JP6356774B2 - Selection device, selection method, and selection program - Google Patents

Description

  The present invention relates to a selection device, a selection method, and a selection program.

  2. Description of the Related Art Conventionally, a technique for presenting search query candidates to a user to assist various web searches by the user is known. As an example of such a technique, a set of keywords and an image are stored in association with each other, and an image associated with a set of keywords including a search query input by the user is displayed as a search query candidate. When an image is selected, a technique for displaying a search result using a set of keywords corresponding to the selected image as a search query is known.

Japanese Patent Laying-Open No. 2006-001406

  However, in the related art, there is a possibility that an appropriate image cannot be displayed as a search query candidate.

  For example, in the prior art, an image registered by an exhibitor is collected as an image related to a product or service related to electronic commerce, and a set of keywords related to the product or service and the collected image are stored in association with each other. However, if the image is not appropriate as an indication of goods or services, the concept may be different between the image displayed as the search query candidate and the search query that is input when the image is selected. There is.

  The present application has been made in view of the above, and an object thereof is to improve the accuracy of selecting images related to products and services.

  The selection device according to the present application includes an acquisition unit that acquires an image associated with a predetermined keyword, and a plurality of models that determine whether the image satisfies a predetermined condition and that satisfy different conditions. And a selection unit for selecting a display image that is an image to be displayed when the predetermined keyword is input from the image acquired by the acquisition unit.

  According to one aspect of the embodiment, it is possible to improve the accuracy of selecting images related to products and services.

FIG. 1 is a diagram illustrating an example of a generation process and a selection process executed by the information distribution apparatus according to the embodiment. FIG. 2 is a diagram illustrating an example of an image displayed on the user terminal by the information distribution apparatus according to the embodiment. FIG. 3 is a diagram illustrating a configuration example of the management server according to the embodiment. FIG. 4 is a diagram illustrating an example of information registered in the search target database according to the embodiment. FIG. 5 is a diagram illustrating a configuration example of the information distribution apparatus according to the embodiment. FIG. 6 is a diagram illustrating an example of information registered in the structured data database according to the embodiment. FIG. 7 is a diagram illustrating an example of information registered in the query candidate database according to the embodiment. FIG. 8 is a diagram illustrating an example of information registered in the model database according to the embodiment. FIG. 9 is a flowchart illustrating an example of a process flow in which the information distribution apparatus according to the embodiment extracts structured data. FIG. 10 is a flowchart illustrating an example of a process flow in which the information distribution apparatus according to the embodiment generates a query candidate. FIG. 11 is a flowchart illustrating an example of a processing flow in which the information distribution apparatus according to the embodiment selects a display image. FIG. 12 is a diagram illustrating an example of a hardware configuration.

  Hereinafter, a mode for implementing a selection device, a selection method, and a selection program according to the present application (hereinafter referred to as “embodiment”) will be described in detail with reference to the drawings. In addition, the selection apparatus, the selection method, and the selection program which concern on this application are not limited by this embodiment. In the following embodiments, the same parts are denoted by the same reference numerals, and redundant description is omitted.

[Embodiment]
[1. Example of information distribution device)
First, an example of generation processing and selection processing executed by the information distribution apparatus will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of a generation process and a selection process executed by the information distribution apparatus according to the embodiment.

  The information distribution apparatus 10 is an information processing apparatus capable of communicating with the management server 100 used by the user U01 and the user terminal 200 used by the user U01 via a predetermined network N such as the Internet. It is realized by a device or a cloud system. Note that the information distribution apparatus 10 may be able to communicate with an arbitrary number of management servers 100 and an arbitrary number of user terminals 200 via the network N. In addition to the management server 100 and the user terminal 200, the information distribution device 10 may be able to communicate with a server device that manages or provides an arbitrary service.

  The management server 100 is an information processing device that provides various services related to electronic commerce, and is realized by a server device, a cloud system, or the like. For example, the management server 100 provides the user U01 with a service in an online shopping mall that sells the provision of arbitrary products and services (hereinafter referred to as “transaction targets”). The management server 100 may be a server that provides services related to auctions.

  The user terminal 200 is a smart device such as a smartphone or a tablet, and is a mobile terminal device that can communicate with an arbitrary server device via a wireless communication network such as 3G (3rd Generation) or LTE (Long Term Evolution). It is. The user terminal 200 may be an information processing apparatus such as a desktop PC (Personal Computer) or a notebook PC as well as a smart device.

For example, the management server 100 distributes various types of information related to a transaction target to the user terminal 200 in response to a request from the user terminal 200, that is, an operation of the user U01. For example, when the management server 100 receives a search query from the user terminal 200, the management server 100 searches for a transaction target having a higher relevance to the search query than a predetermined threshold. Then, the management server 100 searches the content in which the detailed description of the searched transaction object and the link to the web content to be sold are set, and the transaction object content indicating the transaction object photo and price is arranged. It generates as a result and distributes the generated search result to the user terminal 200.

  In the following description, processing executed by the information distribution apparatus 10 in connection with a search for a transaction target will be described. Accordingly, in the following description, products, services, and the like that are managed by the management server 100 are collectively referred to as “search targets”.

[2. About conventional technology]
Here, in the prior art, only the search result corresponding to the search query input by the user and the search result of the genre corresponding to the search query are displayed, and what kind of search query the user inputs I couldn't assist with the search if I didn't know it was good.

  For example, in the prior art, when the user is a long wallet of brand A and wants a wallet made of leather-stitched material, the user can search for the name of the search target, “brand A”, “long wallet” ”,“ Leather ”,“ Embossed ”attributes cannot be recalled, and only some attributes such as“ Brand A ”,“ Long wallet ”are entered as search queries. In addition to the web pages that have been searched, many web pages on which search targets that match the search query are posted are displayed. As a result, in the related art, the user cannot be guided to the web page on which the search target desired by the user is posted. In addition, in the prior art, when a user inputs only a part of a search query, when all related search targets are displayed or when a check box for narrowing down is displayed, search results are displayed. Since it takes time to narrow down, usability deteriorates.

  Therefore, a method of estimating search query candidates based on the input search query and displaying an image indicating the estimated search query candidates is conceivable. For example, in an online shopping mall, there may be a case where classification information indicating an attribute of a search target is given to the search target. In such a case, in the related art, a combination of classification information given to a search target is collected, and corresponding search target images are acquired for all the collected combinations. Further, in the prior art, a combination of classification information including classification information input by a user as a search query is specified as a search query candidate, and images corresponding to all the classification information included in the specified search query candidate are displayed. To do. In the conventional technique, when the user selects an image, a technique can be considered in which all classification information included in search query candidates corresponding to the image is used as a new search query.

  However, in an online shopping mall, classification information having a hierarchical relationship may be assigned to a search target. For this reason, when all combinations of classification information are simply collected, there is a possibility that search query candidates including classification information having no hierarchical relationship may be generated.

  In addition, the search target image is often registered by the search target exhibitor. However, in such an image, an image that appropriately indicates the search target and an image that does not appropriately indicate the search target are mixed. For example, the search target image may be an image of the search target itself, such as an image in which the search target is difficult to see due to the background color, an image including text such as a price or a discount rate, a decorated image, or an image in which a person is reflected. In some cases, an image other than an image indicating a concept is included.

  It is considered that such an image that does not appropriately indicate the search target cannot appropriately indicate the concept indicating the search query candidate such as the concept of the search target and the concept of the classification information given to the search target. For this reason, when such an image is displayed, there is a possibility that the user's search act cannot be appropriately assisted.

  Therefore, the information distribution apparatus 10 generates search query candidates reflecting the hierarchical relationship of the classification information by executing the following generation process, and also generates the search by executing the following selection process. Select an appropriate image that shows the candidate query. Hereinafter, the process executed by the information distribution apparatus 10 will be described separately for the generation process and the selection process.

[3. About generation processing)
First, the generation process executed by the information distribution apparatus 10 will be described. For example, the information distribution apparatus 10 refers to a database that stores classification information having a hierarchical relationship as information indicating the classification of a search target, and sets a classification information set (hereinafter referred to as “structure”) having predetermined classification information as the root of the hierarchical relationship. ")") Is extracted. And the information delivery apparatus 10 produces | generates the candidate of the search query provided when a user inputs predetermined classification information as a search query based on the extracted structured data.

[3-1. (Process to extract a set of classification information)
For example, the information distribution apparatus 10 refers to a database included in the management server 100 and collects classification information registered in the database and having a hierarchical structure. Then, the information distribution apparatus 10 selects one of the collected classification information, sets the selected classification information as the root of the hierarchical relationship, and traces the hierarchical relationship from the predetermined classification information, thereby selecting the selected classification information. To extract structured data that is a set of classification information with the root of the hierarchical relationship.

  For example, in the example illustrated in FIG. 1, the management server 100 stores a search target database 131 in which a search target is associated with a set of classification information that is information indicating a classification of the search target and has a hierarchical relationship. Yes. More specifically, the management server 100 stores, as classification information, category information indicating a category to which the search target belongs and tag information indicating various attributes such as a search target brand, raw material, and color.

  As an example of category information, when the search target is a T-shirt, the management server 100 includes category information “fashion”, category information “tops” subordinate to the category information “fashion”, and category information “tops”. A set of classification information including the category information “T-shirt” belonging to the lower level and including information indicating the hierarchical relationship of each category information is stored. For example, the management server 100 stores a set of classification information such as “fashion> tops> T-shirt”.

  In the following description, conceptual values such as “category # 1”, “child category # 1”, and “grandchild category” are used as the classification information. Here, it is assumed that “child category # 1” is a category belonging to the hierarchy one level lower than “category # 1”, and “grandchild category # 1” is one level lower than “child category # 1”. It is assumed that the category belongs to the hierarchy. The management server 100 may manage the search target using a category that belongs to a lower hierarchy than “grandchild category # 1”.

