Information processing method, information processing apparatus, and program

The method and device address the issue of inappropriate gift recommendations by using product information and labels to determine and estimate attributes, ensuring appropriate item presentation.

JP2024162609AActive Publication Date: 2024-11-21RAKUTEN GROUP INC
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
JP2023078298
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-05-11
Publication Date
2024-11-21
Estimated Expiration
2043-05-11

AI Technical Summary

Technical Problem

Existing technologies fail to effectively exclude inappropriate products as gifts from recommended lists, leading to potential discomfort for users.

Method used

An information processing method and device that extracts and presents items based on product information and labels, determining and estimating attributes to exclude sensitive or inappropriate items from user presentation.

Benefits of technology

Effectively excludes sensitive or inappropriate items from presentation, enhancing user experience by providing appropriate recommendations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024162609000001_ABST
    Figure 2024162609000001_ABST
Patent Text Reader

Abstract

To provide a technique capable of appropriately presenting an item to a user by excluding an item having a specific attribute.SOLUTION: Provided is a method by which an information processing apparatus restricts presentation of an item having a first attribute, including: extracting a plurality of items, each of which includes product-related information and label-related information; selecting, based on the label-related information, any of that each extracted item has the first attribute, that the item does not has the first attribute, and that it is unknown whether or not the item has the first attribute; estimating that an item about which it is unknown whether or not the item has the first attribute whether or not the item has the first attribute, based on at least information on a product of each of the plurality of extracted items; and presenting, to a user, an item determined not to have the first attribute and an item estimated not to have the first attribute among the plurality of items.SELECTED DRAWING: Figure 4
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to an information processing method, an information processing device, and a program for appropriately presenting items. [Background technology]

[0002] In recent years, online electronic commerce services have been provided. Electronic commerce websites recommend products to users. For example, Patent Document 1 discloses a technology for extracting products similar to a product to be purchased based on graph information in which multiple nodes corresponding to each of multiple products are linked according to the similarity of the products, and order information requesting the purchase of the product, and for recommending an appropriate product based on the extracted similar products. Patent Document 2 discloses a technology for excluding products in a list of products not recommended as gifts for the user from a group of products that are candidates for recommended information for the user, and displaying the products to the user. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2019-125257 A [Patent Document 2] JP 2017-228157 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, even if the group of products that are candidates for recommended information for a user includes products that are inappropriate as gifts, the technology described in Patent Document 2 cannot exclude those products if they are not already included in the list of non-recommended gifts.

[0005] Certain embodiments of the present invention aim to provide a technique that can appropriately present items to a user. [Means for solving the problem]

[0006] A method according to one aspect of the present disclosure is a method for an information processing device to limit presentation of items having a first attribute, the method comprising: extracting a plurality of items, each of which includes information related to a product and information related to a label; determining, based on the information related to the label, whether each of the extracted items has the first attribute, does not have the first attribute, or it is unknown whether it has the first attribute; estimating, based at least on information related to the product of each of the extracted plurality of items, whether an item whose first attribute is unknown has the first attribute; and presenting, to a user, items determined not to have the first attribute and items estimated not to have the first attribute among the plurality of items.

[0007] Another aspect of the present disclosure is an information processing device for limiting presentation of items having a first attribute, the information processing device including: an extraction unit that extracts a plurality of items, each of which includes information related to a product and information related to a label; a determination unit that determines, based on the information related to the label, whether each of the extracted items has the first attribute, does not have the first attribute, or it is unknown whether or not it has the first attribute; an estimation unit that estimates, based at least on information related to the product of each of the extracted plurality of items, whether or not an item whose first attribute is unknown has the first attribute; and a presentation unit that presents to a user items determined not to have the first attribute and items estimated not to have the first attribute among the plurality of items.

[0008] Yet another aspect of the present disclosure is a program for limiting presentation of items having a first attribute, the program causing a computer to extract a plurality of items, each of which includes information related to a product and information related to a label, determine whether each of the extracted items has the first attribute, does not have the first attribute, or is unknown whether it has the first attribute, based on the information related to the label, estimate whether an item whose first attribute is unknown has the first attribute, based at least on the information related to the product of each of the extracted plurality of items, and present to a user the items determined to not have the first attribute and the items estimated to not have the first attribute, among the plurality of items. Effect of the Invention

[0009] According to certain embodiments of the present invention, it is possible to provide a technique that can appropriately present items to a user. [Brief description of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram showing an overall configuration of an information processing system 1 according to an embodiment of the present disclosure. [Diagram 2] FIG. 2 is a functional block diagram of the information processing device 100 according to the embodiment of the present disclosure. [Diagram 3] FIG. 3 is a diagram illustrating an example of a physical configuration of the information processing device 100 according to an embodiment of the present disclosure. [Figure 4] FIG. 4 shows an overview of the process executed by the information processing device 100 according to an embodiment of the present disclosure. [Diagram 5] FIG. 5 is a diagram illustrating an example of the estimation process according to an embodiment of the present disclosure. [Figure 6] FIG. 6 is a diagram illustrating an example of a viewing screen according to an embodiment of the present disclosure. [Figure 7] FIG. 7 is a diagram showing a comparative example of a viewing screen according to an embodiment of the present disclosure. [Figure 8]FIG. 8 is a diagram illustrating an example of a viewing screen according to an embodiment of the present disclosure. [Figure 9] FIG. 9 is a diagram illustrating an example of a process for extracting similar items. [Figure 10] FIG. 10 is an example flowchart of information processing executed according to an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] Hereinafter, a method, an information processing device, and a program for limiting the presentation of an item having a first attribute by an information processing device according to the present embodiment will be described with reference to the drawings. In each drawing, the same components are denoted by the same reference numerals as much as possible, and duplicated descriptions will be omitted.

