end, the first artificial intelligence model is trained to identify the ingredient that is contextually the focus of the regulatory judgment as the first raw material by understanding the sentence structure, even when multiple ingredient names are mentioned. Through this, the first artificial intelligence model can confirm through the first question that the first raw material is the target raw material to be analyzed at present.

[0100] In step S302, the first artificial intelligence model can identify the country of analysis as the first country based on the location of the first user terminal (110) or the nationality of the first user.

[0101] Specifically, the first artificial intelligence model can identify the country of analysis as the first country if it is confirmed that the location of the first user terminal (110) is within the first country based on GPS information of the first user terminal (110) separately obtained from the device (200), and can identify the country of analysis as the first country if it is confirmed that the nationality of the first user is the first country based on first user information separately obtained from the device (200). To this end, the device (200) can obtain GPS information of the first user terminal (110) from the first user terminal (110) and can obtain first user information by querying a database.

[0102] In step S303, the first artificial intelligence model can identify the category of the first product containing the first raw material as the first category.

[0103] Specifically, the first artificial intelligence model can confirm that the category of the first product is the first category based on the first product information separately obtained from the device (200). At this time, the first category can be confirmed as any one of a plurality of categories, and the plurality of categories can be classified into cosmetics, health functional foods, medical / beauty treatment products, personal care products, etc.

[0104] In step S304, the first artificial intelligence model can generate a first list by compiling raw materials that are prohibited from being used in products of the first category in the first country.

[0105] Specifically, when the first country and the first category are determined, the first artificial intelligence model obtains a list of prohibited ingredients by category of the first country from the database of the device (200), identifies the types of raw materials prohibited from use in products of the first category in the first country based on the list of prohibited ingredients by category of the first country, and can generate a first list based on the identified results. At this time, the first list includes raw materials explicitly prohibited by laws, notices, guidelines, etc., and can be configured to enable accurate comparison by mapping raw material names, synonyms, INCI names, code systems, etc.

[0106] In step S305, the first artificial intelligence model can check whether the first raw material is included in the first list.

[0107] If it is confirmed in step S305 that the first raw material is included in the first list, in step S306, the first artificial intelligence model may generate a first answer indicating that the first raw material has violated regulations.

[0108] That is, if the first artificial intelligence model determines that the first ingredient is an ingredient prohibited from use in products of the first category according to the regulations of the first country when the first ingredient exists in the first list, it may generate a first response indicating that the first ingredient has violated the law. In this case, the first artificial intelligence model may generate the first response to include the reason for the prohibition or the source of the relevant laws, if necessary.

[0109] If it is confirmed in step S305 that the first raw material is not included in the first list, in step S307, the first artificial intelligence model can confirm the content of the first raw material included in the first product as the first content.

[0110] Specifically, the first artificial intelligence model can verify the actual content of the first raw material included in the first product as the first content based on the first product information separately obtained from the device (200). At this time, the first artificial intelligence model can convert the unit of the first content so that the component information expressed in various units such as %, mg / g, ppm can be normalized to enable comparison of usage amounts.

[0111] In step S308, the first artificial intelligence model can determine the maximum content of the first raw material permitted for use in the first category of products in the first country as the second content.

[0112] Specifically, the first artificial intelligence model obtains a list of permitted ingredients by category of the first country from the database of the device (200), and based on the list of permitted ingredients by category of the first country, checks the value of the maximum content of the first raw material permitted for use in the first category product of the first country, and can check the second content based on the confirmed result. At this time, the first artificial intelligence model can convert the unit of the second content to match the unit of the first content.

[0113] In step S309, the first artificial intelligence model can determine whether the first content is higher than the second content.

[0114] If it is confirmed in step S309 that the first content is higher than the second content, in step S306, the first artificial intelligence model may generate a first answer indicating that the first raw material has violated regulations.

[0115] That is, if the first AI model determines that the first raw material does not meet the permissible standards under the regulations of the first country when the first content is higher than the second content, it may generate a first response indicating that the first raw material has violated the law. In this case, the first AI model may generate a first response that includes both the first content and the second content if necessary.

[0116] If it is confirmed in step S309 that the first content is not higher than the second content, in step S310, the first artificial intelligence model can generate a first answer indicating that the first raw material does not violate regulations.

