Purchase assistance method, purchase assistance program, purchase assistance system, and ai use transboundary ec system

Abstract

This purchase assistance system acquires, via an interface screen image displayed on a user terminal, page identification information for identifying a web page on which a commodity that a user desires to purchase is published. The purchase assistance system automatically extracts, from HTML constituting the web page identified by the page identification information, some necessary information from among items of information pertaining to the commodity published on the web page (S13 to S15). The purchase assistance system displays, on an interface screen image of the user terminal, at least some of the extracted necessary information (S20).

Description

Purchase assistance method, purchase assistance program, purchase assistance system, and AI-based cross-border e-commerce system

[0001] The present disclosure relates to a purchase assistance method, a purchase assistance program, a purchase assistance system, and an AI-based cross-border e-commerce system for assisting users in purchasing overseas products from their country of residence.

[0002] One way for users to purchase overseas products while remaining in their home country is to use an EC (Electronic Commerce) site that handles overseas products. An EC site is a website that enables the buying and selling of products over the Internet. There are two types of EC sites: one that is primarily built by a seller to sell their own products (in-house EC type), and another where multiple sellers offer their products (mall type). However, regardless of the type of EC site used, if the overseas product the user wishes to purchase is not available on the EC site, the user will be unable to purchase the product they desire.

[0003] If the overseas product that a user wishes to purchase is not available on an e-commerce site, the user can use an import agent. In recent years, services have been provided that allow users to request the import of products from an import agent via a website. However, even when using a website, the user still has to browse the overseas website where the product is listed, understand the product information written in a foreign language, and enter it into the import agent's website or fill out a paper form and send it to the import agent.

[0004] In light of the above background, technologies have been proposed that aim to support users in purchasing overseas products. For example, according to the technology described in Patent Literature 1, a product page for a product sold on an e-commerce site is displayed in a web browser on a user's terminal, along with an input field for purchasing the product from overseas. The input field allows users to input information such as the quantity, color, and size of the product.

[0005] JP 2023-100932 A

[0006] However, even with the technology described in Patent Document 1, users still need to manually find the information needed to purchase a product from the information provided on the overseas website and enter it into the appropriate fields. This makes it difficult for users to easily purchase overseas products, and there are cases where input errors or other imperfections in the sales process occur. Furthermore, with conventional services, it can be difficult to understand shipping fees or customs duties imposed when importing products.

[0007] A typical objective of the present disclosure is to provide a purchase assistance method, a purchase assistance program, a purchase assistance system, and an AI-based cross-border e-commerce system that can more appropriately assist users in purchasing overseas products from their country of residence by solving at least one of the above problems.

[0008] A purchase assistance method provided by a typical embodiment of the present disclosure is a purchase assistance method executed in a purchase assistance system that assists a user in purchasing overseas products from their country of residence via an interface screen displayed on a user terminal used by the user, and includes a page identification information acquisition step of acquiring, via the interface screen, page identification information that identifies a web page on which the product that the user wishes to purchase is listed, an extraction step of automatically extracting, from HTML that constitutes the web page identified by the page identification information, some necessary information from the information on the product listed on the web page, and a necessary information display step of displaying, on the interface screen, at least some of the necessary information extracted in the extraction step.

[0009] A purchase assistance program provided by a typical embodiment of the present disclosure is a purchase assistance program executed in a purchase assistance system that assists a user in purchasing overseas products from their country of residence via an interface screen displayed on a user terminal used by the user, and the purchase assistance program is executed by a control unit of the purchase assistance system to cause the purchase assistance system to perform the following steps: a page identification information acquisition step that acquires, via the interface screen, page identification information that identifies a web page on which the product that the user wishes to purchase is listed; an extraction step that automatically extracts, from the HTML that constitutes the web page identified by the page identification information, some necessary information from the product information listed on the web page; and a necessary information display step that displays, on the interface screen, at least some of the necessary information extracted in the extraction step.

[0010] A typical embodiment of the present disclosure provides a purchase assistance system that assists a user in purchasing overseas products from their country of residence via an interface screen displayed on a user terminal used by the user, and executes the following steps: a page identification information acquisition step that acquires, via the interface screen, page identification information that identifies a web page on which the product the user wishes to purchase is listed; an extraction step that automatically extracts, from the HTML that constitutes the web page identified by the page identification information, some necessary information from the product information listed on the web page; and a necessary information display step that displays, on the interface screen, at least some of the necessary information extracted in the extraction step.

[0011] An AI-based cross-border e-commerce system provided by a typical embodiment of the present disclosure is an AI-based cross-border e-commerce system that assists a user in purchasing overseas products from their country of residence via an interface screen displayed on a user terminal used by the user, and executes the following steps: a page-specific information acquisition step of acquiring, via the interface screen, page-specific information that identifies a web page on which the product the user wishes to purchase is posted; an extraction step of automatically extracting the necessary information by inputting the HTML of the web page identified by the page-specific information into a mathematical model trained by a machine learning algorithm so as to extract and output the necessary information formed on the web page when the HTML of the web page is input; and a necessary information display step of displaying at least a portion of the necessary information extracted in the extraction step on the interface screen.

[0012] According to the purchase assistance method, purchase assistance program, purchase assistance system, and AI-based cross-border e-commerce system disclosed herein, purchases of overseas products by users from their country of residence are appropriately subsidized.

[0013] The purchase assistance system exemplified in the present disclosure assists a user in purchasing overseas products from their country of residence via an interface screen displayed on a user terminal used by the user. The purchase assistance method of the present disclosure is executed in the purchase assistance system. The purchase assistance program of the present disclosure is executed by a control unit of the purchase assistance system. The purchase assistance system exemplified in the present disclosure executes a page-specific information acquisition step, an extraction step, and a required information display step. In the page-specific information acquisition step, the purchase assistance system acquires page-specific information that identifies a web page on which a product the user wishes to purchase is listed via an interface screen (user interface screen) displayed on the user terminal. In the extraction step, the purchase assistance system automatically extracts some required information from the information on the product listed on the web page from the HTML that constitutes the web page specified by the page-specific information. In the required information display step, the purchase assistance system displays at least some of the required information extracted in the extraction step on the interface screen of the user terminal.

[0014] According to the purchase assistance system exemplified in the present disclosure, when a user specifies a web page (e.g., a web page of an overseas e-commerce site), necessary information is automatically extracted from the product information listed on the specified web page in the HTML that constitutes the web page and displayed on the interface screen of the user terminal. Therefore, the user can easily and appropriately understand the automatically extracted and displayed necessary information and proceed with the product purchase procedure without having to directly browse the overseas website to understand the necessary information. This also reduces the possibility of defects occurring in the purchase procedure. Therefore, the purchase procedure for overseas products is appropriately assisted.

[0015] The method for displaying the interface screen on the user terminal can be selected as appropriate. For example, the purchase assistance system may display the interface screen on a web browser on the user terminal. Alternatively, a dedicated application for implementing the purchase assistance method exemplified in this disclosure may be installed on the user terminal. In this case, the purchase assistance system may use the application to display the interface screen on the user terminal.

[0016] The necessary information extracted from the web page in the extraction step can be selected as appropriate. For example, at least one of the following may be extracted as the necessary information: an image (e.g., a photo) of the product, a price, a description of the product, and available options (e.g., a color and size selected by the user). In this case, the user can more appropriately proceed with the purchase process by understanding the necessary information displayed on the interface screen.

[0017] In the extraction step, the purchase assistance system may automatically extract the necessary information by analyzing the HTML structure of the web page specified by the page specification information (hereinafter referred to as the "specific web page"). Hereinafter, the process of analyzing the HTML structure and extracting the necessary information will be referred to as the "structure analysis process."

[0018] In this case, it becomes easier to appropriately extract the necessary information in a short time compared to when a mathematical model trained by a machine learning algorithm is used.

[0019] In the extraction step, the purchase assistance system may input the HTML of the web page identified by the page identification information into a mathematical model, and extract (acquire) information output by the mathematical model as the required information. The mathematical model may be trained by a machine learning algorithm so that, when the HTML of the web page is input, the mathematical model extracts and outputs the required information formed on the web page. Hereinafter, the process of extracting the required information using a machine learning algorithm is referred to as "AI-based processing."

[0020] The HTML structures that make up different websites tend to vary greatly, but by using mathematical models trained by machine learning algorithms, it becomes easier to appropriately extract necessary information from various web pages with different HTML structures.

[0021] The mathematical model may be trained using a set of input training data and output training data (training data set). For example, the mathematical model for extracting necessary information may be trained according to a machine learning algorithm using a web page as input training data and correct necessary information actually extracted from the web page that is the input training data as output training data. In this case, the output training data may be correct necessary information constructed by an administrator of the purchase assistance system, or correct necessary information confirmed or corrected by a user.

[0022] Furthermore, the mathematical model for extracting the necessary information may be continuously trained after the system is put into operation, using the web pages identified by the page identification information when the user uses the system and the correct necessary information on the identified web pages as a training data set, which further improves the accuracy of extracting the necessary information using the mathematical model.