[0012] In this embodiment, "item" refers to individual products and services. For example, products and services that are traded in electronic commerce services such as online shopping are examples of items. However, items are not limited to each product and service itself, and may be understood to include data, images, and other information related to the products and services. Below, an example will be described in which a method for limiting the presentation of items according to this embodiment is implemented, mainly targeting products. Therefore, items and products are sometimes used almost synonymously.

[0013] Fig. 1 is a diagram showing an overall configuration of an information processing system 1 according to an embodiment of the present disclosure. As shown in Fig. 1, the information processing system 1 includes an information processing device 100 and a plurality of user terminals 200-1, 200-2, ... 200-n (n is a natural number), and the information processing device 100 and the plurality of user terminals 200-1 to 200-n are connected by wire or wirelessly via a network 20 such as the Internet or a telephone network. Hereinafter, any user terminal among the user terminals 200-1 to 200-n will also be referred to as a user terminal 200. The information processing system 1 may further include a database DB.

[0014] The information processing device 100 may be, for example, an e-commerce server that provides a web page of an online shopping site to the user terminal 200 and allows the user to browse and purchase products based on user operations on the web page, and may be composed of, for example, one or more physical servers, or may be composed of a virtual server running on a hypervisor, or may be composed of a cloud server.

[0015] The user terminal 200 may be configured to display, for example, a web page of an online shopping site provided by the information processing device 100 to the user and accept the user's operation on the displayed web page. For example, an electronic device having a web browser function may be used as the user terminal 200. Specifically, for example, a personal computer, a mobile phone terminal, a mobile phone, etc. may be used as the user terminal 200. Through the user terminal 200 on which the web page of the online shopping site is displayed, the user can perform a purchase procedure for the product, such as inputting search conditions for a product to be purchased, selecting a product to be purchased from multiple products being viewed, and inputting payment and shipping methods.

[0016] The database DB may be, for example, a database that stores, for any item, information about the product of each item and information about the label. Hereinafter, the information about the product of each item and information about the label are collectively referred to as "product information". The database DB may store, for example, product information about all products handled in the electronic commerce platform realized by the information processing system 1 of this embodiment, or may store product information about some of the products. Note that, although an example will be described here in which the product information is stored in an external database DB, the product information may also be saved in, for example, the memory of the information processing device 100 (memory 10b (see FIG. 3) described later).

[0017] 2 is an example of a functional block diagram of an information processing device 100 according to an embodiment of the present disclosure. The information processing device 100 includes an extraction unit 12, a determination unit 14, an estimation unit 16, and a presentation unit 18.

[0018] The extraction unit 12 extracts a plurality of items, for example, from a database DB external to the information processing device 100. Each extracted item includes information related to the product and information related to the label. Note that, although the extraction unit 12 is described here as acquiring items from an external database DB, if the items are stored in the memory of the information processing device 100, the items may be acquired from the memory.

[0019] In an embodiment of the present disclosure, the information about a product may include any information about an individual product, such as the product title, product description, price, condition, product image, etc. The product title may include the product name, brand name, manufacturer, importer, color, shape, size, and other product features. The information about a product may be given when a seller puts the product up for sale on an e-commerce platform, or may be given by artificial intelligence (AI).

[0020] The information on the label may be at least one of a report from a user, a score calculated from an image of the product, or a genre to which the product belongs. The report from a user may be information on whether or not the user requested that the product not be displayed due to reasons such as disgust. The score calculated from the image of the product may be calculated, for example, by a machine learning model, and a score in terms of stimulating strength, vulgarity, etc. is calculated based on the image of the product. When the score of the image is calculated by a machine learning model, for example, the score of the image of the product may be calculated using a learning model trained by learning data including the image of the product and a score previously assigned to the product. The genre to which the product belongs may be information corresponding to a classification of multiple products, such as electronic parts, stationery, cooking utensils, nursing care products, funeral products, etc.

[0021] The determination unit 14 determines whether each extracted item has a first attribute, does not have the first attribute, or whether it has the first attribute is unclear, based on information about the label of each item extracted by the extraction unit 12. In this embodiment, the first attribute may be any attribute, and may be, for example, an attribute indicating that the item is a sensitive item requiring caution when viewed, such as adult, violent, grotesque, or offensive to the user.

[0022] The estimation unit 16 estimates, based on at least information relating to each of the multiple items extracted by the extraction unit 12, whether or not an item, for which it is unknown whether or not it has the first attribute, has the first attribute.

[0023] The presenting unit 18 presents to the user, among the multiple items extracted by the extracting unit 12, the items determined not to have the first attribute and the items estimated not to have the first attribute.

[0024] In the information processing device 100 of this embodiment, for example, when presenting items to a user in electronic commerce, items that are determined or estimated not to have a specific attribute are presented to the user. That is, items that have a specific attribute and items whether they have a specific attribute or not are not known to have a specific attribute are not presented to the user if they are estimated to have a specific attribute, so that, for example, sensitive items that require caution when viewed can be excluded from being presented to the user. Therefore, items can be presented appropriately to the user.