[0117] That is, if the first AI model determines that the first raw material meets the permissible standards under the regulations of the first country when the first content is not higher than the second content, it may generate a first answer indicating that the first raw material does not violate the law. In this case, the first AI model may generate a first answer that includes both the first content and the second content if necessary.

[0118] As described above, since it is possible to perform AI-based automatic analysis of ingredients included in functional products by considering both national regulatory standards and product categories, it is possible to determine whether there is a violation of regulations more quickly and accurately than the method of checking individually through manual work. Furthermore, objectivity and consistency are improved by quantitatively determining whether there is a violation by comparing the actual content of the ingredients with the permissible content by country, and efficiency in global product management can be improved because the analysis reflects regulatory differences by country and product group.

[0119] FIGS. 4 and 5 are flowcharts for explaining the process of analyzing the reliability of raw materials according to one embodiment.

[0120] Referring to FIGS. 4 and 5, first, in step S401, when the second artificial intelligence model inputs the second question, it can confirm through the second question that the first raw material is the subject of analysis.

[0121] Specifically, when the second artificial intelligence model receives the second question from the device (200), it can use a natural language processing-based analysis module to extract ingredient names, efficacy-related queries, and risk-related expressions included in the second question, thereby identifying the ingredient that is the center of the reliability evaluation as the first raw material. To this end, the second artificial intelligence model is trained to identify the ingredient that is the focus of the legal judgment contextually as the first raw material by understanding the sentence structure, even when multiple ingredient names are mentioned. Through this, the second artificial intelligence model can confirm through the second question that the first raw material is the target raw material to be analyzed at present.

[0122] In step S402, the second artificial intelligence model can classify papers that mention the first raw material and were published during a reference period into a first group. Here, the reference period may be set differently depending on the embodiment, for example, it may be set to the last year.

[0123] Specifically, the second AI model can search multiple paper databases for papers mentioning the first material and classify papers published during a reference period into the first group. In this case, the second AI model can classify the first group by performing filtering based on metadata such as the publication date and date of publication, confirming whether the papers were published during the reference period, and then verifying whether the first material is mentioned in the titles, abstracts, and text of the filtered papers. To this end, the second AI model can be configured to access multiple paper databases.

[0124] In step S403, the second artificial intelligence model can classify papers among those classified into the first group that mention efficacy regarding the first raw material into the first-1 group.

[0125] Specifically, the second AI model analyzes the abstracts and text of papers classified into the first group to classify papers that mention the first ingredient exhibiting positive effects, such as skin improvement, antioxidant properties, and anti-inflammatory effects, into the first-1 group. This process can be performed by a natural language processing module that recognizes keywords related to positive effects and descriptions of positive effects within the text.

[0126] In step S404, the second artificial intelligence model can classify papers among those classified into Group 1 that mention side effects regarding the first raw material into Group 1-2.

[0127] Specifically, the second AI model can analyze the abstracts and text of papers classified into Group 1 to classify papers that mention that the first ingredient causes side effects such as irritation, rash, or toxicity into Group 1-2. This process can be performed by a natural language processing module that recognizes keywords related to side effects and descriptions of negative effects within the text.

[0128] In step S405, the second artificial intelligence model can identify the number of papers classified into the first group as the first figure, identify the number of papers classified into the first-1 group as the second figure, and identify the number of papers classified into the first-2 group as the third figure.

[0129] In step S406, the second artificial intelligence model can calculate the first ratio by dividing the second value by the first value, and calculate the second ratio by dividing the third value by the first value.

[0130] In step S407, the second artificial intelligence model may assign a higher score to the first score within the first score range as the first ratio increases. Here, the first score range may be set differently depending on the embodiment, for example, to a range of 0 to 100 points.

[0131] For example, if the second artificial intelligence model confirms that the first ratio is 10%, the first score can be set to 10 points, and if the first ratio is confirmed to be 20%, the first score can be set to 20 points.

[0132] In step S408, the second artificial intelligence model may assign a higher second score within the second score range as the second ratio decreases. Here, the second score range may be set differently depending on the embodiment, for example, to a range of 0 to 100 points.

[0133] For example, if the second artificial intelligence model confirms that the second ratio is 90%, the second score can be set to 10 points, and if the second ratio is confirmed to be 80%, the second score can be set to 20 points.

[0134] In step S409, the second artificial intelligence model can classify patents disclosed during the reference period among the patents mentioning the first raw material into a second group.

