JP2019020980A - Estimation device, estimation method, estimation program, and model - Google Patents

Abstract

To make it possible to improve the accuracy of information in electronic commerce services.SOLUTION: An estimation device according to the present application includes an acquisition unit and an estimation unit. The acquisition unit acquires information on exhibition in an electronic commerce service in which products provided by a plurality of distributors can be purchased. The estimating unit estimates information on an exhibited object based on the information on the exhibition acquired by the acquiring unit.SELECTED DRAWING: Figure 4

Description

  The present invention relates to an estimation device, an estimation method, an estimation program, and a model.

  In recent years, with the rapid spread of the Internet, sales via the Internet have become popular. For example, a technology is known that enables a buyer to place a bid for a product that has been exhibited by a general consumer in an auction system as genuine.

Japanese Patent No. 5140467

  However, with the above-described conventional technology, it may be difficult to improve the accuracy of information in the electronic commerce service. For example, it is possible to improve the accuracy of information in electronic commerce services when sellers such as merchants of merchandise have not entered the category of the merchandise, or have entered an incorrect category. Is difficult.

  The present application has been made in view of the above, and an object thereof is to provide an estimation device, an estimation method, an estimation program, and a model that can improve the accuracy of information in an electronic commerce service.

  The estimation apparatus according to the present application includes an acquisition unit that acquires information about an exhibition in an electronic commerce service that can purchase products provided by a plurality of vendors, and the information about the exhibition acquired by the acquisition unit. And an estimation unit for estimating information related to the object.

  According to one aspect of the embodiment, there is an effect that it is possible to improve the accuracy of information in the electronic commerce service.

FIG. 1 is a diagram illustrating an example of a generation process according to the embodiment. FIG. 2 is a diagram illustrating an example of an estimation process according to the embodiment. FIG. 3 is a diagram illustrating a configuration example of the estimation system according to the embodiment. FIG. 4 is a diagram illustrating a configuration example of the estimation apparatus according to the embodiment. FIG. 5 is a diagram illustrating an example of a learning data storage unit according to the embodiment. FIG. 6 is a diagram illustrating an example of the model information storage unit according to the embodiment. FIG. 7 is a diagram illustrating an example of a product information storage unit according to the embodiment. FIG. 8 is a flowchart illustrating an example of the generation process according to the embodiment. FIG. 9 is a flowchart illustrating an example of the estimation process according to the embodiment. FIG. 10 is a diagram illustrating an example of model generation according to the embodiment. FIG. 11 is a diagram illustrating an example of a counting process according to the embodiment. FIG. 12 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the estimation device.

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

(Embodiment)
[1. Generation process)
First, an example of the generation process according to the embodiment will be described with reference to FIG. FIG. 1 is a diagram illustrating an example of a generation process according to the embodiment. FIG. 1 shows a case where the estimation apparatus 100 generates a model based on information related to an exhibition to which correct information is given by the manager GM1 in the electronic commerce service (hereinafter also referred to as “product information”). Hereinafter, the product information to which the correct answer information is assigned is also referred to as “learning data”. In the example of FIGS. 1 and 2, the case where the electronic commerce service is an online shopping mall service (hereinafter simply referred to as “net shopping mall”) exhibited by a plurality of stores (sales sources) will be described as an example. The electronic commerce service is not limited to the online shopping mall, and may be various services such as an auction service.

  For example, the administrator GM1 may be a provider that provides an electronic commerce service. The product information here means information related to a predetermined product or service (hereinafter collectively referred to as “product”). For example, the product information includes information input by the product provider at the time of listing. In the example of FIG. 1, the product information includes a case where two types of information, image information (hereinafter also simply referred to as “image”) and character information, are included. In the example of FIG. 1, the estimation apparatus 100 generates a model based on product information to which a product category is assigned as correct information. That is, the example of FIG. 1 shows a case where the estimation apparatus 100 generates a model for estimating the product category that is the target indicated by the product information based on the product information to which the product category is assigned as correct information. .

  Moreover, the estimation apparatus 100 estimates the category of the product based on the product information when a certain product is exhibited using the generated model. The estimation apparatus 100 provides a service based on the estimated product category, which will be described with reference to FIG.

[Configuration of estimation system]
First, prior to the description of FIG. 1, the estimation system 1 shown in FIG. 3 will be described. As illustrated in FIG. 3, the estimation system 1 includes an administrator device 10, a sales source device 20, and an estimation device 100. The administrator device 10, the sales source device 20, and the estimation device 100 are connected via a predetermined network N so as to be communicable by wire or wirelessly. FIG. 3 is a diagram illustrating a configuration example of the estimation system according to the embodiment. Note that the estimation system 1 illustrated in FIG. 3 may include a plurality of administrator apparatuses 10, a plurality of vendor apparatuses 20, and a plurality of estimation apparatuses 100.

  The administrator device 10 is an information processing device used by the administrator of the estimation device 100. The administrator device 10 is realized by, for example, a smartphone, a tablet terminal, a notebook PC (Personal Computer), a desktop PC, a mobile phone, a PDA (Personal Digital Assistant), or the like. Hereinafter, the administrator device 10 may be referred to as an administrator. That is, in the following, the administrator can be read as the administrator device 10. The administrator GM1 illustrated in FIG. 1 may display information on the estimation device 100 or request the estimation device 100 to perform processing by operating the administrator device 10.

  For example, the manager GM1 uses the manager device 10 to give correct information to predetermined product information. For example, the manager GM1 uses the manager device 10 to give correct information to product information used as learning data. In the example of FIG. 1, the manager GM1 gives correct information to the product information based on the category classification in the electronic commerce service.

  The administrator device 10 and the estimation device 100 may be integrated. In this case, for example, information included in the estimation device 100 may be displayed on a screen included in the estimation device 100, and a processing instruction or the like may be given to the estimation device 100 using an input interface such as a keyboard or a mouse.

  The sales source device 20 is an information processing device used by a seller who sells products. The seller may be an individual or a corporation. For example, the sales source device 20 is used by the sales source SA1. For example, the sales agency SA1 uses the sales agency device 20 to add products to be sold in the electronic commerce service, manage inventory, and set prices. The vendor apparatus 20 is realized by, for example, a smartphone, a tablet terminal, a notebook PC, a desktop PC, a mobile phone, a PDA, or the like. In the following, the sales source device 20 may be referred to as a sales source. That is, in the following, the seller can be read as the seller device 20. Specifically, FIG. 2 shows a case where the sales source device 20 is a notebook PC used by the sales source SA1.