[0025] FIG. 3 is a diagram showing an example of a physical configuration of the information processing device 100 according to the present embodiment. The information processing device 100 has a processor 10a such as a CPU (Central Processing Unit) corresponding to a calculation unit, a memory 10b, a communication unit 10c, an input unit 10d, and a display unit 10e. These components are connected to each other via a bus so that they can transmit and receive data. In this example, the information processing device 100 is configured by one computer, but the information processing device 100 may be realized by combining multiple computers. The configuration shown in FIG. 3 is an example, and the information processing device 100 may have other configurations than these, or may not have some of these configurations.

[0026] The processor 10a is a control unit that controls the execution of the programs stored in the memory 10b and performs calculations and processing of data. The processor 10a is a calculation unit that executes a program for restricting the presentation of items having a specific attribute (first attribute) using each data, etc. The processor 10a receives various data from the input unit 10d and the communication unit 10c, and displays the calculation results of the data on the display unit 10e and stores them in the memory 10b.

[0027] The memory 10b may have, for example, a RAM (Random Access Memory) and a ROM (Read Only Memory) corresponding to a storage unit. The RAM is a part of the storage unit where data can be rewritten, and may be composed of, for example, a semiconductor memory element. The RAM may store programs executed by the processor 10a and data used for learning (for example, product information). Note that these are merely examples, and the RAM may store data other than these, or may not store some of these.

[0028] The ROM is a storage unit from which data can be read, and may be configured, for example, with a semiconductor memory element. The ROM may store, for example, a learning execution program or data that is not rewritten.

[0029] The communication unit 10c is an interface that connects the information processing device 100 to other devices, and may be connected to a communication network such as the Internet.

[0030] The input unit 10d receives data input from a user, and may include, for example, at least one of a keyboard and a touch panel.

[0031] The display unit 10e visually displays the results of the calculations performed by the processor 10a and may be configured, for example, with an LCD (Liquid Crystal Display). The display unit 10e may display, for example, items output in the item presentation process according to the present embodiment.

[0032] The program for restricting the presentation of items having a specific attribute (first attribute) may be provided by being stored in a computer-readable storage medium such as the memory 10b (for example, a RAM or a ROM), or may be provided via a communication network connected by the communication unit 10c. In the information processing device 100, the processor 10a executes the learning program to realize various operations described with reference to FIGS. 1 and 2. Note that these physical configurations are merely examples, and do not necessarily have to be independent configurations. For example, the information processing device 100 may include an LSI (Large-Scale Integration) in which the processor 10a and the memory 10b are integrated. In addition, the information processing device 100 may include a GPU (Graphical Processing Unit) as the processor 10a, and the GPU may execute the program for restricting the presentation of items having the above-mentioned specific attribute to realize various operations described with reference to FIGS. 1 and 2.

[0033] FIG. 4 shows an overview of the process executed by the information processing device 100 according to this embodiment. In this embodiment, the process shown in FIG. 4 is executed when the information processing device 100 recommends a plurality of items to a user. For example, when a user is browsing a certain product on an online shopping site using the user terminal 200, the information processing device 100 can present a plurality of recommended items related to the product. Also, for example, when a user opens a top page or a personal page of an online shopping site using the user terminal 200, the information processing device 100 can present a plurality of recommended items that are presumed to be of interest to the user.

[0034] As shown in FIG. 4, in the process according to the embodiment of the present disclosure, first, the extraction unit 12 extracts a plurality of items X1, X2, X3, X4, X5, and X6 from a database DB in which information on items is stored (S401). The plurality of items extracted by the extraction unit 12 are candidates for items to be presented to the user. Each of the extracted items X1 to X6 includes information on the product and information on the label. Note that, although FIG. 4 illustrates an example in which the extraction unit 12 extracts six items, the number of extracted items is not limited to this, and any number of items may be extracted.

[0035] In this embodiment, the items extracted by the extraction unit 12 may be, for example, items related to the item being viewed by the user. Specifically, for example, when a user is considering purchasing a smartphone and is viewing smartphones, smartphones of multiple models from multiple manufacturers may be extracted as related items. Also, accessories for the smartphone may be extracted as related items.

[0036] Furthermore, the items extracted by the extraction unit 12 may be extracted based on at least one of the user's browsing history, purchase history, and profile information. For example, items related to items previously viewed by the user or items previously purchased by the user may be extracted. Alternatively, items may be extracted based on the user's profile information, such as the user's age, sex, residential area, occupation, and annual income. In this case, for example, items may be extracted based on information on the browsing history and purchase history of other users with similar profile information.

[0037] Returning to FIG. 4, next, in step S402, for each of the extracted items X1 to X6, the determination unit 14 determines whether the item has the first attribute, does not have the first attribute, or is unclear whether it has the first attribute, based on information about the label of each of the items X1 to X6. In this embodiment, for example, if an item is not suitable for browsing by the user, the item is determined to have the first attribute. In other words, in this embodiment, for example, if an item is suitable for browsing by the user, the item is determined to not have the first attribute. In FIG. 4, X1, X2, and X3 are determined to have the first attribute, X5 and X6 are determined to not have the first attribute, and it is unclear whether X4 has the first attribute.