  As an example of tag information, when the search target is a red T-shirt made of cotton, the management server 100 stores a set of classification information including tag information “red” and tag information “cotton”. . For example, for the search target # 1, the management server 100 stores tag information indicating the color or material of the transaction target such as “red_cotton” in association with the category information individually. Note that the management server 100 may store the tag information as information incorporated in the hierarchical relationship of category information.

  In the following description, conceptual values such as “tag # 1A” and “tag # 2A” are used as tag information. Further, a plurality of tag information indicating common attributes of transaction objects (for example, a plurality of tag information indicating brand names and materials) are set as tag information included in the same tag group. For example, “tag # 1A” and “tag # 1B” are tag information belonging to “tag group # 1”, and “tag # 2A” and “tag # 2B” are tags belonging to “tag group # 2”. It is assumed to be information.

  As described above, when the management server 100 stores the group of classification information having the hierarchical relationship and the search target in association with each other, the information distribution apparatus 10 acquires the classification information having the hierarchical relationship from the management server 100 ( Step S1). That is, the information distribution apparatus 10 acquires the search target database 131. Then, the information distribution apparatus 10 traces the hierarchical relationship, and extracts structured data including certain classification information and classification information or a combination of classification information having the classification information as a root (step) (step S2).

  For example, the information distribution apparatus 10 selects one piece of classification information from the classification information registered in the search target database 131, and identifies a set of classification information including the selected classification information from the search target database 131. For example, when “category # 1” is selected, the information distribution apparatus 10 specifies all the categories belonging to the hierarchy immediately below “category # 1”. As a result, for example, the information distribution apparatus 10 specifies “child category # 1” and “child category # 2”. In addition, the information distribution apparatus 10 extracts tag information associated with the search target together with “category # 1”, and generates a tag group in which tag information indicating common attributes of the transaction target is collected. Then, the information distribution apparatus 10 registers “category # 1” and classification information and a set of classification information with “category # 1” as a root in the structured data database 31 as structured data.

  For example, when “child category # 1” and “child category # 2” are specified as categories having “category # 1” as a root, information distribution apparatus 10 determines “category # 1” and “child category # 1”. ”And“ category # 1 ”and“ child category # 2 ”are registered as structured data. For example, the information distribution apparatus 10 associates “category # 1”, “child category # 1”, and “grandchild category # 1” with a certain search target, and “category # 1” with other search targets. ”,“ Child category # 1 ”, and“ grandchild category # 2 ”are associated with each other, a set of“ category # 1 ”,“ child category # 1 ”, and“ grandchild category # 1 ”is structured data. And a set of “category # 1”, “child category # 1”, and “grandchild category # 2” is registered as structured data. That is, the information distribution apparatus 10 selects certain classification information as a root, and generates structured data including the selected classification information and classification information or a group of classification information belonging to the lower level of the selected classification information.

  Further, for example, the information distribution apparatus 10 determines that “category # 1” when “tag # 1A” or “tag # 1B” indicating a common attribute is associated with the search target together with “category # 1”. And the “tag group # 1” indicating “tag # 1A” and “tag # 1B” are generated. The information distribution apparatus 10 may register information indicating that “tag # 1A” or “tag # 1B” is included in the “tag group # 1” in the structured data database 31, and “tag group # 1”. Information indicating that “tag # 1A” and “tag # 1B” are included in “1” may be registered in another database and managed.

  In addition, the information distribution apparatus 10 selects and selects, as a root, a combination of certain classification information and classification information belonging to a lower level of the classification information, such as “category # 1” and “child category # 1”. Generate structured data including classification information and classification information or a set of classification information belonging to the lower level of the selected classification information. In addition, the information distribution apparatus 10 recursively executes the above-described processing for all classification information and classification information sets registered in the search target database 131, so that all classification information sets having a hierarchical structure are obtained. Generate structured data indicating In other words, the information distribution apparatus 10 traces the hierarchical relationship so that all the structured data including the selected classification information and the classification information or the classification information group belonging to the lower level of the selected classification information are retrieved from the search target database 131. Extract.

  When the user U01 inputs a search query, the structured data extracted in this way can be a search query candidate that can be input subsequently to the search query. Therefore, the information distribution apparatus 10 uses a search query (hereinafter referred to as “invocation query”) that is estimated to be input by the user U01 from the generated structured data, and when the activation query is input. A candidate that is a candidate for a search query provided to the user U01 is generated (step S3).

[3-2. About normalization)
Here, the information distribution apparatus 10 may perform normalization of information when generating structured data. For example, the information distribution apparatus 10 executes normalization processing such as unification of full-width characters and half-width characters, unification of uppercase letters and lowercase letters, etc., on the classification information selected as the root and the text of the classification information included in the structured data. Also good. In addition, the information distribution apparatus 10 may delete additional symbols such as parentheses, spaces, colons, semicolons, midpoints, dashes, columns, and the like. In addition, the information distribution apparatus 10 may delete the range enclosed in parentheses. Further, the information distribution apparatus 10 may exclude classification information including an additional symbol or classification information including a predetermined character string from the extraction source of structured data.

[3-3. (Process to generate search query candidates)
Next, an example of processing for generating query candidates will be described. For example, the classification information that is structured data is considered to correspond to a search query candidate when the root classification information is input as an activation query. Therefore, the information distribution apparatus 10 selects an activation query from the structured data, and when the selected activation query is input, the information distribution apparatus 10 provides the user U01 as a search query candidate (hereinafter collectively referred to as “query candidate”). A set of classification information to be provided is generated.

  Hereinafter, a basic flow of processing in which the information distribution apparatus 10 generates query candidates will be described. For example, the information distribution device 10 refers to the structured data and selects the root classification information or the group of classification information. For example, the information distribution apparatus 10 selects the root classification information “category # 1” and uses the selected classification information as an activation query. Then, the information distribution apparatus 10 extracts structured data including the classification information “category # 1” selected as the activation query as a root, and generates query candidates including the classification information included in the extracted structured data.

  For example, when the structured data including “category # 1” and “child category # 1” exists in the structured data including the classification information “category # 1” as a root, the information distribution apparatus 10 displays “category “# 1” is an activation query, and “category # 1_child category # 1” includes a symbol or character indicating an operator “and” in search processing such as a blank between classification information included in structured data. The added character string is generated as a query candidate.

  Further, for example, when there is structured data including “category # 1” and “tag group # 1”, the information distribution apparatus 10 sets all the tags “tag # 1A” included in “tag group # 1”. ”And“ Tag # 1B ”. Then, the information distribution apparatus 10 uses “category # 1” as an activation query, “category # 1_tag # 1A”, “category # 1_tag # 1B”, and the tags included in the tag group. A query candidate is generated for each combination. Then, the information distribution apparatus 10 registers the activation query and the query candidate in the query candidate database 32 in association with each other.

[3-4. (Processing for selecting structured data as a query candidate)
Here, when query candidates are generated from all structured data extracted from the search target database 131, there is a possibility that query candidates that are unlikely to be selected by the user U01 are included. As a result, query candidates that do not reflect the search intention of the user U01 are provided, and the search for the user U01 may not be appropriately guided. Therefore, the information distribution apparatus 10 selects, from the structured data, structured data that is a source of a query candidate that is highly likely to be selected by the user U01, that is, a process that selects appropriate structured data. Execute.

  For example, the information distribution apparatus 10 generates a query candidate from structured data, and the search target output as a search result when the generated query candidate is input, that is, the generated query candidate is associated as classification information. The search target is specified as a search target corresponding to the structured data. Then, the information distribution apparatus 10 counts the number of times that the specified search target is selected by the user, and selects a pair that satisfies the predetermined number. For example, the information distribution apparatus 10 counts the number of search targets corresponding to the structured data, and selects a set of classification information in which the counted number satisfies a predetermined condition.

  For example, the information distribution apparatus 10 refers to the structured data database 31 and selects one activation query. Further, the information distribution apparatus 10 extracts all structured data corresponding to the selected activation query. And the information delivery apparatus 10 acquires the frequency | count that the user selected the search object corresponding to the extracted structured data from the management server 100 for every extracted structured data. For example, the information distribution apparatus 10 acquires the number of times that the user has selected explanations and photographs for each search target, the detailed description of the search targets, the number of times web content to be sold is distributed, and the like. The information distribution apparatus 10 may collect the number of sales to be searched as the number of times selected by the user.

  Then, the information distribution apparatus 10 selects structured data in which the number of times selected by the user satisfies a predetermined condition. For example, the information distribution apparatus 10 may select a predetermined number of structured data in descending order of the number of times selected by the user, and the structured data whose number of times selected by the user exceeds a predetermined threshold value. Select. That is, the information distribution apparatus 10 narrows down structured data that is a source of search query candidates based on the number of times a user selects a corresponding search target.

  Further, the information distribution apparatus 10 groups the structured data after narrowing down according to the type of classification information one layer below the activation query. More specifically, the information distribution apparatus 10 performs grouping according to whether the classification information one layer below is category information or tag information. Further, the information distribution apparatus 10 performs grouping according to which tag group the tag information belongs to.

