Product verification method, product verification system, and program

Abstract

We provide a product verification method, a product verification system, and a program. [Solution] The product verification method includes: at least one processor acquiring product data for products listed on an e-commerce site, wherein the product data includes a product description entered by the seller of the product; inputting the product data into a machine learning model, wherein the machine learning model is configured to output a confidence level that the product is a specified product when product data is input; acquiring the confidence level output by the machine learning model; and performing one or more processes from among a plurality of processes according to the confidence level.

Description

Technical Field

[0001] The present disclosure relates to a product verification method, a product verification system, and a program.

Background Art

[0002] A wide variety of products are traded on electronic commerce (EC) sites. There is a possibility that illegal products prohibited by laws and regulations are listed on such EC sites. Detecting such inappropriate listings promptly and accurately is important for protecting consumer safety.

[0003] Patent Document 1 discloses a method for checking illegal products at the time of listing, by extracting character strings included in listing information and comparing the character strings with NG character strings. The NG character strings are, for example, "copy CD-ROM" or "user registration not available". By searching listing information using keywords such as NG character strings, illegal products such as illegally copied software products can be detected. As a result, it is possible to stop the listing of illegal products.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] When sellers create product pages to sell their own products, the content varies from seller to seller. For example, product pages generally include the product name, description, or images, making it difficult to detect illegal products solely through keyword searches using prohibited strings. This challenge arises not only when detecting illegal products, but also when detecting specified products designated by the e-commerce site administrator from among the listed products.

[0006] This disclosure aims to provide a product verification method, product verification system, and program that enable more appropriate verification of designated products. [Means for solving the problem]

[0007] A product verification method according to one aspect of the present disclosure includes: at least one processor acquiring product data for a product listed on an e-commerce site, wherein the product data includes a product description entered by the seller of the product; inputting the product data into a machine learning model, wherein the machine learning model is configured to output a degree of confidence that the product is a designated product when the product data is input; acquiring the degree of confidence output by the machine learning model; and performing one or more processes from a plurality of processes according to the degree of confidence.

[0008] A product verification system according to one aspect of the present disclosure comprises at least one memory for storing computer program code and at least one processor, wherein the at least one processor executes the computer program code to acquire product data for products listed on an e-commerce site, the product data including a product description entered by the seller of the product; inputs the product data into a machine learning model, the machine learning model is configured to output a degree of confidence that the product is a designated product when the product data is input; acquires the degree of confidence output by the machine learning model; and performs one or more processes from a plurality of processes according to the degree of confidence.

[0009] A program according to one aspect of the present disclosure causes at least one processor to perform the following actions: acquire product data for products listed on an e-commerce site, wherein the product data includes a product description entered by the seller of the product; input the product data into a machine learning model, wherein the machine learning model is configured to output a confidence level that the product is a specified product when the product data is input; acquire the confidence level output by the machine learning model; and perform one or more processes from a plurality of processes according to the confidence level. [Brief explanation of the drawing]

[0010] [Figure 1] Figure 1 is a schematic diagram illustrating the product verification system according to an embodiment. [Figure 2] Figure 2 is an explanatory diagram illustrating a product sales page. [Figure 3] Figure 3 is a flowchart showing the first example of how to verify a product. [Figure 4] Figure 4 is a flowchart illustrating a method for determining the appropriate action based on the level of confidence. [Figure 5] Figure 5 is a flowchart showing a second example of how to verify a product. [Modes for carrying out the invention]

[0011] Examples of product verification methods, product verification systems 11, and programs (computer program code) will be described with reference to Figures 1 to 5. The present invention is not limited to these examples and is intended to include all modifications within the meaning and scope equivalent to the claims, as shown in the claims. Furthermore, the numerals such as "first," "second," etc., used in the following description are used to distinguish the components to which they are attached and do not rank the components.

[0012] [Overview of the Product Verification System] The product verification system 11 shown in Figure 1 (hereinafter simply referred to as "System 11") includes a verification server 20. The verification server 20 is configured to verify whether or not a product listed on an e-commerce site (hereinafter referred to as "EC site") is a designated product. Hereinafter, verifying whether or not a product is a designated product may simply be referred to as product verification. The EC site may be, for example, a mall-type shopping site or an online auction site, but is not limited to these.

[0013] The verification server 20 may, for example, verify products based on product data entered by a seller to list products on an e-commerce site. The verification server 20 may also perform product verification using one or more machine learning models 17. After the verification server 20 has verified the products, the verification server 20 or another device may perform one or more processes from among several processes according to the verification result.

[0014] Multiple processes may include at least one of the following: one or more prescribed processes specific to the designated product, and one or more common processes. Prescribed processes are processes specific to each designated product when there are multiple designated products. One or more common processes are processes that are executed for all designated products when there are multiple designated products. One or more common processes may include, but are not limited to, at least one of the following: recording the confirmation results in a database (e.g., adding a flag), notifying the seller, requesting a response from the seller, or notifying the EC site administrator.

[0015] Multiple processes may include one or more designation processes executed if the product is found to be a designated product, or one or more non-applicable processes executed if the product is not a designated product. Furthermore, multiple processes may include one or more confirmation processes executed to determine whether the product is a designated product if the confirmation was inconclusive after a single check.

[0016] The designated product may be, for example, a prohibited product whose sale on the e-commerce site is prohibited. A prohibited product is, for example, an illegal product whose sale on the e-commerce site or sale itself is illegal. In this case, the verification server 20 may, as a standard process, perform a prohibition process to prohibit the sale of that product on the e-commerce site. The prohibition process may be, but is not limited to, at least one of the following: deletion of product data, suspension of publication of the product page, notification to the seller, or notification to the e-commerce site administrator.

[0017] The designated product is not limited to a prohibited product or an illegal product. For example, the designated product may be a product that requires display of precautions related to the product at the time of sale. In this case, the confirmation server 20 may perform, as the prescribed process, an additional writing process of adding the precautions for the designated product to the product page. The precautions may be, for example, the expiry date or the best-before date for food, may be information provision for pharmaceuticals, may be age verification or warning display for tobacco, may be age verification or warning display for alcohol, or may be a handling method or a storage method for a specific product, but are not limited thereto.

[0018] The designated product may be a campaign product, a discounted product, or a product with a privilege. In this case, the confirmation server 20 may perform, as the prescribed process, an additional writing process of adding campaign information, discount information, or privilege information to the product page.

[0019] The confirmation of the product may be performed, for example, to determine whether the product belongs to a specific category. In this case, the designated product may be a product of a specific category. For example, when it is necessary to categorize products in order to sell or search for the products on an EC site, the confirmation server 20 may confirm whether the product is a product of the specific category.