[0023] The purchase assistance system may be capable of performing both the structural analysis process and the AI-utilizing process described above in the extraction step. The purchase assistance system may selectively perform the structural analysis process and the AI-utilizing process depending on the web page identified by the page identification information.

[0024] The structural analysis process makes it easier to appropriately extract necessary information in a short time. On the other hand, since the algorithm for executing the structural analysis process needs to be constructed according to the type of website (i.e., HTML structure), the structural analysis process may not be executed appropriately for websites that are not compatible with the algorithm for the structural analysis process. Furthermore, the AI-based process makes it easier to appropriately extract necessary information from various web pages with different HTML structures. On the other hand, the processing time of the AI-based process tends to be longer than the processing time of the structural analysis process. Therefore, the purchase assistance system can automatically extract necessary information more appropriately from various web pages by selectively executing the structural analysis process and the AI-based process according to the web page.

[0025] The method of selectively executing the structural analysis process and the AI-utilizing process depending on the web page can be selected as appropriate. For example, websites capable of executing the structural analysis process (i.e., websites on which the algorithm for the structural analysis process has already been constructed) may be stored in advance in a storage device. The purchase assistance system may execute the structural analysis process if the web page identified by the page identification information is a website capable of executing the structural analysis process, or may execute the AI-utilizing process if the web page is not a website capable of executing the structural analysis process. The purchase assistance system may also determine whether the web page identified by the page identification information is a website capable of executing the structural analysis process based on the HTML structure, etc., of the web page.

[0026] If the web page identified by the page identification information is displayed in a language of a foreign country different from the language of the user's country of residence, the purchase assistance system may automatically translate at least a portion of the necessary information extracted from the HTML (e.g., language excluding numbers and symbols) into the language of the user's country of residence and display it on the interface screen of the user terminal. In this case, even if the user is not familiar with the language displayed on the web page, the necessary information displayed on the interface screen can be easily understood in the language of the user's country of residence. This provides more appropriate assistance with the purchase process for overseas products.

[0027] In the page identification information acquisition step, the purchase assistance system may acquire, as the page identification information, the URL of a web page entered by the user on an interface screen displayed on the user terminal. In this case, the user can easily and appropriately grasp the information necessary to purchase the product simply by entering the URL of the web page on which the product he or she wishes to purchase is posted on the interface screen.

[0028] The specific method for acquiring page identification information may be changed. For example, the purchase assistance system may display one or more product selection sections on an interface screen. Each product selection section may be associated with the URL of a web page on which the product is posted. The purchase assistance system may automatically display product selection sections for products recommended to the user on the interface screen. The purchase assistance system may also search for products in response to a user's search instruction and display the product selection section of the searched product on the interface screen. When a user specifies one of the product selection sections on the interface screen, the purchase assistance system may acquire the URL of the web page associated with the specified product selection section as page identification information. In this case, simply by the user specifying the product selection section of the product they wish to purchase, the information required to purchase the product is automatically extracted from the HTML of the web page and displayed. This provides more appropriate assistance with the purchase process for overseas products.

[0029] Furthermore, the user terminal may launch a dedicated application when a sharing operation to share information on a web page is input to the dedicated application while a web page listing a product is displayed on an interface screen (e.g., a web browser). The purchase assistance system may identify the web page displayed on the interface screen when the sharing operation was performed on the user terminal, extract necessary information from the HTML of the identified web page, and display it in the dedicated application. In this case, the information sharing function of the user terminal is used to smoothly purchase the product.

[0030] The purchase assistance system may further execute a physical information acquisition step, a shipping fee prediction step, and a predicted shipping fee display step. In the physical information acquisition step, the purchase assistance system acquires physical information of the product listed on the website by inputting at least some of the necessary information extracted in the extraction step into a mathematical model. The mathematical model is trained by a machine learning algorithm so as to output physical information of the product when the necessary product information is input. In the shipping fee prediction step, the purchase assistance system predicts shipping fees for the product listed on the website based on the physical information acquired in the physical information acquisition step. In the predicted shipping fee display step, the purchase assistance system displays the shipping fee predicted in the shipping fee prediction step on an interface screen.

[0031] In this case, even if the physical information of the product is not sufficiently posted on the website, the physical information of the product is appropriately acquired. Based on the acquired physical information of the product, the shipping cost of the product is appropriately predicted and displayed on the interface screen of the user terminal. Therefore, regardless of the product information posted on the website, the shipping cost is predicted with high accuracy and can be understood by the user.

[0032] The details of the physical information acquired in the physical information acquisition step can be selected as appropriate. For example, the weight and size of the product listed on the website may be acquired as physical information. In this case, even if the shipping fee is determined based on both the weight and size of the product, it becomes easier to predict the shipping fee with high accuracy.

[0033] The mathematical model for acquiring physical information about a product may be trained according to a machine learning algorithm, for example, using necessary product information as input training data and correct physical information about the product from which the input training data has been extracted as output training data. In this case, the output training data may be correct physical information input by, for example, an administrator of the purchase assistance system.

[0034] Furthermore, the mathematical model for acquiring physical information about products may be continuously trained after the system goes into operation using the necessary product information actually used by users when using the system and the correct physical information about the products as a training data set, which facilitates further improvement in the accuracy of the physical information acquired by the mathematical model.

[0035] In the physical information acquisition step, the purchase assistance system may acquire physical information about the product by inputting the multiple types of necessary information extracted in the extraction step into a mathematical model. In this case, the accuracy of the acquired physical information is more likely to be improved than when a single piece of necessary information about the product is input into the mathematical model.

[0036] The mathematical model for acquiring physical information about a product may be trained using necessary information about multiple types of products as input training data, which further improves the accuracy of the physical information about the product output by the mathematical model.

[0037] The plurality of types of necessary information to be input to the mathematical model may be selected as appropriate. For example, at least two or more of the product title, category, image, detailed information, etc. may be input to the mathematical model.

[0038] In the shipping fee prediction step, the purchase assistance system may apply an algorithm (e.g., a known 3D packing algorithm) that predicts an appropriate packing box from multiple types of packing boxes that are rectangular parallelepipeds to the physical information acquired in the physical information acquisition step. The purchase assistance system may predict a shipping fee for the product based on the packing box predicted by the algorithm.

[0039] In this case, the shipping cost is predicted after predicting the packaging box that will be used when actually delivering the product, which makes it easier to further improve the accuracy of the shipping cost prediction.

[0040] When the physical information acquired is the size of the product, the size information may include, for example, the maximum width, depth, and height of the product. In this case, an algorithm is applied to the maximum width, depth, and height of the product, making it easier to predict an appropriate packaging box with high accuracy.

[0041] When a user selects multiple candidate products for purchase, physical information for each of the selected products may be acquired. Based on the acquired physical information for each of the selected products, a shipping fee for shipping the multiple products together may be predicted.

[0042] In this case, the shipping cost when a user purchases multiple items at once is appropriately predicted based on the physical information of each of the multiple items. Details of the physical information and the method of constructing a mathematical model for acquiring the physical information are as described above. When acquiring the physical information of each item, multiple types of necessary information about each item may be input into the mathematical model, as described above.

[0043] The method for allowing the user to select multiple candidate products for purchase can be selected as appropriate. For example, when a user has added multiple products to their cart, the purchase assistance system may determine that the multiple products in the cart are candidate products for purchase by the user. Furthermore, when the user selects multiple products from the multiple products in the cart, the purchase assistance system may determine that the selected multiple products are candidate products for purchase by the user.

[0044] In the shipping fee prediction step, the purchase assistance system may apply an algorithm (e.g., a known 3D packing algorithm) to the physical information of each of the multiple products acquired in the physical information acquisition step, which predicts an appropriate packing box from multiple types of rectangular parallelepiped packing boxes. The purchase assistance system may execute the process of predicting an overall packing box that can accommodate all of the multiple arranged packing boxes multiple times while changing the arrangement of the multiple predicted packing boxes (note that a packing box prediction algorithm such as the above-mentioned 3D packing algorithm may also be used in the process of predicting the overall packing box). The purchase assistance system may predict the shipping fee for delivering the multiple products together based on the overall packing box with the smallest volume among the multiple predicted overall packing boxes.

[0045] In this case, when multiple products are packed together in a complete packaging box for delivery, the complete packaging box that will have the smallest volume is automatically predicted, and therefore the shipping cost for delivering multiple products together can be more appropriately predicted.

[0046] When a user selects multiple candidate products for purchase, the purchase assistance system may calculate the difference between the estimated shipping cost for shipping the multiple products together and the estimated shipping cost for shipping the multiple products separately, and display this difference on the interface screen. In this case, the user can easily and appropriately understand the amount of money that can be saved by shipping the multiple products together. This appropriately increases the user's motivation to purchase multiple products.

[0047] The purchase assistance system may further execute a tariff prediction step and a predicted tariff display step. In the tariff prediction step, the purchase assistance system predicts a tariff to be levied when importing the product into the user's country of residence based on product information posted on the web page identified by the page identification information. In the predicted tariff display step, the purchase assistance system displays the tariff predicted in the tariff prediction step on an interface screen.

[0048] In this case, the user can decide whether or not to purchase the product after taking into consideration the predicted tariffs that will be imposed when importing the product, thereby providing more appropriate assistance in the process of purchasing overseas products.