  For example, as the structured data after narrowing down, a set of “category # 1” and “child category # 1”, a set of “category # 1” and “child category # 2”, “category # 1” and “tag # 1A” , “Category # 1” and “Tag # 1B”, “Category # 1” and “Tag # 2A”, and “Category # 1” and “Tag # 2B” And In such a case, the information distribution apparatus 10 sets a group of “category # 1” and “child category # 1” and a group of “category # 1” and “child category # 2” corresponding to the category information. Sort out. Further, the information distribution apparatus 10 distributes the set of “category # 1” and “tag # 1A” and the set of “category # 1” and “tag # 1B” to a group corresponding to the tag group # 1. Further, the information distribution apparatus 10 distributes the set of “category # 1” and “tag # 2A” and the set of “category # 1” and “tag # 2B” to a group corresponding to the tag group # 2.

  Note that the information distribution apparatus 10 may distribute query candidates whose category information is category information one level below the activation query to a group corresponding to the type indicated by the category information. That is, the information distribution apparatus 10 may further sort the query candidates whose category information is category information one level below the activation query to the group corresponding to the category information type.

  For example, as classification information one level below the classification information “electric appliances”, classification information indicating household appliances such as “refrigerator” and “washing machine”, and PC-related items such as “desktop PC” and “notebook PC” There may be classification information indicating the device. In such a case, the information distribution device 10 sets the group of the classification information “electric appliance” and the classification information “refrigerator”, the group of the classification information “electric appliance” and the classification information “washing machine”, You may distribute to a corresponding group. In addition, the information distribution apparatus 10 converts a set of the classification information “electric appliance” and the classification information “desktop type PC” and a set of the classification information “electric appliance” and the classification information “node type PC” into the PC related apparatus. May be assigned to the corresponding group.

  Subsequently, the information distribution apparatus 10 aligns the cut points of the structured data distributed to each group. Here, the section is information indicating the commonality of the types of classification information belonging to the hierarchy one level below the activation query for the structured data distributed to the same group. For example, the information distribution apparatus 10 selects the structured data distributed to the same group so that the transaction target corresponding to the structured data distributed to the same group is the same type of product.

  It should be noted that any condition can be adopted for setting the cut end. For example, the information distribution apparatus 10 selects each search target image corresponding to a query candidate generated from structured data distributed to the same group by a selection process described later. The information distribution apparatus 10 deletes the query candidate from the provision target when the search target image cannot be selected for any query candidate, that is, when an appropriate image does not exist. May be. Further, when a predetermined number of search target images cannot be selected for any query candidate, the information distribution apparatus 10 may delete the query candidate from the provision target.

  Subsequently, the information distribution device 10 generates query candidates from the structured data distributed to each group. For example, the information distribution apparatus 10 generates a query candidate such as “category # 1_child category # 1” from a set of “category # 1” and “child category # 1”. Then, the information distribution apparatus 10 indicates a group in which the generated query candidate, the activation query (that is, classification information or a set of classification information that is the root of the query candidate), and structured data that is the basis of the query candidate are distributed. The information is associated and registered in the query candidate database 32.

  Here, the group mentioned above becomes a unit at the time of displaying a query candidate. For example, as a result of the processing described above, the query candidate corresponding to the activation query “category # 1” is any one of a group corresponding to the category information, a group corresponding to the tag group # 1, and a group corresponding to the tag group # 2. It is classified. However, when all of these query candidates are displayed, the user may have difficulty selecting the query candidates.

  Therefore, the information distribution apparatus 10 sets tabs corresponding to each group in the search result, and arranges query candidates corresponding to the corresponding group for each tab. For example, the information distribution apparatus 10 can select any one of a group corresponding to the query information and the group corresponding to the category information, a group corresponding to the tag group # 1 indicating the brand to be traded, and a group corresponding to the tag group # 2 indicating the material to be traded. Correlate with. In such a case, the information distribution apparatus 10 sets tabs such as “category”, “brand”, and “material” corresponding to each group in the search result, as shown in FIG. 2 described later. And the information delivery apparatus 10 arrange | positions the query candidate sorted into the group corresponding to each tab in each tab. As a result, the user U01 can easily select from what viewpoint the search of the search target is performed by selecting the tab.

  The information distribution apparatus 10 includes a character string of predetermined classification information in a character string of classification information belonging to a lower level of the predetermined classification information among classification information included in the classification information or the group of classification information. In this case, only the character string of the classification information included in the classification information belonging to the lower level is set as a search query candidate. For example, it is assumed that the information distribution apparatus 10 specifies the classification information “couch sofa” with respect to the activation query “sofa” as a result of each process described above. In such a case, the information distribution apparatus 10 does not generate the query candidate “sofa_couch sofa”, but generates the query candidate “couch sofa”. Then, the information distribution apparatus 10 registers the activation query “sofa” and the query candidate “couch sofa” in the query candidate database 32 in association with each other.

  Subsequently, the information distribution device 10 is a query candidate and an image indicating the query candidate, which is displayed when the user inputs predetermined classification information, and when the user selects the query candidate, a new query candidate is displayed. The search result is registered in the query candidate database 32 in association with an image displaying a search result as a search query.

  For example, when the information distribution device 10 generates a query candidate “category # 1_child category # 1” for the activation query “category # 1”, an image corresponding to the query candidate “category # 1_child category # 1” is displayed. Obtained from the management server 100. In addition, the information distribution apparatus 10 sets a link to a search result in which the search query is “category # 1_child category # 1” for the acquired image. Then, the information distribution apparatus 10 registers the link-set image in the query candidate database 32 in association with the activation query “category # 1” and the query candidate “category # 1_child category # 1”.

[4. About the selection process)
Here, the image registered in the query candidate database 32 is displayed as an image indicating a query candidate by a distribution process described later. However, if such an image is not properly selected, there is a possibility that the user's search action cannot be properly assisted.

  Therefore, the information distribution apparatus 10 selects an appropriate image indicating the generated search query candidate by executing the following selection process. For example, the information distribution apparatus 10 acquires an image related to a predetermined keyword such as a query candidate. The information distribution apparatus 10 receives a predetermined keyword from the acquired image using a plurality of models that determine whether the image satisfies a predetermined condition and determine whether the image satisfies different conditions. A display image that is an image to be displayed at the time of selection is selected.

[4-1. (Example of selection process)
For example, when a query candidate is generated, the information distribution apparatus 10 acquires an image corresponding to the generated query candidate from the management server 100 (step S4). That is, the information distribution apparatus 10 acquires an image indicating a product or service related to a predetermined keyword. Then, the information distribution apparatus 10 selects a display image to be displayed as a query candidate using a plurality of models that determine different conditions from images corresponding to the query image (step S5). That is, the information distribution apparatus 10 selects an image that is considered appropriate as an image indicating the query candidate from the images searched using the query candidate.

  That is, when a plurality of classification information is included in a query candidate, the information distribution apparatus 10 acquires an image associated with all the classification information included in the query candidate. The information distribution device 10 is an image that is displayed as a query candidate when any classification information (for example, classification information that becomes an activation query) is input as a search query candidate among query candidates. When selected by the user, all the classification information included in the query candidates is selected as an image for a new search query.

  For example, in the example illustrated in FIG. 1, the information distribution apparatus 10 acquires a search target image related to the query candidate from the management server 100. Such an image may include not only a search target photo indicated by a query candidate but also characters such as “half price!” And decorations. Therefore, the information distribution apparatus 10 uses a plurality of models for determining whether the images satisfy different conditions, and does not include characters, decorations, or the like from the acquired images, and uses the concept of the search target indicated by the query candidate. Choose an image that shows appropriately. Then, the information distribution apparatus 10 registers the selected image in the query candidate database 32 as a display image indicating the query candidate.

[4-2. (Example of model)
Here, an image suitable as an image indicating a query candidate indicates the concept of a keyword (that is, a query candidate) that is input as a search query when the image is selected, and an image that does not include other concepts is ideal. It is believed that there is. Therefore, the information distribution apparatus 10 selects, as a display image, an image having the same concept as the query candidate from images acquired from the management server 100 using a plurality of models. Then, when the activation query corresponding to the query candidate is input as a search query, the information distribution apparatus 10 is an image indicating the query candidate for the selected image, and when the user selects the query candidate, Distributes images that display search results as search queries.

  For example, when the classification information included in the query candidate includes classification information indicating a color, the information distribution apparatus 10 desirably displays an image in which a subject having the same color as the color indicated by the classification information is captured. . Therefore, the information distribution apparatus 10 uses the model that determines whether the color indicated by the classification information included in the query candidate and the color of the subject captured in the image are the same as the color indicated by the query candidate. An image in which a colored subject is captured is selected. For example, when the query candidate includes a query corresponding to category information such as “red” or tag information such as “red”, the information distribution apparatus 10 uses a model for determining whether the color of the subject is red or not. An image in which a subject having the same color as that indicated by the candidate is captured is selected.

  For example, when the search target indicated by the query candidate is a shoe, it is preferable to display an image in which two shoes, the right foot and the left foot, are captured rather than displaying an image in which only one shoe is captured. . For example, when the search target indicated by the query candidate is a single item, it is desirable to display an image in which one subject is captured rather than an image in which a plurality of subjects are captured. Therefore, the information distribution apparatus 10 selects an image using a model that determines whether or not a predetermined number of subjects are captured as a model. More specifically, the information distribution apparatus 10 selects an image using a model that determines whether or not the subject is imaged by the number of units to be traded.

  For example, it is desirable that the image displayed as the search query is an image in which the transaction target is easily visible. Therefore, the information distribution apparatus 10 selects a display image using a model for determining whether the background of the image satisfies a predetermined condition or a model for determining the size of the subject in the image as a model. For example, the information distribution apparatus 10 determines whether the background of the image has a predetermined color such as white or gray, or is configured with a single color, or the ratio of the subject in the image exceeds a predetermined threshold. A display image is selected using a model for determining whether or not. Moreover, in order to exclude the concept of the number, the image displayed as the search query is desirably an image in which the transaction target is captured as a single item. Therefore, the information distribution apparatus 10 selects a display image using a model that determines whether a single subject or the like in the image is a single model.

