Specified product detection method, specified product detection system, and computer program product

Abstract

A method for detecting specified products, a system for detecting specified products, and a computer program product are provided. The method for detecting specified products includes at least one processor comprising: acquiring a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions including descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords representing one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval using the one or more keywords on the plurality of product datasets; detecting one or more non-containing products from the one or more suspected products; and outputting a list of containing products by excluding one or more non-containing products from the one or more suspected products.

Description

Technical Field

[0001] This invention relates to a method for testing specified commodities, a system for testing specified commodities, and a computer program product. Prior Technology

[0002] Various goods are traded on online auction and other e-commerce websites. These e-commerce websites may contain illegal or prohibited goods. To protect consumer safety, it is important to detect such inappropriate listings quickly and accurately.

[0003] Japanese Patent Application Publication No. 2006-53809 discloses a method for detecting illegal products during the production process. This method extracts strings from the production information and compares these strings with "NG" strings (e.g., "CD-ROM copied" or "user registration failed"). By searching the production information using keywords such as "NG" strings, illegal products such as illegally copied software can be detected. As a result, the production of illegal products can be stopped. Summary of the Invention

[0004] [The problem the invention aims to solve] For example, if a product is a health food and contains ingredients that are classified as pharmaceuticals, it will become an unapproved and illegal pharmaceutical product, constituting an illegal commodity. In other words, while an ingredient may be legal in a pharmaceutical product, a health food containing that ingredient may be an illegal commodity.

[0005] Therefore, especially when judging the suitability of a product based on its ingredients, simply using ingredient names as keywords to search for product information is insufficient to detect illegal products. This issue is not limited to detecting illegal products; the same problem arises when attempting to detect products containing any specified ingredient.

[0006] The present invention aims to provide a method, system, and computer program for detecting specified products, which can detect specified products containing specified ingredients. [Methods for solving problems]

[0007] The present invention provides a method for detecting designated products, wherein at least one processor comprises: acquiring a plurality of product datasets, each of the plurality of product datasets comprising product descriptions of the plurality of products, the product descriptions comprising descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords representing one or more designated ingredients; performing keyword retrieval on the plurality of product datasets using the one or more keywords, extracting one or more suspected products from the plurality of products, each of the suspected products containing at least one of the one or more designated ingredients in its corresponding aforementioned product description; detecting one or more non-containing products that do not contain the one or more designated ingredients from the one or more suspected products based on the aforementioned product descriptions corresponding to each of the one or more suspected products; and outputting a list of containing products containing at least one of the aforementioned designated ingredients by excluding the one or more non-containing products from the one or more suspected products.

[0008] The present invention discloses a method for detecting a specified product, wherein at least one processor comprises: acquiring a plurality of product datasets, each of the plurality of product datasets comprising product descriptions of the plurality of products, the product descriptions comprising descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords indicating one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the one or more specified ingredients in its corresponding product description; detecting negative statements regarding non-containing ingredients not contained in the product descriptions of each of the one or more suspected products; detecting one or more of the aforementioned non-containing ingredients described in the aforementioned negative statements; and displaying a confirmation screen on a display for determining whether the aforementioned suspected products contain the one or more specified ingredients, the confirmation screen being configured to display the aforementioned product descriptions of the corresponding previously suspected products, and the one or more non-containing ingredients being highlighted in the aforementioned product descriptions of the confirmation screen.

[0009] The present invention provides a designated product detection system comprising: at least one memory storing computer program code; and at least one processor; wherein the at least one processor executes the computer program code and is configured to perform the following: acquiring a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords representing one or more designated ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval using the one or more keywords on the plurality of product datasets, each of the suspected products containing at least one of the one or more designated ingredients in its corresponding product description; detecting one or more non-containing products that do not contain the one or more designated ingredients from the one or more suspected products based on the corresponding product descriptions for each of the one or more suspected products; and outputting a list of containing products containing at least one of the plurality of designated ingredients by excluding the one or more non-containing products from the one or more suspected products.

[0010] The present invention discloses a designated product detection system comprising: at least one memory storing computer program code; and at least one processor; wherein the at least one processor executes the computer program code and is configured to: acquire a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions including descriptions of the ingredients contained in the corresponding products; acquire one or more keywords, each of the one or more keywords indicating one or more designated ingredients; and extract one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords. Each of the aforementioned suspected products contains at least one of the aforementioned designated ingredients in its corresponding aforementioned product description; negative statements regarding non-containing ingredients not contained in the corresponding aforementioned suspected product are detected from the aforementioned product descriptions of each of the aforementioned suspected products; one or more of the aforementioned non-containing ingredients are detected in the aforementioned negative statements; and a confirmation screen for determining whether each of the aforementioned suspected products contains the aforementioned one or more designated ingredients is displayed on the display screen, wherein the aforementioned confirmation screen is configured to display the aforementioned product description of the corresponding aforementioned suspected product, and the aforementioned one or more non-containing ingredients are highlighted in the aforementioned product description of the aforementioned confirmation screen.

[0011] A computer program product of the present invention causes at least one processor to perform the following: obtaining a plurality of product datasets corresponding to a plurality of products, each product dataset containing product descriptions of the ingredients contained in the corresponding product; obtaining one or more keywords, each of the one or more keywords representing one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the aforementioned specified ingredients in its corresponding aforementioned product description; detecting one or more non-containing products that do not contain the aforementioned plurality of specified ingredients from the aforementioned one or more suspected products based on the aforementioned aforementioned product descriptions corresponding to each of the aforementioned one or more suspected products; and outputting a list of containing products containing at least one of the aforementioned plurality of specified ingredients by excluding the aforementioned one or more non-containing products from the aforementioned one or more suspected products.

[0012] A computer program product of the present invention causes at least one processor to perform the following: acquiring a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords indicating one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the one or more specified ingredients in its corresponding product description; detecting negative statements regarding non-containing ingredients not contained in the product descriptions of each of the one or more suspected products; detecting one or more of the aforementioned non-containing ingredients described in the aforementioned negative statements; and displaying a confirmation screen on a display for determining whether each of the suspected products contains the one or more specified ingredients, the confirmation screen being configured to display the aforementioned product descriptions for the corresponding previously suspected products, and in the aforementioned product descriptions of the confirmation screen, the one or more non-containing ingredients are highlighted. [Technical Effects]

[0013] It can provide testing methods, testing systems, and computer programs for testing specified products containing specified ingredients. Simple Explanation of the Diagram

[0014] [Figure 1] Figure 1 is a schematic diagram illustrating a specified commodity detection system in an implementation form. [Figure 2] Figure 2 is an explanatory diagram of a product sales page. [Figure 3] Figure 3 is a table illustrating a portion of a plurality of specified ingredients for a health food. [Figure 4] Figure 4 is a flowchart illustrating the method for generating the classification model. [Figure 5] Figure 5 is a flowchart illustrating the testing method for a specified product. [Figure 6] Figure 6 is an explanatory diagram illustrating a confirmation screen used to detect a specified product. Implementation

[0015] Referring to Figures 1 to 6, examples of a specified commodity inspection method, a specified commodity inspection system 11, and a computer program product are described. This invention is not limited to these examples, but is intended to include all modifications within the scope and equivalent meaning of the claims, as shown in the claims. Furthermore, the numerals "first," "second," etc., used in the following description are used to distinguish the constituent elements to which these numerals are attached, and do not order these constituent elements.