[0020] Such category confirmation may be performed when category data has not been input among product data that should be input by a seller. Alternatively, the product confirmation may be performed to confirm whether the category data input by the seller is correct. In this case, the confirmation server 20 may perform, as the prescribed process, an input process of inputting the category specified by the product confirmation into the data field for the category.

[0021] The confirmation server 20 may perform the product confirmation before starting the sale of the product on the EC site, that is, before the product page is published on the EC site. The confirmation server 20 may perform the product confirmation for some designated products after the product page is published on the EC site.

[0022] In addition to this, when product data of a product on sale is edited by the seller, the confirmation server 20 may perform product confirmation before or after the edited data is published on the EC site. In this case, the confirmation server 20 may target the entire edited data for confirmation, or may target only the edited items for confirmation.

[0023] The confirmation server 20 may be implemented as a computer comprising at least one processor 21, at least one memory 22, and a communication IF 23. In the following, a case where the confirmation server 20 includes one processor 21 and a plurality of memories 22 will be described. The communication IF 23 enables communication with other devices via a network.

[0024] At least one of the plurality of memories 22 may store a program 24 to be executed by the processor 21. The program 24 may include applications and an operating system. The processor 21 implements various functions by executing a plurality of instructions (for example, a plurality of program codes) included in the program 24.

[0025] At least one of the plurality of memories 22 may store a list of designated products to be confirmed (hereinafter referred to as the "confirmation list"). The confirmation list may include names of a plurality of designated products. The confirmation list may include the genre of each designated product. Alternatively or in addition, the confirmation list may include additional items to be added to the product page according to the confirmation result.

[0026] At least one of the plurality of memories 22 may store one or more conditions for determining whether a product is a designated product. At least one of the plurality of memories 22 may store one or more program codes related to processing to be executed by the confirmation server 20 according to the confirmation result.

[0027] At least one of the multiple memories 22 may store one or more keyword sets for detecting a specified product. Each keyword set contains one or more keywords related to the corresponding specified product. For example, a keyword set for the specified product "tobacco" may include nouns such as "tobacco," "cigarette," "tabaco," "tobacco," "cigarette," or it may include brand names, product names, distributor names, or aliases of tobacco products.

[0028] The verification server 20 may perform a search (screening) using one or more keywords, and then verify the products that match the search using one or more machine learning models 17. Alternatively, or in addition to this, when verifying multiple products, the verification server 20 may determine the order in which to verify multiple products based on the screening results. For example, the verification server 20 may verify products that match the search before products that did not.

[0029] System 11 may include an administrator terminal 12 used by the administrator of the e-commerce site. The administrator terminal 12 may be implemented as a computer having at least one processor, at least one memory, and a communication interface, similar to the verification server 20. The administrator terminal 12 may have a display as an output device, or a display may be connected externally.

[0030] System 11 may include a database 13. The database 13 may be maintained in a data server 14. The data server 14 may be implemented as a computer having at least one processor, at least one memory, and a communication interface, similar to the verification server 20.

[0031] Database 13 may include a seller table for storing seller data, a buyer table for storing buyer data, and a product table for storing product data. The seller table stores multiple seller datasets for each of the multiple sellers (listers) who provide products. The buyer table stores multiple buyer datasets for each of the multiple registrants (buyers and potential buyers) on the e-commerce site. The product table stores multiple product datasets for each of the multiple products. One product dataset contains one or more data items for the corresponding one product. Each of the multiple product datasets corresponds to multiple products listed on (sold on) the e-commerce site.

[0032] System 11 may include a seller database, a buyer database, and a product database, each storing seller data, buyer data, and product data, respectively. These databases may be maintained on a single data server 14, or on multiple different data servers 14.

[0033] System 11 may include a web server 30 for providing an e-commerce site. The web server 30 may be implemented as a computer having at least one processor 31, at least one memory 32, and a communication interface 33. Below, an example is described in which the web server 30 has one processor 31 and one memory 32. The communication interface 33 enables communication with other devices via a network.

[0034] Memory 32 may store a program 34 executed by the processor 31. Memory 32 may also store part or all of the database 13. The program 34 may include an application and an operating system. The processor 31 implements various functions by executing multiple instructions (e.g., multiple program codes) contained in the program 34.

[0035] For example, the web server 30 retrieves multiple product datasets stored in the database 13 by communicating with the data server 14. The web server 30 is then configured to display each of the retrieved product datasets on multiple product pages. The web server 30 may also provide a product search function.

[0036] System 11 may include multiple buyer terminals 15 that can access an e-commerce site (web server 30) via a network. The multiple buyer terminals 15 may include personal computers or mobile devices such as smartphones or tablets. The personal computers or mobile devices may include at least one processor, at least one memory, and a communication interface. Each buyer terminal 15 may have a display as an output device, or a display may be connected externally. The display may have a touch panel as an input device.

[0037] System 11 may include a machine learning server 18 that holds one or more machine learning models 17. If System 11 uses multiple machine learning models 17, the multiple machine learning models 17 may be held on multiple different machine learning servers 18. One or more machine learning models 17 may be configured to output an index value as a result of a judgment when product data for a certain product is input. The index value may be, for example, a confidence level CL (Confidence Level) of whether a product is a specified product.

[0038] In this case, system 11 is configured to verify the product by having one or more machine learning models 17 output a confidence score CL. For example, if the confidence score CL satisfies the conditions (e.g., threshold conditions) defined for determining that the product is a specified product, the product is confirmed to be a specified product. Outputting the confidence score CL is part of the verification process for verifying the product.

[0039] System 11 may include a model generator 19 for generating one or more machine learning models 17. If System 11 uses multiple machine learning models 17, the multiple machine learning models 17 may be generated by multiple different model generators 19.

[0040] The machine learning server 18 and the model generator 19 may be implemented as computers having at least one processor, at least one memory, and a communication interface, similar to the verification server 20. The memory of the machine learning server 18 may store one or more machine learning models 17. The memory of the model generator 19 may store pre-trained or untrained learning models and training data.

[0041] System 11 may include one or more seller terminals 40 used by sellers (product vendors) who list products on the e-commerce site. The following describes an example in which System 11 includes one seller terminal 40. The seller terminal 40 may be a personal computer or a mobile device such as a smartphone or tablet.

[0042] The seller terminal 40 may be implemented as a computer comprising at least one processor 41, at least one memory 42, and a communication interface 43. Below, an example is described in which the seller terminal 40 comprises one processor 41 and one memory 42. The communication interface 43 enables communication with other devices via a network.