[0135] Specifically, the second artificial intelligence model can search for patents mentioning the first raw material in multiple patent databases and select patents disclosed during a reference period to classify into a second group. In this case, the second artificial intelligence model can perform filtering based on metadata such as the patent's publication date and announcement date to determine whether it was disclosed during the reference period, and then classify the second group based on the results of verifying whether the first raw material is mentioned in the title, abstract, effect, etc., of the filtered patents. To this end, the second artificial intelligence model may be configured to access multiple patent databases.

[0136] In step S410, the second artificial intelligence model can classify patents among the patents classified into the second group that mention efficacy for the first raw material into the second-1 group.

[0137] Specifically, the second artificial intelligence model can analyze the summaries, effects, etc. of patents classified into the second group and classify patents that mention the first ingredient exhibiting positive efficacy into the second-1 group.

[0138] In step S411, the second artificial intelligence model can classify patents among the patents classified into the second group that mention side effects regarding the first raw material into the second-2 group.

[0139] Specifically, the second artificial intelligence model analyzes the summaries, effects, etc. of patents classified into the second group, and can classify patents that mention that the first ingredient causes side effects into the second-2 group.

[0140] In step S412, the second artificial intelligence model can verify the number of patents classified into the second group as the fourth figure, verify the number of patents classified into the second-1 group as the fifth figure, and verify the number of patents classified into the second-2 group as the sixth figure.

[0141] In step S501 following step S412, the second artificial intelligence model can calculate the third ratio by dividing the fifth value by the fourth value, and calculate the fourth ratio by dividing the sixth value by the fourth value.

[0142] In step S502, the second artificial intelligence model may assign a higher third score within the third score range as the third ratio increases. Here, the third score range may be set differently depending on the embodiment, for example, to a range of 0 to 100 points.

[0143] In step S503, the second artificial intelligence model may assign a higher score to the fourth score within the fourth score range as the fourth ratio decreases. Here, the fourth score range may be set differently depending on the embodiment, for example, to a range of 0 to 100 points.

[0144] In step S504, the second artificial intelligence model may classify clinical data disclosed during the reference period among the clinical data from public institutions mentioning the first raw material into a third group. Here, public institutions can be understood as a concept that includes various types of institutions that generate, manage, and provide information or data for public purposes, such as institutions performing regulatory duties in the fields of health, medicine, and food; research institutions established or operated by the state; public medical institutions that produce and manage information related to clinical trials or safety; and institutions under the government that operate databases for public purposes.

[0145] Specifically, the second AI model can search for clinical data mentioning the first ingredient in multiple public institution clinical databases and classify clinical data disclosed during a reference period into a third group. In this case, the third AI model can classify the third group by performing filtering based on metadata such as the disclosure date and publication date of the clinical data to determine whether it was disclosed during the reference period, and then verifying whether the first ingredient is mentioned in the trial summaries, clinical results, reports, etc., of the filtered clinical data. To this end, the third AI model can be configured to access multiple public institution clinical databases.

[0146] In step S505, the second artificial intelligence model can classify the clinical data mentioned as having efficacy for the first raw material among the clinical data classified into the third group into group 3-1.

[0147] Specifically, the second artificial intelligence model can analyze the summary, results, etc. of clinical data classified into the third group and classify the clinical data in which the first ingredient is mentioned as exhibiting positive efficacy into the third-1 group.

[0148] In step S506, the second artificial intelligence model can classify clinical data that mentions side effects regarding the first raw material among the clinical data classified into the third group into the third-2 group.

[0149] Specifically, the second artificial intelligence model analyzes the summary, results, etc. of clinical data classified into the third group, and can classify clinical data in which the first ingredient is mentioned as causing side effects into the third-2 group.

[0150] In step S507, the second artificial intelligence model can verify the number of clinical data classified into the third group as the seventh figure, the number of clinical data classified into the third-1 group as the eighth figure, and the number of clinical data classified into the third-2 group as the ninth figure.

[0151] In step S508, the second artificial intelligence model can calculate the 5th ratio by dividing the 8th value by the 7th value, and calculate the 6th ratio by dividing the 9th value by the 7th value.

[0152] In step S509, the second artificial intelligence model may assign a higher fifth score within the fifth score range as the fifth ratio increases. Here, the fifth score range may be set differently depending on the embodiment, for example, to a range of 0 to 100 points.