[0038] An example of a label is a report from a user. For example, an item to which information that a user has reported that the item should not be displayed is attached as a label may be determined to have the first attribute. Another example of a label is an image of a product. For example, an item to which a score calculated from an image of the product exceeds (or falls below) a predetermined threshold may be determined to have the first attribute.

[0039] Yet another example of the label is the genre to which each item belongs. That is, the determination unit 14 may determine whether or not each item has the first attribute based on the genre of the item attached to the item. For example, the database DB may store information about the genre to which each item belongs for each item, and may store information about whether the genre is appropriate or inappropriate for presentation to the user for each genre. At this time, the determination unit 14 refers to the database DB to obtain the genre of each item and information about whether the genre is appropriate or inappropriate for presentation to the user, and determines that an item belonging to a genre that is inappropriate for presentation to the user has the first attribute. For example, for a user who does not like to view nursing care products and funeral products, the genre to which these products belong may be determined to be an inappropriate genre, and electronic parts, stationery, cooking utensils, etc. other than these nursing care products and funeral products may be determined to be an appropriate genre that does not cause problems when viewed.

[0040] In this embodiment, the genre may be set on the user's personal setting screen of the online shopping website. For example, a toggle button, a check box, a radio button, a drop-down list, or the like may be attached to each genre so that the user can label each genre as appropriate or inappropriate for viewing. In addition, when genres are prepared hierarchically, the appropriateness of viewing may be set at a predetermined level, or may be set at different levels. For example, when it is desired to set the entire genre of a certain higher level as appropriate or inappropriate, the setting may be made for the genre of the higher level, and when it is desired to set each genre of a lower level as appropriate or inappropriate, the setting may be made for the genre of the lower level without making a setting for the genre of the higher level.

[0041] The genres appropriate or inappropriate for presentation to the user may be preset by the administrator, in addition to being specified by the user. For example, the administrator who operates an online shopping site may set the genres appropriate or inappropriate for presentation to the user in advance. The administrator may also set the genres appropriate or inappropriate for presentation to the user according to the user's attributes, such as the user's gender, age, occupation, etc. The information set by the administrator as appropriate or inappropriate for presentation to the user may be configured to be changeable by the user. Similarly, in the aspect in which the administrator sets the genres, the administrator may be configured to accept the selection from the administrator on the administrator's setting page or the like. Note that both the appropriate genres and the inappropriate genres may be set, or either one of them may be set.

[0042] Furthermore, when the determination unit 14 is unable to determine whether the item has the first attribute based on the information on the label, it determines that it is unclear whether the item has the first attribute. For example, when a user reports the same item multiple times and the multiple reports include both appropriate and inappropriate items, it may be determined that it is unclear whether the item has the first attribute. For example, when the image score is higher than a first threshold and it is determined that the item does not have the first attribute when it is lower than a second threshold, it may be determined that the item has the first attribute when the image score is lower than the first threshold and higher than the second threshold. For example, when the genre attached to the item is not indicated as being appropriate or inappropriate for viewing in the database DB, it may be determined that it is unclear whether the item has the first attribute. In addition, for example, an item to which no information on a label is attached may be determined to have the first attribute. When the determination result of a certain label differs from the determination result of another label, the determination unit 14 may determine whether the item has the first attribute based on a predetermined rule. For example, the presence or absence of the first attribute may be determined according to the judgment result of any one of the labels based on a predetermined priority order between the labels, or may be determined by majority vote between multiple labels. Also, when the judgment results between the labels differ, it may be determined that it is unclear whether or not the first attribute is present.

[0043] Returning to Fig. 4, next, in step S403, the estimation unit 16 estimates whether an item, for which it is unclear whether it has the first attribute, has the first attribute based at least on information about each of the extracted multiple items. In the example shown in Fig. 4, it is determined that it is unclear whether item X4 has the first attribute, and in step S403, it is estimated whether item X4 has the first attribute.

[0044] Next, among the multiple items, the items determined not to have the first attribute and the items estimated not to have the first attribute are presented to the user by the presentation unit 18. In the example shown in FIG. 4, when the item X4 determined to have the first attribute is estimated to have the first attribute (S404), the items X5 and X6 determined not to have the first attribute among the items X1 to X6 are presented to the user (S405). That is, at this time, the items X1, X2, and X3 determined to have the first attribute and the item X4 estimated to have the first attribute are not presented to the user. On the other hand, when the item X4 determined to have the first attribute is estimated not to have the first attribute (S406), the items X5 and X6 determined not to have the first attribute among the items X1 to X6 and the item X4 estimated not to have the first attribute are presented to the user (S407). That is, at this time, the items X1, X2, and X3 determined to have the first attribute are not presented to the user.

[0045] In this embodiment, for example, when recommending an item to a user in electronic commerce, items that are determined or estimated not to have a specific attribute are presented to the user from among a plurality of items. Therefore, even if it is unclear whether an item has a specific attribute, if it is estimated to have the specific attribute, it is not presented to the user. This makes it possible to exclude, for example, sensitive items that require caution when viewed from the targets to be presented to the user.