  The estimation apparatus 100 is an information processing apparatus that estimates information related to an exhibition target based on information related to an exhibition (product information). In the example of FIG. 2, the estimation apparatus 100 is a model used for estimating a product category, and estimates the product category based on the output of the model to which product information is input.

  The estimation apparatus 100 provides a service based on information related to the estimated exhibition target. In the example of FIG. 2, the estimation apparatus 100 provides information related to the estimated category of the product to the seller that has exhibited the product. Moreover, in the estimation system 1, the estimation apparatus 100 provides an electronic commerce service that enables the purchase of products provided by a plurality of vendors. In other words, the estimation system 1 provides an electronic commerce service that can purchase products provided by a plurality of vendors.

  Further, the estimation apparatus 100 may provide various services such as an electronic commerce service and a search service. Further, when an external information processing device provides a search service or an electronic commerce service, the estimation device 100 may not provide a search service or an electronic commerce service. In this case, the estimation device 100 acquires various types of information from an external information processing device or the like that provides a search service or an electronic commerce service, and performs estimation processing or information provision to the external information processing device based on the acquired various types of information. You may go.

  In the example shown in FIG. 1, the administrator GM1 gives correct information to the product information GD101 including the image IM101, the character information CH101, and the like (step S11-1). For example, the administrator GM1 gives correct information indicating that the category of the product information GD101 is “A category” based on the contents of the image IM101 and the character information CH101 included in the product information GD101. In the example of FIG. 1, the product category is represented by an abstract symbol such as “A category”. For example, “A category” is a specific category such as “food” or “drinking water”. It is assumed that the information is information.

  As described above, when “product GD * (* is an arbitrary numerical value)” is described, it indicates that the product is a product identified by the product ID “GD *”. For example, when “product GD1” is described, the product is a product identified by the product ID “GD1”. Hereinafter, the product information of “product GD * (* is an arbitrary numerical value)” may be expressed as “product GD * (* is an arbitrary numerical value)”. For example, hereinafter, the product GD1 can be read as product information of the product GD1.

  Further, the manager GM1 gives correct information to the product information GD102 including the image IM102, the character information CH102, and the like (step S11-2). For example, the administrator GM1 gives correct information indicating that the category of the product information GD102 is “B category” based on the contents of the image IM102 and the character information CH102 included in the product information GD102.

  Further, the administrator GM1 gives correct information to the product information GD103 including the image IM103, the character information CH103, and the like (step S11-3). For example, the administrator GM1 gives correct information indicating that the category of the product information GD103 is “C category” based on the contents of the image IM103 and the character information CH103 included in the product information GD103.

  Further, the manager GM1 gives correct information to the product information GD104 including the image IM104, the character information CH104, and the like (step S11-4). For example, the administrator GM1 gives correct information indicating that the category of the product information GD104 is “D category” based on the contents of the image IM104, the character information CH104, and the like included in the product information GD104.

  Note that steps S11-1 to S11-4 are for explaining the processing, and any of steps S11-1 to S11-4 may be performed first, and hereinafter steps S11-1 to S11- 4 will be collectively referred to as step S11. For example, the manager GM1 performs the assignment of correct information as shown in step S11 on a large number of product information that is used as learning data. Thereby, in the estimation system 1, learning data LD1 in which product information (exhibition information) and correct answer information (category) are associated with each other is generated. In the example of FIG. 1, the case where the administrator GM1 assigns the correct information is illustrated in order to explain the correct information, but the correct information may be mechanically performed by the estimation device 100 or the like.

  And the estimation apparatus 100 acquires the learning data LD1 to which the correct information was given in step S11 (step S12). Specifically, the estimating apparatus 100 acquires the product information GD101 to GD104 to which correct information is assigned in step S11.

  And the estimation apparatus 100 adds the merchandise information contained in the learning data LD1 acquired by step S12 as learning data (step S13). Specifically, the estimating apparatus 100 adds the learning data LD1 to the learning data storage unit 121.

  And the estimation apparatus 100 produces | generates a model based on learning data (step S14). For example, the estimation apparatus 100 generates a model by performing learning using the product information GD101 to GD104 in the learning data storage unit 121 as learning data (teacher data).

  Here, a model (learning device) generated by the estimation apparatus 100 will be briefly described. In the example of FIG. 1, the model M11 generated by the estimation device 100 is a model that outputs a plurality of scores SC corresponding to each of a plurality of categories when product information is input. For example, the model M11 is a model that outputs 10 scores corresponding to each of the A category to the J category when the category to be classified is 10 categories of the A category to the J category. Note that the number of outputs may be set as appropriate according to the content to be classified. For example, when the estimation device 100 wants to classify products into 100 categories, the estimation device 100 may generate a model that outputs 100 scores corresponding to each of the 100 categories. In addition, the estimation apparatus 100 may generate a model that performs classification related to various objects without being limited to the category. For example, the estimation apparatus 100 may generate a model that performs classification for each product name.

  For example, in the example of FIG. 1, it is assumed that the estimation apparatus 100 generates a model M11 that estimates which of the 26 categories from category A to category Z corresponding to the product information input. In the example of FIG. 1, the model M11 generated by the estimation apparatus 100 is a model that outputs a score SC indicating the probability of belonging to each of the categories A to Z when product information is input. In this case, the model M11 outputs a value from 0 to 1 as each score. For example, when the score corresponding to the A category is “0.65”, the probability that the target (product) corresponding to the input product information belongs to the A category is “65%”. Note that the model M11 generated by the estimation apparatus 100 is not limited to the above, and may be any model as long as the score is larger, the model M11 is more likely to belong to the category corresponding to the score. The estimation apparatus 100 may use information obtained by converting the output of the model M11 into probabilities using various functions such as a softmax function.

  From here, the structure of the model M11 shown in FIG. 1 is demonstrated. The model M11 includes a partial model M11-1 having an input layer to which image information is input, a partial model M11-2 having an input layer to which character information is input, and a portion having a layer for outputting a score, among product information Includes model M11-3.

  For example, the partial model M11-1 is a partial model that extracts the feature information FE1 from the input image information IMD. In the example of FIG. 1, the partial model M11-1 is configured by a so-called convolutional neural network (CNN) that repeats convolution and pooling of a local region of image information. The partial model M11-1 is not limited to CNN, but may be various neural networks such as RNN (Recurrent Neural Network) and LSTM (Long Short-Term Memory).