[0016] [Overview of the Designated Product Inspection System] The designated product detection system (hereinafter referred to as the "detection system") 11 shown in Figure 1 includes a designated product detection device (hereinafter referred to as the "detection device") 20. The detection device 20 is configured to detect designated products containing one or more designated ingredients from a plurality of products produced on an EC website or sold on an EC website. When one or more designated ingredients contain a plurality of designated ingredients, the plurality of designated ingredients may also be referred to as a group of designated ingredients. When a designated ingredient is an illegal ingredient, the detection device 20 detects illegal products containing illegal ingredients from the plurality of products.

[0017] The detection device 20 can be implemented as a computer having at least one processor 21, at least one memory 22, and a communication interface 23. An example of the detection device 20 having one processor 21 and one memory 22 is described below. The communication interface 23 is configured to communicate with other devices via a network.

[0018] The memory 22 stores the program 24 executed by the processor 21 and the detection data 25. The program 24 includes the application program and the operating system. The processor 21 performs various functions by executing processing based on the program 24. The detection data 25 contains various data used when detecting a specified product.

[0019] The detection system 11 may include a database 13. The database 13 may be stored in a web server 14 used to provide the EC website. The web server 14 may be a computer with the same components as the detection device 20 (e.g., at least one processor, at least one memory, and a communication interface). Alternatively, part or all of the database 13 may be stored in the memory 22 of the detection device 20.

[0020] The detection system 11 may have a plurality of buyer terminals 15, which can access an EC website (web server 14) via the network. The EC website may be, for example, an online shopping website or a standalone store website, or an online auction website, but is not limited thereto. The detection system 11 may have a plurality of seller terminals 16, which provide goods for sale on the EC website via the network.

[0021] The buyer terminal 15 and the seller terminal 16 can be, for example, personal computers, or mobile terminals such as smartphones or tablets. Each of the buyer terminal 15 and the seller terminal 16 can have a display as an output device. Each display can have a touch panel as an input device.

[0022] Database 13 may contain seller data tables, buyer data tables, and product data tables. The seller data tables store multiple seller datasets for multiple sellers offering products. The buyer data tables store multiple buyer datasets for multiple buyers purchasing products through the EC website and multiple registered users (potential buyers) on the EC website. The product data tables store multiple product datasets for multiple products. A product dataset contains one or more data items corresponding to a single product. Each of the multiple product datasets corresponds to a multiple number of products produced on the EC website, or a multiple number of products sold on the EC website.

[0023] The detection system 11 may include a learning device 30, which is capable of communicating with the detection device 20 via a network. Like the detection device 20, the learning device 30 can be implemented as a computer having at least one processor 31, at least one memory 32, and a communication interface 33. The following describes an example of the learning device 30 having one processor 31 and one memory 32.

[0024] The memory 32 stores a program 34. The program 34 includes an application and an operating system. The processor 31 performs various functions by executing processing based on the program 34. The memory 32 can store learning data 35 used to train one or more learning models. The generated learning data 35 can be stored in the memory 22 of the detection device 20.

[0025] The detection system 11 may include a language model 17, which can communicate with the learning device 30 via a network. The language model 17 can be either a large language model (LLM) or a small language model (SLM). A large language model is a language model trained using a large amount of text data. A small language model is a language model smaller in scale (e.g., with fewer parameters) than a large language model. Depending on the input prompt, the language model 17 can be a general-purpose natural language processing (NLP) model capable of adapting to various natural language processing tasks such as information extraction, article summarization, text generation, or question response.

[0026] The language model 17 is configured to generate data based on a prompt containing a certain instruction when input, and output it as a completion. In this example, the learning device 30 generates the prompt and inputs it into the language model 17, and then the learning device 30 obtains the completion generated by the language model 17. Alternatively, the detection device 20 can utilize the language model 17, which is not included in the detection system 11, via a network.

[0027] The detection system 11, which can replace the learning device 30, has another information processing device for performing at least one of the following: generating a prompt, inputting the prompt into the language model 17, and obtaining completion. The other information processing device can be a computer with the same components as the detection device 20 (e.g., at least one processor, at least one memory, and a communication interface).

[0028] The detection system 11 may include an operator terminal 40, which can communicate with the detection device 20 via a network. The operator operating the operator terminal 40 may be, for example, an operator of an e-commerce website or a manager of goods produced on an e-commerce website.

[0029] Like the detection device 20, the operator terminal 40 can be implemented as a computer having at least one processor 41, at least one memory 42, and a communication interface 43. An example of the operator terminal 40 having one processor 41 and one memory 42 is described below. The operator terminal 40 may have a display 46. Alternatively, an external display 46 may be connected to the operator terminal 40.

[0030] [Product Dataset] Each of the multiple goods can be classified into one of the multiple goods categories. The multiple goods categories may include, for example, at least one of health foods, pharmaceuticals, quasi-pharmaceuticals, and cosmetics, but are not limited to these.

[0031] Each product dataset contains information about the corresponding product, such as the product name, product category, production date, sales start date, sales price, and product description, but is not limited to these. Product categories can be manually entered in advance by the seller or operator. Product descriptions can include information about the ingredients of the corresponding product.

[0032] As shown in Figure 2, the product description may include a plurality of descriptive texts for display on the product sales page 50 of the EC website. The sales page 50 may include, for example, a slogan 51, a title 52, a detailed description 53, and one or more images 54. The images 54 may be still images or animated images.

[0033] When at least a portion of one or more images 54 contains a string, the detection device 20 can be configured to extract the string from one or more images 54. The detection device 20 can then append the extracted string from one or more images 54 to the product description in the product dataset. In this case, the product description includes the strings contained in the tagline 51, title 52, and detailed description 53, as well as the extracted string.

[0034] At least some of the multiple product categories may have more than one designated ingredient for each product category. A product category with more than one designated ingredient is called a designated product category. For example, when multiple designated ingredients (referred to as the "first designated ingredient group") are designated for use in one designated product category, namely health foods, the first designated ingredient group may be illegal ingredients that are prohibited in the health food product category. Illegal ingredients, for example, are ingredients prohibited by pharmaceutical laws and belong to pharmaceutical products. When another group of multiple designated ingredients (referred to as the "second designated ingredient group") is designated for use in pharmaceutical products, the second designated ingredient group may be, for example, pharmaceutical ingredients that are not certified in the country or region where the pharmaceutical product is sold.

[0035] Figure 3 illustrates a portion of the first designated ingredient group for health foods. The first designated ingredient group, when present in health foods, represents pharmaceutical ingredient names that violate the Pharmaceutical Affairs Law. Multiple designated ingredients can each be set as multiple keywords. In this case, multiple keywords can be set for a single designated ingredient. For example, for a single designated ingredient, two or more of the following keywords can be set: katakana name, English name, kanji name, and alternative names. The set keywords can be included in the detection data 25 as a set of search keywords.

[0036] Even if one or more designated ingredients are not illegal, they can still be prohibited ingredients that are banned from trading on EC websites. Products containing prohibited ingredients can include illegal goods. Furthermore, the designation of one or more designated ingredients is not limited to the detection of illegal ingredients; it can also be set for the detection of specific goods. Specific goods could be, for example, goods containing specific allergens or goods containing ingredients included for a specific purpose. In this case, for a designated product category (e.g., health foods), multiple groups of designated ingredients can be designated, each corresponding to multiple purposes.