[0153] In step S510, the second artificial intelligence model may assign a higher score to the 6th score within the 6th score range as the 6th ratio decreases. Here, the 6th score range may be set differently depending on the embodiment, for example, to a range of 0 to 100 points.

[0154] In step S511, the second artificial intelligence model can sum the first through sixth scores to calculate the total sum.

[0155] In step S512, the second artificial intelligence model can generate a second answer as the total reliability of the first raw material.

[0156] As described above, by automatically collecting papers, patents, and clinical data within a reference period based on an artificial intelligence model, and quantitatively analyzing and scoring the mention of efficacy and side effects, the reliability of raw materials can be calculated based on objective and consistent standards. Consequently, the speed of evaluation is significantly improved compared to the existing method where humans reviewed individual data, and reliability judgments based on the latest evidence become possible, thereby increasing the accuracy of raw material safety verification.

[0157] FIG. 6 is a flowchart illustrating the process of recommending a substitute raw material for a raw material that violates regulations according to one embodiment.

[0158] Referring to FIG. 6, in step S601, the device (200) can classify raw materials that have a main efficacy matching the first raw material into a first group if it is confirmed that the first raw material among the raw materials included in the first product has violated regulations. Here, the main efficacy of the first raw material may refer to the main function that the first raw material performs within the first product, such as whitening, wrinkle improvement, UV protection, moisturizing, soothing, antioxidant, etc., and the device (200) can identify which of the main efficacy of the first raw material is by referring to a functional raw material database, a dictionary of efficacy by ingredient, papers, patents, and functional descriptions included in clinical data.

[0159] The device (200) can verify the main efficacy of each other external raw material not included in the first product using the same criteria as the method used to verify the main efficacy of the first raw material, and can select and classify into a first group the raw materials that are judged to have the same or substantially similar main efficacy as the first raw material. At this time, the first group may include alternative candidate raw materials that overlap with the functional role of the first raw material.

[0160] That is, the device (200) can form a candidate group capable of replacing the function of the first raw material by comparing the main efficacy of various functional raw materials registered in the raw material database, and by selecting raw materials that have the same or substantially similar efficacy as the first raw material and classifying them into a first group.

[0161] In step S602, the device (200) can classify raw materials classified into the first group that are not included in the first list into the second group as a result of comparing the first list and the first group. Here, the first list is a list of raw materials prohibited from use in products of the first category in the first country, which can be obtained through the first artificial intelligence model, and among the raw materials included in the first group, the raw materials included in the first list may correspond to raw materials prohibited from use in products of the first category in the first country, just like the first raw materials.

[0162] The device (200) can form a set of suitable candidate raw materials that can replace the first raw materials within the scope of not violating regulations by excluding the raw materials included in the first list from the replacement candidates among the raw materials included in the first group and classifying only the raw materials not included in the first list into the second group.

[0163] In step S603, the device (200) can determine the maximum content allowed for use in a product of the first category in the first country for each of the raw materials classified into the second group.

[0164] Specifically, the device (200) can verify the value of the maximum content permitted for use in the first category product in the first country for each raw material classified into the second group, based on the list of permitted ingredients by category of the first country stored in the database. At this time, if the maximum content for each raw material classified into the second group is set in different units, the device (200) can normalize them into a single common unit to match the units.

[0165] In step S604, the device (200) can identify the raw material with the highest maximum content among the raw materials classified into the second group as the second raw material.

[0166] That is, the device (200) can determine the second raw material by comparing the maximum content confirmed for each raw material belonging to the second group and selecting the raw material having the largest value. If there are multiple raw materials with the same maximum content, the device may be configured to select one of them as the second raw material by additionally applying predefined priority criteria such as reliability score, market usage history, raw material cost, and additional efficacy. However, basically, the device may be designed so that the functional effect of the product can be maintained or sufficiently supplemented even if the first raw material is replaced, by setting the raw material that can be used up to the highest regulatory content within the second group as the second raw material.

[0167] In step S605, the device (200) may add recommendation information recommending a second raw material as a substitute for the first raw material to the analysis report on the raw material of the first product. Here, the recommendation information may include the name of the second raw material, whether the main efficacy matches that of the first raw material, the maximum content allowed in the first country and first category, and a summary description regarding the expected effect of maintaining or improving function when the first raw material is replaced with the second raw material.