  For example, the partial model M11-2 is a partial model that extracts the feature information FE2 from the input character information CHD. In the example of FIG. 1, the partial model M11-2 is assumed to be configured by CNN. The partial model M11-2 is not limited to CNN, and may be various neural networks such as RNN, LSTM, and the like. As described above, the model M11 illustrated in FIG. 1 extracts the feature information FE1 from the image information included in the product information, and extracts the feature information FE2 from the character information. That is, the model M11 extracts features independently from image information and character information.

  Further, for example, the partial model M11-3 is a partial model that outputs a score in accordance with the input of the feature information FE1 extracted by the partial model M11-1 and the feature information FE2 extracted by the partial model M11-2. For example, the partial model M11-3 is a partial model to which the output from the partial model M11-1 and the output from the partial model M11-2 are input.

  For example, the output layer in the partial model M11-1 and the output layer in the partial model M11-2 are connected to the input layer of the partial model M11-3. For example, as for the input in the input layer of the partial model M11-3, the node (neuron) in the output layer in the partial model M11-1 and the node (neuron) in the output layer in the partial model M11-2 may be fully coupled.

  As described above, the partial model M11-2 is a partial model that extracts the features of the entire product from the feature information FE1 and the feature information FE2 and outputs a score. For example, the partial model M11-3 may be configured by CNN. The partial model M11-3 is not limited to CNN, but may be various neural networks such as RNN and LSTM.

  In addition, the partial model M11-1, the partial model M11-2, and the partial model M11-3 are codes assigned to the toilet in order to explain the configuration of each unit in the model M11. That is, the model M11 is one model that outputs a score corresponding to each category when image information and character information, which are product information, are input when viewed from outside the model M11.

  For example, the estimating apparatus 100 generates a model M11 using the learning data LD1. For example, when the product information GD101 is input to the model M11, the estimation device 100 performs a learning process so that the score corresponding to the A category approaches “1”. For example, when the image IM101 of the merchandise information GD101 is input to the partial model M11-1 of the model M11 and the character information CH101 of the merchandise information GD101 is input to the partial model M11-2, the estimation apparatus 100 is partially model M11- The learning process is performed so that the score corresponding to the A category output by 3 approaches “1”.

  Further, for example, when the product information GD102 is input to the model M11, the estimation apparatus 100 performs a learning process so that the score corresponding to the B category approaches “1”. For example, when the product information GD103 is input to the model M11, the estimation apparatus 100 performs a learning process so that the score corresponding to the C category approaches “1”. For example, when the product information GD104 is input to the model M11, the estimation apparatus 100 performs a learning process so that the score corresponding to the D category approaches “1”. That is, the estimation apparatus 100 performs backpropagation (error back propagation method) for correcting a parameter (connection coefficient) so that an error between an output (score) in the model M11 and correct information corresponding to the input product information is reduced. A model M11 is generated by such processing as described above. For example, the estimating apparatus 100 generates the model M11 by performing processing such as back propagation so as to minimize a predetermined loss function. By repeating the processing as described above, the estimating apparatus 100 can generate a model M11 that outputs a score that makes it possible to estimate a product category corresponding to the input product information more accurately. Note that the model learning technique is not limited to the technique described above, and any known technique can be applied. That is, if the estimation apparatus 100 can learn the model M11 so as to output a score corresponding to the correct answer information when the product information included in the learning data is input, the estimation apparatus 100 can output M11 by any method. Generation may be performed.

  With the processing as described above, in the example of FIG. 1, the estimation apparatus 100 generates a model (model M11) identified by the model ID “M11” as shown in the model information storage unit 122. As described above, when “model M * (* is an arbitrary numerical value)” is described, it indicates that the product is a model identified by the product ID “M *”. For example, when “model M11” is described, the model is a model identified by the model ID “M11”. As shown in the model information storage unit 122 in FIG. 1, the model M11 is a model used for “category”, that is, category classification, and the specific model data is “model data MTD11”. It shows that. For example, the estimation apparatus 100 inputs product information to the model M11, and causes the model M11 to output a score indicating which category of the input product information is likely to be the category, and the model M11 outputs the score. Estimate the category based on the score.

  As described above, the estimation apparatus 100 can generate a model that can appropriately estimate a product category by learning using learning data in which product information and correct answer information are associated with each other. In addition, the estimation apparatus 100 can appropriately estimate the category of the product by generating a model that extracts the features of the entire product after individually extracting the features of the image and character information included in the product information. A model can be generated. Thereby, the estimation apparatus 100 can generate a model used to enable improvement of information accuracy in the electronic commerce service. Therefore, the estimation apparatus 100 can improve the accuracy of product information displayed in the electronic commerce service, for example, by using the model generated as described above.

[2. (Estimation process)
An example of the estimation process according to the embodiment will be described with reference to FIG. FIG. 2 is a diagram illustrating an example of an estimation process according to the embodiment. FIG. 2 shows a case where the estimation apparatus 100 provides information for assisting information input in the listing by estimating the category that is the attribute information of the listing when the selling agency performs the listing.

  First, the estimating apparatus 100 acquires information related to the exhibition from the sales source device 20 used by the sales source SA1 (step S21). For example, the seller SA1 transmits information related to the exhibition to the estimation device 100 by performing a predetermined operation on a site for the seller of the electronic commerce service displayed on the seller device 20. In the example of FIG. 2, the sales source device 20 transmits product information GD11 including the image IM11 and character information CH11.

  The estimation apparatus 100 that has acquired the information related to the exhibition from the sales source apparatus 20 inputs the information related to the exhibition into the model (step S22). In the example of FIG. 2, the estimation apparatus 100 inputs the product information GD11 acquired from the sales agency SA1 to the model M11. Specifically, the estimating apparatus 100 inputs the image IM11 and the character information CH11 to the model M11.

  For example, the estimating apparatus 100 inputs the image IM11 to the partial model M11-1 of the model M11. Then, the partial model M11-1 extracts feature information FE1-11 from the image IM11. For example, the estimation apparatus 100 inputs the character information CH11 to the partial model M11-2 of the model M11. Then, the partial model M11-2 extracts feature information FE2-11 from the character information CH11. The partial model M11-3 calculates a score according to the feature information FE1-11 and the feature information FE2-11.