[0037] [Machine Learning Model for Classification] When the product dataset does not contain data representing product categories, the detection device 20 or the learning device 30 can generate data representing product categories. This data can be generated using a classification machine learning model (hereinafter referred to as the "classification model"). The generated data representing product categories can be appended to the product description.

[0038] The learning device 30 can generate a classification model (a machine learning model for classification) by training an untrained learning model or a pre-trained learning model using learning data. The classification model can be configured to be a binary classification model that, when given a product description, outputs whether the product belongs to one of a plurality of product categories.

[0039] For example, when multiple product categories include one category (the first category being health foods) and another category (the second category being pharmaceuticals), the first classification model can be configured to output whether a product is a health food, and the second classification model can be configured to output whether a product is a pharmaceutical. Alternatively, one classification model can be configured to classify which of the multiple product categories a product belongs to. Or, another classification model can be configured to classify which of the multiple product categories a product belongs to, or whether it does not belong to any of these product categories.

[0040] When multiple products include sets of products that combine multiple categories of products, the classification model can be a multi-category classification model. A multi-category classification model can be configured to output, when a product description is input, which of the following three groups (categories) the product belongs to: 1) to 3).

[0041] 1) A group of objects belonging to a product category (object product group). 2) Groups of non-object goods belonging to a product category other than a product category (non-object goods groups).

[0042] 3) Groups of bundled products that include the target product (bundled product groups). The group of bundled products mentioned in 3) above can be any group of all bundled products, regardless of whether it includes the target product.

[0043] The learning device 30 can use multiple product datasets obtained from the database 13 as learning data. The multiple product datasets used as learning data are referred to as multiple training product datasets (hereinafter referred to as "training datasets"), and the product descriptions corresponding to the products in the training datasets are referred to as training product descriptions (hereinafter referred to as "training descriptions"). The training descriptions contain descriptions of the ingredients contained in the corresponding products.

[0044] The classification model can be, for example, a supervised learning model based on the decision tree algorithm used to perform the classification task described above. As an example, the classification model could be one using random forest, GBDT, XGBoost, or lightGBM, but is not limited to these.

[0045] The training data consists of multiple positive response datasets used as positive response data. Each positive response dataset contains training instructions for the corresponding product, as well as a positive response label indicating whether the corresponding product is an object product, a non-object product, or a bundled product.

[0046] When the number of positive solution datasets used as learning data is insufficient, the detection device 20 or the learning device 30 can generate positive solution datasets. For example, when a portion of the training dataset obtained from the database 13 does not contain data representing product categories that serve as positive solution labels, the detection device 20 or the learning device 30 can generate data representing product categories.

[0047] The following describes an example of how the learning device 30 uses the language model 17 to generate data representing product categories. In this case, the learning device 30 generates prompts to be input into the language model 17 in order to generate multiple correct answer datasets. In this example, the prompts are generated to provide information that is insufficient in the correct answer datasets, and for a given product category, to indicate whether the corresponding product belongs to the category of object-oriented products, non-object-oriented products, or bundled products.

[0048] The prompt can contain training instructions as information for obtaining an answer. In this case, language model 17 uses the training instructions as judgment material to determine whether the corresponding product belongs to an object product of a certain product category, and whether it is a package product containing object products. The prompt can contain multiple training instructions to obtain a unified answer for multiple products, or it can contain a single training instruction to obtain an answer for a single product.

[0049] In one example, the prompt contains multiple training descriptions for each of the multiple products, and an instruction for the language model 17 to output a response. The instruction should output a response indicating which of the multiple products described by the multiple training descriptions belongs to: target product (health food), non-target product (other than health food), and package product (including health food). The instruction could be, for example, stated as: "Based on the product descriptions for the product, please answer whether the product belongs to health food, a package product containing health food, or any other type of product."

[0050] When the learning device 30 inputs the prepared prompts into the language model 17, the language model 17 outputs complete responses to the prompts. The learning device 30 obtains the output of the corresponding prompts from the language model 17. Then, based on the output of the language model 17, the learning device 30 extracts one or more items from a plurality of items that are judged as the target item. Next, the learning device 30 appends item classification (e.g., "health food") data to the training dataset for each extracted item, thus completing the generation of the correct answer dataset.

[0051] Methods for generating classification models Referring to Figure 4, an example of a method for the learning device 30 to generate a classification model is illustrated. In step S11, the learning device 30 retrieves a plurality of product datasets from the database 13 for use as learning data. Next, in step S12, the learning device 30 prepares learning data to be input into the untrained learning model.

[0052] Preparing training data can include preprocessing. Preprocessing can include identifying missing data in each training dataset. Identifying missing data can include, for example, confirming the presence or absence of correct labels, i.e., whether product categories are included.

[0053] The preparation of learning data may involve extracting product classifiers containing positive labels from a plurality of obtained product datasets as a plurality of training datasets. Furthermore, as a check for omissions, the operator can verify whether the positive labels of at least a portion of the training datasets are correct. Additionally, the operator can generate positive labels for some product datasets that do not contain positive labels.

[0054] The preparation of learning data may involve splitting multiple training datasets into training data, validation data, and test data. At this point, at least one of the training datasets, from which the operator generated the dataset with the correct answer labels and the dataset whose correct answer labels the operator confirmed, can be used as test data.

[0055] In step S13, the learning device 30 determines whether the amount of learning data (e.g., the number of training datasets) is above a threshold. The threshold can be set by statistically calculating the amount of data required to train the learning model. When the amount of data is less than the threshold, the learning device 30 determines that the amount of data is insufficient (step S13 = NO) and proceeds to step S14 to add training data. The additional training data can be obtained, for example, by the learning device 30 retrieving a new plurality of product datasets from the database 13.

[0056] Alternatively, or furthermore, the additional training data can be generated as described above by adding data representing product categories to the product descriptions of the multiple product datasets that do not contain product categories as positive labels. For example, the learning device 30 can generate product category data using the language model 17 as described above.

[0057] The learning device 30 can acquire, as at least a portion, of a virtual product dataset generated by the operator for learning purposes. For example, the operator can generate a virtual product description by including the description of one or more specified ingredients in any product description. This is particularly useful when the product dataset of non-compliant products is small, as it allows for the generation of correct data for non-compliant products from the virtual product dataset.

[0058] After adding training data in step S14, the learning device 30 can return to step S12 to prepare learning data. After the learning data is prepared, the learning device 30 proceeds to step S13. In step S13, when the amount of data is above a threshold, the learning device 30 determines that the amount of data is sufficient (step S13 = YES) and proceeds to step S15.

[0059] In step S15, the learning device 30 trains the learning model by inputting training data into it, thereby updating the weights in the learning model. In the next step S16, the learning device 30 adjusts the parameters of the learning model using validation data.

[0060] In step S17, the learning device 30 uses test data to test the learning model. For example, the learning device 30 calculates the accuracy of the learning model as a result of the test. The accuracy metric can be, for example, AUC (Area Under the Curve), but is not limited to this. Then, the learning device 30 can determine whether the accuracy of the learning model is above a predefined accuracy threshold. The accuracy threshold can be stored in memory 32. When the accuracy of the learning model is less than the accuracy threshold, the learning device 30 determines that the learning model is underlearned and can proceed to step S14 to add training data.