[0049] In the tariff prediction step, the purchasing assistance system may input product information listed on the web page identified by the page-specific information into a mathematical model to obtain a predicted result of the tariff to be levied on the product. The mathematical model is trained using a machine learning algorithm so that the mathematical model outputs a predicted result of the tariff when the product information is input. In this case, completion of various products can be predicted with higher accuracy.

[0050] The mathematical model for predicting the tariff to be levied on a product may be trained according to a machine learning algorithm, for example, using product information as input training data and the correct tariff to be levied on the product as output training data, where the output training data may be the correct tariff, etc., input by, for example, an administrator of the purchasing assistance system.

[0051] In the tariff prediction step, the purchase assistance system may obtain a prediction result of the tariff to be levied on the product by inputting multiple pieces of information about the product listed on the web page identified by the page identification information into the mathematical model. In this case, the accuracy of the prediction of the tariff to be obtained is more likely to be improved than when one piece of information about the product is input into the mathematical model.

[0052] The types of information input to the mathematical model for predicting the tariff to be levied on a product may also be selected as appropriate. For example, the information input to the mathematical model may include textual information describing the product and an image of the product.

[0053] In the tariff prediction step, the purchasing assistance system may use RAG to obtain text that indicates product characteristics from a collection of text related to the product posted on the web page identified by the page identification information. The purchasing assistance system may input the obtained text into a mathematical model to obtain a prediction result of the tariff to be levied on the product.

[0054] Retrieval-Augmented Generation (RAG) is a technique used in the field of natural language processing. Using RAG, specific text can be generated based on a large set of text. This makes it easier to further improve the accuracy of the tariff predictions that will be obtained.

[0055] In the tariff prediction step, the purchasing assistance system may obtain a predicted HS code for the product by inputting product information listed on the web page identified by the page identification information. The mathematical model for HS code prediction is trained by a machine learning algorithm so as to output a predicted HS code when the product information is input. The purchasing assistance system may predict a tariff based on the predicted HS code.

[0056] HS codes are codes used to classify international trade products. Customs duty rates are predetermined for each product's HS code. Therefore, predicting customs duties after predicting the HS code appropriately improves the accuracy of customs duty predictions.

[0057] The method for constructing a mathematical model for predicting an HS code can be selected as appropriate. For example, the mathematical model may be trained according to a machine learning algorithm using product information as input training data and a correct HS code designated by an operator who understands the product information as output training data. Furthermore, if the prediction result output by the mathematical model for predicting an HS code differs from the correct HS code, the mathematical model may be retrained based on the incorrect prediction result. In this case, the accuracy of HS code prediction is further improved.

[0058] The purchase assistance system may identify rules that may be problematic in customs procedures based on the predicted HS code and notify the user of the identified rules. Depending on the classification of a product by HS code, customs procedures may include rules that limit export volume or prohibit export. Therefore, by notifying the user of customs procedures based on the predicted HS code, the user can more appropriately decide whether or not to purchase the product.

[0059] Furthermore, the purchase assistance system may obtain a predicted result of a product attribute (e.g., at least one of attributes such as second-hand, refrigerated, frozen, large, short-life, or detained at customs) by inputting product information listed on the web page identified by the page identification information into a mathematical model. The mathematical model for predicting product attributes is trained using a machine learning algorithm so as to output a predicted result of the product attribute when product information is input. The purchase assistance system may notify the user of the obtained product attribute information. In this case, the user can appropriately determine whether or not to purchase the product after appropriately understanding the product attributes. Note that the mathematical model for predicting product attributes may be trained according to a machine learning algorithm, for example, using product information as input training data and correct attributes specified by an operator who has understood the product information as output training data.

[0060] The purchase assistance system may calculate the total cost required for the user to purchase the product using the shipping fee predicted in the shipping fee prediction step and the customs duty predicted in the customs duty prediction step, and may complete payment of the total cost to the user in advance before the product is delivered. In this case, the user can grasp the total cost required to purchase the product at an early stage before receiving the product, making it easier for them to decide whether or not to purchase the product. Note that the purchase assistance system may not charge or refund the difference to the user even if the predicted total cost differs from the actual total cost. In this case, the number of payments and receipts between the user and the system is unlikely to increase, simplifying the product purchase process.

[0061] The purchase assistance system may further execute a link display step of displaying on the interface screen a link designated by the user to display the web page identified by the page identification information. In this case, by designating the link displayed on the interface screen, the user can easily compare the necessary information displayed on the interface screen with the original web page from which the necessary information was extracted. Thus, the user can easily confirm the accuracy of the extracted necessary information before deciding whether to purchase the product.

[0062] In the necessary information display step, the purchase assistance system may convert the price of the product extracted as necessary information into the currency of the user's country of residence based on the currency of the user's country of residence and the exchange rate information of the currency listed on the web page, and display the converted price on the interface screen. In this case, the user can determine whether or not to purchase the product after understanding the price of the product in the currency of the user's country of residence.

[0063] In the necessary information display step, the purchase assistance system may display the price of the product on the interface screen in both the currency of the user's country of residence and the currency listed on the web page. In this case, the user can more easily determine the value of the product by knowing the price of the product in both currencies.

[0064] The purchase assistance system may further execute a dedicated icon display step of displaying, on the interface screen, a dedicated icon that the user can operate to display product information, along with a web page listing the product the user desires to purchase. When the dedicated icon is operated by the user, the purchase assistance system may identify the web page that was displayed on the interface screen when the dedicated icon was operated. The purchase assistance system may automatically extract necessary information from HTML that constitutes the identified web page. The purchase assistance system may display an extended browser on the interface screen and display product information including the extracted necessary information on the displayed extended browser. In this case, the user can view product information including the extracted necessary product information on the extended browser while the web page listing the product the user desires to purchase is displayed on the interface screen. This allows the user to more appropriately determine whether or not to purchase the product.

[0065] The product information to be displayed on the extended browser can be selected as appropriate. For example, the purchase assistance system may display at least one of the product price, estimated shipping fee, estimated customs duty, etc. on the extended browser. The purchase assistance system may also display at least one of an "Add to Cart" button and a "View on Dedicated Screen" button on the extended browser along with product information. When the "Add to Cart" button is operated, the purchase assistance system may add the product whose information is displayed on the extended browser to the cart. When the "View on Dedicated Screen" button is operated, the purchase assistance system may also display a new dedicated interface screen for implementing the purchase assistance method exemplified in the present disclosure. The purchase assistance system may also display price trend information indicating the trend in product prices on the extended browser.

[0066] The purchase assistance system may further execute an auxiliary browser display step of, when a web page listing a product is displayed on the interface screen, displaying an auxiliary browser on the interface screen together with the web page, the auxiliary browser displaying information to assist the user in purchasing the product. When the auxiliary browser is displayed, the purchase assistance system may identify the web page that was displayed on the interface screen when the auxiliary browser was displayed. The purchase assistance system may automatically extract necessary information from HTML constituting the identified web page. The purchase assistance system may display product information including the necessary information on the displayed auxiliary browser.

[0067] In this case, the user can view product information, including the necessary information for the extracted product, on the automatically displayed auxiliary browser while the web page on which the product is posted is displayed on the interface screen, thereby enabling the user to more appropriately decide whether or not to purchase the product.

[0068] A specific method for detecting that a web page listing a product (hereinafter referred to as a "product listing page") has been displayed on an interface screen can be selected as appropriate. For example, a mathematical model that outputs the probability that a web page displayed on an interface screen is a product listing page may be constructed in advance according to a machine learning algorithm. The purchase assistance system may obtain the probability that a web page is a product listing page by inputting information about the displayed web page (e.g., HTML, etc.) into the mathematical model. The purchase assistance system may detect that the displayed web page is a product listing page based on the obtained probability.

[0069] The product information to be displayed on the auxiliary browser can be selected as appropriate. For example, the purchase assistance system may display at least one of the product price, estimated shipping costs, estimated customs duties, etc. on the auxiliary browser. As described above, the purchase assistance system may also display at least one of an "Add to Cart" button and a "View on Dedicated Screen" button on the auxiliary browser along with the product information. The purchase assistance system may also display price trend information indicating the price trend of the product on the auxiliary browser.

[0070] 2 is a block diagram showing the schematic configuration of a purchase assistance system 1 and a user terminal 20. FIG. 2 is a flowchart of purchase assistance processing executed by the purchase assistance system 1 of this embodiment. FIG. 3 is a diagram showing an example of an interface screen (home screen 40) displayed on the display unit 26 of the user terminal 20. FIG. 4 is a flowchart of product-specific information display processing executed during purchase assistance processing. FIG. 5 is a diagram showing an example of an interface screen (product-specific screen 50) displayed on the display unit 26 of the user terminal 20. FIG. 6 is a flowchart of bundled delivery shipping fee prediction processing executed by the purchase assistance system 1 of the first modified example. FIG. 7 is a diagram showing an example of a state in which a dedicated icon 70 and an extended browser 71 are displayed on a general-purpose interface screen displayed on the display unit 26 of the user terminal 20. FIG. 8 is a diagram showing an example of a state in which an auxiliary browser 73 is displayed on a general-purpose interface screen displayed on the display unit 26 of the user terminal 20. FIG. 9 is a diagram showing a state in which an information sharing instruction button 74 and an information sharing icon field 75 are displayed on the general-purpose interface screen displayed on the display unit 26 of the user terminal 20.