[0043] Memory 42 may store a program executed by the processor 41. The program may include an application and an operating system. The processor 41 implements various functions by executing multiple instructions (e.g., multiple program codes) contained in the program. The exhibitor terminal 40 may be equipped with a display 46 which is an output device, or a display may be connected externally.

[0044] [Product data and product page] The confirmation server 20 may be configured to display a registration screen 50 for registering products to be listed on the e-commerce site on the display 46 of the seller terminal 40 in response to a display request from the seller terminal 40. The registration screen 50 may include multiple input fields corresponding to multiple data items included in the product dataset.

[0045] Multiple input fields may include, but are not limited to, one or more image fields 51, number fields 52, link fields 53, product name fields 54, catchphrase fields 55, or description fields 56. The catchphrase field 55 may be a free-form text input field with a character limit. One or more keywords can be entered in the catchphrase field 55 to help the product be found in search results. The description field 56 may be a free-form text input field.

[0046] Multiple input fields may further include a category field for entering the product category, a price field for entering the product price, or a sales period field for entering the sales period (start date and end date). The seller enters the corresponding information (product description) in each of the multiple input fields.

[0047] The registration screen 50 may include a preview button 57 and a registration button 58. After the seller enters a product description in multiple input fields, if the seller operates the preview button 57 or the registration button 58, the confirmation server 20 may generate a product page to be displayed on the e-commerce site based on the input data. If the seller operates the preview button 57, the confirmation server 20 may display the generated product page on the display 46.

[0048] Figure 2 illustrates a generated product page 60. The product page 60 may, but is not limited to, include one or more images 61, a product number 62, a product name 64, a catchphrase 65, a description 66, and a product price 67. The one or more images 61 may be a still image or a video. The one or more images 61, product number 62, product name 64, catchphrase 65, and description 66 correspond to the input data in the image field 51, number field 52, product name field 54, catchphrase field 55, and description field 56, respectively. Therefore, the product page 60 includes the product description entered by the seller of the product.

[0049] The product description included on product page 60 is not limited to information directly entered on registration screen 50. For example, in one or more image fields 51, the seller may select an image from one or more images already registered in the database. The product description may also include information stored separately from the information entered via registration screen 50, not limited to images. For example, product data registered in the product table via registration screen 50 may be read and reused on the new product page 60.

[0050] When the registration button 58 is pressed, the confirmation server 20 may save the generated product page 60 or the product data contained in the product page as new product data in one of the multiple memories 22 or in the database 13. The new product data may be saved as temporary registration data in the product table of the database 13. The temporary registration data may be, for example, a product dataset with a temporary registration flag. Alternatively, the system 11 may have a separate database for storing temporary registration data. Hereinafter, products that have been temporarily registered by saving temporary registration data may be referred to as temporary registration products.

[0051] [Machine learning models] The one or more machine learning models 17 may include a trained model that has been trained on training data containing ground truth data, or it may include a training model used for general-purpose applications. The one or more machine learning models 17 may include, for example, an object detection model using image recognition, such as YOLO (You Only Look Once), or it may include one or more language models that take strings such as descriptive text as input and output them. The machine learning model 17 that performs image recognition may take the product page itself as an image as input, or it may take product images contained on the product page as input.

[0052] One or more language models may include large language models (LLMs) or small language models (SLMs). A large language model is a language model trained using a large amount of text data. A small language model is a language model that is smaller in scale than a large language model (for example, with fewer parameters). One or more language models may include a general-purpose natural language processing (NLP) model that can adapt to various natural language processing tasks such as information extraction, text summarization, text generation, or question and answer in response to input prompts. One or more language models may also be Vision-Language Models (VLMs) that can accept images in addition to language as input.

[0053] If the machine learning model 17 is a language model configured to take prompts, then at least one of the multiple memories 22 may store a prompt template. The prompt template may include instructions such as, for example, "Based on the following product description, output the probability that the described product is *** (name of the specified product)." The instructions are not limited to this and can be changed as desired. In this case, the confirmation server 20 may input a prompt to the machine learning model 17 that includes product data in addition to the prompt template.

[0054] If system 11 includes multiple machine learning models 17, the verification server 20 may verify the product based on multiple confidence levels CL output by the multiple machine learning models 17. For example, the verification server 20 may verify the product based on the maximum value among the multiple confidence levels CL. Alternatively, if the multiple confidence levels CL do not match, the verification server 20 may consider the result unclear and perform additional verification processing, or it may add a flag to the product data indicating that additional verification is required.

[0055] The verification server 20 verifies the product based on the confidence score CL output by the first machine learning model 17, depending on the specified product. Furthermore, if a definitive result cannot be obtained from this verification, it may refer to the confidence score CL output by one or more other machine learning models 17. For example, for specified products where features appear in images, the first machine learning model may be an image recognition model, and for specified products where features appear in descriptions, the first machine learning model 17 may be a language model.

[0056] If the verification server 20 has a pre-trained machine learning model 17 for a specified product, it may use that pre-trained machine learning model 17 to verify that product. On the other hand, for specified products for which there is no pre-trained model, the verification server 20 may use a general-purpose machine learning model 17.

[0057] [Criteria for determining designated products] One or more conditions for determining whether a product is a designated product may include, for example, comparing a specified index value with one or more thresholds. The index value may be, for example, a confidence level CL, the probability that a product is a designated product, or an arbitrarily set frequency. The confidence level CL may be, for example, a frequency where the percentage of the probability is an integer value. For example, the confidence level CL corresponding to a probability of 100% may be 100 (the highest value). Alternatively, the index value may be the confidence level, probability, or frequency of a product not being a designated product.

[0058] The one or more thresholds may be a plurality of thresholds, or may be three or more thresholds N1, N2, N3 (N1<N2<N3). The one or more conditions may include determining which of a plurality of (e.g., four) numerical ranges defined by three or more thresholds a prescribed index value (e.g., confidence level CL) belongs to. For example, when the first threshold N1, the second threshold N2, and the third threshold N3 are set, four numerical ranges can be set, such as CL<N1 (first range), N1≦CL<N2 (second range), N2≦CL<N3 (third range), and N3≦CL (fourth range). The threshold serving as a boundary value may be included in either the upper range or the lower range.

[0059] The number of thresholds and the number of set numerical ranges are not limited thereto, and may be arbitrarily changed. When the system 11 uses a plurality of machine learning models 17, the thresholds, numerical ranges, or one or more conditions may differ according to the machine learning models 17. In addition to or instead of this, when there are a plurality of designated products, the thresholds, numerical ranges, or one or more conditions may differ according to the designated products.