[0061] When the accuracy of the learning model reaches or exceeds a threshold, the learning device 30 determines that the learning model has been properly trained. In this case, the learning device 30 proceeds to step S18, stores the trained learning model as a learned classification model in memory 32, and ends the process.

[0062] [Specific Product Testing Methods] Referring to Figure 5, an example is described where the detection device 20 detects an illegal product containing a plurality of specified ingredients for health foods from a plurality of products belonging to a plurality of product categories.

[0063] In step S21, the detection device 20 retrieves a plurality of product datasets from the database 13 or the memory 22. The plurality of product datasets may include data on products before they are manufactured or data on products currently on sale.

[0064] As shown in Figure 6, the detection device 20 can extract text information contained in each product dataset into product descriptions. Product descriptions may include, for example, strings 61 extracted from the slogan 51 contained in the sales page 50 shown in Figure 2, strings 62 extracted from the title 52, strings 63 extracted from the detailed description 53, and strings 64 extracted from the image 54.

[0065] In step S22, the detection device 20 retrieves one or more keywords from the memory 22. In this example, the detection device 20 retrieves a set of search keywords containing multiple keywords representing multiple specified ingredients. The set of search keywords may include, for example, the names of various ingredients that would be illegal to contain in health foods, as shown in FIG3.

[0066] In step S23, the detection device 20 performs a keyword search on the plurality of product datasets obtained in step S21 using the plurality of keywords obtained in step S22. More specifically, the detection device 20 searches for the plurality of product descriptions extracted from each of the plurality of product datasets using the plurality of keywords, and extracts product descriptions containing at least one keyword. The product corresponding to the extracted product description is called a suspected product. Therefore, through the keyword search in step S23, one or more suspected products whose product descriptions contain at least one of the specified ingredients are extracted.

[0067] In step S24, the detection device 20 will classify one or more suspected items extracted into one of three groups based on their corresponding product descriptions: target products (health food), non-target products (products other than health food), and packaged products.

[0068] More specifically, the detection device 20 inputs the product descriptions of each suspected item into a classification model trained using the method shown in Figure 4, which categorizes products as health foods. Then, based on the input product descriptions, the classification model outputs the classification result for each suspected item. By obtaining the classification result from the classification model for each of the more than one suspected item, the detection device 20 obtains a list of items extracted as health foods.

[0069] Furthermore, the classification in step S24 can be performed before the keyword search in step S22. In this case, a keyword search will be performed on a plurality of object products classified as health foods, and one or more suspected products containing at least one specified ingredient will be extracted from the health foods.

[0070] In step S25, the detection device 20 performs Named Entity Recognition (NER) on the product descriptions of each suspected product, namely, those classified as health foods, and extracts ingredient names from the extracted proper nouns. In this way, more than one ingredient name is detected from the product descriptions of each suspected product.

[0071] The detection device 20 performs steps S26 to S30 on each of the plurality of suspected products. The following example illustrates how the detection device 20 can detect multiple ingredient names from a single product description of a suspected product.

[0072] In step S26, the detection device 20 extracts one or more designated ingredients that match the keywords from a plurality of ingredient names in a product description. In Figure 6, the shaded proper nouns in the product description are the extracted designated ingredients.

[0073] In step S27, the detection device 20 detects a negative statement in a product description regarding a non-containing ingredient that the product does not contain. A negative statement is an explicit description of an ingredient that the product does not contain, but it may also include implied descriptions of an ingredient that is not present. In the example shown in Figure 6, the statement "Does not use silicone, mineral oil, ethanol, petroleum-based surfactants, wheat, egg, or crustacean-derived ingredients, cetyl alcohol, chlorophenol glycerol ether, methylisothiazolinones, UV absorbers, UV scattering agents, parabens, fragrances, colorants, etc." is a negative statement 66.

[0074] The detection of negative statements can be achieved, for example, by including a pre-defined plurality of negative expressions, such as "not used," "not used," and "does not include," in the detection data 25, and then having the detection device 20 perform a search on the product description using the negative expressions as keywords. Alternatively, the detection device 20 can use a language model for detecting negative sentences to detect negative statements.

[0075] In step S28, the detection device 20 detects the non-containing components described in the negative statement 66 from a plurality of component names. In Figure 6, among the designated components that are shaded, the component names enclosed in boxes are the non-containing components.

[0076] In step S29, the detection device 20 detects a positive statement regarding the ingredients contained in a product description. A positive statement is an explicit description of the ingredients contained in the product, but may also include implied descriptions of the ingredients. In the example shown in Figure 6, the article "[Latest popular ingredient addition] Carefully selected addition of retinol, arbutin, placental extract, fullerene, plant stem cells, human ceramide (synergist), vitamin A derivative, collagen, centella asiatica extract, glycyrrhizic acid" is a positive statement 67.

[0077] The detection of affirmative statements can be achieved, for example, by including a pre-defined plurality of ingredient labels, such as "formula" or "raw materials," in the detection data 25, and having the detection device 20 perform a search on the product description using the ingredient labels as keywords. Alternatively, the detection device 20 can use a language model for detecting the ingredient labels of the product to detect affirmative statements.

[0078] In step S30, the detection device 20 detects the component listed in the affirmative description 67 from a plurality of component names. In Figure 6, the component names shown in the affirmative description 67 that are shaded are the components that are contained.

[0079] In step S31, the detection device 20, based on the product descriptions corresponding to each of the more than one suspected product, detects one or more non-containing products that do not contain one or more specified ingredients from among the more than one suspected product. More specifically, based on the results of steps S28 and S30, the detection device 20 determines that the corresponding suspected product is a non-containing product when all of the more than one specified ingredients contained in each product description are included as non-containing ingredients in the negative description 66, and the more than one specified ingredient contained in the product description is not included as a containing ingredient in the positive description 67.

[0080] Furthermore, steps S29 and S30 can also be omitted. In this case, when all of the specified ingredients included in each product description are non-containing ingredients only included in the negative description 66, the detection device 20 can determine that the corresponding suspected product is a non-containing product. Furthermore, when the specified ingredient is not included in the negative description, but the positive description includes the specified ingredient, the corresponding suspected product can be determined to be a containing product.

[0081] In step S32, the detection device 20 excludes the non-containing goods detected in step S31 from one or more suspected items. Then, in step S33, the detection device 20 outputs a list of non-compliant goods as non-compliant goods (illegal goods in this example) from the remaining suspected items after excluding non-containing goods, and ends the process. In this example, the list of non-compliant goods is a list of containing goods that include at least one of a plurality of specified ingredients.

[0082] The detection device 20 can be configured to display a confirmation screen 60 on the display 46 of the operator terminal 40, which includes the detection and extraction results of steps S25 to S30, based on the operator's operation. The confirmation screen 60 can be configured, for example, to display a product description of one of a plurality of product datasets.

[0083] In the confirmation screen 60, one or more non-containing and containing ingredients can be highlighted by different markers. These different markers can be distinguished by color, such as red and blue, or by different text styles, such as underlines and bold text. The confirmation screen 60 may include a specification 65 explaining the display of markers for non-containing and containing ingredients.

[0084] In Figure 6, the names of components other than those not present in the negative description are highlighted as present components. Alternatively, different markers can be used to highlight the non-present components in the negative description and the present components in the positive description. Alternatively, different markers can be used to display the non-present components in the negative description, the present components in the positive description, and other component names.