  Thereby, the estimation apparatus 100 outputs the score corresponding to the product information GD11 from the model M11 (step S23). In the example of FIG. 2, the estimation apparatus 100 causes the model M11 to output scores corresponding to each of a plurality of categories such as the A category to the D category, as indicated by the score SC11.

  In the example of FIG. 2, as shown in the score SC11, the model M11 has a “A” category score of “0.01”, a B category score of “0.65”, and a C category score of “ “0” and the score of the D category are output as “0.05”.

  Then, the estimating apparatus 100 estimates a product category corresponding to the product information GD11 based on the output of the model M11 (step S24). Here, in the example of FIG. 2, the score “0.65” corresponding to the B category is the maximum value, as indicated by the score SC11. Therefore, as illustrated in the estimation information ES11, the estimation apparatus 100 estimates the category of the product corresponding to the product information GD11 as “B category”.

  After that, the estimating apparatus 100 provides information related to the category estimated in step S24 to the selling apparatus 20 (step S25). For example, the estimating apparatus 100 provides information indicating that the category to be listed by the seller SA1 corresponds to the B category. In the example of FIG. 2, the estimation device 100 provides information indicating that the category of the product corresponding to the product information GD11 corresponds to the B category to the selling device 20.

  The vendor apparatus 20 to which the information related to the category is provided from the estimation apparatus 100 displays the information related to the category (step S26). For example, the sales source device 20 displays information indicating that the category of the subject of listing by the sales source SA1 corresponds to the B category. In the example of FIG. 2, the sales source device 20 displays information indicating that the category of the product corresponding to the product information GD11 corresponds to the B category.

  And sales agency SA1 registers the information regarding exhibition into the estimation apparatus 100 based on the information provided from the estimation apparatus 100 (step S27). For example, the vendor apparatus 20 transmits information requesting to register the B category as the category of the product corresponding to the product information GD11 to the estimation device 100.

  And the estimation apparatus 100 which acquired the request | requirement of information registration from the sales origin apparatus 20 registers the information regarding exhibition (step S28). For example, the estimation apparatus 100 registers information related to the object that the seller SA1 has submitted. In the example of FIG. 2, the estimation apparatus 100 registers information related to the product information GD11 in the product information storage unit 123.

  For example, as illustrated in the product information storage unit 123 in FIG. 1, the estimation apparatus 100 registers the B category as the category of the product (product GD11) identified by the product ID “GD11”. Further, the estimating apparatus 100 registers the image IM11 as the image of the product GD11, and registers the character information CH11 as the character information of the product GD11.

  As described above, the estimation apparatus 100 suppresses the registration of a wrong category by the sales agency by estimating the target category exhibited by the sales agency and providing the seller with information on the estimated category. For example, in an electronic commerce service that can purchase products provided by a plurality of vendors as described above, the accuracy and amount of information input by each vendor is often different. For example, there is a case where a certain vendor has extremely little information to input. Even in such a case, the estimation apparatus 100 can increase the information related to the exhibition without imposing a load on the seller by estimating the information related to the subject of the exhibition by appropriately using various models. As described above, the estimation apparatus 100 can improve the accuracy of information in the electronic commerce service. In addition, the estimation apparatus 100 can increase the amount of listing in the electronic commerce service by reducing the load of the seller in the exhibition by providing information that assists the seller to input information in the exhibition. Thereby, the estimation apparatus 100 can improve an electronic commerce service and can raise a user's satisfaction.

(2-1. Use of estimated information)
In the example of FIGS. 1 and 2, the case in an online shopping mall is illustrated, but when an individual makes an exhibition in an auction service, there are many cases where information to be input is insufficient as compared with the case where an operator performs. Even in such a case, the estimation apparatus 100 generates models for various uses, and performs estimation processing on information related to listings acquired using the generated models, thereby obtaining the estimated information. Can be provided. Thereby, the estimation apparatus 100 can improve the accuracy of information in the electronic commerce service.

  The same problem arises not only in the auction service but also in an online shopping mall or the like that accepts listings from a plurality of sellers, not a form in which one seller sells products. For example, if the product information such as JAN (Japan Article Number) code is not included in the information related to the exhibition, it is often unclear which product corresponds to which product, such as the sales status of the product. Is difficult to manage. As described above, even when the identification information of the product is not included in the information related to the exhibition, the information related to the product is estimated by using a model for estimating the product name, for example, and the sales of the product based on the information is performed. The situation can be managed. Details of this point will be described later.

(2-2. Information provision service)
In the example of FIG. 1, the estimation apparatus 100 provides the seller with information for assisting information input in the exhibition. However, the estimation apparatus 100 may perform various services using the estimated information. For example, the estimation apparatus 100 may provide information related to a recommended product to a predetermined user based on information related to the estimated exhibition target. For example, when the product purchased by the user can be specified using the model M21 (see FIG. 10) for estimating the product name, the estimation apparatus 100 may provide the user with information related to the product related to the product. For example, the estimation apparatus 100 improves the accuracy of the recommended information provided to the user by using the model M21 (see FIG. 10) that estimates the product name to improve the specific accuracy of the product purchased by each user. be able to. For example, the estimation apparatus 100 inputs the product information included in the purchase history of the user into the model M21 (see FIG. 10), thereby accurately identifying the product purchased by the user in the past, and the identified user purchased in the past. The recommended information may be provided to the user based on the purchased product.

  For example, when the list of products is displayed, the estimation apparatus 100 may improve the diversity of products included in the list by using the product names estimated using the model M21 (see FIG. 10). For example, the estimation apparatus 100 provides a user with a list in which product information is arranged so that product information having the same product name estimated using the model M21 (see FIG. 10) is not displayed. It is possible to suppress providing a user with a list of products. For example, the estimation apparatus 100 may provide product ranking information based on information related to the estimated exhibition target. For example, the estimation apparatus 100 can accurately collect the sales results of the product by specifying the product purchased by the user using the model M21 (see FIG. 10) that estimates the product name. Based on this, a sales ranking of the product may be generated. This point will be described in detail in FIG.