[0060] The smallest threshold (first threshold N1) is, for example, the lower limit value of the index value indicating that a product is a designated product. The smallest threshold may be a lower limit value at which it is determined that the product is not the designated product, and in this case, CL≦N1 may be set as the target range. When the index value is less than the lower limit value (e.g., probability = 1 to 10%, confidence level CL = 1 to 10) (when belonging to the first range), the confirmation server 20 can determine that the product is not the designated product. In this case, the confirmation server 20 may perform non-conformity processing for the case where the product is not the designated product. The non-conformity processing may be, for example, adding a confirmed flag indicating that product confirmation is completed to the product data.

[0061] An example of a non-applicable process when product verification is performed before the product data is published is the publication process. The publication process may be, for example, the process of publishing the product page including the product description to the e-commerce site (starting sales of the product) or the process of preparing for publication. The publication process may also be, for example, the process of changing the temporary registration flag of the temporary registration data from on to off, thereby making the temporary registration data product data for publication.

[0062] When publishing the product data, at least one of the administrator or seller may be notified that the product data will be posted on the e-commerce site. The product data for publication may be published on the e-commerce site as product page 60 after approval is obtained from the administrator or seller who has received the publication notification.

[0063] If the index value falls within the second or third range, the corresponding product may be a designated product, but the result is not definitive. In this case, the confirmation server 20 may perform a confirmation process to determine whether the product is a designated product. In the confirmation process, one or more of several different confirmation methods may be adopted, such as the first confirmation process and the second confirmation process, depending on the range to which the index value belongs (second or third range).

[0064] The confirmation process may include, for example, a notification process that sends a notification to the terminal used by the person in charge requesting confirmation. In this case, the person in charge may be at least one of the seller of the corresponding product and the administrator of the e-commerce site. The confirmation process may also involve adding a confirmation flag to the product data to indicate that additional confirmation is required in order to request confirmation from the person in charge.

[0065] If the person in charge is the seller, the confirmation server 20 may send a notification to the seller's terminal 40 used by the seller informing them that the product may be a designated product. The notification process for making this notification may include an inquiry process that includes an inquiry requesting whether the product is a designated product. If the seller responds that the product is a designated product, the confirmation server 20 may confirm that the product is a designated product.

[0066] If the designated product is a prohibited product, the verification server 20 may suspend the sale of the product (suspension process) or delete the product data (deletion process) if it does not receive a response from the seller within a specified period (e.g., 24 hours). It is advisable to include an explanation or notice of such suspension or deletion process in the notification sent to the seller's terminal 40.

[0067] Alternatively, if the seller does not respond within the specified period, the verification server 20 may send a notification to the administrator terminal 12. The administrator may then check the corresponding product data and determine whether the product is the designated product. If a disposal process is to be carried out, the administrator or at least one of the sellers may be notified that the product data will be disposed of.

[0068] If the person in charge is an administrator, the confirmation server 20 may send a notification to the administrator terminal 12 used by the administrator to inform them that the product may be a designated product. The administrator who receives the notification may confirm whether the product is a designated product by checking the corresponding product data. Either the notification to the administrator or the notification to the seller may be designated as the first confirmation process, and the other as the second confirmation process.

[0069] If the administrator determines that a product is a prohibited product, they may stop the sale of that product by destroying the corresponding product data. If the administrator determines that a product is not a prohibited product, they may proceed with the procedure to publish the corresponding product page on the e-commerce site. Alternatively, if the administrator determines that a product is not a prohibited product, they may add a registration flag to the corresponding product data. In this case, the web server 30 may publish the product page containing the product data with the registration flag on the e-commerce site.

[0070] The largest threshold (third threshold N3) is, for example, the confirmed value that confirms a product is a designated product. The confirmation server 20 may determine that a product is a designated product if the indicator value is greater than or equal to the confirmed value (for example, probability = 80-100%, confidence level CL = 80-100) (i.e., it falls within the fourth range). In this case, the confirmation server 20 may perform the designation process for products that are designated products.

[0071] The designation process may include, for example, a cancellation process to stop the publication (sale of the product) of the product data on the e-commerce site if the designated product is a prohibited product. The designation process may also include a destruction process to discard the product data. The cancellation and destruction processes may be performed before the product data is published on the e-commerce site.

[0072] The cancellation process may, for example, involve adding a prohibition flag to the product data. In this case, the administrator may delete the product data after confirming that it has the prohibition flag. Alternatively, the administrator may decide not to publish the confirmed product data without deleting it. The product data of the designated products (prohibited products) that were not deleted may be used as training data (ground truth data) when the model generation device 19 generates the machine learning model 17.

[0073] The verification server 20 may perform notification processing and publication processing if the confidence level CL is less than the confirmed value, or if the confidence level CL is between the lower limit and the confirmed value. In particular, if the specified product is not a prohibited product, the server may perform product verification after publishing the product data on the e-commerce site and perform processing according to the verification result.

[0074] [Product verification process] Referring to Figure 3, the verification process for the first example of a product will be explained. The verification server 20 may perform the verification process for each of the multiple provisionally registered products sequentially, or in any order. The data subject to the verification process may be product data entered by the seller, product data saved as provisional registration data, or product data included in the edited page edited by the seller. In the case of product data included in the edited page, the verification server 20 may only verify the edited edited data.

[0075] In step S11, the verification server 20 obtains product data for the product to be verified. The obtained product data includes the product description entered by the product seller. In the following step S12, the verification server 20 may process the product data as needed to make it suitable for the machine learning model 17 to be used.

[0076] Data processing may include, but is not limited to, at least one of the following: processing images, extracting text from images, or extracting multiple words (e.g., at least one of nouns and verbs) from text. Multiple words can be obtained, for example, by performing named entity recognition on a descriptive text.

[0077] Image processing may include, for example, extracting or cropping a portion of an image, or preprocessing an image. Image preprocessing may include, but is not limited to, grayscale conversion, binarization, or normalization to highlight distinctive features. Image preprocessing may include, but is not limited to, image transformation, dimensionality reduction, or resizing to remove noise. Image preprocessing may include, but is not limited to, inversion, smoothing, or brightness adjustment to increase feature density.

[0078] In step S13, the confirmation server 20 transmits the product data acquired in step S11 or the product data processed in step S12 (processed data) to the machine learning server 18, so as to input the same into the machine learning model 17. When the product data is input, the machine learning model 17 outputs a confidence level CL indicating that the product related to the product data is the designated product. In step S14, the confirmation server 20 acquires the confidence level CL output by the machine learning model 17 from the machine learning server 18. Thereafter, the confirmation server 20 executes a determination process (S50), and terminates the process.