[0071] A typical embodiment of the present disclosure will be described below with reference to the drawings. First, with reference to FIG. 1 , an example of the configuration of a purchase assistance system (AI-based cross-border e-commerce system) 1 and a user terminal 20 according to the present embodiment will be outlined.

[0072] The purchase assistance system 1 provides a service to assist users in purchasing overseas products. As an example, a server, which is a type of information processing device, is used as the purchase assistance system 1 of this embodiment. More specifically, in this embodiment, a server of a manufacturer that provides cloud services (a so-called cloud server) is used as the purchase assistance system 1. However, a server other than a cloud server may also be used as the purchase assistance system 1. An information processing device other than a server (for example, a personal computer (hereinafter referred to as a "PC")) may also be used as the purchase assistance system 1. Furthermore, the device that constitutes the purchase assistance system 1 may be one or more. For example, an information processing device and a database may work together to function as the purchase assistance system 1, or multiple information processing devices may work together to function as the purchase assistance system 1.

[0073] The purchase assistance system 1 includes a control unit 11 that controls various processes and a communication I / F 14. The control unit 11 includes a CPU 12, which is a controller responsible for control, and a storage device 13 that can store programs, data, and the like. The storage device 13 stores a purchase assistance program for executing the purchase assistance process (see FIG. 2 ), which will be described later. The storage device of this embodiment also stores data necessary for displaying a dedicated interface screen (described in detail below) on the user terminal 20. The communication I / F 14 connects the purchase assistance system 1 to external devices (e.g., multiple user terminals 20 and multiple web servers 30) via a network 5 (e.g., the Internet, etc.).

[0074] The user terminal 20 (20A, 20B) is used by a user who utilizes the services provided by the purchase assistance system 1. The user terminal 20 illustrated in this embodiment is a personal computer. However, a mobile terminal such as a smartphone or a tablet terminal may also be used as the user terminal 20. The user terminal 20 includes a control unit 21 (21A, 21B) that performs various control processes and a communication I / F 24 (24A, 24B). The control unit 21 includes a CPU 22 (22A, 22B) that is a controller responsible for control, and a storage device 23 (23A, 23B) that can store programs, data, and the like. The communication I / F 24 also connects the user terminal 20 to external devices (e.g., the purchase assistance system 1 and multiple web servers 30) via the network 5.

[0075] The user terminal 20 is connected to an operation unit 25 and a display unit 26. The operation unit 25 is operated by the user to input various instructions to the user terminal 20. The operation unit 25 may be, for example, at least one of a keyboard, a mouse, a touch panel, etc. Note that a microphone or the like for inputting various instructions may be used together with or instead of the operation unit 25. The display unit 26 displays various images. The display unit 26 may be any of various devices for displaying images (for example, at least one of a monitor, a projector, a head-mounted display, etc.). It goes without saying that the operation unit 25 and the display unit 26 externally connected to the user terminal 20 may be replaced by an operation unit and a display unit included in the user terminal 20.

[0076] The web server 30 (30A, 30B) stores data necessary for displaying a web page on an information processing device, such as HTML and image data that constitute the web page. When a web page is accessed, the web server 30 provides the data for constructing the accessed web page to the information processing device. As a result, the information processing device (e.g., user terminal 20) can display the accessed website on a display unit (e.g., display unit 26).

[0077] An example of the purchase assistance process executed by the purchase assistance system 1 of this embodiment will be described with reference to Figures 2 to 5. In the purchase assistance process, necessary information is automatically extracted from a web page on which a product is posted (e.g., a web page written in a language different from the language of the user's country of residence) and displayed on an interface screen on the display unit 26 of the user terminal 20. The interface screen of this embodiment is configured according to a predetermined format. Therefore, the user can easily and appropriately understand the necessary information that is automatically extracted and displayed on the interface screen before proceeding with the product purchase, without having to directly browse an overseas website to understand the necessary information themselves. This also reduces the possibility of defects occurring in the sales procedure.

[0078] The purchase assistance process of this embodiment is executed by the CPU 12 of the purchase assistance system 1. When a user operates the operation unit 25 of the user terminal 20 and inputs an instruction to display the dedicated interface screen exemplified in this embodiment into the user terminal 20, the CPU 12 executes the purchase assistance process shown in FIG.

[0079] First, the CPU 12 displays a dedicated interface screen (home screen 40 in S1) for providing the user with the purchase assistance service on the user terminal 20 (in this embodiment, the display unit 26, the display of which is controlled by the user terminal 20) (S1). In this embodiment, an example is shown in which the purchase assistance system 1 displays a dedicated interface screen on the web browser of the user terminal 20. However, it is also possible to change the method for displaying the interface screen on the user terminal 20. For example, a dedicated application for providing the user with the purchase assistance service may be installed on the user terminal 20. In this case, the purchase assistance system 1 may use the application to display the interface screen on the user terminal 20.

[0080] An example of the display mode of a home screen 40, which is one of the interface screens, will be described with reference to Fig. 3. The home screen 40 illustrated in Fig. 3 includes a URL input section 41, a product selection section 42, and a login button 43. The login button 43 is operated when the user logs in to the purchase assistance service.

[0081] The URL of a web page listing the product that the user wishes to purchase is input to the URL input unit 41. When the URL is input to the URL input unit 41, the CPU 12 automatically extracts necessary information to be presented to the user from the web page specified by the input URL, and displays the information on an interface screen (in this embodiment, a product-specific screen 50, which will be described later).

[0082] The product selection section 42 is displayed to present product information to the user as a purchase candidate. Each product selection section 42 includes product overview information (e.g., a product image, a brief description, and a price). In the example shown in FIG. 3 , the purchase assistance system 1 automatically displays a product selection section 42 for each of multiple products recommended to the user as a "PICK UP" on the interface screen 40. However, the display method of the product selection section 42 can be changed. For example, when a user inputs a search instruction via the user terminal 20, the purchase assistance system 1 can search for products in accordance with the input search instruction and display the product selection section 42 of the searched product on the interface screen. Each product selection section 42 is associated with the URL of a web page on which the product is listed. When a user selects one of the product selection sections 42 on the interface screen, the purchase assistance system 1 obtains the URL of the web page associated with the selected product selection section 42. The purchase assistance system 1 automatically extracts the necessary information to be presented to the user from the web page specified by the acquired URL and displays it on an interface screen (in this embodiment, a product-specific screen 50 described later).

[0083] Returning to the description of FIG. 2 , the CPU 12 determines whether a login instruction has been input by the user (S2). If a login instruction has been input (S2: YES), the CPU 12 executes login processing in accordance with the input instruction (S3). If a login instruction has not been input (S2: NO), the CPU 12 determines whether a URL has been input into the URL input unit 41 (S5). If a URL has not been input (S5: NO), the CPU 12 determines whether any of the product selection units 42 has been designated by the user (S6). If no product selection unit 42 has been designated (S6: NO), the CPU 12 determines whether a logout instruction has been input by the user (S8). If a logout instruction has not been input (S8: NO), the process returns to S2, and steps S2 to S8 are repeated. Note that if a logout instruction has been input by the user (S8: YES), the CPU 12 executes logout processing (S9) and terminates the purchase assistance process.

[0084] When a URL is input to the URL input section 41 (S5: YES), the CPU 12 executes a product-specific information display process (S7) using the input URL as page identification information that identifies the web page on which the product the user wishes to purchase is posted. Also, when any product selection section 42 is designated (S6: YES), the CPU 12 executes a product-specific information display process (S7) using the URL of the web page associated with the designated product selection section 42 as page identification information. In other words, the processes of S5 and S6 can also be expressed as processes for obtaining page identification information.

[0085] The product-specific information display process will be described with reference to FIG. 4. First, the CPU 12 acquires HTML that constitutes the web page identified by the page identification information (in this embodiment, a URL) (S11). From the HTML acquired in S11, the CPU 12 automatically extracts some necessary information from the product information listed on the web page (S13-S15). The type of necessary information to be automatically extracted can be selected as appropriate. As an example, in this embodiment, at least one of the following is extracted as necessary information: a product image (e.g., a photo), price, product description, and available options (e.g., a color and size selected by the user). Therefore, by understanding the automatically extracted necessary information, the user can more appropriately proceed with the purchase process.

[0086] The purchase assistance system 1 of this embodiment can execute a structure analysis process (S14) and an AI utilization process (S15) as processes for automatically extracting necessary information from a website.

[0087] In the structural analysis process (S14), the CPU 12 automatically extracts the necessary information by analyzing the HTML structure of the web page (hereinafter sometimes referred to as the "specific web page") identified by the page identification information (URL in this embodiment).

[0088] According to the structural analysis process (S14), it is possible to automatically extract necessary information with higher accuracy while reducing the amount of processing.