[0046] Next, the estimation process in step S403 will be described. In step S403, the estimation unit 16 performs the estimation process based on, for example, graph information generated based on a plurality of extracted items. At this time, the graph information can be made up of nodes corresponding to each item and edges connecting any two nodes. In this embodiment, the graph information may be handled by a GNN (Graph Neural Network). A GNN is a neural network for expressing the feature amounts of nodes and edges in a graph structure. A GNN learns a feature representation incorporating information on adjacent nodes and edges of each node in a graph structure. In this embodiment, for example, by performing node classification using a GNN, it is possible to predict the label of each node for data having a graph structure, taking into consideration the feature amount of the target node and information (feature amount, etc.) of the surrounding nodes. For example, the estimation unit 16 performs node classification using a GNN for data having a graph structure including a node including information on the presence or absence of a distributed representation and a first attribute, and an edge including information on the similarity between items corresponding to the nodes connected to both ends. Based on the result of this node classification, the estimation unit 16 may estimate whether or not a node corresponding to an item, for which it is unknown whether or not it has the first attribute, has the first attribute.

[0047] The process of estimating the presence or absence of a first attribute of an item based on graph information will be described with reference to Fig. 5. Fig. 5 illustrates a case in which, as in the example of Fig. 4, for items X1 to X6 extracted as items related to item X, X1, X2, and X3 are determined to have the first attribute, X5 and X6 are determined not to have the first attribute, and it is determined that X4 is unknown as to whether it has the first attribute, and whether or not item X4 has the first attribute is estimated.

[0048] As shown in FIG. 5, the estimation unit 16 first creates an adjacency matrix A for expressing a graph structure (S501). The adjacency matrix A is constructed based on the similarity between items X1 to X6, for example. In the example shown in FIG. 5, the rows of the adjacency matrix A represent items X1, X2, X3, X4, X5, and X6 from the top, and similarly, the columns of the adjacency matrix A represent items X1, X2, X3, X4, X5, and X6 from the left. The information on the similarity between items may be, for example, a relevance score calculated based on at least one of the browsing history and the purchase history of a plurality of users, and the adjacency matrix A may be constructed based on the relevance score. An example of the relevance score is shown in FIG. 9, and the details will be described later. In the example shown in FIG. 5, an edge is connected between nodes corresponding to items whose relevance score is higher than a predetermined threshold. 5, in the graph, edges connect between the nodes corresponding to items X1, X2, and X3, between the nodes corresponding to items X3 and X4, between the nodes corresponding to items X4 and X5, and between the nodes corresponding to items X5 and X6. According to the adjacency matrix A, for example, 0 or 1 is placed at the intersections of rows and columns, and 1 is placed in the adjacency matrix A at the intersections of nodes that are highly related to items and connected by edges, and 0 is placed at the other intersections.

[0049] A node corresponding to each item may include a feature vector generated from information about a product included in the corresponding item. In this embodiment, the feature vector included in the node may include, for example, at least one of a distributed representation generated from text information about the product and a distributed representation generated from image information about the product. The distributed representation generated from the text information about the product may be generated using a text embedding technique such as fastText. Furthermore, the distributed representation generated from image information about the product may be generated by performing deep learning such as a convolutional neural network.

[0050] Next, as shown in Fig. 5, a feature matrix F is created to express the features assigned to the nodes corresponding to the items X1 to X6 (S502). In the example shown in Fig. 5, for the nodes corresponding to the items X1 to X6, x 11 ~x 1m , x 21 ~x 2m , x 31 ~x 3m , x 41 ~x 4m , x 51 ~x 5m , x 61 ~x 6m A feature vector including m features represented by F1...Fm (m is a natural number) is assigned to each item. The features F1...Fm of each item may be information related to the product, such as the product title, price, etc. Note that, although an example is described here in which information related to the product of each item is used as the features F1...Fm, information related to the label of each item may be used as the features F1...Fm instead of or in addition to the information related to the product. The features F1...Fm may include a distributed representation generated based on the information related to the product.

[0051] In this way, graph information of the relationships between the items X1 to X6, which is configured with the items X1 to X6 as nodes and edges connecting the nodes corresponding to the items, is expressed based on the generated adjacency matrix A and feature matrix F.

[0052] Next, the feature amounts from the adjacent nodes are aggregated using an aggregation function (S503). For example, the aggregation function may be a function that calculates the sum or average of the features F1...Fm of the features of the adjacent nodes.

[0053] Subsequently, the feature matrix F is updated using the update function (S504). The feature quantities aggregated in the previous step are updated, for example, using an activation function such as the sigmoid function or the softmax function as the update function. As shown in FIG. 5, the updated feature matrix F' has features F'1... F'k (where k is a natural number such that k < m). Therefore, in the updated feature matrix F', the feature vectors for the nodes corresponding to each item X1 to X6 are updated, and each node has x' 11 ~x' 1k 、x' 21 ~x' 2k 、x' 31 ~x' 3k 、x' 41 ~x' 4k 、x' 51 ~x' 5k 、x' 61 ~x' 6k including k features represented by.

[0054] In the aggregation of the feature quantities and the update of the feature matrix in the present embodiment, for example, message passing is executed with the graph structure and the features of the nodes as inputs. By executing message passing to aggregate the feature quantities and update the feature matrix, the distributed representations included in the features F1... Fm of each node of the graph information are converted into lower-dimensional distributed representations. Note that in the present embodiment, the distributed representation may be reduced in dimension by executing message passing a plurality of times. For example, message passing may be executed three or four times.