[0079] The determination process (S50) will be described with reference to FIG. 4. In the determination process (S50), the confirmation server 20 performs one or more processes among a plurality of processes according to the confidence level CL. First, in step S51, the confirmation server 20 determines whether or not the confidence level CL falls within a first range. The first range is, for example, CL<N1. If the confidence level CL falls within the first range (step S51=YES), the confirmation server 20 proceeds to step S52.

[0080] In step S52, the confirmation server 20 executes a disclosure process, and terminates the confirmation process. The disclosure process may be, for example, turning off a temporary registration flag of temporary registration data. If the confidence level CL does not fall within the first range (step S51=NO), the confirmation server 20 proceeds to step S53.

[0081] In step S53, the confirmation server 20 determines whether or not the confidence level CL falls within a second range. The second range is, for example, N1≦CL<N2. If the confidence level CL falls within the second range (step S53=YES), the confirmation server 20 proceeds to step S54.

[0082] In step S54, the confirmation server 20 executes a confirmation process, and terminates the confirmation process. The confirmation process may be, for example, transmitting a notification for informing that there is a possibility that the product is the designated product to the administrator terminal 12. The administrator who has received the notification confirms the corresponding product data, so as to confirm whether or not the product is the designated product.

[0083] When the confidence level CL does not belong to the second range (step S53=NO), the confirmation server 20 proceeds to step S55. In step S55, the confirmation server 20 determines whether or not the confidence level CL belongs to the third range. The third range is, for example, N2≦CL<N3. When the confidence level CL belongs to the third range (step S55=YES), the confirmation server 20 proceeds to step S56.

[0084] In step S56, the confirmation server 20 executes a notification process and ends the confirmation process. The notification process may be, for example, transmitting a notification for informing that the product may be the designated product to the seller terminal 40. This notification may include a request for a reply as to whether the product is the designated product. In this case, the confirmation server 20 may determine whether or not the product is the designated product based on the reply from the seller that the product is the designated product or is not the designated product.

[0085] When the confidence level CL does not belong to the third range (step S55=NO), the confirmation server 20 proceeds to step S57. In step S57, the confirmation server 20 executes a designation process. The designation process may be, for example, if the designated product is a prohibited product, a suspension process for suspending the publication of product data (sale of the product), or a discarding process for discarding the product data. The discarding process may be, for example, deleting the temporarily registered data.

[0086] Steps S52, S54, S56, and S57 may be executed by another execution device. In this case, the confirmation server 20 may transmit a processing request to the other execution device to execute the corresponding process instead of steps S52, S54, S56, and S57, or may add a flag for executing the corresponding process to the product data.

[0087] With reference to FIG. 5, the product confirmation process of the second example will be described. Similarly to the first example, the confirmation process of the second example may target both the temporarily registered data and the edited data. In step S21, the verification server 20 retrieves the product data. The product data includes a string and one or more images. The following example describes a case where the image data includes one image, but the product data may include multiple images.

[0088] In the following step S22, the verification server 20 extracts images from the product data. The verification server 20 may perform the preprocessing described in step S12 on the extracted images. In step S23, the verification server 20 extracts strings from the images and other product data. The strings may be extracted as multiple words.

[0089] In step S24, the verification server 20 performs a keyword search on the extracted string using one or more keywords included in the keyword set. In step S25, the verification server 20 determines, through the search, whether the string contains characters corresponding to at least one of the one or more keywords. If the string contains one or more keywords (step S25 = YES), the verification server 20 proceeds to step S26.

[0090] In step S26, the verification server 20 inputs the image extracted in step S22 into the machine learning server 18, thereby inputting the image into the machine learning model 17 that performs image recognition. Upon receiving the image, the machine learning model 17 outputs a confidence level CL indicating that the product in the image is the specified product.

[0091] In step S27, the verification server 20 obtains the confidence score CL output by the machine learning model 17 from the machine learning server 18. Then, the verification server 20 performs the judgment process (S50) described above and terminates the process. If the image data includes multiple images, the verification server 20 may obtain a confidence score CL for each of the multiple images and perform the judgment process based on the maximum value among those multiple confidence scores CL.

[0092] On the other hand, if the string does not contain one or more keywords in step S25 (step S25=NO), the verification server 20 proceeds to step S28. The verification server 20 may terminate the verification process after executing the publication process in step S28. In other words, if the product description does not contain one or more keywords related to the specified product, there is a high possibility that the product related to the product data is not the specified product, so the corresponding product page 60 may be published on the EC site as is.

[0093] However, for the product data published here, the confidence level CL obtained by performing steps S26 and S27 may be used to perform the judgment process (S50). In particular, if there are many products to be checked, performing the machine learning model 17 to check all products and then the subsequent manual confirmation process would take a long time before the products could be put on sale. Therefore, products that do not meet the keyword screening criteria may be published first, and then the check process may be performed.

[0094] [Effects of this disclosure] System 11 selectively executes one or more of several processes, such as publication processing, confirmation processing, notification processing, or specified processing, based on the confidence score CL output by the machine learning model 17. Therefore, for products for which a reasonably reliable confidence score CL has been obtained, publication processing or specified processing can be automatically executed.

[0095] On the other hand, if a definitive result cannot be obtained, for example, when the confidence level CL is in the range of above the lower limit but below the confirmed value, the product can be confirmed through a manual confirmation process or notification process. This makes it possible to achieve both labor savings in the confirmation process and reliable detection of the specified product.

[0096] [Effects of this disclosure] According to this disclosure, the following effects can be achieved. (1) The verification server 20 performs processing according to the confidence score CL output from the machine learning model 17. This makes it possible to verify the specified product more appropriately.

[0097] (2) If the confidence level CL is less than the confirmed value, the notification process can inform the EC site administrator or seller that the product may be a designated product. This allows the administrator to confirm whether the product is a designated product by visually checking the product page, etc. Alternatively, the administrator can confirm whether the product is a designated product by obtaining a response from the seller to the notification.

[0098] (3) If the confidence level CL is less than the confirmed value, the product can be verified and sales can be started promptly by performing the publication process along with the notification process. (4) If the confidence level CL falls within the third range, a notification can be sent to the seller terminal 40 to allow the seller to confirm whether the product is a designated product. After sending the notification to the seller terminal 40, if the seller does not respond that the product is not a prohibited product, the transaction can be canceled, thereby preventing suspected prohibited products from being sold on the e-commerce site.