[0089] In the AI ​​utilization process (S15), the CPU 12 inputs the HTML of the web page identified by the page identification information (URL in this embodiment) into a mathematical model for necessary information extraction, and extracts (acquires) the information output by the mathematical model as necessary information. The mathematical model for necessary information extraction is trained using a machine learning algorithm so that, when the HTML of the web page is input, it extracts and outputs the necessary information formed on the web page. As an example, in this embodiment, the mathematical model for necessary information extraction is trained using a set of input training data and output training data (training data set). In detail, the mathematical model for necessary information extraction is trained according to the machine learning algorithm using the web page as input training data and correct necessary information actually extracted from the web page as input training data. The output training data may be, for example, correct necessary information constructed by an administrator of the purchase assistance system 1, or correct necessary information confirmed or corrected by a user.

[0090] As mentioned above, different websites tend to have significantly different HTML structures. However, by using a mathematical model for extracting necessary information trained by a machine learning algorithm, it becomes easier to appropriately extract necessary information from various web pages with different HTML structures.

[0091] The purchase assistance system 1 of this embodiment selectively executes a structural analysis process (S14) and an AI-based process (S15) depending on the web page identified by the page identification information. As described above, the structural analysis process (S14) facilitates appropriate extraction of necessary information in a short time. However, since the algorithm for executing the structural analysis process must be constructed depending on the type of website (i.e., the HTML structure), the structural analysis process may not be executed appropriately for websites that are not compatible with the structural analysis process algorithm. Furthermore, the AI-based process (S15) facilitates appropriate extraction of necessary information from various web pages with different HTML structures. However, the processing time of the AI-based process tends to be longer than the processing time of the structural analysis process. Therefore, the purchase assistance system 1 selectively executes the structural analysis process and the AI-based process depending on the web page, thereby automatically extracting necessary information more appropriately from various web pages.

[0092] As an example, in the purchase assistance system 1 of this embodiment, websites capable of executing the structural analysis process (S14) (i.e., websites for which an algorithm for the structural analysis process has already been constructed) are pre-stored in the storage device 13. If the web page identified by the page identification information is a website capable of executing the structural analysis process (S13: YES), the CPU 12 executes the structural analysis process. On the other hand, if the web page identified by the page identification information is not a website capable of executing the structural analysis process (S13: NO), the CPU 12 executes the AI ​​utilization process (S15).

[0093] When the AI-based process (S15) is executed, the CPU 12 determines whether the accuracy of the necessary information extracted by the AI-based process meets the standard (whether the extraction accuracy is acceptable) (S16). The determination in S16 may be made, for example, based on the determination result of an administrator of the purchase assistance system 1 or a user. If the extraction accuracy meets the standard (S16: YES), the CPU 12 adopts the information automatically extracted in S15 as the necessary information, and the process proceeds to S20. On the other hand, if the extraction accuracy does not meet the standard (S16: NO), the CPU 12 acquires the correct necessary information to be extracted from the identified web page. The CPU 12 trains a mathematical model for necessary information extraction using the specific web page identified by the page-identifying information and the correct necessary information on the specific web page as a training data set (S17). This facilitates further improvement in the accuracy of the output from the mathematical model for necessary information extraction. The correct necessary information acquired in S17 may be input by, for example, an administrator of the purchase assistance system 1 or by a user. If the extraction accuracy does not meet the standard (S16: NO), the correct information acquired in S17 is set as the necessary information instead of the information automatically extracted in S15, and the process proceeds to S20.

[0094] When the necessary information is extracted (acquired), the CPU 12 displays at least a part of the extracted necessary information on an interface screen (in this embodiment, the product-specific screen 50 exemplified in FIG. 5) (S20).

[0095] The product-specific screen 50 illustrated in FIG. 5 includes, in addition to the URL input section 41 described above, an image display section 51, a country of residence language description display section 52, a local language description display section 53, an option selection section 54, a product price display section 55, a predicted shipping fee display section 56, a predicted customs duty display section 57, a subtotal display section 58, a purchase button 59, and a link 60 for displaying the original page.

[0096] The image display section 51 displays images of products extracted as necessary information from the web page.

[0097] The country-of-residence language explanation display unit 52 displays product descriptions extracted as necessary information from the web page in the language of the user's country of residence of the user of the user terminal 20. In other words, in the process of S20 (see FIG. 4), if the web page specified by the user is written in a foreign language different from the language of the user's country of residence, the purchase assistance system 1 automatically translates at least a portion of the necessary information extracted from the specified web page (in the example shown in FIG. 5, the product description and product options described below) into the language of the user's country of residence and displays it on the interface screen (product-specific screen 50). Therefore, even if the user is not familiar with the language displayed on the web page, the user can easily understand the necessary information displayed on the interface screen in the language of the user's country of residence.

[0098] The local language description display section 53 displays the product description extracted as necessary information from the web page in the language used on the web page (i.e., the local language). Therefore, the user can more accurately understand the contents of the product by comparing the product description in the language of the country where the user lives with the product description in the local language.

[0099] The option selection unit 54 displays information about product options extracted as necessary information from the web page. In the process of S20 (see FIG. 4), the purchase assistance system 1 displays information about multiple product options on the option selection unit 54 in a state where the user can select one. When the user selects one of the multiple options, the purchase assistance system 1 executes other processes (e.g., the purchase process (S33) described below) for the product of the selected option as the product that the user wishes to purchase. As an example, the option selection unit 54 of this embodiment displays multiple options in a pull-down format, allowing the user to select an option. However, it goes without saying that the way the options are displayed by the option selection unit 54 can be changed.

[0100] The product price display unit 55 displays product price information extracted as necessary information from the web page. In the process of S20, the purchase assistance system 1 converts the product price extracted as necessary information into the currency of the user's country of residence based on the currency of the user's country of residence using the user terminal 20 and the exchange rate information for the currency listed on the web page, and displays the converted price on the product price display unit 55. Therefore, the user can determine whether or not to purchase the product after understanding the product price in the currency of their country of residence. The purchase assistance system 1 may also display the exchange rate along with the product price. Furthermore, the purchase assistance system 1 of this embodiment displays the product price on the product price display unit 55 in both the currency of the user's country of residence and the currency listed on the web page. Therefore, by understanding the product price in both currencies, the user can more accurately determine the value of the product.

[0101] The predicted shipping cost display section 56 displays the predicted shipping cost (predicted shipping cost) for delivering the product listed on the web page to the user. As will be described in detail later, the purchase assistance system 1 can predict shipping costs based on necessary information extracted from the web page.

[0102] The predicted customs duty display unit 57 displays the predicted customs duties (predicted customs duties) that will be levied when importing products listed on the web page into the user's country of residence. As will be described in detail later, the purchase assistance system 1 can predict customs duties based on product information listed on the web page.

[0103] The subtotal display section 58 displays the total amount of the product price, estimated shipping fee, and estimated customs duty. The purchase button 59 is operated by the user when carrying out the purchase procedure for the product displayed on the product-specific screen 50. The original page display link 60 is a link operated by the user to display the web page (original page) on which the product is posted.

[0104] Returning to the description of FIG. 4 , the CPU 12 executes a process for predicting and displaying shipping costs for delivering products listed on a webpage to a user (S21-S23). First, the CPU 12 inputs the necessary product information acquired in S11-S17 into a mathematical model for acquiring physical product information, thereby acquiring information output by the mathematical model as physical product information (S21). The mathematical model for acquiring physical information is trained using a machine learning algorithm so that it outputs physical product information in response to the necessary product information. As an example, in this embodiment, the mathematical model for acquiring physical information is trained according to the machine learning algorithm using the necessary product information as input training data and the correct physical product information extracted from the input training data as output training data. The output training data may be, for example, correct physical information input by an administrator of the purchase assistance system. Furthermore, the mathematical model for acquiring physical information is continuously trained even after the service begins, using the necessary product information actually used by users when using the purchase assistance service and the correct physical product information as a training data set. This facilitates further improvement in the accuracy of the physical information acquired by the mathematical model.

[0105] In detail, in the process of S21, the CPU 12 inputs the multiple types of necessary information acquired in S11 to S17 into a mathematical model for acquiring physical information about the product, thereby acquiring physical information about the product listed on the website. Therefore, the accuracy of the acquired physical information is more likely to be improved than when only one type of necessary information about the product is input into the mathematical model. The mathematical model for acquiring physical information is trained using multiple types of necessary information about the product as input training data. As an example, in this embodiment, the product's title, category, image, and detailed information are input into the mathematical model for acquiring physical information, thereby acquiring the product's physical information with high accuracy.

[0106] In this embodiment, at least the weight and size of the product are acquired as physical information, so that even if the shipping fee is determined based on both the weight and size of the product, it becomes easier to predict the shipping fee with high accuracy.

[0107] Next, the CPU 12 predicts the shipping cost of the product based on the physical information of the product acquired in the process of S21 (S22). The CPU 12 displays the shipping cost predicted in S22 on the interface screen (in this embodiment, the predicted shipping cost display section 56 of the product-specific screen 50) (S23). Therefore, even if the physical information of the product is not sufficiently listed on the website, the shipping cost can be predicted with high accuracy and presented to the user.