  In addition, it is desirable that an image displayed as a search query does not capture a person in order to show the concept of a transaction target. In addition, when a person or a person's face is included in an image displayed as a search query, the portrait right may be a problem. Therefore, the information distribution apparatus 10 selects a display image using a model for determining whether a person is included in the image as a model. In addition, the information distribution apparatus 10 selects a display image using a model that determines whether a human face is reflected in the image.

  Moreover, in order to show the concept of transaction object, it is desirable that the image displayed as a search query does not include characters and is not decorated. Therefore, the information distribution apparatus 10 selects a display image using a model for determining whether a character is included in the image or a model for determining whether the image is decorated.

  The information distribution apparatus 10 uses a model in which deep learning such as DNN (Deep Neural Network) is performed, or a model in which conditions are learned using an arbitrary learning technique such as SVM (Support Vector Machine). It may be determined whether the image satisfies a predetermined condition. In the following description, the information distribution apparatus 10 is a model that deeply learns the characteristics of an image that satisfies a predetermined condition, and uses a plurality of models that deeply learn the characteristics of an image that satisfy different conditions. An example of the process to select is demonstrated.

[4-3. About model variations)
In addition, the information delivery apparatus 10 may use only a part of the models described above, or may select a display image using another model. For example, when a subject that is not related to a query candidate is captured in the acquired image, it is not possible to appropriately assist the user search. Therefore, the information distribution apparatus 10 may select a display image using a model that determines whether a subject in the image is related to a predetermined keyword, that is, a query candidate, as a model.

  Further, when the query candidate includes classification information indicating an attribute such as the size of a search target, the information distribution apparatus 10 determines whether a subject having the same attribute as the attribute indicated by the classification information is captured in the image. A display image may be selected using a model to be determined.

  In addition, for example, when a product to be searched is a white home appliance, if the background color is white, the search target may be difficult to visually recognize. Therefore, the information distribution apparatus 10 may select a display image using a model that determines whether or not a condition corresponding to a search target attribute associated with a predetermined keyword is satisfied. For example, the display image may be selected using the information distribution device 10 and a model for determining whether or not the search target category related to the query candidate is satisfied and satisfying the condition according to the characteristics of the specified category. . To give a more specific example, the information distribution apparatus 10 uses a color other than white as the background when the search target related to the query candidate is a product mainly in white color such as a refrigerator or a washing machine. A display image may be selected using a model for determining whether or not there is.

  Here, the information distribution apparatus 10 calculates a value indicating the degree to which the acquired image satisfies each condition by using a plurality of models that output values indicating the degree to which the image satisfies a predetermined condition. An image in which each value exceeds a predetermined threshold is selected as a display image. That is, the information distribution apparatus 10 selects a display image using a plurality of models that calculate scores indicating the probability that the image satisfies the condition. More specifically, the information distribution apparatus 10 inputs an image to each model, and determines whether or not the score output from each model exceeds a predetermined threshold value. Then, the information distribution apparatus 10 sets the image as a display image when all the scores exceed a predetermined threshold, and excludes the image from the display image when any score is smaller than the predetermined threshold. To do.

[4-4. (Selected images)
Here, when there are a plurality of query candidates for a certain activation query, the information distribution device 10 provides a display image indicating each query candidate when the activation query is input as a search query. . However, when the appearances of the plurality of provided display images deviate, there is a possibility that the information included in each display image, that is, the query candidate information cannot be appropriately provided to the user.

  Therefore, when selecting a display image for each of a plurality of query candidates including common classification information, the information distribution apparatus 10 may select images having similar appearances for each of the query candidates. For example, the information distribution apparatus 10 may select, for each of the query candidates, an image in which the transaction target is captured in the same form, such as an image in which a commodity to be traded is laid flat.

  Further, for example, the information distribution apparatus 10 may change the model to be used according to the search target category indicated by the query candidate, and changes the threshold value of the score calculated by the model for the image to be used as the display image. May be. To give a more specific example, when the search target indicated by the query target is clothing such as a jacket, the information distribution apparatus 10 is an image in which a person who takes a similar pose wears clothing to be traded. May be selected.

  That is, if the information distribution apparatus 10 can select an image that appropriately provides the user with the concept of the search target indicated by the query candidate, the information distribution apparatus 10 uses a model that determines an arbitrary condition to display the display image. You may make a selection.

[5. About distribution processing)
Subsequently, a distribution process for distributing an image indicating a query candidate will be described. First, the user terminal 200 used by the user U01 transmits the search query input by the user U01 to the management server 100 (step S6). In such a case, the management server 100 delivers the search result corresponding to the search query input by the user U01 to the user terminal 200 (step S7).

  Here, in the search result distributed by the management server 100, the information distribution device 10 is requested to distribute the query candidate, and the image distributed as the query candidate from the information distribution device 10 is arranged and displayed in the search result. Instructions to do so are included. Such an instruction can be realized by a technique such as Java Script (registered trademark). Therefore, the user terminal 200 requests a query candidate distribution request to the information distribution apparatus 10 together with the search query input by the user U01 (step S8).

  In such a case, the information distribution apparatus 10 selects an activation query that matches or relates to the search query from the activation queries registered in the query candidate database 32. And the information delivery apparatus 10 delivers the image matched with the selected activation query, ie, the image which shows the query candidate corresponding to the selected activation query, to the user terminal 200 (step S9). As a result, the user terminal 200 can display an image indicating a query candidate corresponding to the search query, that is, an image indicating a search query that is more likely to be input when the search query is input. (Step S10).

[5-1. (About an example of distribution processing)
Here, a link to a search result having a query candidate indicated by the image as a new search query is set in the image indicating the query candidate. Therefore, when the user U01 selects an image indicating a query candidate, the user terminal 200 requests the management server 100 for a search result with the query candidate indicated by the selected image as a new search query. In such a case, the management server 100 requests the information distribution apparatus 10 to distribute a new query candidate corresponding to the new search query, which is a search result for the new search query. A search result including an instruction to arrange and display an image distributed as a query candidate in the search result is distributed. As a result, the user terminal 200 can assist the search of the search target by the user U01 with the image.

  For example, FIG. 2 is a diagram illustrating an example of an image displayed on the user terminal by the information distribution apparatus according to the embodiment. In the following description, the group “category”, “brand”, and “material” are set for the query candidate “fashion”, and the query candidates “fashion_tops”, “fashion” are set for the group “category”. “_Pants” and “Fashion_Skirt” are assigned.

  For example, when the query candidate “fashion” is input, the information distribution apparatus 10 sets tabs corresponding to the groups “category”, “brand”, and “material”, respectively. Further, when displaying the query candidates in a state where the tab corresponding to the group “category” is first selected, the information distribution apparatus 10 displays the query candidates “fashion_tops”, “fashion__” distributed to the group “category”. “Pants” and “Fashion_Skirt” are specified.

  Then, the information distribution apparatus 10 distributes the images selected for the query candidates “fashion_tops”, “fashion_pants”, and “fashion_skirt” to the user terminal 200. As a result, as shown in the first state in FIG. 2, the information distribution apparatus 10 uses the query candidates “fashion_tops”, “fashion_pants”, and “fashion_skirt” as search results for the search query “fashion”. Are displayed on the user terminal 200. Further, the information distribution apparatus 10 displays, together with each image, content in which text indicating “tops”, “pants”, and “skirt”, which is the lowest classification information among the classification information included in the query candidates, is arranged. .

  When the user U01 selects an image indicating the query candidate “fashion_tops”, the user terminal 200 acquires a search result using the query candidate “fashion_tops” as a new search query. Further, the information distribution apparatus 10 may use, for example, query candidates “fashion_tops_T-shirt”, “fashion_tops_blouson”, “fashion_tops_coat” as query candidates having the query candidate “fashion_tops” as an activation query. Is specified.

  Then, the information distribution apparatus 10 distributes the images selected for the query candidates “fashion_tops_T-shirt”, “fashion_tops_blouson”, and “fashion_tops_coat” to the user terminal 200. As a result, as shown in the second state in FIG. 2, the information distribution apparatus 10 uses the query candidates “fashion_tops_T-shirt” and “fashion_tops_blouson” as search results of the search query “fashion_tops”. , Along with an image showing each of “Fashion_Tops_Coat”, text indicating “T-shirt”, “Blouson”, and “Coat”, which is the lowest classification information among the classification information included in the query candidates, is arranged Display content.

  When the user U01 selects an image indicating the query candidate “fashion_tops_T-shirt”, the user terminal 200 acquires a search result using the query candidate “fashion_tops_T-shirt” as a new search query. . On the other hand, the information distribution device 10 generates the query candidates “fashion_tops_T-shirt_black”, “fashion_tops_T-shirt_white”, “fashion” generated for the activation query “fashion_tops_T-shirt” by the generation process described above. _Tops_T-shirt_Red "is specified.

  Then, the information distribution apparatus 10 selects the images selected for the query candidates “fashion_tops_T-shirt_black”, “fashion_tops_T-shirt_white”, and “fashion_tops_T-shirt_red” by the selection process described above. Delivered to the user terminal 200. As a result, as shown in the third state in FIG. 2, the information distribution apparatus 10 uses the query candidates “fashion_tops_T-shirt_black”, “fashion__” as search results for the search query “fashion_tops_T-shirt”. “Black”, “White”, and “Red” which are the lowest classification information among the classification information included in the query candidate, together with images indicating each of “Tops_T-shirt_white” and “Fashion_Tops_T-shirt_red”. Displays the content where the text indicating is placed.