[0168] The device (200) structures and includes recommendation information recommending a second raw material as a substitute for the first raw material as part of an analysis report on the raw material of the first product, and transmits the analysis report to the first user terminal (110), thereby enabling the user to identify the substitute raw material that complies with regulations in addition to identifying the raw material that violates the regulations.

[0169] As described above, even if a raw material violates regulations, the system automatically selects compliant ingredients among raw materials with the same primary efficacy and recommends the most suitable alternative raw material by comparing the maximum allowable content of such raw materials. This allows users to quickly prepare a revised formulation that preserves the functional effects of the product while maintaining regulatory compliance, and by systematically presenting alternative raw materials based on regulatory standards, efficacy, and allowable content, the efficiency and quality of the product renewal process can be improved.

[0170] FIG. 7 is a flowchart illustrating the process of recommending a substitute raw material for a raw material with low reliability according to one embodiment.

[0171] Referring to FIG. 7, in step S701, if the device (200) confirms that the reliability of the third raw material among the raw materials included in the first product is lower than a reference value, the raw material that matches the main efficacy of the third raw material can be classified into a third group. Here, the reference value may be set differently depending on the embodiment, and for example, it may be set in the form of 60 points or 70 points in a reliability score system based on a maximum of 100 points.

[0172] The device (200) can verify the main efficacy of each other external raw material not included in the first product using the same criteria as the method used to verify the main efficacy of the third raw material, and can select and classify into a third group the raw materials among the external raw materials that are judged to have the same or substantially similar main efficacy as the third raw material. At this time, the third group may include alternative candidate raw materials that overlap with the functional role of the third raw material.

[0173] That is, the device (200) can form a candidate group capable of replacing the function of the third raw material by comparing the main efficacy of various functional raw materials registered in the raw material database, and by selecting raw materials that have the same or substantially similar efficacy as the third raw material and classifying them into a third group.

[0174] In step S702, the device (200) can establish a reliability for each raw material classified into a third group through a second artificial intelligence model. At this time, the second artificial intelligence model can calculate the reliability by calculating the ratio of mentions regarding the efficacy and side effects of the raw material based on various public evidence materials such as papers, patents, and clinical data, and by scoring this ratio. To this end, steps S401 through S512 may be performed repeatedly.

[0175] In step S703, the device (200) can classify raw materials classified into the third group that have been confirmed to have a reliability higher than the standard value into the fourth group.

[0176] That is, the device (200) can select a highly reliable raw material suitable for replacing the third raw material by comparing the reliability value of each raw material classified into the third group with a reference value and classifying only the raw materials that exceed the reference value into the fourth group.

[0177] In step S704, the device (200) can determine the unit price per reference unit for each of the raw materials classified into the fourth group. Here, the reference unit can be set to weight (g), volume (ml), or other standardized units, and the unit price per reference unit can be determined through a raw material unit price database, an external supply chain management server, information posted on the web, etc.

[0178] In step S705, the device (200) can identify the raw material with the lowest unit price among the raw materials classified into the fourth group as the fourth raw material.

[0179] That is, the device (200) can select the raw material having the lowest value by comparing unit price information for each raw material within the fourth group, and if multiple raw materials with the same unit price exist, it can be configured to select one of them as the fourth raw material by applying additional priority criteria such as reliability score, supply stability, or compliance with regulatory standards. However, it can be designed to prioritize cost efficiency and enable an economical replacement while maintaining the function of the third raw material within the fourth group.

[0180] In step S706, the device (200) may add recommendation information recommending a fourth raw material as a substitute for the third raw material to the analysis report on the raw materials of the first product. Here, the recommendation information may include the name of the fourth raw material, whether the main efficacy matches that of the third raw material, a reliability score, a standard unit price, and a summary description regarding the expected effect of maintaining or improving function when the third raw material is replaced with the fourth raw material.

[0181] The device (200) structures and includes recommendation information recommending a fourth raw material as a substitute for a third raw material as part of an analysis report on the raw material of the first product, and transmits the analysis report to the first user terminal (110), thereby enabling the user to check at a glance which raw material can replace the raw material with which reliability is lacking and whether the substitute raw material is suitable in terms of regulation, efficacy, and economic feasibility, in addition to supporting practical decision-making support functions for product renewal, raw material optimization, or prescription improvement.