[0108] Specifically, in the process of S22, the CPU 12 applies an algorithm (in this embodiment, a known 3D packing algorithm) to the physical information of the product acquired in S21 to predict an appropriate packaging box from among multiple types of rectangular parallelepiped packaging boxes. The CPU 12 predicts the shipping cost of the product based on the packaging box predicted by the algorithm. Therefore, the shipping cost is predicted after predicting the packaging box that will actually be used to deliver the product, which makes it easier to further improve the accuracy of the shipping cost prediction. Note that the product size information acquired as necessary information in S11 to S17 includes information on the maximum width, maximum depth, and maximum height of the product. Therefore, applying the algorithm to the maximum width, maximum depth, and maximum height of the product makes it easier to predict an appropriate packaging box with high accuracy.

[0109] The CPU 12 predicts the customs duty based on the product information posted on the web page (S25). The CPU 12 displays the customs duty predicted in S25 on the interface screen (in this embodiment, the predicted customs duty display section 57 of the product-specific screen 50) (S26). Therefore, the user can decide whether or not to purchase the product after taking into consideration the predicted customs duty that will be levied when importing the product.

[0110] In the process of S25, the CPU 12 inputs product information posted on the web page into a mathematical model for predicting customs duties, and obtains the amount output by the mathematical model as a predicted customs duty. The mathematical model for predicting customs duties is trained using a machine learning algorithm so that it outputs a predicted customs duty result in response to input product information. This makes it easier to predict the completion of various products with higher accuracy. As an example, the mathematical model for predicting customs duties in this embodiment is trained according to a machine learning algorithm using product information as input training data and the correct customs duties to be levied on the product as output training data. The output training data may be, for example, the correct customs duties input by an administrator of the purchase assistance system 1.

[0111] Specifically, in the process of S25, the CPU 12 inputs multiple pieces of product information listed on the web page into a mathematical model for tariff prediction to obtain a predicted result of the tariff to be levied on the product. Therefore, compared to when a single piece of product information is input into the mathematical model, the accuracy of the obtained tariff prediction is more likely to be improved. As an example, in this embodiment, the information input into the mathematical model for tariff prediction includes text information describing the product and an image of the product.

[0112] Furthermore, in the process of S25, the CPU 12 uses RAG to acquire text indicating product characteristics from a collection of product-related text posted on the web page. RAG (Retrieval-Augmented Generation) is a technology used in the field of natural language processing. By using RAG, specific text can be generated based on a large collection of text. The CPU 12 inputs the acquired text into a mathematical model for tariff prediction to acquire a predicted result of the tariff to be levied on the product. As a result, tariffs can be more accurately predicted based on the product's characteristics.

[0113] The CPU 12 displays, on the interface screen (the product-specific screen 50 in this embodiment), a source page display link 60 designated by the user to display the web page identified by the page identification information (a URL in this embodiment). Therefore, by designating a link displayed on the product-specific screen 50, the user can easily compare the necessary information displayed on the product-specific screen 50 with the original web page from which the necessary information was extracted. Therefore, the user can easily confirm the accuracy of the extracted necessary information before deciding whether or not to purchase the product.

[0114] The CPU 12 executes various processes while displaying the interface screen (product-specific screen 50) illustrated in FIG. 5 on the display unit 26 of the user terminal 20 (S30). For example, when the user operates the quantity change unit on the product-specific screen 50, the CPU 12 changes the quantity of the product in accordance with the operation instruction. The CPU 12 recalculates the amounts to be displayed in the product price display unit 55, the estimated shipping fee display unit 56, the estimated customs duty display unit 57, and the subtotal display unit 58 in accordance with the changed quantity, and displays the recalculated amounts. Furthermore, when the user operates the "Add to Cart" button on the product-specific screen 50, the CPU 12 adds the product information displayed on the product-specific screen 50 to the user's cart. When the user operates the "Favorites" button on the product-specific screen 50, the CPU 12 adds the product information displayed on the product-specific screen 50 to the user's favorites. When the user operates the "Share" button on the product-specific screen 50, the CPU 12 executes a process to share the product information displayed on the product-specific screen 50 with other users designated by the user.

[0115] The CPU 12 determines whether the user has operated the purchase button 59 on the product-specific screen 50 (S31). If the purchase button 59 has not been operated (S31: NO), the CPU 12 determines whether the user has input an instruction to end the display of the product-specific screen 50 (S32). If an end instruction has not been input (S32: NO), the processes of S30 to S32 are repeated and the system enters a standby state. If the purchase button 59 is operated (S31: YES), the CPU 12 executes product purchase processing in accordance with the instruction input by the user (S33), and the process returns to the purchase assistance process (see FIG. 2). If an end instruction has been input, the process returns directly to the purchase assistance process (see FIG. 2).

[0116] <First Modification> A first modification of the above embodiment will be described with reference to Figure 6. The purchase assistance system 1 of the first modification can predict the shipping cost when multiple items are delivered together by executing the consolidated delivery shipping fee prediction process shown in Figure 6. For example, when a user has added multiple items to their cart, the purchase assistance system 1 may start the consolidated delivery shipping fee prediction process for the multiple items added to the cart as candidate items for purchase by the user. Furthermore, when the user selects multiple items from the multiple items added to the cart, the purchase assistance system 1 may start the consolidated delivery shipping fee prediction process for the selected multiple items as candidate items for purchase by the user.

[0117] As shown in FIG. 6 , the CPU 12 of the purchase assistance system 1 acquires physical information for each of the multiple products based on the necessary information for each of the multiple products selected as candidates for purchase by the user (S41). The process of acquiring the necessary information for each product can be similar to S11 to S17 (see FIG. 4 ) in the above embodiment. The process of acquiring the physical information for each product can be similar to S21 (see FIG. 4 ) in the above embodiment. Next, the CPU 12 applies an algorithm (in this embodiment, a known 3D packing algorithm) to the physical information for each of the multiple products acquired in S41 to predict an appropriate packing box from among multiple types of rectangular parallelepiped packing boxes. As a result, a packing box appropriate for packing (containing) each of the multiple products is predicted (S42). The process of predicting an appropriate packing box for packing the products can be similar to S22 (see FIG. 4 ) in the above embodiment.

[0118] Next, the CPU 12 determines an initial arrangement for delivering the multiple packaging boxes predicted in S42 together (S43). The CPU 12 predicts an overall packaging box that can pack (contain) all of the multiple packaging boxes arranged according to the determination in S43 (S45). In S45 of the first modified example, the CPU 12 applies an algorithm (in the first modified example, a known 3D packing algorithm) for predicting an appropriate overall packaging box for containing all of the multiple packaging boxes from multiple types of overall packaging boxes that are rectangular parallelepipeds. As a result, an appropriate overall packaging box is predicted. The predicted overall packaging box is stored in the storage device 13. Note that if an appropriate overall packaging box is not predicted in S45, the process proceeds directly to S46.

[0119] Next, the CPU 12 determines whether the arrangement of the multiple packaging boxes predicted in S42 can be changed (S46). When changing the arrangement of the multiple packaging boxes, the orientation (angle) of at least one of the packaging boxes is also changed. If the arrangement can be changed (S46: YES), the CPU 12 changes the arrangement of the multiple packaging boxes to a new arrangement (S47) and predicts an overall packaging box that can pack (contain) all of the multiple packaging boxes arranged according to the determination of S47 (S45). In other words, the CPU 12 executes the process of predicting an overall packaging box that can contain all of the multiple arranged packaging boxes multiple times while changing the arrangement of the multiple packaging boxes (S43 to S47).

[0120] When the arrangement of multiple packaging boxes can no longer be changed (S46: NO), the CPU 12 identifies the overall packaging box with the smallest volume from the one or more overall packaging boxes predicted in S45.The CPU 12 predicts the shipping cost when multiple products are shipped together based on the overall packaging box with the smallest volume, and displays this on the interface screen (S48).By performing the above process, the shipping cost when multiple products are shipped together in an overall packaging box is automatically and appropriately predicted.

[0121] Next, the CPU 12 calculates the difference between the estimated shipping cost for shipping multiple items together (the result estimated in S48) and the estimated total shipping cost for shipping multiple items separately. The CPU 12 displays the calculated difference on the interface screen. This allows the user to easily and appropriately understand the amount of money that can be saved by shipping multiple items together. Note that the method for estimating shipping costs for shipping multiple items separately can be the same as S22 (see FIG. 4) in the above embodiment.

[0122] <Second Modification> A second modification of the above embodiment will be described. The purchase assistance system 1 of the second modification also takes into account the predicted results of the HS code of a product when performing the customs duty prediction process (S25, see FIG. 4). HS codes are codes used to classify international trade products. Customs duty rates are predetermined for each HS code of a product. Therefore, by predicting customs duties while taking into account the predicted results of the HS code, the accuracy of customs duty prediction is appropriately improved.