[5-2. (Variation of distribution processing)
Here, when distributing a plurality of images indicating query candidates, the information distribution device 10 may set the order in which the images are arranged using an arbitrary listing technique. For example, the information distribution apparatus 10 holds the number of times selected by the user as a selection history for each image indicating a query candidate. Then, when distributing the image indicating the query candidate, the information distribution apparatus 10 may instruct the user terminal 200 to arrange and display the distribution target images in the order corresponding to the number of times the user has selected. Good. For example, the information distribution apparatus 10 may instruct the user terminal 200 to arrange the images from the left side of the screen in descending order of the number of times selected by the user.

  Further, the information distribution apparatus 10 may target only images whose number of selections by the user exceeds a predetermined threshold. For example, the information distribution apparatus 10 specifies query candidates generated based on the same cut at predetermined time intervals, specifies the number of times an image indicating each specified query candidate is selected, and the specified number of times is predetermined. When the value is lower than the threshold, an image indicating such a query candidate may be excluded from the distribution target.

[6. Configuration of management server 100]
Hereinafter, an example of the functional configuration of the management server 100 described above will be described. FIG. 3 is a diagram illustrating a configuration example of the management server according to the embodiment. As illustrated in FIG. 3, the management server 100 includes a communication unit 120, a storage unit 130, and a control unit 140.

  The communication unit 120 is realized by, for example, a NIC (Network Interface Card). The communication unit 120 is connected to the network N in a wired or wireless manner, and transmits and receives information to and from the information distribution apparatus 10 and the user terminal 200.

  The storage unit 130 is realized by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 130 also stores a search target database 131.

  In the search target database 131, search target classification information is registered. For example, FIG. 4 is a diagram illustrating an example of information registered in the search target database according to the embodiment. In the example illustrated in FIG. 4, items such as “search target”, “image”, and “classification information” are registered in the search target database 131. In addition to the information shown in FIG. 4, arbitrary information related to the search target is registered in the search target database 131.

  Here, the “search target” is an identifier indicating a transaction target such as a product or service to be traded. The “image” is a search target image indicated by the associated “search target”. The “classification information” is classification information indicating a category, an attribute, and the like of the associated “search target”, and is classification information having a hierarchical relationship. For example, the search target database 131 stores category information indicating the category to which the search target belongs and tag information indicating the attribute of the search target as the classification information.

  For example, in the example illustrated in FIG. 4, information such as a search target “search target # 1”, an image “image # 1”, and classification information “fashion> tops” is registered. In such information, the search target image indicated by the search target “search target # 1” is the image “image # 1”, and the search target classification indicated by the search target “search target # 1” is the classification information “fashion>. Indicates that the category is “Tops”. In the example shown in FIG. 4, conceptual values such as “search target # 1” and “image # 1” are described. However, actually, a character string of an identifier indicating the search target, image data, and the like are registered. . In addition to the information shown in FIG. 4, the search target database 131 includes various types of search targets such as the number of times content related to each search target has been selected, the number of times the content has been browsed, and the number of times the search target has been purchased. It is assumed that information is registered.

  Returning to FIG. 3, the description will be continued. The control unit 140 is a controller, and various programs stored in a storage device inside the information distribution apparatus 10 are stored in a RAM or the like by a processor such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). Is implemented as a work area. The control unit 140 is a controller, and may be realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

  As illustrated in FIG. 3, the control unit 140 includes a reception unit 141, a search unit 142, and a search result distribution unit 143. The accepting unit 141 accepts a search query from the user terminal 200. For example, the reception unit 141 distributes web content for receiving a search query to be searched to the user terminal 200 and receives a search query input by the user U01 from the user terminal 200.

  The search unit 142 searches for a search target corresponding to the search query. For example, the search unit 142 searches the search target database 131 for a search target corresponding to the search query received by the receiving unit 141. And the search part 142 produces | generates the search result which is the web content which put in order the content containing the image of the search object, description, etc. corresponding to a search query.

  Here, a predetermined area is set in the search result, and an instruction for causing the user terminal 200 to execute processing for arranging the query candidate in such a predetermined area is included. When accepting such distribution of search results, the user terminal 200 transmits a search query to the information distribution device 10 and displays the query candidates received from the information distribution device 10 in a predetermined area. Note that the search unit 142 may generate search results and instructions for arranging a plurality of images that are query candidates in a predetermined area so as to be scrollable or displaying a carousel. Moreover, the search part 142 may change the order of the content of search object using arbitrary listing techniques, and may arrange | position the advertisement selected using arbitrary matching techniques in a search result.

  The search result distribution unit 143 distributes the search result to the user terminal 200. For example, when the search result distribution unit 143 receives the search result generated by the search unit 142, the search result distribution unit 143 distributes the received search result to the user terminal 200.

[7. Configuration of Information Distribution Device 10]
Hereinafter, an example of the functional configuration of the information distribution apparatus 10 described above will be described. FIG. 5 is a diagram illustrating a configuration example of the information distribution apparatus according to the embodiment. As illustrated in FIG. 5, the information distribution apparatus 10 includes a communication unit 20, a storage unit 30, and a control unit 40.

  The communication unit 20 is realized by a NIC or the like, for example. The communication unit 20 is connected to the network N by wire or wirelessly and transmits / receives information to / from the management server 100 and the user terminal 200.

  The storage unit 30 is realized by, for example, a semiconductor memory element such as a RAM or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 30 also stores a structured data database 31, a query candidate database 32, and a model database 33.

  In the structured data database 31, structured data is registered in association with each other. For example, FIG. 6 is a diagram illustrating an example of information registered in the structured data database according to the embodiment. In the example shown in FIG. 6, the structured data database 31 registers a set of classification information serving as a reference for structured data, that is, classification information that is the root and classification information that is subordinate to the classification information that is the root. Has been.

  For example, in the example illustrated in FIG. 6, structured data in which the root classification information “category # 1” and the subordinate classification information “child category # 1” are associated is registered in the structured data database 31. Has been. In the structured data database 31, the root classification information “category # 1” and the subordinate classification information “tag group # 1 (tag # 1A, tag # 1B...)” Are associated with each other. Structured data is registered. Here, the classification information “tag group # 1 (tag # 1A, tag # 1B...)” Includes the tag information “tag # 1A” and “tag # 1B” in the tag information “tag group # 1”. This information indicates that the Note that the information distribution apparatus 10 separately manages a database in which “tag group # 1” and tag information “tag # 1A” and “tag # 1B” are associated with each other, for example. It may be stored that tag information “tag # 1A” and “tag # 1B” are included in “1”.

  Returning to FIG. 5, the description will be continued. In the query candidate database 32, images indicating classification information sets and classification information sets that are query candidates are registered. For example, FIG. 7 is a diagram illustrating an example of information registered in the query candidate database according to the embodiment. In the example illustrated in FIG. 7, items such as “invocation query”, “group”, “query candidate”, and “display image” are registered in the query candidate database 32. In addition to the information shown in FIG. 7, arbitrary information such as the number of times a display image displayed as a query candidate is selected by the user may be registered in the query candidate database 32.

  Here, the “group” is information for identifying a group of associated query candidates, and more specifically, a group in which structured data that is a source of a query candidate is classified. In other words, each query candidate associated with the same “group” is a query candidate generated from structured data that is classification information of a type with common classification information lower than the classification information that is the root.

  In addition, the “query candidate” is a search query provided to the user when the classification information that becomes the associated “execution query” or a search query related to such classification information is input. Is a candidate. The “display image” is an image showing the associated “query candidate”, and a link to a search result having the “query candidate” associated when selected as a new search query is set. It is an image.

  For example, in the example illustrated in FIG. 7, the groups “group # 1” and “group # 2” are associated with the activation query “category # 1”. The group “group # 1” includes query candidates “category # 1_child category # 1...”, “Category # 1_child category # 2...”, Display images “image # 1”, “image # 1”. 2 ”is associated. Further, query candidates “category # 1_tag # 1A...” And “category # 1_tag # 1B...” Are associated with the group “group # 2”.

  When the activation query “category # 1” is input, different information is displayed for each group “group # 1” and “group # 2”. Indicates that each sorted query candidate is displayed. For example, the tab corresponding to the group “group # 1” includes the display image “image # 1” corresponding to the query candidate “category # 1_child category # 1...” And the query candidate “category # 1_child category #”. The display image “image # 2” corresponding to “2.

  Returning to FIG. 5, the description will be continued. In the model database 33, models used when selecting a display image, that is, a plurality of models for determining whether the images satisfy different conditions are registered. For example, FIG. 8 is a diagram illustrating an example of information registered in the model database according to the embodiment. In the example illustrated in FIG. 8, items such as “model” and “model data” are registered in the model database 33. Here, “model” is information indicating how the model is to be determined, and “model data” is data of the associated model.

  For example, in the example illustrated in FIG. 8, the model database 33 includes a “background determination model” that determines whether the background satisfies a predetermined condition, and a “number determination model” that determines whether a predetermined number of subjects are captured. “Person determination model” for determining whether a person is captured, “Character determination model” for determining whether a character is included in an image, “Decoration determination model” for determining whether a decoration is included in an image, an image "Face determination model" that determines whether a person's face is included in the image, "Subject area determination model" that determines the area occupied by the subject in the image, whether the subject is related to each classification information included in the query candidate In addition, a “subject condition determination model” for determining whether the subject has the same attribute as the classification information included in the query candidate, such as whether the subject has a predetermined color, or the like is registered. In the model database 33, model data “model data # 1” to “model data # 8” of each model is registered.