[0099] (5) If the confidence level CL of whether a product is a prohibited product is greater than or equal to the confirmed value, the prohibited product can be prevented from being sold on the e-commerce site by automatically canceling the process.

[0100] (6) If the confidence level CL of whether a product is a prohibited product is greater than or equal to the confirmed value, the prohibited product can be prevented from being sold on the e-commerce site by performing cancellation and destruction processes before the product data is published on the e-commerce site.

[0101] (7) By setting three thresholds for the confidence level CL, four numerical ranges can be set. Then, by determining which numerical range the confidence level CL belongs to, the process to be executed can be automatically performed according to the confidence level CL. Furthermore, by setting multiple numerical ranges according to the process to be executed, more detailed verification can be performed according to the specified product.

[0102] (8) When the seller edits the product page, a confidence level CL is obtained based on the editing data. This allows for efficient verification of whether the product has been changed to a specified product when the product page is edited.

[0103] (9) The verification server 20 can perform a publication process, a confirmation process, or a specified process depending on the confidence level CL. Therefore, the administrator's work in verifying products can be streamlined. As a result, it is possible to quickly start selling registered products on the e-commerce site while suppressing the sale of prohibited products on the e-commerce site.

[0104] (10) The verification server 20 can perform publication processing, cancellation processing, or notification processing according to the confidence level CL. Therefore, the administrator can reduce the effort required to verify products. As a result, it is possible to quickly start selling registered products on the e-commerce site while suppressing the sale of prohibited products on the e-commerce site.

[0105] (11) In the second verification method, the verification server 20 performs keyword-based screening and inputs images of products whose product descriptions contain the keywords into the machine learning model 17. This allows verification of whether a product is the specified product using both text and images. Furthermore, by performing keyword-based screening, the time and computational resources required for verification by the machine learning model 17, as well as the effort required for subsequent confirmation processing, can be reduced.

[0106] This embodiment can be implemented with the following modifications. This embodiment and the following modifications can be combined with each other to the extent that they do not contradict each other technically. [Example of changing the verification server] System 11 may include a registration server separate from the confirmation server 20. This registration server may, on behalf of the confirmation server 20, display a registration screen 50 on the display 46 of the seller terminal 40, retrieve product descriptions entered via the registration screen 50, and generate product pages 60 based on the product descriptions.

[0107] [Verification process for the third example] In the verification method shown in Figure 5, the verification server 20 may, instead of steps S24, S25, and S28, input the extracted string into the machine learning model 17, which is a language model, and obtain a confidence score CL from this machine learning model 17. In this case, the determination process may be performed based on the higher of the confidence score CL output by the language model and the confidence score CL output by the image recognition model.

[0108] [Verification process for the 4th example] In the verification method shown in Figure 5, the verification server 20 may obtain the product page 60 in step S21, omit steps S22 and S23, and perform a keyword search on the product page 60 in step S24. Subsequently, in step S26, the verification server 20 may input the product page 60, including the image, into the machine learning model 17, which is a VLM.

[0109] [Verification process for the 5th case] The verification server 20 may obtain a confidence score CL for multiple products using one of the multiple machine learning models 17 (for example, the model with the shortest processing time or the model with the lowest computational load). The verification server 20 may then determine the order in which to perform additional verification processing based on the confidence score CL of each of the multiple products. For example, additional verification processing may be performed on products with a high confidence score CL before products with a low confidence score CL. Alternatively, if the confidence score CL in the initial verification is greater than or equal to the confirmed value, a specified processing may be performed, and if the confidence score CL is less than the lower limit, a publication processing may be performed, while additional verification processing may be performed on products where the confidence score CL is greater than or equal to the lower limit but less than the confirmed value. The additional verification processing may be performed using one or more machine learning models 17 with different input data than the machine learning model 17 used in the initial verification processing.

[0110] [Examples of changes to flowcharts and configuration diagrams] The flowcharts and configuration diagrams of this disclosure illustrate the architecture, functionality, and operation of the apparatus, system, method, and program according to embodiments of this disclosure. Each step included in these flowcharts and each component included in the configuration diagrams may correspond to a part of a program containing one or more instructions for realizing a logical functional unit. In other embodiments, some of the illustrated steps may be omitted, other steps may be included, the order of the steps may be different, or some steps may be executed simultaneously. Furthermore, a flowchart described as a series of actions may be divided into several parts and executed, or multiple flowcharts may be executed sequentially or in relation to each other. Also in other embodiments, some of the illustrated components may be omitted, other components may be included, or the arrangement of components may be changed. Furthermore, the functions realized by these steps and components may be realized by hardware, software, or a combination of hardware and software.

[0111] [Example of system changes] The statement "at least one processor executes multiple instructions to perform the specified operation" means that if system 11 comprises multiple servers, at least one processor among the servers executes the corresponding instructions (program code). If system 11 is implemented by a single server, it means that at least one processor on that server executes the corresponding instructions. Furthermore, if system 11 includes one or more client terminals (e.g., seller terminals or administrator terminals), it includes the execution of at least one processor on these terminals by executing at least a portion of the multiple instructions (e.g., a script for controlling the registration screen 50). In other words, "at least one processor" may be multiple processors included in either one or more servers or one or more terminals, and which server or terminal executes which operation can be arbitrarily changed. The one or more instructions executed by a processor may be stored in the memory of the server that has the executing processor, or in the memory of another server, terminal, or hardware.

[0112] [Example of memory modification] The memory included in the apparatus of this disclosure is a computer-readable storage medium, including non-transitory computer-readable medium. The memory may include, but is not limited to, ROM, hard disks, storage, removable media, flash memory, memory sticks, optical media, magneto-optical media, and CD-ROMs.

[0113] [Example of processor change] One or more processors included in the apparatus of this disclosure may include, but are not limited to, a CPU (central processing unit), a GPU (graphics processing unit), an APU (accelerated processing unit), an NPU (Neural network Processing Unit), a microprocessor, a microcontroller, a DSP (digital signal processor), an FPGA (field programmable gate array), a CPLD (Complex Programmable Logic Device), an application-specific integrated circuit (ASIC), a general-purpose processor, or any combination thereof designed to perform the functions described herein.

[0114] [Example of network changes] Communication between multiple devices or systems may be carried out via one or more communication networks in accordance with well-known communication protocols. The communication network may be, but is not limited to, an intranet, the internet, a local area network, a wide area network, a wireless network, a wired network, a virtual network, a software-defined network, or any other type of network, or a combination thereof.