[0123] In the second modified example, a mathematical model for predicting HS codes is used to predict HS codes. The mathematical model for predicting HS codes is trained using a machine learning algorithm so as to input product information (e.g., at least a portion of the product's necessary information extracted in S11 to S17 of FIG. 4 ) and output a predicted HS code for the product. As an example, the mathematical model of the second modified example is trained according to the machine learning algorithm using the product information as input training data and the correct HS code designated by an operator who understands the product information as output training data. Furthermore, if the prediction result output by the mathematical model for predicting HS codes differs from the correct HS code, the mathematical model is retrained based on the incorrect prediction result. As a result, the accuracy of HS code prediction is further improved. The CPU 12 inputs product information (necessary information) listed on a web page into the mathematical model to obtain the predicted HS code output by the mathematical model. The CPU 12 predicts the tariff when exporting the product based on the predicted HS code.

[0124] Furthermore, in the purchase assistance system 1 of the second modified example, the CPU 12 identifies rules that may be problematic in customs clearance procedures based on the predicted HS code. The CPU 12 notifies the user of the identified rules. Depending on the classification of a product by HS code, customs clearance rules may include rules that limit export volume and rules that prohibit export. Therefore, by notifying the user of customs clearance rules based on the predicted HS code, the user can more appropriately decide whether or not to purchase the product.

[0125] In the second modified example, a mathematical model for predicting product attributes is constructed. The mathematical model for predicting product attributes is trained by a machine learning algorithm so that, when product information is input, the mathematical model outputs a predicted result of the product's attributes (e.g., at least one of attributes such as used products, refrigerated products, frozen products, large products, products with a short expiration date, products that are detained at customs, etc.). The CPU 12 inputs the product information into the mathematical model for predicting product attributes to obtain the predicted result of the product attributes output by the mathematical model. The CPU 12 notifies the user of the obtained product attribute information. Therefore, the user can appropriately understand the product's attributes and then appropriately decide whether or not to purchase the product.

[0126] Furthermore, the purchase assistance system 1 of the second modified example calculates the total cost required for the user to purchase the product using the shipping fee predicted in S22 and the customs duty predicted in S25, and completes payment of the total cost to the user in advance before the product is delivered. This allows the user to grasp the total cost required to purchase the product at an early stage, before receiving the product, making it easier for them to decide whether or not to purchase the product. Furthermore, the purchase assistance system 1 of the second modified example does not charge or refund the difference to the user even if the predicted total cost differs from the actual total cost. As a result, the number of payments and receipts between the user and the system is less likely to increase, simplifying the product purchasing process.

[0127] <Third Modification> A third modification of the above embodiment will be described with reference to FIG. 7 . As shown in FIG. 7 , the purchase assistance system 1 of the third modification displays a dedicated icon 70 on a general-purpose interface screen, allowing the user to instruct the display of product information listed on a web page. The dedicated icon 70 is displayed on the general-purpose interface screen along with the web page on which the product is listed. The user operates the dedicated icon 70 when wanting to check product information listed on the web page on the interface screen. When the user operates the dedicated icon 70, the CPU 12 of the purchase assistance system 1 identifies the web page that was displayed on the general-purpose interface screen when the dedicated icon 70 was operated. The CPU 12 automatically extracts necessary information about the product listed on the web page from the HTML that constitutes the identified web page. The process of automatically extracting the necessary information can be similar to steps S11 to S17 of the above embodiment (see FIG. 4 ). Furthermore, the CPU 12 executes a process for estimating shipping costs for the product (S21-S22, see FIG. 4) and a process for estimating customs duties (S25, see FIG. 4). The CPU 12 displays the extended browser 71 on the general-purpose interface screen. The CPU 12 displays product information, including necessary information, on the extended browser 71. Therefore, the user can grasp product information on the extended browser while the web page listing the product they wish to purchase is displayed on the interface screen.

[0128] 7, the CPU 12 displays the price of the product, the estimated shipping cost, and the estimated customs duty on the extended browser 71. Also, in the example shown in FIG. 7, the CPU 12 displays an "Add to Cart" button and a "View in SAZO" button on the extended browser 71 along with product information. When the "Add to Cart" button is operated, the CPU 12 adds the product whose information is displayed on the extended browser 71 to the cart. Also, when the "View in SAZO" button is operated, the CPU 12 newly displays the dedicated interface screen exemplified in the above embodiment.

[0129] <Fourth Modification> A fourth modification of the above embodiment will be described with reference to FIG. 8 . The purchase assistance system 1 of the fourth modification determines whether a product is listed on the displayed web page (i.e., whether the displayed page is a product listing page) each time the web page displayed on the general-purpose interface screen is updated. As an example, in the fourth modification, a mathematical model that outputs the probability that the web page displayed on the general-purpose interface screen is a product listing page is constructed in advance according to a machine learning algorithm. The CPU 12 inputs information about the displayed web page (e.g., HTML, etc.) into the mathematical model to obtain the probability that the web page is a product listing page. The CPU 12 determines whether the displayed web page is a product listing page based on the obtained probability.

[0130] As shown in FIG. 8 , if the displayed web page is a product listing page, the CPU 12 displays an auxiliary browser 73, which displays information to assist the user in purchasing the product, on the interface screen along with the web page. The CPU 12 identifies the web page displayed on the interface screen when displaying the auxiliary browser 73. The CPU 12 automatically extracts necessary information about the product listed on the web page from the HTML that constitutes the identified web page. The process of automatically extracting the necessary information can be similar to steps S11 to S17 (see FIG. 4 ) in the above embodiment. Furthermore, the CPU 12 executes a process of estimating shipping costs for the product (S21 to S22, see FIG. 4 ) and a process of estimating customs duties (S25, see FIG. 4 ). The CPU 12 displays product information, including the necessary information, on the extended browser 71. Therefore, while the web page listing the product is displayed on the interface screen, the user can view product information, including the extracted necessary information, on the automatically displayed auxiliary browser 73.

[0131] In the example shown in Fig. 8, the CPU 12 displays the product price, estimated shipping cost, and estimated customs duty on the auxiliary browser 78. Also, in the example shown in Fig. 8, the CPU 12 displays an "Add to Cart" button and a "Buy Now" button on the auxiliary browser 73 along with product information. Also, in the example shown in Fig. 8, the CPU 12 displays price trend information (a price trend graph in Fig. 8) showing the price trend of the product on the auxiliary browser 73. Furthermore, in the example shown in Fig. 8, the CPU 12 displays information on "no customs clearance issues," which is one of the product attributes described in the second modified example, on the auxiliary browser 73.

[0132] <Fifth Modification> A fifth modification of the above embodiment will be described with reference to FIG. 9 . The display unit 26 of the user terminal 20 shown in FIG. 9 displays a web page displaying a product, along with an information sharing instruction button 74 that is operated to instruct the execution of an information sharing function of the user terminal 20. The user operates the information sharing instruction button 74 when sharing information on the currently displayed web page with another application. When the information sharing instruction button 74 is operated, the user terminal 20 displays an information sharing icon field 75 on the display unit 26. The information sharing icon field 75 displays icons corresponding to multiple applications that are candidates for sharing information. When the user wants to share information on the currently displayed web page with the dedicated application described in the above embodiment, the user operates an icon 76 corresponding to the dedicated application among the multiple icons displayed in the information sharing icon field 75. As a result, a sharing operation for sharing information with the dedicated application is input to the user terminal 20. When the sharing operation is input to the user terminal 20, the user terminal 20 launches the dedicated application. When a sharing operation is performed on the user terminal 20, the purchase assistance system 1 identifies the web page displayed on the interface screen of the user terminal 20. The purchase assistance system 1 automatically extracts necessary information about the product listed on the identified web page from the HTML that constitutes the web page. The process of automatically extracting the necessary information can be similar to steps S11 to S17 (see FIG. 4) in the above embodiment. Furthermore, the purchase assistance system 1 executes a process of estimating the shipping cost of the product (S21 to S22, see FIG. 4) and a process of estimating the customs duty (S25, see FIG. 4). The purchase assistance system 1 displays product information, including the necessary information, on the interface screen of the user terminal 20. According to the fifth modified example, the information sharing function provided in the user terminal 20 is utilized to smoothly purchase the product.

[0133] The techniques disclosed in the above embodiments and modifications are merely examples. Therefore, it is possible to modify the techniques exemplified in the above embodiments and modifications. For example, it is possible to execute only a part of the techniques exemplified in the above embodiments and modifications. Specifically, it is possible to omit the customs duty prediction and display process shown in S25 and S26 of FIG. 4 and execute other processes. It is also possible to omit the shipping fee prediction and display process shown in S21 to S23 of FIG. 4 and execute other processes. It is also possible to combine and use multiple techniques exemplified in the above embodiments and modifications.

[0134] The process of acquiring page-specific information in S5 and S6 of FIG. 2 is an example of a "page-specific information acquisition step." The process of automatically extracting necessary information in S11 to S15 of FIG. 4 is an example of an "extraction step." The process of displaying necessary information on an interface screen in S20 of FIG. 4 is an example of a "necessary information display step." The structural analysis process executed in S14 of FIG. 4 is an example of a "structural analysis step." The AI-utilizing process executed in S15 of FIG. 4 is an example of an "AI-utilizing step." The process of acquiring physical information about a product in S21 of FIG. 4 is an example of a "physical information acquisition step." The process of predicting shipping costs for a product in S22 of FIG. 4 is an example of a "shipping cost prediction step." The process of displaying predicted shipping costs on an interface screen in S23 of FIG. 4 is an example of a "predicted shipping cost display step." The process of predicting customs duties in S25 of FIG. 4 is an example of a "tariff prediction step." The process of displaying predicted customs duties on an interface screen in S26 of FIG. 4 is an example of a "predicted customs duty display step." The process of displaying the link in S27 of FIG. 4 is an example of a "link display step."