  In the example shown in FIG. 8, a conceptual value such as “model data # 1” is described, but in reality, data constituting the model is registered in the model database 33. For example, when the model is configured by a neural network, information indicating the nodes of the neural network, data indicating the connection relationship between the nodes, the connection coefficient between the nodes, and the like are registered as model data.

  Returning to FIG. 5, the description will be continued. The control unit 40 is a controller, and is realized, for example, by executing various programs stored in a storage device inside the information distribution apparatus 10 using a RAM or the like as a work area by a processor such as a CPU or MPU. The control unit 40 is a controller, and may be realized by an integrated circuit such as an ASIC, for example.

  As illustrated in FIG. 5, the control unit 40 includes a collection unit 41, an extraction unit 42, a generation unit 43, an acquisition unit 44, a selection unit 45, and a provision unit 46. The collection unit 41, the extraction unit 42, and the generation unit 43 execute the generation process for generating the query candidate described above, and the acquisition unit 44 and the selection unit 45 execute the selection process for selecting the display image described above. . Also, the providing unit 46 executes the above-described distribution process.

  The collection unit 41 is information indicating the classification information of the search target, and collects the classification information with reference to the search target database 131 that stores the classification information having a hierarchical relationship and the search target in association with each other. For example, the collection unit 41 accesses the management server 100 at a predetermined timing and collects classification information included in the search target database 131 of the management server 100. That is, the collection unit 41 collects classification information from the search target database 131 that stores category information indicating the category to which the search target belongs and tag information indicating the attribute of the search target as the classification information.

  The extraction unit 42 extracts a set of classification information having predetermined classification information as the root of the hierarchical relationship, that is, structured data. More specifically, the extraction unit 42 selects classification information registered in the search target database 131 as predetermined classification information, and sets classification information or a group of classification information with the selected classification information as the root of the hierarchical relationship. Are extracted by following the hierarchical relationship from predetermined classification information.

  For example, the extraction unit 42 selects, from the classification information registered in the search target database 131, a group of classification information not selected as a root or classification information having a hierarchical structure as an activation query. Here, when selecting a group of classification information as a root, the extraction unit 42 selects a group of certain classification information and classification information belonging to a lower level of the classification information as an activation query. Then, the extraction unit 42 identifies the classification information belonging to the lower level of the activation query, and extracts the activation query and the selected classification information as structured data.

  For example, the extraction unit 42 identifies the classification information that matches the activation query from the classification information registered in the search target database 131, and extracts all the classification information that is one level lower than the identified classification information. . Here, when the classification information is a tag group, all tag information belonging to the tag group is also specified. Then, the extraction unit 42 uses the combination of the activation query and each of the extracted classification information as hierarchical data.

  In this way, the extraction unit 42 traces the hierarchical relationship based on the activation query step by step, and extracts all of the activation query and the classification information belonging to the lower level as structured data. Then, the extraction unit 42 registers the structured data in the structured data database 31. Further, the extraction unit 42 recursively executes such processing, and extracts a set of all classification information having a hierarchical structure as structured data.

  Note that the extraction unit 42 may execute various normalization processes together when extracting structured data. Further, the extraction unit 42 uses, as structured data, an identifier that identifies each piece of classification information (for example, “category 133” or the like, an identifier that indicates which category information is category information among the category information, or “tag 230”). Etc.) may be registered in the structured data database 31.

  The generation unit 43 generates query candidates that are provided when the user inputs an activation query as a search query based on the extracted structured data. More specifically, the generation unit 43 generates query candidates from the extracted structured data.

  Here, the generation unit 43 selects structured data satisfying a predetermined condition from the structured data, and generates search query candidates based on the selected structured data. For example, the generation unit 43 counts the number of search targets associated with the extracted structured data selected by the user, and the counted number satisfies a predetermined condition among the set of structured data. The structured data is selected, and query candidates are generated based on the selected structured data. In addition, for example, the generation unit 43 counts the number of search targets associated with the structured data, and selects structured data in which the counted number satisfies a predetermined condition from the structured data. Query candidates are generated based on the structured data.

  Further, the generation unit 43 distributes the structured data to a plurality of groups serving as units for providing search query candidates to the user based on the commonality of the classification information belonging to the lower hierarchy of the activation query. For example, the still picture unit 43 uses the tag information belonging to which tag group whether the category information included in the structured data is category information, tag information, or the category information that belongs to the lower hierarchy of the dynamic query. Depending on whether or not there is, the structured data is distributed to a plurality of groups.

  Subsequently, the generation unit 43 generates a query candidate from the structured data distributed to each group. For example, the generation unit 43 generates a character string obtained by combining character strings of classification information included in structured data. Here, when the character string of the classification information that becomes the activation query is included in the character string of the classification information that belongs to the lower level of the activation query among the classification information included in the structured data, the generation unit 43 Only the character string of the classification information included in the digitized data is used as a query candidate. Then, the generation unit 43 registers the activation query, the group to which the structured data including the activation query is allocated, and the query candidate generated from the structured data in association with each other in the query candidate database 32.

  For example, the generation unit 43 refers to the structured data database 31 and selects the root classification information “category # 1” as the activation query. In such a case, the generation unit 43 stores “child category # 1”, “child category # 2”, “tag group # 1”, and “tag group # 2” associated with the classification information “category # 1”. Extract. The generation unit 43 also includes “tag # 1A”, “tag # 1B” included in “tag group # 1”, and “tag # 2A” and “tag # 2B” included in “tag group # 2”. To extract.

  As a result, the generation unit 43 extracts the extracted “child category # 1”, “child category # 2”, “tag group # 1”, “tag group # 2”, “tag” for the classification information “category # 1”. “# 1A”, “tag # 1B”, “tag # 2A”, and “tag # 2B” are identified as structured data. Subsequently, the generation unit 43 groups the specified structured data based on the type of classification information one level lower than the activation query.

  For example, “child category # 1” and “child category # 2” are category information. Therefore, the generation unit 43 sets the structured data in which the extracted “child category # 1” and “child category # 2” are respectively associated with the classification information “category # 1” as “group #” corresponding to the category information. 1 ”. For example, “tag # 1A” and “tag # 1B” are tag information included in the tag group # 1. Therefore, the generation unit 43 sets the structured data in which the extracted “tag # 1A” and “tag # 1B” are associated with the classification information “category # 1” in the “group # 1” corresponding to the tag group # 1. 2 ”. For example, “tag # 2A” and “tag # 2B” are tag information included in the tag group # 2. Therefore, the generation unit 43 sets the structured data in which the extracted “tag # 2A” and “tag # 2B” are respectively associated with the classification information “category # 1” to the “group # 2” corresponding to the tag group # 2. 3 ”.

  Further, the generation unit 43 selects structured data satisfying a predetermined condition from the specified structured data. For example, for each extracted structured data, the generation unit 43 inquires the management server 100 of the number of search targets associated with the classification information including the structured data selected by the user. Then, the generation unit 43 queries the structured data in which the number of selected search targets associated with the classification information including the structured data is equal to or less than a predetermined threshold among the extracted structured data. Exclude from candidate generation.

  Here, the generation unit 43 determines whether or not the structured data classified into each group satisfies a predetermined condition regarding the cut end, and excludes the structured data determined to be not satisfied from the generation targets of the query candidates. For example, the generation unit 43 excludes the structured data from the generation targets of the query candidates when the selection unit 45 described later does not select the transaction target image corresponding to the structured data classified into each group. . It should be noted that any condition can be adopted as the predetermined condition regarding such a cut as long as it is related to the commonality of structured data distributed within the same group.

  And the production | generation part 43 produces | generates a query candidate from the structured data classified into each group. For example, the generation unit 43 generates, as a query candidate, text in which texts of classification information included in structured data are arranged. Further, when the text of the activation query is included in the text of the lower classification information, the generation unit 43 generates a query candidate using only the text of the lower classification information. Then, the generation unit 43 associates the activation query, the generated query candidate, and the group to which the structured data that is the basis of the query candidate is distributed, that is, the group corresponding to the query candidate and registers them in the query candidate database 32. To do.

  The acquisition unit 44 acquires an image associated with a predetermined keyword. More specifically, the acquisition unit 44 acquires an image indicating a product or service associated with a predetermined keyword. Further, the acquisition unit 44 acquires an image related to a set of keywords including a plurality of keywords as the predetermined keyword.

  For example, the acquisition unit 44 refers to the query candidate database 32 and acquires query candidates that are not registered in association with display images, that is, a plurality of pieces of classification information. Then, the acquisition unit 44 acquires search target images corresponding to the acquired plurality of classification information. For example, the acquisition unit 44 transmits the acquired plurality of classification information to the management server 100 as a search query, and acquires a transaction target image corresponding to the search query.

  The selection unit 45 is a model that determines whether an image satisfies a predetermined condition and uses a plurality of models that determine whether the image satisfies different conditions, and a predetermined keyword, that is, a query candidate is obtained from an acquired image. A display image which is an image to be displayed when input is selected. Specifically, the selection unit 45 selects, as a display image, an image having the same concept as the query candidate from the acquired images using a plurality of models. Specifically, the selection unit 45 is an image that is displayed as a query candidate when classification information that becomes an activation query is input as a search query candidate among classification information included in the query candidate as a display image. Thus, an image in which all classification information included in the query candidate is selected as a new search query when selected by the user is selected.

  For example, the selection unit 45 uses, as models, a plurality of models that output values indicating the degree to which the image satisfies a predetermined condition, that is, the models registered in the model database 33, and the acquired image is displayed in each condition. Each value indicating the degree of satisfaction is calculated. Then, the selection unit 45 selects an image in which each calculated value exceeds a predetermined threshold as a display image, and registers the selected display image in the query candidate database 32 as a display image indicating a query candidate. In other words, the selection unit 45 is an image showing a query candidate and a query candidate, which is displayed when the user inputs an activation query, and when the user selects the search, uses the query candidate as a new search query. The display image on which the result is displayed is registered in the query candidate database 32 in association with the display image.