(2-3. Use of information between multiple services)
Moreover, the estimation system 1 may provide information among a plurality of services. For example, the estimation apparatus 100 uses an information on an electronic commerce service (for example, an internet shopping mall) different from an electronic commerce service (for example, an auction service), and based on information about an estimated exhibition target, Information for assisting information input in the service may be provided. For example, the estimation apparatus 100 estimates information related to an exhibition target in the auction service using a model generated based on information in an online shopping mall different from the auction service, and based on the estimated information related to the exhibition target. Information for assisting information input in the auction service may be provided. In this case, the estimation device 100 provides information for assisting information input in the auction service as shown in FIG. 2 using a model generated based on information in the online shopping mall as shown in FIG. May be. The estimation device 100 may acquire information from a device that provides an online shopping mall outside the estimation system 1, or may acquire information from a device that provides an online shopping mall in the estimation system 1. Moreover, the estimation apparatus 100 may provide services of both an auction service and an online shopping mall. Thereby, the estimation apparatus 100 appropriately estimates information in a service (auction service) having relatively low information accuracy using information in a service (net shopping mall) having relatively high information accuracy. be able to. Therefore, the estimation apparatus 100 may provide information that assists information input appropriately even in a service (auction service) in which the accuracy of information is relatively low.

[3. Configuration of estimation device]
Next, the configuration of the estimation apparatus 100 according to the embodiment will be described with reference to FIG. FIG. 4 is a diagram illustrating a configuration example of the estimation apparatus 100 according to the embodiment. As illustrated in FIG. 4, the estimation device 100 includes a communication unit 110, a storage unit 120, and a control unit 130. The estimation device 100 includes an input unit (for example, a keyboard and a mouse) that receives various operations from an administrator of the estimation device 100 and a display unit (for example, a liquid crystal display) for displaying various types of information. May be.

(Communication unit 110)
The communication unit 110 is realized by, for example, a NIC (Network Interface Card). The communication unit 110 is connected to the network by wire or wireless, and transmits / receives information to / from the administrator device 10.

(Storage unit 120)
The storage unit 120 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. As shown in FIG. 4, the storage unit 120 according to the embodiment includes a learning data storage unit 121, a model information storage unit 122, and a product information storage unit 123.

(Learning data storage unit 121)
The learning data storage unit 121 according to the embodiment stores various types of information related to learning data. FIG. 5 is a diagram illustrating an example of a learning data storage unit according to the embodiment. For example, the learning data storage unit 121 stores teacher data used for generating a model. The learning data storage unit 121 illustrated in FIG. 5 includes items such as “product ID”, “correct information”, “image”, and “character information”.

  “Product ID” indicates identification information for identifying a product. For example, the product information of the product identified by the product ID “GD101” corresponds to the product information GD101 illustrated in the example of FIG. “Correct answer information” indicates the classification of the product identified by the product ID. “Image” indicates an image included in the product information. FIG. 5 shows an example in which conceptual information such as “IM101” is stored in “image”, but actually image information or a file path name indicating the storage location is stored.

  “Character information” indicates character information of product information. FIG. 5 shows an example in which conceptual information such as “CH101” is stored in “character information”, but in practice, various character information related to product information such as titles and explanations of product information is stored. .

  For example, in the example illustrated in FIG. 5, the product information (product information GD101) identified by the product information ID “GD101” includes the image IM101 and the character information CH101. For example, in the example illustrated in FIG. 5, the product information (product information GD102) identified by the product information ID “GD102” includes the image IM102 and the character information CH102.

  Note that the learning data storage unit 121 is not limited to the above, and may store various information according to the purpose. For example, the learning data storage unit 121 may store information related to the date and time when learning data was added. Further, for example, the learning data storage unit 121 may store information indicating what kind of determination processing each learning data has been added. For example, the learning data storage unit 121 may store information indicating whether each learning data has been determined by an administrator's selection.

(Model information storage unit 122)
The model information storage unit 122 according to the embodiment stores information about the model. For example, the model information storage unit 122 stores model information (model data) generated by the generation process. FIG. 6 is a diagram illustrating an example of the model information storage unit according to the embodiment. The model information storage unit 122 illustrated in FIG. 6 includes items such as “model ID”, “use”, and “model data”. In FIG. 6, only the model M11 is illustrated, but a plurality of model information such as models M12 and M13 may be stored.

  “Model ID” indicates identification information for identifying a model. For example, the model identified by the model ID “M11” corresponds to the model M11 illustrated in the example of FIG. “Use” indicates the use of the corresponding model. “Model data” indicates data of a corresponding model associated with the model data. For example, “model data” includes information including nodes in each layer, functions adopted by the nodes, connection relationships between the nodes, and connection coefficients set for connections between the nodes.

  For example, in the example illustrated in FIG. 6, the model (model M11) identified by the model ID “M11” has a usage of “category” and is used for estimation of a target category corresponding to input information. Show. Further, the model data of the model M11 is model data MDT11.

  The model M11 (model data MDT11) is an input layer to which image information and character information related to a predetermined object are input, an output layer, and any layer from the input layer to the output layer, and a layer other than the output layer. And a second element whose value is calculated based on the first element and the weight of the first element, and for image information and character information input to the input layer, other than the output layer A computer is used to output information about a value obtained by quantifying the estimation of a predetermined target from the output layer by performing an operation based on the first element and the weight of the first element, with each element belonging to each layer as a first element. Is a model for making

  For example, the first element included in the model M11 corresponds to any node of the input layer or the intermediate layer. The second element corresponds to the next node, which is a node to which a value is transmitted from the node corresponding to the first element. The weight of the first element corresponds to a connection coefficient that is a weight considered for a value transmitted from a node corresponding to the first element to a node corresponding to the second element.

  The model information storage unit 122 is not limited to the above, and may store various model information according to the purpose.

(Product information storage unit 123)
The product information storage unit 123 according to the embodiment stores various types of information related to products. FIG. 7 shows an example of the product information storage unit 123 according to the embodiment. The product information storage unit 123 illustrated in FIG. 7 includes items such as “product ID”, “classification”, “image”, and “character information”.

  “Product ID” indicates identification information for identifying a product. “Classification” indicates the classification of the corresponding product. For example, “classification” indicates the classification of the product estimated by the model. “Image” indicates an image of product information that is product information. Although FIG. 7 shows an example in which conceptual information such as “IM1” is stored in “image”, actually, image information or a file path name indicating the storage location is stored.

  “Character information” indicates text information included in the product information of the product identified by the product ID. The “character information” includes various character information such as a title and a description included in the product information of the product identified by the product ID. In the product information storage unit 123 shown in FIG. 7, character information is indicated by an abstract symbol such as character information “CH1”. For example, “new bag A ...”, “free shipping”, etc. Each title is a specific title of the corresponding product.

  For example, in the example illustrated in FIG. 7, the image of the product (product GD1) identified by the product ID “GD1” is the image “IM1”. Further, the character information of the product GD1 indicates that the character information is “CH1”.