[0055] Next, node classification is executed based on the updated feature vectors of each node (S505). In the present embodiment, for the item X4 whose first attribute is unknown, the lower-dimensional feature vector x' 41 ~x' 4kNode classification is performed using the above. By applying node classification, it is possible to estimate whether a node corresponding to an item, whether it is unknown whether it has the first attribute, belongs to the node category of an item determined to have the first attribute or to the node category of an item determined not to have the first attribute. In this way, after the feature vector included in each node of the graph information is converted into a lower dimensional vector, an estimate is made as to whether an item X4, whether it is unknown whether it has the first attribute, has the first attribute.

[0056] In this embodiment, instead of node classification, a community detection method may be applied to the graph information. The community detection method can detect a plurality of communities to which nodes with high similarity to each other belong for a plurality of nodes. In this embodiment, for example, by applying the community detection method, it is possible to estimate whether a node corresponding to an item that is unknown whether it has a first attribute belongs to a community of items determined to have the first attribute or a community of items determined not to have the first attribute.

[0057] Next, an application example of this embodiment will be described with reference to FIGS.

[0058] Fig. 6 is a diagram showing an example of a viewing screen when this embodiment is applied to a case where a user is viewing a certain product on an online shopping site and multiple items related to the product are presented. In contrast to this, Fig. 7 is a diagram showing a comparative example of a viewing screen when a conventional technique is applied. In both the examples of Figs. 6 and 7, when a user is viewing a product X, p items (X1 to Xp) related to the product X being viewed by the user are extracted, and among them, items X1 to X3 are determined or estimated to have a first attribute.

[0059] In this case, when the presentation technique of this embodiment is used, as shown in FIG. 6, items X1 to X3 determined or estimated to have the first attribute are excluded from the recommendations and are not presented to the user, and a predetermined number of items determined or estimated not to have the first attribute (here, six items) are presented to the user as recommended items. Note that when the number p of extracted items is large and the number of items determined or estimated not to have the first attribute is greater than the predetermined number, the recommended items may be presented in order by user operation, etc. In contrast, in the comparative example using the conventional technique, items X1 to X3 determined or estimated to have the first attribute are also presented to the user as recommended items, as shown in FIG. 7.

[0060] 8 is a diagram showing an example of a viewing screen when this embodiment is applied in the case where, when a user opens a top page, personal page, etc. of an online shopping site, a plurality of items that are estimated to be of interest to the user are presented. As shown in the figure, a plurality of items recommended to the user are presented based on the history of products that the user has viewed so far. In the figure, items X1 to X3 that are determined or estimated to have the first attribute are excluded from the recommended items and are not presented to the user, and a predetermined number of items that are determined or estimated not to have the first attribute are presented to the user as recommended items.

[0061] 6 and 8, the number of items presented to the user as recommended items may be set by a provider or administrator of an electronic commerce service, or may be set by the user. For example, when presenting recommended products to a user on an electronic commerce webpage, a setting may be made to display up to five products, and if the total number of items determined or estimated not to have the first attribute is less than five, all items may be presented, and if the total number is five or more, five items may be presented.

[0062] Next, an example of a process for extracting an item, which is executed in the extraction unit 12, will be described with reference to FIG.

[0063] 9, first, information on the browsing history and purchase history of multiple users is collected (S901). In the figure, for example, a case is illustrated in which user A browses items X1, X2, X3, X4, and X5 in this order, user B browses items X6, X1, X2, X7, and X3 in this order, and user C browses items X1, X8, X2, X3, and X9 in this order.

[0064] Next, items viewed or purchased by multiple users A to C are identified (S902). In the example shown in Fig. 9, since all of users A to C viewed items X1, X2, and X3, items X1, X2, and X3 are identified.

[0065] Next, a relevance score between a plurality of items is calculated (S903). The relevance score may be obtained, for example, by calculating the ratio of the total number of users who viewed or purchased both any two items to the total number of users who viewed or purchased each of the items. By using the score exemplified here, the higher the ratio of users who viewed both of two items, the higher the relevance score.

[0066] For example, in the example shown in Fig. 9, among items X1 to X9, the relevance scores between items X1 and X2, X2 and X3, and X1 and X3 are higher than the relevance scores between any two other items. Therefore, when another user D browses item X1, the extraction unit 12 extracts items X2 and X3, which have a higher relevance score between item X1 and item X1 than the others, as items highly related to item X1. Note that, although the above example shows a case where each user has five browsing histories, it goes without saying that this is not limited to this.

[0067] The browsing history or purchase history of the user may be, for example, a browsing history or purchase history within a predetermined period. The period of the browsing history or purchase history used may be, for example, within the same session or the same day. Since products viewed within the same session are presumed to be highly related to each other, the accuracy of the calculated relevance can be improved by calculating the relevance based on the browsing history within the same session. In addition, even if products are viewed not within the same session but during, for example, an interruption in communication or a temporary termination and resumption of the browser, etc., products viewed within the same day are presumed to be highly related to each other, so the accuracy of the relevance can also be improved by calculating a relevance score based on the browsing history within the same day.

[0068] Furthermore, when presenting extracted items to a user, the presentation unit 18 may determine the order in which the items are presented based on the relevance scores. For example, when items X1, X2, and X3 are extracted as items related to the item X being viewed, the relevance scores between the item X and the items X1, X2, and X3 may be calculated, and the items X1, X2, and X3 may be presented by the presentation unit 18 in order of relevance to the item X. Note that in this embodiment, the items X1, X2, and X3 presented in order of relevance are items determined or estimated not to have the first attribute.