[0085] In the confirmation screen 60 for confirming the presence of ingredients, or in a confirmation screen for confirming keywords (different from the confirmation screen 60), the ingredient names extracted in step S25 using named entity recognition, and one or more specified ingredients extracted using keywords in step S26, can be displayed using different markers. In this case, for example, the operator can visually identify the names of specified ingredients not included in the keyword set, or ingredient names that do not match the keywords due to misrepresentation. Thus, by extracting ingredient names using named entity recognition, the appropriateness of the keyword settings can be confirmed, or omissions in the keyword settings can be detected.

[0086] If the designated product detected above is a non-compliant or illegal product, the operator may notify the manufacturer of the designated product and request a halt to its production, or the operator may directly stop the production or sale of the designated product. If the designated product contains a specific ingredient, the operator may add information appropriate to its intended use to the designated product and continue production or sale. For example, it may be labeled as a product containing a specific ingredient, or the specific ingredient may be noted as a contained ingredient in the product description (e.g., a tagline) to facilitate product searches for products containing that specific ingredient.

[0087] [Function of the Invention] When a product description includes information about the ingredients contained in the product, this information may include not only affirmative statements about the ingredients that are present, but also negative statements about non-present ingredients that are not included in the product. When searching for such product descriptions using keywords, the keywords will match not only the present ingredients, but also the non-present ingredients included in the negative statements. As a result, products that do not contain the specified ingredients will be incorrectly detected as products containing the specified ingredients.

[0088] In this respect, in this invention, after extracting the suspected product from a plurality of products by keyword retrieval, non-containing products that do not contain the specified ingredient are detected based on the product description of the suspected product. Then, by excluding non-containing products from the suspected products, it is possible to detect only products containing the specified ingredient.

[0089] In product descriptions, for example, as shown in Figure 6, it might state, "Effective ingredients for dullness include nicotinic acid, retinoic acid, and retinol, but this product specifically adds high-purity hydroquinone from a carefully selected range of ingredients." However, sometimes product descriptions may list ingredients that the product does not contain, but include other ingredient names to illustrate its efficacy. In this regard, if, in addition to testing negative statements, positive statements are also tested, and products explicitly containing the ingredient are tested, then the specific product can be tested more accurately. On the other hand, if positive statements are not tested, and only negative statements are used to exclude products that do not contain the ingredient from the suspected product pool, then products that may contain the specified ingredient can be identified more broadly.

[0090] When a product is a bundle, its description includes details about each of the individual items included. Therefore, by categorizing bundles from the multiple individual items, different detection processes can be applied to bundles compared to the target product. This improves the detection accuracy of the target product. For example, for bundles, the operator can check whether they contain a specified item.

[0091] [Effects of the Invention] According to the present invention, the following effects can be achieved. (1) By excluding non-containing products that do not contain the specified ingredients from the suspected products that contain the specified ingredients in the product description, products containing the specified ingredients can be detected more accurately.

[0092] (2) By detecting negative statements in the product description, it can be determined that the ingredient name included in the negative statement is a non-containing ingredient that is not present in the product. Therefore, when all the specified ingredients in the product description are only included in the negative statement, it can be determined that the corresponding suspected product is a non-containing product.

[0093] (3) By detecting negative and positive statements in the product description, it can be determined whether the specified ingredient contained in the product description is a contained ingredient in the suspected product or a non-contained ingredient not contained in the suspected product. Therefore, when the specified ingredient contained in the product description is only included in the negative statement and not in the positive statement, the corresponding suspected product can be determined to be a non-contained product. Conversely, when the specified ingredient is not included in the negative statement but is included in the positive statement, the corresponding suspected product can be determined to be a contained product.

[0094] (4) By performing named entity recognition on the product description, more than one ingredient name can be detected from the product description. This makes it possible to detect a specific ingredient from the ingredient name.

[0095] (5) By displaying the confirmation screen 60 on the monitor 46, the operator can confirm the detected negative statements, positive statements, and the presence and absence of ingredients. In this way, the operator can confirm whether the detection device 20 is performing the detection appropriately.

[0096] (6) When a specified ingredient is an illegal ingredient in a certain product category, the illegal product can be detected from multiple products by outputting a list of products containing the illegal ingredient from products belonging to that product category. As a result, the production and sale of illegal products on the EC website can be stopped.

[0097] (7) By setting specified ingredients for each product category, products containing specified ingredients can be detected for each product category. (7) By extracting strings from image 54, the names of components contained in image 54 can also be detected.

[0098] (8) By classifying multiple products based on their corresponding product descriptions into object products belonging to a certain product category or non-object products belonging to other product categories, object products can be detected for each product category. This allows non-object products to be excluded from keyword searches or from ingredient name extraction. Furthermore, by classifying packaged products from multiple products, additional detection processing can be performed on packaged products.

[0099] (9) By using a classification model, the classification of goods can be carried out more efficiently. (10) By using language model 17, the correct answer dataset can be generated more efficiently.

[0100] This embodiment can be modified and implemented as follows. This embodiment and the following modifications can be combined and implemented to the extent that they are not technically contradictory. [Example of Change 1] In the method illustrated in Figure 5 and the system and program used to implement the method, steps S31 to S33 can be replaced by displaying the confirmation screen 60 illustrated in Figure 6. In this case, the operator can replace the detection device 20 to determine whether the suspected item contains or does not contain the product. Alternatively, the operator can use the confirmation screen 60 to confirm at least a portion of the plurality of suspected items.

[0101] [Example of Amendment 2] In the method illustrated in Figure 5 and the system and program used to implement the method, steps S27, S28, and S32 can be omitted and step S31 can be replaced. The detection device 20 will only detect products containing the specified ingredients in the affirmative statement as non-compliant products.

[0102] [Example 3 of the amendment] When the detection device 20 detects more than one suspected product by keyword retrieval in step S23, it can mark the ingredient names in the product description of each suspected product that match the keywords. In this case, the extraction of ingredient names from named entity recognition in step S25 and the extraction of specified ingredients from ingredient names in step S26 can be omitted.

[0103] [Amendment Example 4] Products tested based on specified ingredients are not limited to those containing undesirable ingredients such as illegal ingredients or allergens. For example, the technology of this invention can also be used to detect products containing specific desired ingredients, or products containing selectively used ingredients (e.g., antipyretics containing specific antipyretic and analgesic ingredients, or antihistamines containing specific anti-allergic ingredients).

[0104] [Amendment Example 5] When testing for affirmative or negative statements, the testing is not limited to the ingredient name; other elements such as the part of the plant used can also be included. This is especially important because plant-derived substances sometimes use only specific parts of the plant as pharmaceuticals.

[0105] [Amendment Example 6] The detection methods for negative and positive statements can vary depending on the product category. For example, when there is a descriptive method that is characteristic of a particular product category, a detection method based on that descriptive method can be used.

[0106] [Other Change Examples] The flowcharts and structural diagrams of this invention illustrate the architecture, function, and operation of apparatus, systems, methods, and programs (computer program products) according to embodiments of the invention. Each step in these flowcharts and each component in the structural diagrams can be equivalent to a portion of computer program code containing one or more commands for implementing logical functions. In other embodiments, a portion of the illustrated steps may be omitted, or other steps may be included; the order of the steps may differ, or some steps may be executed simultaneously. Furthermore, flowcharts illustrating a series of actions can be divided into several parts for execution, or multiple flowcharts can be executed consecutively or in a mutually related manner. Also, in other embodiments, a portion of the illustrated components may be omitted, or other components may be included, or the configuration of the components may be changed. Moreover, the functions implemented by these steps and components can be implemented by hardware, software, or a combination of hardware and software.