[0182] As described above, since it is possible to automatically detect raw materials with low reliability and recommend the most suitable alternative raw material among candidate raw materials with the same primary efficacy by considering both reliability and unit price, it is possible to secure economic efficiency while maintaining the functional stability and quality of the product. Furthermore, as the reliability-based alternative raw material selection process is automated, the efficiency of formulation improvement work is enhanced, and product safety and market competitiveness can be simultaneously increased.

[0183] FIGS. 8 and 9 are flowcharts for explaining the process of displaying an analysis report according to one embodiment.

[0184] Referring to FIGS. 8 and 9, first, in step S801, the device (200) may generate a first page for displaying an analysis report on the raw material of the first product and transmit it to the first user terminal (110). At this time, the first page may have a structure that dynamically arranges components of an interface so as to include detailed information on the raw material, basic information on the raw material, ranking information, visualization information, etc. Here, the detailed information may include detailed information on the raw material, such as the name of the raw material, main efficacy, content, reason for violation of regulations, degree of exceedance compared to permissible standards, reliability score, etc., and the basic information may include only basic information on the raw material, such as the name of the raw material, etc.

[0185] In step S802, the device (200) can classify raw materials included in the first product that are found to be in violation of regulations and raw materials that are found to have a reliability lower than the standard value as problematic raw materials.

[0186] In step S803, the device (200) can determine the number of raw materials classified as problematic raw materials as the number of raw materials for the first time.

[0187] In step S804, the device (200) can verify the entire area of ​​the first page as the first area.

[0188] In step S805, the device (200) can identify the minimum display area required to display detailed information of a raw material as the second area. At this time, the second area may refer to a unit area for displaying detailed information of a single raw material and may be set in advance.

[0189] In step S806, the device (200) can calculate a third area by multiplying the second area by the number of first raw materials.

[0190] In step S807, the device (200) can check whether the third area is larger than the first area.

[0191] If it is confirmed in step S807 that the third area is larger than the first area, in step S808, the device (200) can set the first rank using the quotient part of the value obtained by dividing the first area by the second area.

[0192] That is, the device (200) can check the quotient portion of the value obtained by dividing the first area by the second area, and then set the ranking corresponding to the checked quotient portion as the first ranking.

[0193] For example, if the device (200) confirms that the first area is 30 cm² and the second area is 4 cm², it can determine the quotient part of the value obtained by dividing 30 by 4 as 7 and set the 7th rank as the first rank.

[0194] In step S809, the device (200) may assign a higher priority to each raw material classified as a problem raw material as the amount included in the first product increases. At this time, the amount of each raw material included in the first product can be verified through the first product information.

[0195] In step S810, the device (200) can classify raw materials that have been identified as having a priority of at least 1st rank among the raw materials classified as problematic as important raw materials.

[0196] In step S811, the device (200) can control the display of detailed information of a raw material classified as an important raw material on the first page in the entire area. At this time, if there are multiple raw materials classified as important raw materials, the device (200) can control the display of detailed information of each raw material classified as an important raw material separately within the entire area.

[0197] If it is confirmed in step S807 that the third area is not larger than the first area, in step S901, the device (200) can calculate the fourth area by subtracting the third area from the first area.

[0198] In step S902, the device (200) can identify the minimum display area required to display basic information of a raw material as the fifth area. At this time, the fifth area may refer to a unit area for displaying basic information of a single raw material and may be set in advance.

[0199] In step S903, the device (200) can set the second rank using the quotient portion of the value obtained by dividing the fourth area by the fifth area.

[0200] That is, the device (200) can check the quotient portion of the value obtained by dividing the fourth area by the fifth area, and then set the ranking corresponding to the checked quotient portion as the second ranking.

[0201] In step S904, the device (200) can classify the remaining materials, excluding the problematic materials among the materials included in the first product, as normal materials.

[0202] In step S905, the device (200) may assign a higher priority to each raw material classified as a normal raw material as the amount included in the first product increases. At this time, the amount of each raw material included in the first product can be verified through the first product information.

[0203] In step S906, the device (200) can generate a first list based on raw materials that are identified as having a priority of second rank or higher among the raw materials classified as normal raw materials.