[0069] An information process for restricting presentation of an item having a specific attribute (first attribute) executed by the information processing device 100 according to this embodiment will be described with reference to Fig. 10. Fig. 10 is an example of a flowchart of the information process executed by this embodiment.

[0070] First, the extraction unit 12 extracts a plurality of items (S1001). Each of the plurality of items includes information about the product and information about the label.

[0071] Next, the determination unit 14 determines whether each of the extracted items has the first attribute, does not have the first attribute, or it is unclear whether the item has the first attribute, based on the information about the label (S1002).

[0072] Next, the estimation unit 16 estimates whether or not the item, for which it is unknown whether or not it has the first attribute, has the first attribute, based at least on the information related to each of the extracted multiple items (S1003).

[0073] Next, the presenting unit 18 presents, from among the multiple items, the items determined not to have the first attribute and the items estimated not to have the first attribute to the user (S1004).

[0074] Therefore, in this embodiment, when an item having a specific attribute is included among a plurality of items extracted as candidates for items recommended to the user, the item can be excluded from being presented to the user. In addition, when an item is uncertain as to whether it has a specific attribute, it can be excluded from being presented to the user if it is estimated that the item has the specific attribute. As a result, when presenting recommended items to the user, it is possible to refrain from presenting sensitive items that require caution when viewing to the user. In conventional electronic commerce services, there is a possibility that products that the user does not want to view are included in the recommended items and presented, which may induce a deterioration in the user's impression of the electronic commerce service and cause a decrease in the number of users. By applying the process of limiting the presentation of items according to the embodiment of the present disclosure, it is possible to limit the presentation of products that the user does not want to view, making it possible to provide an attractive electronic commerce service to the user.

[0075] The present embodiment has been described above with reference to specific examples. However, the present disclosure is not limited to these specific examples. Any design modifications made by a person skilled in the art to these specific examples are also included within the scope of the present disclosure as long as they have the features of the present disclosure. The elements of each of the above-mentioned specific examples and their arrangements, conditions, shapes, etc. are not limited to those exemplified and can be changed as appropriate. The combination of each of the elements of each of the above-mentioned specific examples can be changed as appropriate as long as no technical contradiction occurs.

[0076] (Appendix 1) A method for limiting presentation of an item having a first attribute by an information processing device, comprising: extracting a plurality of items, each of the items including information related to a product and information related to a label; determining whether each of the extracted items has a first attribute, does not have the first attribute, or is unclear whether each of the extracted items has the first attribute based on information about the label; estimating whether the item, the presence or absence of which is unknown, has the first attribute based at least on information about each of the extracted plurality of items; presenting to a user, among the plurality of items, items determined to not have the first attribute and items estimated to not have the first attribute; A method for providing the above.

[0077] (Appendix 2) 2. The method of claim 1, wherein the information about the label includes at least one of a report from a user, a score calculated from an image of the product, or a genre to which the product belongs.

[0078] (Appendix 3) As the plurality of items, a plurality of items related to the item being viewed by the user are extracted; presenting to the user, on a page on which items being viewed by the user are displayed, items determined to not have the first attribute and items estimated to not have the first attribute; 3. The method according to claim 1 or 2.

[0079] (Appendix 4) The plurality of items are extracted based on at least one of a browsing history, a purchasing history, and profile information of the user. 4. The method according to any one of claims 1 to 3.

[0080] (Appendix 5) the estimating includes estimating whether or not an item, the presence or absence of which is unknown, has the first attribute based on graph information generated based on the extracted plurality of items; the graph information includes a node corresponding to each of the items and an edge connecting any two nodes; The node includes a feature vector generated from information about the products included in the corresponding item; The edge includes information about the similarity between items corresponding to the nodes connected to both ends of the edge. 5. The method according to any one of claims 1 to 4.

[0081] (Appendix 6) the feature vector includes at least one of a distributed representation generated from text information about the product and a distributed representation generated from image information about the product; The method described in Appendix 5.

[0082] (Appendix 7) The estimation includes converting a distributed representation included in each node of the graph information into a lower dimensional distributed representation, and then estimating whether or not the item, which is unknown whether or not it has the first attribute, has the first attribute. 7. The method according to claim 5 or 6.

[0083] (Appendix 8) The inferring includes applying a community extraction technique or node classification to the graph information to infer whether the item, which is unknown whether it has the first attribute, has the first attribute. 8. A method according to any one of claims 5 to 7.

[0084] (Appendix 9) The information regarding the similarity between the items is calculated based on at least one of browsing histories and purchase histories of a plurality of users. 9. A method according to any one of claims 5 to 8.

[0085] (Appendix 10) The presenting includes not presenting to the user an item determined to have the first attribute and an item estimated to have the first attribute among the plurality of items. 10. The method according to any one of claims 1 to 9.

[0086] (Appendix 11) The presenting includes presenting to the user a predetermined number of items among the items determined not to have the first attribute and the items estimated not to have the first attribute. 11. The method according to any one of claims 1 to 10.

[0087] (Appendix 12) 12. The method of claim 1, wherein the first attribute is an attribute of a product that requires caution when viewed.