Claims

1. A purchase assistance method executed in a purchase assistance system that assists a user in purchasing overseas products from their country of residence via an interface screen displayed on a user terminal used by the user, comprising: a page identification information acquisition step of acquiring, via the interface screen, page identification information that identifies a web page on which the product the user wishes to purchase is listed; an extraction step of automatically extracting, from the HTML that constitutes the web page identified by the page identification information, some necessary information from the product information listed on the web page; and a necessary information display step of displaying, on the interface screen, at least some of the necessary information extracted in the extraction step.

2. A purchasing assistance method as described in claim 1, characterized in that in the extraction step, the necessary information is automatically extracted by analyzing the HTML structure of the web page identified by the page identification information.

3. A purchasing assistance method as described in claim 1, characterized in that in the extraction step, the HTML of the web page identified by the page identification information is input into a mathematical model trained by a machine learning algorithm so as to extract and output the necessary information formed on the web page when the HTML of the web page is input, thereby extracting the necessary information.

4. A purchase assistance method as described in claim 1, wherein the extraction step is capable of executing: a structural analysis step of automatically extracting the necessary information by analyzing the HTML structure of the web page specified by the page specification information; and an AI utilization step of extracting the necessary information by inputting the HTML of the web page specified by the page specification information into a mathematical model trained by a machine learning algorithm so as to extract and output the necessary information formed on the web page when the HTML of the web page is input; wherein the purchase assistance method is characterized in that the structural analysis step and the AI ​​utilization step are selectively executed depending on the web page specified by the page specification information.

5. A purchase assistance method as described in claim 1, characterized in that, when the web page identified by the page identification information is displayed in a language of a foreign country different from the language of the user's country of residence, in the necessary information display step, at least a portion of the necessary information extracted from HTML in the extraction step is automatically translated into the language of the user's country of residence and displayed on the interface screen.

6. A purchase assistance method as described in claim 1, characterized in that in the page specification information acquisition step, the URL of a web page entered by the user on an interface screen displayed on the user terminal is acquired as the page specification information.

7. A purchasing assistance method as described in claim 1, further comprising: a physical information acquisition step of acquiring physical information of a product posted on the website by inputting at least some of the necessary information extracted in the extraction step into a mathematical model trained by a machine learning algorithm so as to output physical information of the product when the necessary information of the product is input; a shipping fee prediction step of predicting shipping fees for the product posted on the website based on the physical information acquired in the physical information acquisition step; and a predicted shipping fee display step of displaying the shipping fee predicted in the shipping fee prediction step on the interface screen.

8. A purchasing assistance method as described in claim 7, characterized in that in the physical information acquisition step, the physical information of the product listed on the website is acquired by inputting the multiple types of necessary information extracted in the extraction step into the mathematical model.

9. A purchasing assistance method as described in claim 7, wherein in the shipping cost prediction step, an algorithm is applied to the physical information acquired in the physical information acquisition step, which predicts an appropriate packing box from among multiple types of packing boxes having a rectangular parallelepiped shape, and the shipping cost of the product is predicted based on the packing box predicted by the algorithm.

10. A purchasing assistance method as described in claim 7, characterized in that, when a user selects multiple candidate products for purchase, physical information for each of the selected multiple products is acquired in the physical information acquisition step, and in the shipping cost prediction step, the shipping cost for delivering the multiple products together is predicted based on the physical information for each of the multiple products acquired in the physical information acquisition step.

11. A purchasing assistance method according to claim 10, wherein in the shipping cost prediction step, an algorithm is applied to the physical information of each of the plurality of products acquired in the physical information acquisition step, which predicts an appropriate packing box from among a plurality of types of packing boxes having a rectangular parallelepiped shape; the process of predicting an overall packing box capable of containing all of the plurality of arranged packing boxes is executed multiple times while changing the arrangement of the plurality of predicted packing boxes; and the shipping cost for delivering the plurality of products together is predicted based on the overall packing box with the smallest volume among the plurality of predicted overall packing boxes.

12. A purchasing assistance method as described in claim 10, characterized in that, when a user selects multiple candidate products for purchase, the difference between the estimated shipping cost for shipping the multiple products together and the estimated total shipping cost for shipping the multiple products separately is calculated and displayed on the interface screen.

13. A purchasing assistance method as described in claim 1, further comprising: a tariff prediction step for predicting a tariff to be levied when importing the product into the user's country of residence based on product information posted on the web page identified by the page identification information; and a predicted tariff display step for displaying the tariff predicted in the tariff prediction step on the interface screen.

14. A purchasing assistance method as set forth in claim 13, characterized in that in the tariff prediction step, information about the product listed on the web page identified by the page identification information is input into a mathematical model trained by a machine learning algorithm so as to output a predicted result of the tariff when information about the product is input, thereby obtaining a predicted result of the tariff to be levied on the product.

15. A purchasing assistance method as set forth in claim 14, characterized in that in the tariff prediction step, a plurality of pieces of information about the product listed on the web page identified by the page identification information are input into the mathematical model to obtain a predicted result of the tariff to be levied on the product.

16. A purchasing assistance method as set forth in claim 14, wherein the tariff prediction step uses a RAG that searches for specific information from a collection of texts and generates new text based on the searched information to obtain text that indicates the characteristics of a product from a collection of texts related to the product listed on the web page identified by the page identification information, and inputs the obtained text into the mathematical model to obtain a prediction result of the tariff to be levied on the product.

17. A purchasing assistance method as set forth in claim 13, wherein in the tariff prediction step, information about the product listed on the web page identified by the page identification information is input into a mathematical model trained by a machine learning algorithm to output a predicted result of the HS code that classifies international trade products, thereby obtaining a predicted result of the HS code for the product, and predicting the tariff based on the predicted result of the HS code.

18. A purchase assistance method as described in claim 1, further comprising a link display step of displaying on the interface screen a link designated by the user to display the web page identified by the page identification information.

19. A purchase assistance method as described in claim 1, further comprising a dedicated icon display step of displaying on the interface screen a dedicated icon for instructing the display of information about the product, including the necessary information extracted in the extraction step, together with a web page listing the product that the user wishes to purchase; when the dedicated icon is operated by the user, the page identification information acquisition step identifies the web page that was displayed on the interface screen when the dedicated icon was operated; the extraction step automatically extracts the necessary information from HTML that constitutes the identified web page; and the necessary information display step displays an extension browser on the interface screen, and displays information about the product, including the necessary information, on the displayed extension browser.

20. A purchasing assistance method as recited in claim 1, further comprising an auxiliary browser display step of, when a web page listing a product is displayed on the interface screen, displaying an auxiliary browser on the interface screen together with the web page, the auxiliary browser displaying information to assist the user in purchasing the product; when the auxiliary browser is displayed, the page identification information acquisition step identifies the web page that was displayed on the interface screen when the auxiliary browser was displayed; the extraction step automatically extracts the necessary information from HTML that constitutes the identified web page; and the necessary information display step displays information about the product, including the necessary information, on the displayed auxiliary browser.

21. A purchase assistance program executed in a purchase assistance system that assists a user in purchasing overseas products from their country of residence via an interface screen displayed on a user terminal used by the user, wherein the purchase assistance program is executed by a control unit of the purchase assistance system, causing the purchase assistance system to perform the following steps: a page identification information acquisition step that acquires, via the interface screen, page identification information that identifies a web page on which the product the user wishes to purchase is listed; an extraction step that automatically extracts, from the HTML that constitutes the web page identified by the page identification information, some necessary information from the product information listed on the web page; and a necessary information display step that displays, on the interface screen, at least some of the necessary information extracted in the extraction step.

22. A purchase assistance system that assists a user in purchasing overseas products from their country of residence via an interface screen displayed on the user's terminal, the purchase assistance system comprising: a page identification information acquisition step that acquires, via the interface screen, page identification information that identifies a web page on which the product the user wishes to purchase is posted; an extraction step that automatically extracts, from the HTML that constitutes the web page identified by the page identification information, some necessary information from the product information posted on the web page; and a necessary information display step that displays, on the interface screen, at least some of the necessary information extracted in the extraction step.

23. An AI-based cross-border e-commerce system that assists users in purchasing overseas products from their country of residence via an interface screen displayed on the user's terminal, comprising: a page-specific information acquisition step of acquiring, via the interface screen, page-specific information that identifies a web page on which the product the user wishes to purchase is posted; an extraction step of automatically extracting the necessary information by inputting the HTML of the web page identified by the page-specific information into a mathematical model trained by a machine learning algorithm so as to extract and output the necessary information formed on the web page when the HTML of the web page is input; and a necessary information display step of displaying at least a portion of the necessary information extracted in the extraction step on the interface screen.