  The product information storage unit 123 is not limited to the above, and may store various information according to the purpose. The product information storage unit 123 may store a specific name or the like of the product.

(Control unit 130)
Returning to the description of FIG. 4, the control unit 130 is a controller and is stored in a storage device inside the estimation device 100 by, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like. Various programs (corresponding to an example of an estimation program) are implemented by using the RAM as a work area. The control unit 130 is a controller, and is realized by an integrated circuit such as an application specific integrated circuit (ASIC) or a field programmable gate array (FPGA). The control unit 130 performs an information process according to the model M11 and the like stored in the model information storage unit 122, an input layer to which image information and character information related to a predetermined object are input, an output layer, and an output from the input layer A first element belonging to any layer other than the output layer and a second element whose value is calculated based on the first element and the weight of the first element, and an input layer A predetermined target is estimated by performing an operation based on the first element and the weight of the first element, with each element belonging to each layer other than the output layer as the first element for the image information and character information input to Information on the value obtained by quantifying is output from the output layer.

  As shown in FIG. 4, the control unit 130 includes an acquisition unit 131, a generation unit 132, an estimation unit 133, and a provision unit 134, and implements or executes the information processing functions and operations described below. . Note that the internal configuration of the control unit 130 is not limited to the configuration illustrated in FIG. 4, and may be another configuration as long as information processing described later is performed. The control unit 130, by information processing according to the model M11 (model data MDT11) stored in the storage unit 120, a second element whose value is calculated based on the first element and the weight of the first element, By performing an operation based on the first element and the weight of the first element, with each element belonging to each layer other than the output layer as the first element for image information and character information input to the input layer, Information about a value obtained by quantifying the estimation of a predetermined target is output from the output layer.

(Acquisition part 131)
The acquisition unit 131 acquires various types of information. For example, the acquisition unit 131 acquires various types of information from the learning data storage unit 121, the model information storage unit 122, the product information storage unit 123, and the like. The acquisition unit 131 may acquire various types of information from an external information processing apparatus. Further, the acquisition unit 131 may acquire various types of information from the administrator device 10, the selling device 20, or the like.

  For example, the acquisition unit 131 acquires information related to an exhibition in an electronic commerce service that can purchase products provided by a plurality of vendors. For example, the acquisition unit 131 acquires a plurality of types of information related to the exhibition. For example, the acquisition unit 131 acquires image information and character information related to the exhibition. For example, the acquisition unit 131 acquires a plurality of types of information related to the exhibition. For example, the acquisition unit 131 acquires image information and character information related to the exhibition.

  In the example of FIG. 1, the acquisition unit 131 acquires learning data LD1 to which correct information is assigned. For example, the acquisition unit 131 acquires the product information GD101 to GD104 to which the correct answer information is assigned.

  In the example of FIG. 2, the acquisition unit 131 acquires information related to the exhibition from the sales source device 20 used by the sales source SA1. For example, the acquisition unit 131 acquires the product information GD11 including the image IM11 and the character information CH11 from the sales source device 20.

(Generator 132)
The generation unit 132 generates various types of information. For example, the generation unit 132 generates a model as shown in the model information storage unit 122 using the learning data stored in the learning data storage unit 121. For example, the generation unit 132 generates a model used for target estimation in the electronic commerce service based on the learning data acquired by the acquisition unit 131. The generation unit 132 generates a model and stores the generated model in the model information storage unit 122. Specifically, the generation unit 132 is an input layer to which image information and character information related to a predetermined target are input, an output layer, and any layer from the input layer to the output layer and other than the output layer. And a second element whose value is calculated based on the first element and the weight of the first element, and other than the output layer for image information and character information input to the input layer A model that outputs information about a value obtained by quantifying the estimation of a predetermined target from the output layer by performing an operation based on the first element and the weight of the first element, with each element belonging to each layer of the first element as a first element Generate.

  In the example of FIG. 1, the generation unit 132 generates a model based on the learning data. For example, the generation unit 132 generates a model by performing learning using the product information GD101 to GD104 and the like in the learning data storage unit 121 as learning data (teacher data).

  In the example of FIG. 1, the generation unit 132 generates a model M11 using the learning data LD1. For example, when the product information GD101 is input to the model M11, the generation unit 132 performs a learning process so that the score corresponding to the A category approaches “1”. For example, when the image IM101 of the product information GD101 is input to the partial model M11-1 of the model M11 and the character information CH101 of the product information GD101 is input to the partial model M11-2, the generation unit 132 receives the partial model M11- The learning process is performed so that the score corresponding to the A category output by 3 approaches “1”.

  In the example of FIG. 1, when the product information GD102 is input to the model M11, the generation unit 132 performs a learning process so that the score corresponding to the B category approaches “1”. For example, when the product information GD103 is input to the model M11, the generation unit 132 performs a learning process so that the score corresponding to the C category approaches “1”. For example, when the product information GD104 is input to the model M11, the generation unit 132 performs a learning process so that the score corresponding to the D category approaches “1”. For example, the generation unit 132 performs backpropagation (error back propagation method) that corrects a parameter (connection coefficient) so that an error between the output (score) in the model M11 and correct information corresponding to the input product information is reduced. A model M11 is generated by such processing as described above. For example, the generation unit 132 generates the model M11 by performing processing such as back propagation so as to minimize a predetermined loss function.

(Estimation unit 133)
The estimation unit 133 estimates various information. The estimation unit 133 estimates various information using information stored in the learning data storage unit 121, the model information storage unit 122, the product information storage unit 123, and the like. For example, the estimation unit 133 estimates various information based on various information acquired by the acquisition unit 131.

  For example, the estimation unit 133 estimates information related to the subject of the exhibition based on information related to the exhibition acquired by the acquisition unit 131. For example, the estimation unit 133 estimates the attribute information of the exhibition target as information related to the exhibition target. For example, the estimation unit 133 estimates the listing target as information related to the listing target.

  For example, the estimation unit 133 is a model used for estimation of an object, and estimates information related to an exhibition target based on an output of a model to which information related to exhibition is input. For example, the estimation unit 133 estimates information related to the subject of the exhibition based on the output of the model in which a plurality of types of information related to the exhibition are input. For example, the estimation unit 133 estimates information related to an exhibition target based on an output of a model in which image information and character information are input. For example, the estimation unit 133 estimates information related to an exhibition target using a model including a neural network generated based on predetermined information. For example, the estimation unit 133 estimates information related to an exhibition target using a model including a neural network that performs convolution processing and pooling processing.