[0088] (Appendix 13) An information processing device for limiting presentation of an item having a first attribute, an extractor configured to extract a plurality of items, each of the items including product-related information and label-related information; a determination unit that determines whether each of the extracted items has a first attribute, does not have the first attribute, or is unclear whether each of the extracted items has the first attribute, based on information about the label; an estimation unit that estimates whether or not the item, which is unknown whether or not it has the first attribute, has the first attribute based at least on information about each of the extracted multiple items; a presentation unit that presents to a user, among the plurality of items, an item that is determined not to have the first attribute and an item that is estimated not to have the first attribute; An information processing device comprising:

[0089] (Appendix 14) A program for limiting presentation of an item having a first attribute, comprising: On the computer, extracting a plurality of items, each of the items including information related to a product and information related to a label; determining whether each of the extracted items has a first attribute, does not have the first attribute, or is unclear whether each of the extracted items has the first attribute based on information about the label; estimating whether the item, the presence or absence of which is unknown, has the first attribute based at least on information about each of the extracted plurality of items; presenting to a user, among the plurality of items, items determined to not have the first attribute and items estimated to not have the first attribute; A program for executing the above. [Explanation of symbols]

[0090] 1. Information Processing Systems 10a processor 10b Memory 10c Communication Department 10d Input section 10e Display 12 Extraction part 14 Judgment section 16 Estimation part 18 Presentation section 20 Network 100 Information processing device 200 User terminals 200-1, 200-2, 200-n User terminals 200-1 User terminal 200-2 User terminal DB Database

Claims

1. 1. A method for an information processing device to limit presentation of an item having a first attribute, comprising: extracting a plurality of items, each of the items including information related to a product and information related to a label; determining whether each of the extracted items has a first attribute, does not have a first attribute, or is unclear whether each of the extracted items has a first attribute based on information about the label; estimating whether or not the item, which is unknown whether or not it has the first attribute, has the first attribute based at least on information about each of the extracted multiple items; presenting to a user, among the plurality of items, items determined to not have the first attribute and items estimated to not have the first attribute; A method for providing the above.

2. The method of claim 1 , wherein the information about the label includes at least one of a report from a user, a score calculated from an image of the product, or a genre to which the product belongs.

3. As the plurality of items, a plurality of items related to the item being viewed by the user are extracted; presenting to the user, on a page on which items being viewed by the user are displayed, items determined to not have the first attribute and items estimated to not have the first attribute; The method of claim 1.

4. The plurality of items are extracted based on at least one of a browsing history, a purchasing history, and profile information of the user. The method of claim 1.

5. the estimating includes estimating whether an item, the presence or absence of which is unknown, has the first attribute based on graph information generated based on the extracted plurality of items; the graph information includes a node corresponding to each of the items and an edge connecting any two nodes; The node includes a feature vector generated from information about the products included in the corresponding item; The edge includes information about the similarity between items corresponding to the nodes connected to both ends of the edge. The method of claim 1.

6. the feature vector includes at least one of a distributed representation generated from text information about the product and a distributed representation generated from image information about the product; The method according to claim 5.

7. The estimation includes converting the feature vector included in each node of the graph information into a lower-dimensional feature vector, and then estimating whether or not the item, which is unknown whether or not it has the first attribute, has the first attribute. The method according to claim 5.

8. The inferring includes applying a community extraction technique or node classification to the graph information to infer whether the item, which is unknown whether it has the first attribute, has the first attribute. The method according to claim 5.

9. The information regarding the similarity between the items is calculated based on at least one of browsing histories and purchase histories of a plurality of users. The method according to claim 5.

10. The presenting includes not presenting to the user an item determined to have the first attribute and an item estimated to have the first attribute among the plurality of items. The method of claim 1.

11. The presenting includes presenting to the user a predetermined number of items among the items determined not to have the first attribute and the items estimated not to have the first attribute. The method of claim 1.

12. The method of claim 1 , wherein the first attribute is a sensitive product attribute.

13. An information processing device for limiting presentation of an item having a first attribute, an extractor configured to extract a plurality of items, each of the items including product-related information and label-related information; a determination unit that determines whether each of the extracted items has a first attribute, does not have the first attribute, or is unclear whether each of the extracted items has the first attribute, based on information about the label; an estimation unit that estimates whether or not an item, the presence or absence of which is unknown, has a first attribute based on at least information related to each of the extracted multiple items; a presentation unit that presents to a user, among the plurality of items, an item that is determined not to have the first attribute and an item that is estimated not to have the first attribute; An information processing device comprising:

14. 1. A program for limiting presentation of an item having a first attribute, comprising: On the computer, extracting a plurality of items, each of the items including information related to a product and information related to a label; determining whether each of the extracted items has a first attribute, does not have a first attribute, or is unclear whether each of the extracted items has a first attribute based on information about the labels; estimating whether or not the item, which is unknown whether or not it has the first attribute, has the first attribute based at least on information about each of the extracted multiple items; presenting to a user, among the plurality of items, items determined to not have the first attribute and items estimated to not have the first attribute; A program for executing.

Citation Information

Patent Citations

  • Product selection method and system and storage medium

    CN111738785A

  • Ambiguous post identification method based on keyword retrieval and computer equipment

    CN112800321A

  • Information processing device, information processing method, and information processing program

    JP7133674B1

  • Machine tool having a balancing device

    WO2020161026A1

  • Hospital stay assistance apparatus, system, and method, and computer-readable medium

    WO2022201757A1