[0107] The system of the present invention can be implemented as a single device or distributed across multiple devices or subsystems that co-execute a program. Examples of processing systems include general-purpose central processing units, purpose-specific processors, and logic devices, as well as other types of processing devices, combinations thereof, or variations thereof.

[0108] The method of this invention is realized by software containing various programs and data, and executed by a computing system (information processing system) that executes the software. The computing system has a non-transitory computer-readable medium for storing various commands contained in the programs, and one or more processors for executing these software commands. The computing system can be implemented using a standalone computer or using a client-server architecture that distributes software across multiple computers.

[0109] Memory is a computer-readable medium, including non-transitory computer-readable medium. Non-transitory computer-readable medium may include, but is not limited to, ROM, hard disk, storage, removable media, flash memory, Memory Stick, optical media, magneto-optical media, and CD-ROM.

[0110] At least one processor may include a CPU (Central Processing Unit), GPU (Graphics Processing Unit), APU (Accelerated Processing Unit), NPU (Neural Network Processing Unit), microprocessor, microcontroller, DSP (Digital Signal Processor), FPGA (Field Programmable Gate Array), CPLD (Complex Programmable Logic Device), ASIC (Application-Specific Integrated Circuit), other processors containing general-purpose processors, or any combination of these designed to perform the functions described in this specification, but not limited thereto.

[0111] Communication between a plurality of devices or systems may be performed via one or more communication networks in accordance with well-known communication protocols. Communication networks may be, for example, intranets, the Internet, local area networks, wide area networks, wireless networks, wired networks, virtual networks, software-defined networks, or any other type of network, or combinations thereof, but are not limited thereto.

[0112] A communication interface (IF) enables a device to communicate with other devices via a communication network. A communication IF can be, for example, a LAN (Local Area Network), Wi-Fi (registered trademark), Bluetooth (registered trademark), NFC (Near Field Communication), or other wireless communication interfaces, but is not limited to these.

[0113] The following description is based on the above implementation forms and variations. [1] A method for detecting specified products, wherein at least one processor comprises: acquiring a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords representing one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the one or more specified ingredients in its corresponding product description; detecting one or more non-containing products that do not contain the one or more specified ingredients from the one or more suspected products based on the corresponding product descriptions for each of the one or more suspected products; and outputting a list of containing products containing at least one of the plurality of specified ingredients by excluding the one or more non-containing products from the one or more suspected products.

[0114] [2] As described in the above [1] method for testing designated commodities, the testing of the aforementioned one or more non-containing commodities is for each of the aforementioned one or more suspected commodities and includes the following: detecting negative statements in the aforementioned commodity descriptions regarding non-containing ingredients that are not contained in the corresponding commodities; detecting one or more of the aforementioned non-containing ingredients described in the aforementioned negative statements; and determining that the corresponding aforementioned suspected commodity is the aforementioned non-containing commodity when all of the aforementioned one or more designated ingredients contained in the aforementioned commodity descriptions are the aforementioned non-containing ingredients that are only contained in the aforementioned negative statements.

[0115] [3] As described in the above [1] method for testing designated commodities, the testing of the aforementioned one or more non-containing commodities is for each of the aforementioned one or more suspected commodities and includes the following: detecting negative statements from the aforementioned commodity descriptions regarding non-containing ingredients not contained in the corresponding commodity; detecting positive statements from the aforementioned commodity descriptions regarding containing ingredients contained in the corresponding commodity; detecting one or more of the aforementioned non-containing ingredients described in the aforementioned negative statements; detecting one or more of the aforementioned containing ingredients described in the aforementioned positive statements; and determining that the corresponding suspected commodity is the aforementioned non-containing commodity when all of the aforementioned one or more designated ingredients contained in the aforementioned commodity descriptions are the aforementioned non-containing ingredients included in the aforementioned negative statements, and the aforementioned one or more designated ingredients contained in the aforementioned commodity descriptions are not included as the aforementioned containing ingredients in the aforementioned positive statements.

[0116] [4] The specified commodity detection method described in any of [1] to [3] above includes displaying a confirmation screen on a display; the confirmation screen is configured to display the commodity description of one of the plurality of commodity datasets; in the commodity description of the confirmation screen, the aforementioned non-containing ingredients and the aforementioned one or more containing ingredients are highlighted by different markings.

[0117] [5] A method for detecting a specified product, wherein at least one processor comprises: acquiring a plurality of product datasets, each of the plurality of product datasets containing product descriptions of the plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords indicating one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the one or more specified ingredients in its corresponding product description; detecting negative statements from the product descriptions of each of the one or more suspected products regarding non-containing ingredients not contained in the corresponding previously stated suspected product; detecting one or more of the aforementioned non-containing ingredients stated in the aforementioned negative statements; and displaying a confirmation screen on a display for determining whether the aforementioned suspected products contain the one or more specified ingredients, the confirmation screen being configured to display the aforementioned product description of the corresponding previously stated suspected product, and in the aforementioned product description of the confirmation screen, the one or more non-containing ingredients are highlighted.

[0118] [6] The method for detecting a designated commodity as described in any of [1] to [5] above, wherein the detection of one or more non-containing commodities is for each of the one or more suspected commodities, includes the following: detecting one or more ingredient names from the product description by performing named entity recognition on the product description; and extracting one or more designated ingredients from the one or more ingredient names.

[0119] [7] The specified commodity testing method described in any of [1] to [6] above, wherein the aforementioned plurality of commodities are each classified into one of the plurality of commodity categories; the aforementioned one or more specified components are specified for one of the aforementioned plurality of commodity categories; the aforementioned one or more specified components are illegal components determined to be illegal in the aforementioned commodity category.

[0120] [8] The specified commodity testing method described in any of [1] to [7] above, wherein each of the aforementioned plurality of commodities is classified into one of the plurality of commodity categories; the aforementioned plurality of commodity categories include at least one of health food, pharmaceuticals, quasi-pharmaceuticals, and cosmetics; and the aforementioned one or more specified ingredients are set for each of the aforementioned commodity categories.

[0121] [9] The specified commodity detection method described in any of [1] to [8] above, wherein in the case that one of the aforementioned plurality of commodity datasets contains one or more images, and at least a portion of the aforementioned one or more images contains a string, further includes extracting the aforementioned string from the aforementioned one or more images, and the aforementioned commodity description of the aforementioned commodity dataset contains the aforementioned string extracted from the aforementioned one or more images, i.e., extracting the string.

[0122]

[10] The specified commodity testing method described in any of [1] to [9] above, wherein each of the aforementioned plurality of commodities is classified into one of the plurality of commodity categories; the aforementioned one or more specified components are specified for one of the aforementioned plurality of commodity categories; includes classifying the aforementioned plurality of commodities into three groups based on the aforementioned commodity description corresponding to each of the aforementioned plurality of commodities, the aforementioned three groups including: a group of object commodities belonging to the aforementioned commodity category; a group of non-object commodities belonging to commodity categories other than the aforementioned commodity category; and a group of packaged commodities containing the aforementioned object commodities.