  Note that the selection unit 45 may select a display image from the acquired images using a model that determines whether the condition according to the search target attribute related to the query candidate is satisfied as a model. For example, the selection unit 45 may change the threshold value of the value output by each model for the image selected as the display image, according to the search target category related to the query candidate. For example, when the search target related to the query candidate belongs to a category with a high probability that the main body is white, such as a refrigerator, the selection unit 45 determines the value output by the model that determines whether the background is white. The threshold value may be set to “0”.

  The selection unit 45 may select a plurality of images having similar appearances when selecting a display image for each of a plurality of query candidates having commonality. For example, when selecting a display image for a plurality of query candidates associated with the same section, the selection unit 45 uses a model registered in the model database 33 and uses a plurality of images as display image candidates. Is selected for each query candidate. Then, the selection unit 45 compares the appearances of the plurality of images selected for each query candidate, and selects similar images as display images of the query candidates based on the comparison result.

  The providing unit 46 arranges and provides the query candidates in the order corresponding to the number of selections by the user, together with the search result that uses the activation query as the search query. More specifically, the providing unit 46 selects a query candidate having a number of selections by the user that exceeds a predetermined threshold among the query candidates according to the number of selections by the user, along with the search result using the activation query as a search query. Provide in order.

  For example, when the search result of a certain search query is received from the management server 100, the user terminal 200 transmits the search query to the information distribution apparatus 10 according to the instruction received from the management server 100. In such a case, the providing unit 46 identifies an activation query that matches the search query or an activation query that is highly relevant to the search query from the query candidate database 32. In other words, the providing unit 46 specifies the activation query corresponding to the search query in consideration of the query fluctuation.

  Subsequently, the providing unit 46 sets a tab for each group of query candidates associated with the specified activation query, and arranges display images of the respective query candidates in the corresponding group tabs. For example, for the display image associated with the selected query candidate, the providing unit 46 sets a link to a search result using the selected query candidate as a new search query, and uses the display image with the link as a tab. Deploy. Then, the providing unit 46 distributes information on each tab to the user terminal 200.

  Here, the providing unit 46 collects the number of times each display image has been selected by the user. Then, when distributing the display image, the providing unit 46 instructs the user terminal 200 to display each display image in the order corresponding to the number of times selected so far. Further, the providing unit 46 refers to the number of times each display image is selected at a predetermined time interval, and excludes display images in which the selected number is less than a predetermined threshold from the distribution targets.

[8. Example of flow of processing executed by information distribution apparatus]
Next, the flow of processing executed by the information distribution apparatus 10 will be described with reference to FIGS. First, an example of processing for extracting structured data will be described with reference to FIG. FIG. 9 is a flowchart illustrating an example of a process flow in which the information distribution apparatus according to the embodiment extracts structured data.

  For example, the information distribution apparatus 10 refers to the search target database 131 stored in the management server 100 (step S101), and selects root classification information from the registered classification information (step S102). Then, the information distribution device 10 traces the hierarchical relationship of the classification information, and the structured information corresponding to the root classification information is a set of the classification information that is the root and the classification information that belongs to the lower level of the classification information. Obtained as data (step S103). Then, the information distribution apparatus 10 registers the structured data in the structured data database 31 (step S104) and ends the process.

  Next, an example of a process flow for generating a query candidate among the processes executed by the information distribution apparatus 10 will be described. FIG. 10 is a flowchart illustrating an example of a process flow in which the information distribution apparatus according to the embodiment generates a query candidate. For example, the information distribution apparatus 10 selects one of the root classification information as the activation query from the structured data database 31 (step S201), and the appropriate structured data among the structured data corresponding to the selected activation query. Is selected (step S202). For example, the information distribution apparatus 10 selects structured data in which the number of corresponding search targets satisfies a predetermined condition or structured data having an appropriate image as a transaction target image. Then, the information distribution apparatus 10 selects structured data that is a source of query candidates from the selected structured data (step S203).

  Here, the information distribution apparatus 10 determines whether the text of the classification information that becomes the activation query is included in the classification information that is one level lower than the classification information that becomes the activation query among the classification information included in the selected structured data. It is determined whether or not (step S204). If the information distribution apparatus 10 determines that it is included (step S204: Yes), it sets a query candidate that does not include the activation query (step S205), and determines that it is not included (step S205). S204: No), query candidates including the activation query are set (step S206).

  The information distribution apparatus 10 executes a selection process for selecting a display image corresponding to the query candidate (step S207), and associates the selected display image with the query candidate and registers them in the query candidate database 32 (step S208). ), The process is terminated.

  Next, an example of a process flow for selecting a display image among the processes executed by the information distribution apparatus 10 will be described. FIG. 11 is a flowchart illustrating an example of a processing flow in which the information distribution apparatus according to the embodiment selects a display image. Note that the processing in steps S301 to S303 shown in FIG. 11 corresponds to the processing in step S207 shown in FIG.

  For example, the information distribution apparatus 10 acquires a transaction target image related to the query candidate (step S301). The information distribution apparatus 10 selects a display image from the acquired images using a plurality of models for determining different conditions (step S302), sets the display image as a registration target (step S303), and ends the process. .

[9. (Modification)
In the above, an example of generation processing and selection processing by the information distribution apparatus 10 has been described. However, the embodiment is not limited to this. Hereinafter, variations of the generation process and the selection process executed by the information distribution apparatus 10 will be described.

[9-1. About search processing)
In the example described above, the information distribution apparatus 10 selects a display image using a plurality of models that determine different conditions in order to select an appropriate image as an image indicating a query candidate. However, the embodiment is not limited to this. That is, the above-described selection process can be applied when selecting an arbitrary image other than the process of selecting a query candidate image. For example, the information distribution apparatus 10 may acquire an image related to the search query in the image search, and select an image to be included in the search result from the acquired images using a plurality of models that determine different conditions. .

[9-2. (Search target)
In the example described above, the information distribution apparatus 10 sets a transaction target that is an object of electronic commerce as a search target. However, the embodiment is not limited to this. The information distribution apparatus 10 is a target to be searched using a search query input by the user, such as web content in wiki format, and any information may be used as long as classification information having a hierarchical structure is associated with the information distribution apparatus 10. The target can be a search target.

[9-3. Device configuration〕
The information distribution device 10 may be connected to an arbitrary number of user terminals 200 so as to be communicable. The information distribution apparatus 10 may be realized by a front-end server that exchanges information with the user terminal 200 and a back-end server that selects an input item. In such a case, the providing unit 46 shown in FIG. 5 is arranged in the front-end server, and the collecting unit 41, the extracting unit 42, the generating unit 43, the obtaining unit 44, and the selecting unit 45 are arranged in the back-end server. Is done. The information distribution apparatus 10 includes, for example, a collection unit 41, an extraction unit 42, and a generation unit 43. The information distribution device 10 includes a generation device that executes the generation process described above, an acquisition unit 44, and a selection unit 45. It may be realized by a selection device that performs selection processing and a distribution device that includes the providing unit 46 and performs the above-described distribution processing operating in cooperation.

[9-4. Others]
In addition, among the processes described in the above embodiment, all or part of the processes described as being automatically performed can be performed manually, and conversely, the processes described as being performed manually. All or a part of the above can be automatically performed by a known method. In addition, the processing procedures, specific names, and information including various data and parameters shown in the document and drawings can be arbitrarily changed unless otherwise specified. For example, the various types of information illustrated in each drawing is not limited to the illustrated information.

  Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution / integration of each device is not limited to that shown in the figure, and all or a part thereof may be functionally or physically distributed or arbitrarily distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured.

  In addition, the above-described embodiments can be appropriately combined within a range in which processing contents do not contradict each other.

[9-5. program〕
In addition, the information distribution apparatus 10 according to the above-described embodiment is realized by a computer 1000 configured as shown in FIG. 12, for example. FIG. 12 is a diagram illustrating an example of a hardware configuration. The computer 1000 is connected to an output device 1010 and an input device 1020, and an arithmetic device 1030, a primary storage device 1040, a secondary storage device 1050, an output IF (Interface) 1060, an input IF 1070, and a network IF 1080 are connected via a bus 1090. Have

  The arithmetic device 1030 operates based on a program stored in the primary storage device 1040 and the secondary storage device 1050, a program read from the input device 1020, and the like, and executes various processes. The primary storage device 1040 is a memory device such as a RAM that temporarily stores data used by the arithmetic device 1030 for various arithmetic operations. The secondary storage device 1050 is a storage device in which data used by the arithmetic device 1030 for various calculations and various databases are registered, such as ROM (Read Only Memory), HDD (Hard Disk Drive), flash memory, and the like. It is realized by.

  The output IF 1060 is an interface for transmitting information to be output to an output device 1010 that outputs various types of information such as a monitor and a printer. For example, USB (Universal Serial Bus), DVI (Digital Visual Interface), This is realized by a standard connector such as HDMI (registered trademark) (High Definition Multimedia Interface). The input IF 1070 is an interface for receiving information from various input devices 1020 such as a mouse, a keyboard, and a scanner, and is realized by, for example, a USB.

  The input device 1020 includes, for example, an optical recording medium such as a CD (Compact Disc), a DVD (Digital Versatile Disc), and a PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), and a tape. It may be a device that reads information from a medium, a magnetic recording medium, a semiconductor memory, or the like. The input device 1020 may be an external storage medium such as a USB memory.