[0204] In step S907, the device (200) may set a first area with the size of a third area within the first page and a second area with the size of a fourth area within the first page. At this time, if there is no raw material classified as a problem raw material, the device (200) may calculate the third area as 0, so that the first area is not set separately and the entire area of ​​the first page is set as the second area.

[0205] In step S908, the device (200) can control the display of detailed information of raw materials classified as problematic raw materials on the first page in the first area and the display of the first list in the second area. At this time, if there are multiple raw materials classified as problematic raw materials, the device (200) can control the display of detailed information of each raw material classified as problematic raw material separately within the first area. Additionally, if there are no raw materials classified as problematic raw materials, the device (200) can control the display of only the first list in the second area.

[0206] As described above, by dynamically arranging detailed information on problematic ingredients and basic information on normal ingredients within a page, and automatically adjusting the display area according to page area, number of ingredients, and priority, users can grasp the product's regulatory risk factors and key ingredient composition at a glance. Furthermore, by prioritizing and highlighting ingredients that violate regulations or lack credibility, the readability and effectiveness of analysis results are enhanced, and rapid and accurate decision-making can be supported during product improvement and formulation renewal processes.

[0207] FIG. 10 is an example diagram of the configuration of a device according to one embodiment.

[0208] A device (200) according to one embodiment includes a processor (210) and a memory (220). The processor (210) may include at least one device described with reference to FIGS. 1 through 9 or may perform at least one method described with reference to FIGS. 1 through 9. A person or organization using the device (200) may provide services related to some or all of the methods described with reference to FIGS. 1 through 9.

[0209] The memory (220) may store information related to the methods described above or store a program in which the methods described below are implemented. The memory (220) may be volatile memory or non-volatile memory.

[0210] The processor (210) can execute a program and control the device (200). The code of the program executed by the processor (210) can be stored in memory (220). The device (200) is connected to an external device (e.g., a personal computer or a network) through an input / output device (not shown in the drawing) and can exchange data through wired or wireless communication.

[0211] The device (200) can be used to train an artificial intelligence model or to use a trained artificial intelligence model. The memory (220) may contain an artificial intelligence model that is being trained or has been trained. The processor (210) can train or execute an artificial intelligence model algorithm stored in the memory (220). The training device for training the artificial intelligence model and the device (200) for using the trained artificial intelligence model may be the same or separate.

[0212] The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0213] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0214] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0215] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can apply various technical modifications and variations based on the above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0216] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