  In the example of FIG. 2, the estimation unit 133 estimates a product category corresponding to the product information GD11 based on the output of the model M11. For example, the estimation unit 133 estimates the category of the product corresponding to the product information GD11 as “B category” as indicated by the estimation information ES11.

  For example, the estimation unit 133 calculates a score using a model having an arbitrary structure such as the above-described neural network. Specifically, the model M11 is a value obtained by quantifying the estimation of a predetermined target when product information including image and text information (that is, each element used for calculating the score described above) is input (that is, the model M11). The coefficient is set so as to output a score indicating which category the product corresponding to the product information is likely to belong to. Using the model M11, the estimation unit 133 calculates a score related to each exhibition target.

  In the above example, the model M11 is an example of a model that outputs a value obtained by quantifying the estimation of a predetermined target when product information including image and character information is input. However, the model (model X) according to the embodiment may be a model generated based on a result obtained by repeatedly inputting / outputting data to / from the model M11. For example, the model X may be a model (model Y) that is learned so that product information including image and character information is input and a score output by the model M11 is output. Alternatively, the model M11 may be a model that has been learned so that product information including image and character information is input and the output value of the model Y is output. Further, when the estimation unit 133 performs an estimation process using GAN (Generative Adversarial Networks), the model M11 may be a model that constitutes a part of the GAN.

(Providing unit 134)
The providing unit 134 provides various information. For example, the providing unit 134 provides various information to the administrator device 10 and the sales source device 20. For example, the providing unit 134 provides information related to the subject of listing estimated by the estimating unit 133.

  For example, the providing unit 134 provides information related to a recommended product to a predetermined user based on the information related to the exhibition target estimated by the estimating unit 133. For example, the providing unit 134 provides information for assisting information input in the exhibition based on the information regarding the subject of the exhibition estimated by the estimation unit 133. For example, the providing unit 134 provides product ranking information based on information related to the subject of listing estimated by the estimating unit 133.

  In the example of FIG. 2, the providing unit 134 provides information related to the category estimated by the estimating unit 133 to the selling device 20. For example, the providing unit 134 provides information indicating that the category to be listed by the seller SA1 corresponds to the B category. For example, the providing unit 134 provides the sales source apparatus 20 with information indicating that the category of the product corresponding to the product information GD11 corresponds to the B category.

[4. Generation process flow)
Next, the procedure of the generation process by the estimation system 1 according to the embodiment will be described with reference to FIG. FIG. 8 is a flowchart illustrating an example of the generation process according to the embodiment.

  As shown in FIG. 8, the estimation apparatus 100 acquires learning data (step S101). For example, the estimation apparatus 100 acquires learning data from the learning data storage unit 121.

  Thereafter, the estimating apparatus 100 generates a model based on the learning data (step S102). In the example of FIG. 1, the estimation device 100 generates a model M <b> 11 using learning data from the learning data storage unit 121.

[5. (Estimation process flow)
Next, the procedure of the generation process by the estimation system 1 according to the embodiment will be described using FIG. FIG. 9 is a flowchart illustrating an example of the estimation process according to the embodiment.

  As illustrated in FIG. 9, the estimation apparatus 100 acquires exhibition information from a seller (step S201). In the example of FIG. 2, the estimation device 100 acquires information related to exhibition from the sales source device 20 used by the sales source SA1.

  Moreover, the estimation apparatus 100 estimates the category of the subject of exhibition using a model (step S202). In the example of FIG. 2, the estimation apparatus 100 estimates the category of the product corresponding to the product information GD11 using the model M11 as “B category”.

  Further, the estimating apparatus 100 provides the estimated information to the seller (step S203). In the example of FIG. 2, the estimation device 100 provides information indicating that the category of the product corresponding to the product information GD11 corresponds to the B category to the selling device 20.

[6. Estimated granularity)
Although the case where the category is estimated has been described in the above-described example, the estimation apparatus 100 may estimate the classification of various granularities (hierarchies) without being limited to the category. For example, the estimation device 100 may estimate a product name or the like. This point will be described with reference to FIG. FIG. 10 is a diagram illustrating an example of model generation according to the embodiment. In the example of FIG. 10, the estimation apparatus 100 will be described as an example in which a model M21 that classifies the same product as a product regardless of the sales form or the difference in characteristics is generated. For example, the estimating apparatus 100 generates a model M21 that classifies the same product as a product regardless of the number, color, size (internal capacity), or the like. In addition, about the point similar to FIG. 1, the same code | symbol is attached | subjected to the same site | part and description is abbreviate | omitted suitably.

  In the example illustrated in FIG. 10, the administrator GM1 gives correct information to the product information GD201 including the image IM201, character information CH201, and the like (step S31-1). For example, the administrator GM1 gives correct information indicating that the product of the product information GD201 is “water A” based on the contents of the image IM201 and the character information CH201 included in the product information GD201. In the example of FIG. 10, “water A” is information indicating a product name.

  Further, the manager GM1 gives correct information to the product information GD202 including the image IM202, the character information CH202, and the like (step S31-2). For example, the administrator GM1 gives correct information indicating that the category of the product information GD202 is “water A” based on the contents of the image IM202, the character information CH202, and the like included in the product information GD202.

  As described above, in the example of FIG. 10, the product information GD201 and the product information GD202 are different in the number of items even if they are related to the same product name “water A”. And the same correct answer information is given as an exhibition for the same product “water A”. Note that the manager GM1 is not limited to the difference in the number of items, and the same correct answer information may be assigned to different objects such as colors and sizes (contents) as long as the product names are the same.

  Note that steps S31-1 and S31-2 are for explaining the processing, and any of steps S31-1 and S31-2 may be performed first. Hereinafter, steps S31-1 and S31- When the description is made without distinguishing the two, they are collectively referred to as step S31. For example, the manager GM1 performs the assignment of correct information as shown in step S31 on a large number of product information using learning data. Thereby, in the example of FIG. 10, learning data LD2 in which product information (exhibition information) and correct information (category) are associated with each other is generated.

  And the estimation apparatus 100 acquires the learning data LD2 to which correct information was given in step S31 (step S32). Specifically, the estimating apparatus 100 acquires the product information GD201, GD202, etc. to which the correct answer information is assigned in step S31.