  The network IF 1080 receives data from other devices via the network N and sends the data to the arithmetic device 1030, and transmits data generated by the arithmetic device 1030 to other devices via the network N.

  The arithmetic device 1030 controls the output device 1010 and the input device 1020 via the output IF 1060 and the input IF 1070. For example, the arithmetic device 1030 loads a program from the input device 1020 or the secondary storage device 1050 onto the primary storage device 1040, and executes the loaded program.

  For example, when the computer 1000 functions as the information distribution device 10, the arithmetic device 1030 of the computer 1000 implements the function of the control unit 40 by executing a program loaded on the primary storage device 1040.

[10. effect〕
As described above, the information distribution apparatus 10 acquires an image associated with a predetermined keyword such as query information. Then, the information distribution apparatus 10 is a model that determines whether an image satisfies a predetermined condition, and inputs a predetermined keyword from the acquired image using a plurality of models that determine whether the image satisfies different conditions. A display image, which is an image to be displayed when selected, is selected.

  For example, the information distribution apparatus 10 acquires an image indicating a product or service related to a predetermined keyword. In addition, the information distribution apparatus 10 selects, as a display image, an image having a common concept with a predetermined keyword from the acquired images using a plurality of models. For this reason, the information distribution apparatus 10 can improve the selection accuracy of an image that reminds the user of a predetermined keyword, that is, an image presented to the user as an image indicating the predetermined keyword.

  Further, the information distribution apparatus 10 selects a display image from the acquired images using a model that determines whether a condition corresponding to an attribute of a product or service related to a predetermined keyword is satisfied as a model. For this reason, since the information delivery apparatus 10 can select an image having a background color according to the attribute of the product, for example, it is possible to select a more appropriate image as an image indicating a predetermined keyword.

  In addition, the information distribution apparatus 10 acquires an image associated with a set of keywords including a plurality of keywords such as query candidates including a plurality of classification information as the predetermined keyword. The information distribution device 10 is an image that is displayed as a query candidate when one of the keywords included in the set of keywords is input as a search query, and is selected by the user. All the keywords included in the keyword set are selected as new search queries. For example, the information distribution device 10 is an image that is displayed as a query candidate when classification information that becomes an activation query is input as a search query among classification information included in the query candidate, and is selected by the user At this time, an image whose query candidate is a new search query is selected. For this reason, the information delivery apparatus 10 can display an appropriate image as a query candidate when assisting a user's search.

  The information distribution apparatus 10 also has similar appearance when selecting an image to be displayed as a new search query candidate when a keyword set is input for each of a plurality of keyword sets having commonality. Select multiple images. For example, when selecting display images of query candidates having the same cut end, the information distribution apparatus 10 selects images having similar appearances. For this reason, the information delivery apparatus 10 can select the image which assists a user's search appropriately as a display image.

  Further, the information distribution apparatus 10 calculates a value indicating the degree to which the acquired image satisfies each condition using a plurality of models that output values indicating the degree to which the image satisfies a predetermined condition as a model, An image in which each calculated value exceeds a predetermined threshold is selected as a display image. For example, the information distribution apparatus 10 uses a model for determining whether the background of the image satisfies a predetermined condition as a model, a model for determining whether or not a predetermined number of subjects of the image are captured, and a person in the image. A model for determining whether a character is included in the image, a model for determining whether the image is decorated, a model for determining whether a human face is captured in the image, and a subject The display image is selected using at least one of a model for determining whether or not the color is a predetermined color and a model for determining the size of the subject in the image. In addition, the information distribution apparatus 10 selects a display image using a model that determines whether or not a subject in the image is associated with a predetermined keyword as a model. In addition, the information distribution apparatus 10 selects a display image using a model that determines whether or not the subject in the image has the same attribute as the attribute indicated by the predetermined keyword.

  As a result of the processing using the various models described above, the information distribution apparatus 10 can improve the selection accuracy of an image presented to the user as an image indicating a predetermined keyword.

  As described above, some of the embodiments of the present application have been described in detail with reference to the drawings. However, these are merely examples, and various modifications, including the aspects described in the disclosure section of the invention, based on the knowledge of those skilled in the art, It is possible to implement the present invention in other forms with improvements.

  Moreover, the above-mentioned “section (module, unit)” can be read as “means”, “circuit”, and the like. For example, the distribution unit can be read as distribution means or a distribution circuit.

DESCRIPTION OF SYMBOLS 10 Information delivery apparatus 20 Communication part 30 Storage part 31 Structured data database 32 Query candidate database 33 Model database 40 Control part 41 Collection part 42 Extraction part 43 Generation part 44 Acquisition part 45 Selection part 46 Provision part 100 Management server 131 Search object database 200 User terminal

Claims (15)

An acquisition unit for acquiring a plurality of images related to a predetermined keyword;
It is a plurality of models for determining whether or not images satisfy different conditions, and at least whether or not the transaction target is imaged by the number of units in which the transaction target corresponding to the predetermined keyword is traded. A selection unit that selects a display image, which is an image to be displayed when the predetermined keyword is input, from an image acquired by the acquisition unit using a plurality of models including a model to be determined. Selection device. Reference is made to a database that stores an image to be searched and a set of classification information having a hierarchical relationship as information indicating the classification of the search target , and is associated with a set of classification information corresponding to a predetermined keyword. An acquisition unit for acquiring images;
Corresponding to the predetermined keyword from the image acquired by the acquisition unit using a plurality of models for determining whether the image satisfies a predetermined condition and determining whether the image satisfies different conditions. A selection unit that selects a display image displayed as one of a plurality of images displayed when a keyword corresponding to a part of the classification information included in the classification information set is input. Feature selection device. The selection device according to claim 1, wherein the acquisition unit acquires an image indicating a transaction target related to the predetermined keyword. The selection unit selects the display image from the image acquired by the acquisition unit using a model that determines whether a condition corresponding to a transaction target attribute associated with the predetermined keyword is satisfied as the model. The selection device according to claim 3. The selection unit, using the plurality of models, selects an image having the same concept as the predetermined keyword from the images acquired by the acquisition unit as the display image. The selection apparatus as described in any one of them. The acquisition unit acquires an image associated with a set of keywords including a plurality of keywords as the predetermined keyword,
The selection unit is an image that is displayed as a candidate for a new search query when any one of keywords included in the keyword set is input as a search query as the display image. The selection device according to any one of claims 1 to 5, wherein an image in which all keywords included in the set of keywords are used as a new search query when selected by the selection is selected. When the selection unit selects an image to be displayed as a candidate for the new search query when each of the plurality of keyword sets having commonality is input, the appearance is similar. The selection device according to claim 6, wherein a plurality of images are selected. The selection unit, as the model, using a plurality of model outputs a value indicating the satisfying degree of image Jo Tokoro, images acquired by the acquisition unit is a value that indicates each satisfying degree respectively The selection device according to any one of claims 1 to 7, wherein the selection device calculates and selects an image in which each of the calculated values exceeds a predetermined threshold as the display image. The selection unit determines, as the model, a model that determines whether a background of the image satisfies a predetermined condition, a model that determines whether a person is included in the image, and whether a character is included in the image A model, a model for determining whether the image is decorated, a model for determining whether a human face is captured in the image, a model for determining whether the subject has a predetermined color, or the image The selection apparatus according to any one of claims 1 to 8, wherein the display image is selected using at least one of models for determining the size of a subject. The said selection part selects the said display image using the model which determines whether the to-be-photographed object reflected in the said image is related with the said predetermined keyword as said model. The selection apparatus as described in any one of them. The selection unit selects the display image using a model for determining whether a subject in the image has an attribute similar to the attribute indicated by the predetermined keyword as the model. The selection device according to any one of claims 1 to 10. A selection method performed by a selection device, comprising:
An acquisition step of acquiring a plurality of images associated with a predetermined keyword;
It is a plurality of models for determining whether or not images satisfy different conditions, and at least whether or not the transaction target is imaged by the number of units in which the transaction target corresponding to the predetermined keyword is traded. A selection step of selecting a display image which is an image to be displayed when the predetermined keyword is input from the images acquired by the acquisition step using a plurality of models including a model to be determined. The selection method. On the computer,
An acquisition procedure for acquiring a plurality of images related to a predetermined keyword;
It is a plurality of models for determining whether or not images satisfy different conditions, and at least whether or not the transaction target is imaged by the number of units in which the transaction target corresponding to the predetermined keyword is traded. A selection procedure for selecting a display image that is an image to be displayed when the predetermined keyword is input from an image acquired by the acquisition procedure using a plurality of models including a model to be determined. Selection program. A selection method performed by a selection device, comprising:
Reference is made to a database that stores an image to be searched and a set of classification information having a hierarchical relationship as information indicating the classification of the search target , and is associated with a set of classification information corresponding to a predetermined keyword. An acquisition process for acquiring images;
Corresponding to the predetermined keyword from the image acquired by the acquisition step using a plurality of models for determining whether the image satisfies a predetermined condition and determining whether the image satisfies different conditions. A selection step of selecting a display image displayed as one of a plurality of images displayed when a keyword corresponding to a part of the classification information included in the classification information set is input. Feature selection method. On the computer,
Reference is made to a database that stores an image to be searched and a set of classification information having a hierarchical relationship as information indicating the classification of the search target , and is associated with a set of classification information corresponding to a predetermined keyword. An acquisition procedure to acquire an image;
Corresponding to the predetermined keyword from the image acquired by the acquisition procedure using a plurality of models for determining whether the image satisfies a predetermined condition and determining whether the image satisfies different conditions. A selection procedure for selecting a display image to be displayed as one of a plurality of images displayed when a keyword corresponding to a part of the classification information included in the classification information set is input. Selection program.

Images