[0123]

[11] The specified commodity detection method described in any of [1] to

[10] above, wherein the classification of the aforementioned three groups includes the following: inputting a commodity description of a commodity into a multi-category classification model, wherein the multi-category classification model is configured to classify the aforementioned commodity into one of the aforementioned three groups based on the aforementioned commodity description; and obtaining the classification result of the aforementioned commodity from the aforementioned multi-category classification model.

[0124]

[12] The specified commodity detection method described in any one of [1] to

[11] above includes the following: training a multi-class classification model using learning data; and generating a plurality of positive solution datasets for use as the learning data; the learning data includes a plurality of training commodity datasets; each training commodity dataset includes a training commodity description for the corresponding commodity; the training commodity description includes a description of the ingredients contained in the corresponding commodity; each positive solution dataset includes: the aforementioned training commodity description for the corresponding commodity; and a description representing the corresponding commodity. The product is a positive answer label for one of the aforementioned object products, the aforementioned non-object products, and the aforementioned bundled products; the generation of the aforementioned plurality of positive answer datasets includes the following: inputting prompts into a language model; and obtaining the output results corresponding to the aforementioned prompts from the aforementioned language model; the aforementioned prompts include: a plurality of aforementioned training product descriptions for each of the aforementioned plurality of products; and an instruction text for causing the aforementioned language model to output an answer from the aforementioned plurality of training product descriptions that the aforementioned plurality of products belong to one of the aforementioned object products, the aforementioned non-object products, and the aforementioned bundled products.

[0125]

[13] A designated product detection system includes: at least one memory storing computer program code; and at least one processor; the at least one processor executes the computer program code and is configured to: acquire a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquire one or more keywords, each of the one or more keywords representing one or more designated ingredients; extract one or more suspected products from the plurality of products by performing keyword retrieval using the one or more keywords on the plurality of product datasets, each of the suspected products containing at least one of the one or more designated ingredients in its corresponding product description; detect one or more non-containing products that do not contain the one or more designated ingredients from the one or more suspected products based on the corresponding product descriptions for each of the one or more suspected products; and output a list of containing products containing at least one of the plurality of designated ingredients by excluding the one or more non-containing products from the one or more suspected products.

[0126]

[14] A designated product detection system includes: at least one memory storing computer program code; and at least one processor; the at least one processor executes the computer program code and is configured to: acquire a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions including descriptions of the ingredients contained in the corresponding products; acquire one or more keywords, each of the one or more keywords indicating one or more designated ingredients; and extract one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords. The suspected product contains at least one of the aforementioned designated ingredients in its corresponding aforementioned product description; negative statements regarding non-containing ingredients not contained in the corresponding aforementioned suspected product are detected from the aforementioned product descriptions of each of the aforementioned suspected products; one or more of the aforementioned non-containing ingredients are detected in the aforementioned negative statements; and a confirmation screen is displayed on the display for determining whether the aforementioned suspected products contain the aforementioned one or more designated ingredients, the aforementioned confirmation screen being configured to display the aforementioned product description of the corresponding aforementioned suspected product, and in the aforementioned product description of the aforementioned confirmation screen, the aforementioned one or more non-containing ingredients are highlighted.

[0127]

[15] A computer program product is configured to cause at least one processor to perform the following: obtaining a plurality of product datasets corresponding to a plurality of products, each product dataset containing product descriptions of the ingredients contained in the corresponding product; obtaining one or more keywords, each of the one or more keywords representing one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the aforementioned specified ingredients in its corresponding aforementioned product description; detecting one or more non-containing products that do not contain the aforementioned plurality of specified ingredients from the one or more suspected products based on the aforementioned aforementioned product descriptions corresponding to each of the one or more suspected products; and outputting a list of containing products containing at least one of the aforementioned specified ingredients by excluding the one or more non-containing products from the one or more suspected products.

[0128]

[16] A computer program product is configured to cause at least one processor to perform the following: acquiring a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords indicating one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the one or more specified ingredients in its corresponding product description; detecting negative statements from the product descriptions of each of the one or more suspected products regarding non-containing ingredients not contained in the corresponding previously stated suspected product; detecting one or more of the aforementioned non-containing ingredients stated in the aforementioned negative statements; and displaying a confirmation screen on a display for determining whether the aforementioned suspected products contain the one or more specified ingredients, the confirmation screen being configured to display the aforementioned product descriptions for the corresponding previously stated suspected products, and in the aforementioned product descriptions of the confirmation screen, the one or more non-containing ingredients are highlighted.

[0129] 11: Designated Product Inspection System 13: Database 14: Web Server 15: Buyer Terminal 16: Retailer Terminal 17: Language Model 20: Designated commodity detection device 21, 31, 41: Processor 22, 32, 42: Memory 23,33,43: Communication IF 24,34: Program 25: Data for testing 30: Learning Device 35: Learning Data 40: Operator Terminal 46: Monitor 50: Sales Page 51: Slogan 52: Title 53: Detailed Explanation 54: Image 60: Confirmation screen 61,62,63,64: string 66: Negative Explanation 67: Affirmative Statement

Claims

1. A method for detecting a specified commodity, wherein, The process involves at least one processor performing the following steps: acquiring a plurality of product datasets, each of which contains product descriptions for a plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of which represents one or more specified ingredients; performing keyword retrieval on the plurality of product datasets using the one or more keywords, extracting one or more suspected products from the plurality of products, each of the suspected products containing at least one of the one or more specified ingredients in its corresponding product description; detecting one or more non-containing products that do not contain the one or more specified ingredients from the one or more suspected products based on the corresponding product descriptions for each of the one or more suspected products; and outputting a list of containing products that contain at least one of the one or more specified ingredients by excluding the one or more non-containing products from the one or more suspected products.

2. As specified in Request Item 1, the designated commodity testing method, wherein, The aforementioned detection of one or more non-containing products refers to each of the aforementioned one or more suspected products and includes the following: detecting a negative statement in the aforementioned product description regarding a non-containing ingredient that is not contained in the corresponding product; detecting one or more of the aforementioned non-containing ingredients stated in the aforementioned negative statement; and determining that the corresponding aforementioned suspected product is a aforementioned non-containing product when all of the aforementioned one or more specified ingredients contained in the aforementioned product description are the aforementioned non-containing ingredients that are only included in the aforementioned negative statement.

3. As specified in Request Item 1, the designated commodity testing method, wherein, The aforementioned detection of one or more non-containing products refers to each of the aforementioned one or more suspected products and includes the following: detecting a negative statement in the aforementioned product description regarding a non-containing ingredient not contained in the corresponding product; detecting a positive statement in the aforementioned product description regarding a contained ingredient contained in the corresponding product; detecting one or more of the aforementioned non-containing ingredients described in the aforementioned negative statement; detecting one or more of the aforementioned contained ingredients described in the aforementioned positive statement; and determining that the corresponding aforementioned suspected product is a non-containing product when all of the aforementioned one or more specified ingredients included in the aforementioned product description are the aforementioned non-containing ingredients included in the aforementioned negative statement, and the aforementioned one or more specified ingredients included in the aforementioned product description are not included as the aforementioned contained ingredients in the aforementioned positive statement.

4. The specified product testing method as described in request item 3, wherein, Includes a confirmation screen displayed on a monitor; the confirmation screen is configured to display the product description of one of the aforementioned plurality of product datasets; in the aforementioned product description of the confirmation screen, the aforementioned one or more non-containing ingredients and the aforementioned one or more containing ingredients are highlighted by different markings.