[0115] [Example of changing the communication interface] A communication interface (IF) enables one device to communicate with other devices via a communication network. The communication interface may be, but is not limited to, a LAN (Local Area Network), Wi-Fi (registered trademark), Bluetooth (registered trademark), NFC (Near Field Communication), or other wireless communication interfaces.

[0116] The embodiments and modifications described above are listed below. [1] At least one processor, The acquisition of product data for products listed on an e-commerce site, wherein the product data includes the product description entered by the seller of the product. The process involves inputting the aforementioned product data into a machine learning model, wherein the machine learning model is configured to output a confidence level that the product is a specified product when the product data is input. Obtaining the confidence level output by the aforementioned machine learning model, Among multiple processes, one or more processes are performed according to the confidence level, Product verification methods, including those mentioned above.

[0117] [2] One or more processes corresponding to the confidence level include a notification process that is executed when the confidence level is less than the confirmed value that confirms the product is the designated product. The notification process includes sending a notification to at least one of the administrator terminal used by the administrator of the e-commerce site or the seller terminal used by the seller. Product verification method as described in [1] above.

[0118] [3] One or more processes corresponding to the confidence level include a disclosure process and a notification process, which are performed when the confidence level is less than the confirmation value that confirms the product is the designated product. The notification process includes sending a notification to at least one of the administrator terminal used by the administrator of the e-commerce site or the seller terminal used by the seller, The aforementioned publication process includes publishing the product page containing the product description on the e-commerce site. Product verification method as described in [1] or [2] above.

[0119] [4] The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, One or more processes corresponding to the confidence level include a notification process that is performed when the confidence level is less than the confirmed value that determines the product is the prohibited product. The notification process includes sending a notification to the seller's terminal used by the seller requesting a response that the product is not a prohibited product. If the seller does not respond within the specified period that the product is not one of the prohibited items, the following actions may be taken: The product verification method described in any of the above [1] to [3].

[0120] [5] The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, One or more processes corresponding to the confidence level include a termination process when the confidence level is equal to or greater than the confirmed value that confirms the product is the designated product. The aforementioned discontinuation process includes discontinuing the publication of the product page containing the product description on the e-commerce site. The product verification method described in any of the above [1] to [4].

[0121] [6] The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, One or more processes corresponding to the confidence level include a cancellation process and a disposal process when the confidence level is equal to or greater than the confirmed value that confirms the product is the designated product. The aforementioned discontinuation process includes discontinuing the publication of the product page containing the product description on the e-commerce site, The aforementioned disposal process includes discarding the product data, The aforementioned cancellation process and the aforementioned destruction process are performed before the product page is published on the e-commerce site. The product verification method described in any of the above [1] to [5].

[0122] [7] The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, The process further includes determining which of the four numerical ranges defined by the three thresholds the aforementioned confidence level belongs to, The three thresholds include the smallest first threshold, a second threshold greater than the first threshold, and a third threshold greater than the second threshold. The four numerical ranges include a first range less than the first threshold, a second range greater than or equal to the first threshold but less than the second threshold, a third range greater than or equal to the second threshold but less than the third threshold, and a fourth range greater than or equal to the third threshold. If the confidence level falls within the fourth range, one or more actions corresponding to the confidence level include a discontinuation action to stop the sale of the product. If the confidence level falls within the third range, one or more processes corresponding to the confidence level include an inquiry process that sends an inquiry to the seller's terminal used by the seller requesting a response as to whether the product is a prohibited product. If the confidence level falls within the second range, one or more processing steps corresponding to the confidence level include a notification process that sends a notification to an administrator terminal used by the administrator of the e-commerce site informing them that the product may be a prohibited product. The product verification method described in any of the above [1] to [6].

[0123] [8] The acquisition of editing data, wherein the editing data includes editing information made by the seller of a product sold on the e-commerce site, editing the product page of that product. Inputting the aforementioned editing information into the machine learning model, Obtaining the confidence level based on the editing information output by the machine learning model, Performing one or more of the above processes according to the confidence level, including, The product verification method described in any of the above [1] to [7].

[0124] [9] The aforementioned multiple processes are, A publishing process for publishing the product page containing the aforementioned product description to the aforementioned e-commerce site, A confirmation process to determine whether the aforementioned product is the aforementioned designated product, The designation process when the aforementioned product is the aforementioned designated product, including at least one of the following: The product verification method described in any of the above [1] to [8].

[0125]

[10] The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, One or more processes corresponding to the aforementioned confidence level are: A publishing process for publishing a product page containing the product description to the e-commerce site when the confidence level is less than the lower limit indicating the possibility that the product is the designated product, If the confidence level is equal to or greater than the confirmed value that confirms the product is the designated product, a cancellation process is performed to stop the publication of the product page on the e-commerce site. Notification process when the confidence level is greater than or equal to the lower limit value and less than the confirmed value, wherein the notification process includes sending a notification to at least one of the administrator terminal used by the administrator of the e-commerce site or the seller terminal used by the seller, including, The product verification method described in any of the above [1] to [9].

[0126]

[11] The aforementioned product data includes a string and one or more images. Extracting one or more images from the aforementioned product data, Extracting the string from the aforementioned product data, A keyword search is performed on the aforementioned string using one or more keywords related to the specified product. If the string contains one or more of the keywords, then one or more images are input to the machine learning model. including, The product verification method described in any of the above [1] to

[10] .

[0127]

[12] At least one memory for storing computer program code, At least one processor, The system comprises, and the at least one processor executes the computer program code, The acquisition of product data for products listed on an e-commerce site, wherein the product data includes the product description entered by the seller of the product. The process involves inputting the aforementioned product data into a machine learning model, wherein the machine learning model is configured to output a confidence level that the product is a specified product when the product data is input. Obtaining the confidence level output by the aforementioned machine learning model, Among multiple processes, one or more processes are performed according to the confidence level, A product verification system configured to perform the following actions.

[0128]

[13] At least one processor, The acquisition of product data for products listed on an e-commerce site, wherein the product data includes the product description entered by the seller of the product. The process involves inputting the aforementioned product data into a machine learning model, wherein the machine learning model is configured to output a confidence level that the product is a specified product when the product data is input. Obtaining the confidence level output by the aforementioned machine learning model, Among multiple processes, one or more processes are performed according to the confidence level, A program to execute. [Explanation of Symbols]

[0129] 11…Product verification system, 12…Administrator terminal, 13…Database, 14…Data server, 15…Buyer terminal, 17…Machine learning model, 18…Machine learning server, 19…Model generator, 20…Verification server, 21…Processor, 22…Memory, 23…Communication interface, 24…Program, 30…Web server, 31…Processor, 32…Memory, 33…Communication interface, 34…Program, 40…Seller terminal, 41…Processor, 42 ...Memory, 43...Communication Interface, 46...Display, 50...Registration Screen, 51...Image Field, 52...Number Field, 53...Link Field, 54...Product Name Field, 55...Catchphrase Field, 56...Description Field, 57...Preview Button, 58...Register Button, 60...Product Page, 61...Image, 62...Product Number, 64...Product Name, 65...Catchphrase, 66...Description, 67...Product Price, CL...Confidence Level, N1...First Threshold, N2...Second Threshold, N3...Third Threshold.