Claim 1 A method for evaluating the legal violation status and reliability of an artificial intelligence-based functional ingredient, performed by a device, comprising: receiving a request for analysis of a first product from a first user terminal; identifying each ingredient included in the first product based on an image of the first product or information of the first product; selecting a first ingredient, which is one of the ingredients included in the first product, as an analysis target; generating a first question requesting an analysis of whether the first ingredient violates regulations; inputting the first question into a first artificial intelligence model trained to analyze and answer whether the ingredient violates regulations; when the first artificial intelligence model generates a first answer as an answer to the first question, obtaining the first answer as an output result from the first artificial intelligence model; determining whether the first ingredient violates regulations based on the first answer; generating a second question requesting a reliability analysis of the first ingredient; inputting the second question into a second artificial intelligence model trained to analyze and answer the reliability of the ingredient; when the second artificial intelligence model generates a second answer as an answer to the second question, the second The method comprises: a step of obtaining the second answer as an output result from an artificial intelligence model; a step of setting the reliability of the first raw material based on the second answer; a step of generating an analysis report on the raw material of the first product based on the legal violation status and reliability of each raw material included in the first product when the legal violation status and reliability for each raw material included in the first product are set; and a step of transmitting the analysis report on the raw material of the first product to the first user terminal, wherein the first artificial intelligence model, when the first question is input, confirms that the first raw material is the subject of analysis through the first question, and identifies the country of analysis as the first country based on the location of the first user terminal or the nationality of the first user,Identify the category of the first product containing the first raw material as the first category, and collect raw materials prohibited from use in products of the first category in the first country to generate a first list; if it is confirmed that the first raw material is included in the first list, generate the first answer indicating that the first raw material has violated regulations; if it is confirmed that the first raw material is not included in the first list, identify the content of the first raw material included in the first product as the first content, and identify the maximum content of the first raw material permitted for use in products of the first category in the first country as the second content; if it is confirmed that the first content is higher than the second content, generate the first answer indicating that the first raw material has violated regulations; if it is confirmed that the first content is not higher than the second content, generate the first answer indicating that the first raw material has not violated regulations; and when the second question is input, the second artificial intelligence model confirms that the first raw material is the subject of analysis through the second question, and mentions the first raw material Among the papers, those published during a predetermined reference period are classified into Group 1; among the papers classified into Group 1, those that mention efficacy regarding the first ingredient are classified into Group 1-1; among the papers classified into Group 1, those that mention side effects regarding the first ingredient are classified into Group 1-2; the number of papers classified into Group 1 is identified as a first value; the number of papers classified into Group 1-1 is identified as a second value; the number of papers classified into Group 1-2 is identified as a third value; a first ratio is calculated by dividing the second value by the first value; a second ratio is calculated by dividing the third value by the first value; and the higher the first ratio, the higher the first score is assigned within the predetermined first score range.The lower the second ratio mentioned above, the higher the second score is assigned within the preset second score range; among the patents mentioning the first raw material, patents disclosed during the reference period are classified into a second group; among the patents classified into the second group, patents mentioning efficacy regarding the first raw material are classified into a second-1 group; among the patents classified into the second group, patents mentioning side effects regarding the first raw material are classified into a second-2 group; the number of patents classified into the second group is confirmed as a fourth value; the number of patents classified into the second-1 group is confirmed as a fifth value; the number of patents classified into the second-2 group is confirmed as a sixth value; the third ratio is calculated by dividing the fifth value by the fourth value; the fourth ratio is calculated by dividing the sixth value by the fourth value; the higher the third ratio, the higher the third score is assigned within the preset third score range; the lower the fourth ratio, the higher the fourth score is assigned within the preset fourth score range; and the Among the clinical data from public institutions mentioning the first raw material, the clinical data disclosed during the above reference period is classified into the third group; among the clinical data classified into the third group, the clinical data mentioned as having efficacy for the first raw material is classified into the third-1 group; among the clinical data classified into the third group, the clinical data mentioned as having side effects for the first raw material is classified into the third-2 group; the number of clinical data classified into the third group is confirmed as the seventh figure; the number of clinical data classified into the third-1 group is confirmed as the eighth figure; the number of clinical data classified into the third-2 group is confirmed as the ninth figure; the fifth ratio is calculated by dividing the eighth figure by the seventh figure; and the sixth ratio is calculated by dividing the ninth figure by the seventh figure.The step of assigning a higher fifth score within a preset range of fifth scores as the fifth ratio is higher, assigning a higher sixth score within a preset range of sixth scores as the sixth ratio is lower, summing the first to sixth scores to calculate a total sum, generating a second answer with the reliability of the first raw material being the total sum, and generating an analysis report on the raw material of the first product comprises: a step of classifying raw materials that have the same main efficacy as the first raw material into a first group when it is confirmed that the first raw material among the raw materials included in the first product violates regulations; a step of classifying raw materials that are not included in the first list among the raw materials classified into the first group into a second group as a result of comparing the first list and the first group; a step of confirming the maximum content permitted for use in the first category of products in the first country for each raw material classified into the second group; and a step of confirming the raw material with the highest maximum content among the raw materials classified into the second group as the second raw material. The method includes the step of adding recommendation information recommending the second raw material as a substitute for the first raw material to the analysis report on the raw material of the first product, and the step of generating the analysis report on the raw material of the first product comprises: a step of classifying raw materials that have a main efficacy matching the third raw material into a third group when it is confirmed that the reliability of the third raw material among the raw materials included in the first product is lower than a preset threshold; a step of setting the reliability for each of the raw materials classified into the third group through the second artificial intelligence model; a step of classifying raw materials among the raw materials classified into the third group that are confirmed to have a reliability higher than the threshold into a fourth group; and for each of the raw materials classified into the fourth group,A method for evaluating the reliability and legal compliance of an AI-based functional ingredient, comprising: a step of verifying a preset standard unit price; a step of identifying the ingredient with the lowest unit price among the ingredients classified into the fourth group as the fourth ingredient; and a step of adding recommendation information recommending the fourth ingredient as a substitute ingredient for the third ingredient to an analysis report on the ingredients of the first product. Claim 2 delete Claim 3 delete

Images