5. A method for detecting a specified commodity, wherein, The method comprises, by at least one processor: acquiring a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquiring one or more keywords, each of the one or more keywords indicating one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the one or more specified ingredients in its corresponding aforementioned product description; detecting negative statements from the aforementioned product descriptions of the one or more suspected products regarding non-containing ingredients not contained in the corresponding aforementioned suspected products; detecting one or more of the aforementioned non-containing ingredients stated in the aforementioned negative statements; and displaying a confirmation screen on a display for determining whether the aforementioned suspected products contain the one or more specified ingredients, the confirmation screen being configured to display the aforementioned product description of the corresponding aforementioned suspected product, and in the aforementioned product description of the confirmation screen, the one or more non-containing ingredients are highlighted.

6. The specified commodity testing method as described in any of requests 1 to 4, wherein, The aforementioned detection of one or more non-product-containing items refers to each of the aforementioned one or more suspected items, and includes the following: detecting one or more ingredient names from the aforementioned product description by performing named entity recognition; and extracting the aforementioned one or more designated ingredients from the aforementioned one or more ingredient names.

7. The specified commodity testing method as described in any of items 1 to 5 of the request, wherein, Each of the aforementioned plurality of goods is classified into one of the plurality of goods categories; the aforementioned one or more designated ingredients are designated for one of the aforementioned plurality of goods categories; the aforementioned one or more designated ingredients are illegal ingredients determined to be illegal in the aforementioned one goods category.

8. The specified commodity testing method described in any of requests 1 to 5, wherein, Each of the aforementioned plurality of goods is classified into one of the plurality of goods categories; the aforementioned plurality of goods categories include at least one of health foods, pharmaceuticals, quasi-pharmaceuticals, and cosmetics; the aforementioned one or more designated ingredients are set for each of the aforementioned goods categories.

9. The specified commodity testing method described in any of requests 1 to 5, wherein, In the case where one of the aforementioned plurality of product datasets contains more than one image, and at least a portion of the aforementioned one or more images contains a string, the extraction of the aforementioned string from the aforementioned one or more images is further included; the extraction of the string is defined as the aforementioned product description in the aforementioned product dataset containing the aforementioned string extracted from the aforementioned one or more images.

10. The specified commodity testing method described in any of claims 1 to 5, wherein, Each of the aforementioned plurality of goods is classified into one of the plurality of goods categories; the aforementioned one or more designated components are designated for one of the aforementioned plurality of goods categories; includes classifying the aforementioned plurality of goods into three groups based on the aforementioned goods description corresponding to each of the aforementioned plurality of goods, the aforementioned three groups including: a group of object goods belonging to the aforementioned goods category; a group of non-object goods belonging to goods categories other than the aforementioned goods category; and a group of packaged goods containing the aforementioned object goods.

11. The specified commodity testing method as described in request item 10, wherein, The classification of the aforementioned three groups includes the following: inputting a product description of a product into a multi-category classification model, wherein the multi-category classification model is configured to classify the aforementioned product into one of the aforementioned three groups based on the aforementioned product description; and obtaining the classification result of the aforementioned product from the aforementioned multi-category classification model.

12. The specified commodity testing method as described in request item 11, wherein, The method comprises the following: training a multi-class classification model using learning data; and generating multiple positive answer datasets for use as the learning data; the learning data comprises multiple training product datasets; each training product dataset comprises a training product description for the corresponding product; the training product description comprises a description of the ingredients of the corresponding product; each positive answer dataset comprises: the training product description for the corresponding product; and a positive answer label indicating that the corresponding product is one of the aforementioned target product, the aforementioned non-target product, or the aforementioned bundled product; the generation of the multiple positive answer datasets comprises: inputting a prompt into a language model; and obtaining the output corresponding to the prompt from the language model; the prompt comprises: multiple training product descriptions for each of the multiple products; and an instruction for the language model to output an answer indicating that the multiple products described by the multiple training product descriptions belong to one of the aforementioned target product, the aforementioned non-target product, or the aforementioned bundled product.

13. A designated product detection system, comprising: at least one memory storing computer program code; and at least one processor; wherein the at least one processor executes the computer program code and is configured to: acquire a plurality of product datasets, each of the plurality of product datasets containing product descriptions for the plurality of products, the product descriptions containing descriptions of the ingredients contained in the corresponding products; acquire one or more keywords, each of the one or more keywords representing one or more designated ingredients; extract one or more suspected products from the plurality of products by performing keyword retrieval using the one or more keywords on the plurality of product datasets, each of the suspected products containing at least one of the one or more designated ingredients in its corresponding product description; detect one or more non-containing products that do not contain the one or more designated ingredients from the one or more suspected products based on the corresponding product descriptions for each of the one or more suspected products; and output a list of containing products containing at least one of the one or more designated ingredients by excluding the one or more non-containing products from the one or more suspected products.

14. A designated product detection system comprising: at least one memory storing computer program code; and at least one processor; wherein the at least one processor executes the computer program code and is configured to: acquire a product dataset, the product dataset including product descriptions of the product, the product descriptions including descriptions of the ingredients contained in the product; acquire one or more keywords, each of the one or more keywords indicating one or more designated ingredients; determine whether the product is a suspected product by performing a keyword search on the product description using the one or more keywords, the suspected product containing at least one of the one or more designated ingredients in the product description; detect negative statements from the product description of the suspected product regarding non-containing ingredients not contained in the suspected product; detect one or more of the aforementioned non-containing ingredients stated in the negative statements; and display a confirmation screen on a display for determining whether the suspected product contains the one or more designated ingredients, the confirmation screen being configured to display the product description of the suspected product, and in the product description of the confirmation screen, the one or more non-containing ingredients are highlighted.

15. A computer program product for causing at least one processor to perform the following: obtaining a plurality of product datasets corresponding to a plurality of products, each product dataset containing product descriptions of the ingredients contained in the corresponding product; obtaining one or more keywords, each of the one or more keywords representing one or more specified ingredients; extracting one or more suspected products from the plurality of products by performing keyword retrieval on the plurality of product datasets using the one or more keywords, each of the suspected products containing at least one of the one or more specified ingredients in its corresponding product description; detecting one or more non-containing products that do not contain the one or more specified ingredients from the one or more suspected products based on the corresponding product descriptions of the one or more suspected products; and outputting a list of containing products containing at least one of the one or more specified ingredients by excluding the one or more non-containing products from the one or more suspected products.

16. A computer program product for causing at least one processor to perform the following: acquiring a product dataset, the product dataset including product descriptions of products, the product descriptions including descriptions of the ingredients contained in the products; acquiring one or more keywords, each of the one or more keywords indicating one or more specified ingredients; determining whether the product is a suspected product by performing a keyword search on the product description using the one or more keywords, the suspected product containing at least one of the one or more specified ingredients in the product description; detecting, from the product description of the suspected product, a negative statement regarding a non-containing ingredient not contained in the suspected product; detecting one or more of the aforementioned non-containing ingredients stated in the negative statement; and displaying a confirmation screen on a display for determining whether the suspected product contains the one or more specified ingredients, the confirmation screen being configured to display the product description of the suspected product, and in the product description of the confirmation screen, the one or more non-containing ingredients are highlighted.

Images