Claims

1. At least one processor, The acquisition of product data for products listed on an e-commerce site, wherein the product data includes the product description entered by the seller of the product. The process involves inputting the aforementioned product data into a machine learning model, wherein the machine learning model is configured to output a confidence level that the product is a specified product when the product data is input. Obtaining the confidence level output by the aforementioned machine learning model, Among multiple processes, one or more processes are performed according to the confidence level, Product verification methods, including those mentioned above.

2. One or more processes corresponding to the confidence level include a notification process that is executed when the confidence level is less than the confirmed value that confirms the product is the designated product. The notification process includes sending a notification to at least one of the administrator terminal used by the administrator of the e-commerce site or the seller terminal used by the seller. The product verification method described in claim 1.

3. The one or more processes corresponding to the confidence level include a disclosure process and a notification process, which are performed when the confidence level is less than the confirmation value that confirms the product is the designated product. The notification process includes sending a notification to at least one of the administrator terminal used by the administrator of the e-commerce site or the seller terminal used by the seller, The aforementioned publication process includes publishing the product page containing the product description on the e-commerce site. The product verification method described in claim 1.

4. The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, One or more processes corresponding to the confidence level include a notification process that is executed when the confidence level is less than the confirmed value that confirms the product is the prohibited product. The notification process includes sending a notification to the seller's terminal used by the seller requesting a response that the product is not a prohibited product. If the seller does not respond within the specified period that the product is not one of the prohibited items, the following actions may be taken: The product verification method described in claim 1.

5. The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, One or more processes corresponding to the confidence level include a termination process when the confidence level is equal to or greater than the confirmed value that confirms the product is the designated product. The aforementioned discontinuation process includes discontinuing the publication of the product page containing the product description on the e-commerce site. The product verification method described in claim 1.

6. The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, The one or more processes corresponding to the confidence level include a cancellation process and a disposal process when the confidence level is equal to or greater than the confirmed value that confirms the product is the designated product. The aforementioned discontinuation process includes discontinuing the publication of the product page containing the product description on the e-commerce site, The aforementioned disposal process includes discarding the product data, The aforementioned cancellation process and the aforementioned destruction process are performed before the product page is published on the e-commerce site. The product verification method described in claim 1.

7. The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, The process further includes determining which of the four numerical ranges defined by the three thresholds the aforementioned confidence level belongs to, The three thresholds include the smallest first threshold, a second threshold greater than the first threshold, and a third threshold greater than the second threshold. The four numerical ranges include a first range less than the first threshold, a second range greater than or equal to the first threshold but less than the second threshold, a third range greater than or equal to the second threshold but less than the third threshold, and a fourth range greater than or equal to the third threshold. If the confidence level falls within the fourth range, one or more actions corresponding to the confidence level include a discontinuation action to stop the sale of the product. If the confidence level falls within the third range, one or more processes corresponding to the confidence level include an inquiry process that sends an inquiry to the seller's terminal used by the seller requesting a response as to whether the product is a prohibited product. If the confidence level falls within the second range, one or more processing steps corresponding to the confidence level include a notification process that sends a notification to an administrator terminal used by the administrator of the e-commerce site informing them that the product may be a prohibited product. The product verification method described in claim 1.

8. The acquisition of editing data, wherein the editing data includes editing information made by the seller of a product sold on the e-commerce site, editing the product page of that product. Inputting the aforementioned editing information into the machine learning model, Obtaining the confidence level based on the editing information output by the machine learning model, Performing one or more of the above processes according to the confidence level, including, A method for verifying a product as described in any one of claims 1 to 7.

9. The aforementioned multiple processes are, A publishing process for publishing the product page containing the aforementioned product description to the aforementioned e-commerce site, A confirmation process to determine whether the aforementioned product is the aforementioned designated product, The designation process when the aforementioned product is the aforementioned designated product, Including at least one of the following: A method for verifying a product as described in any one of claims 1 to 7.

10. The aforementioned designated goods are prohibited goods whose sale on the aforementioned e-commerce site is prohibited, One or more processes corresponding to the aforementioned confidence level are: A publishing process for publishing a product page containing the product description to the e-commerce site when the confidence level is less than the lower limit indicating the possibility that the product is the designated product, If the confidence level is equal to or greater than the confirmed value that confirms the product is the designated product, a cancellation process is performed to stop the publication of the product page on the e-commerce site. Notification process when the confidence level is greater than or equal to the lower limit value and less than the confirmed value, wherein the notification process includes sending a notification to at least one of the administrator terminal used by the administrator of the e-commerce site or the seller terminal used by the seller, including, A method for verifying a product as described in any one of claims 1 to 7.

11. The aforementioned product data includes a string and one or more images. Extracting one or more images from the aforementioned product data, Extracting the string from the aforementioned product data, A keyword search is performed on the aforementioned string using one or more keywords related to the specified product. If the string contains one or more of the keywords, then one or more images are input to the machine learning model. including, A method for verifying a product as described in any one of claims 1 to 7.

12. At least one memory for storing computer program code, At least one processor, The system comprises, and the at least one processor executes the computer program code, The acquisition of product data for products listed on an e-commerce site, wherein the product data includes the product description entered by the seller of the product. The process involves inputting the aforementioned product data into a machine learning model, wherein the machine learning model is configured to output a confidence level that the product is a specified product when the product data is input. Obtaining the confidence level output by the aforementioned machine learning model, Among multiple processes, one or more processes are performed according to the confidence level, A product verification system configured to perform the following actions.

13. At least one processor, The acquisition of product data for products listed on an e-commerce site, wherein the product data includes the product description entered by the seller of the product. The process involves inputting the aforementioned product data into a machine learning model, wherein the machine learning model is configured to output a confidence level that the product is a specified product when the product data is input. Obtaining the confidence level output by the aforementioned machine learning model, Among multiple processes, one or more processes are performed according to the confidence level, A program to execute.

Images