  And the estimation apparatus 100 adds the merchandise information contained in the learning data LD2 acquired by step S32 as learning data (step S33). Specifically, the estimation device 100 adds the learning data LD2 to the learning data storage unit 121A.

  And the estimation apparatus 100 produces | generates a model based on learning data (step S34). For example, the estimation apparatus 100 generates a model by performing learning using the product information GD201, GD202, and the like in the learning data storage unit 121A as learning data (teacher data). Note that “1” and “12” illustrated in the product ID column in the learning data storage unit 121A in FIG. 10 indicate the number included in the corresponding product, and the number is based on the input by the seller. It may be specified or estimated from image information or character information.

  In the example of FIG. 10, the model M21 generated by the estimation apparatus 100 is a model that outputs a plurality of scores SC corresponding to each of a plurality of product names when product information is input. For example, the model M21 is a model that outputs 100 scores corresponding to each of 1000 types of product names when there are 1000 types of product names to be classified.

  From here, although the model M21 shown in FIG. 10 is the structure similar to the model M11 in FIG. 1, the structure of the model M21 is demonstrated easily. The model M21 is a partial model M21-1 having an input layer to which image information is input, a partial model M21-2 having an input layer to which character information is input, and a portion having a layer for outputting a score among the product information Includes model M21-3.

  For example, the partial model M21-1 is a partial model that extracts the feature information FE1 from the input image information IMD. In the example of FIG. 10, it is assumed that the partial model M21-1 includes a so-called convolutional neural network that repeats convolution and pooling of a local area of image information. The partial model M21-1 is not limited to CNN, but may be various neural networks such as RNN, LSTM, and the like.

  Further, for example, the partial model M21-2 is a partial model that extracts the feature information FE2 from the input character information CHD. In the example of FIG. 10, the partial model M21-2 is assumed to be composed of CNN. The partial model M21-2 is not limited to CNN, and may be various neural networks such as RNN, LSTM, and the like. As described above, the model M21 illustrated in FIG. 10 extracts the feature information FE1 from the image information included in the product information, and extracts the feature information FE2 from the character information. That is, the model M21 extracts features independently from image information and character information.

  Further, for example, the partial model M21-3 is a partial model that outputs a score in accordance with the input of the feature information FE1 extracted by the partial model M21-1 and the feature information FE2 extracted by the partial model M21-2. For example, the partial model M21-3 is a partial model to which the output from the partial model M21-1 and the output from the partial model M21-2 are input.

  For example, the output layer in the partial model M21-1 and the output layer in the partial model M21-2 are connected to the input layer of the partial model M21-3. For example, as for the input in the input layer of the partial model M21-3, the node (neuron) in the output layer in the partial model M21-1 and the node (neuron) in the output layer in the partial model M21-2 may be fully coupled.

  As described above, the partial model M21-2 is a partial model that extracts features of the entire product from the feature information FE1 and feature information FE2 and outputs a score. For example, the partial model M21-3 may be configured by CNN. The partial model M21-3 is not limited to CNN, but may be various neural networks such as RNN and LSTM.

  In addition, the partial model M21-1, the partial model M21-2, and the partial model M21-3 are codes given as a toilet for explaining the configuration of each unit in the model M21. That is, when viewed from outside the model M21, the model M21 is one model that outputs a score corresponding to each category when image information and character information, which are product information, are input.

  For example, the estimating apparatus 100 generates a model M21 using the learning data LD2. For example, when the product information GD201 is input to the model M21, the estimation device 100 performs a learning process so that the score corresponding to the water A approaches “1”. For example, when the image IM201 of the product information GD201 is input to the partial model M21-1 of the model M21 and the character information CH201 of the product information GD201 is input to the partial model M21-2, the estimation apparatus 100 is configured to display the partial model M21- The learning process is performed so that the score corresponding to the water A output by 3 approaches “1”. For example, when the product information GD202 is input to the model M21, the estimation device 100 performs a learning process so that the score corresponding to the water A approaches “1”. That is, the estimating apparatus 100 performs the model M21 by processing such as back propagation for correcting the parameter (connection coefficient) so that an error between the output (score) in the model M21 and the correct answer information corresponding to the input product information is reduced. Is generated. For example, the estimating apparatus 100 generates the model M21 by performing a process such as back propagation so as to minimize a predetermined loss function. By repeating the processing as described above, the estimating apparatus 100 can generate a model M21 that outputs a score that enables the product name corresponding to the input product information to be estimated more accurately. Note that the model learning technique is not limited to the technique described above, and any known technique can be applied. In other words, when the product information included in the learning data is input, the estimating apparatus 100 can generate the model M21 by any method as long as the model M21 can be learned so as to output a score corresponding to the correct answer information. May be performed.

  Through the processing as described above, the estimation apparatus 100 generates a model (model M21) identified by the model ID “M21” as shown in the model information storage unit 122 in FIG. Further, as shown in the model information storage unit 122 in FIG. 10, the model M21 is a use “product”, that is, a model used for product classification, and its specific model data is “model data MTD21”. It shows that. For example, the estimation apparatus 100 inputs product information to the model M21, and causes the model M21 to output a score indicating what is likely to be the product name of the input product information, and the model M21 outputs a score. Estimate the product name based on.

  As described above, the estimation apparatus 100 generates a model M21 that classifies the same product as a product regardless of the number, color, size (contents), etc. It can be grasped as sales of the same product. Thereby, the estimation apparatus 100 can improve the accuracy of information in the electronic commerce service.

[7. Example of aggregation processing)
Aggregation based on the product names estimated using the model M21 shown in FIG. 10 will be described with reference to FIG. FIG. 11 is a diagram illustrating an example of a counting process according to the embodiment. As shown in the product information storage unit 123A in FIG. 11, the estimating apparatus 100 totals the number of sales for each product. In the example of FIG. 11, as shown in the product information storage unit 123 </ b> A, the product name estimated by the model M <b> 21 or the like in FIG. 10 is stored in the classification. In the example of FIG. 11, the sales number of the product identified by the product ID “GD201” is “100”. In the example of FIG. 11, the sales number of the product identified by the product ID “GD202” is “10”.

  The estimation apparatus 100 performs aggregation based on the product names estimated using the model M21. In the example of FIG. 11, the estimating apparatus 100 totals the number of sales of the product name “water A” estimated using the model M21 (step S41). Here, in the example of FIG. 11, it is estimated that the product identified by the product ID “GD201” is the product name “water A” and the number included in the product is “1”. In the example of FIG. 11, the product identified by the product ID “GD202” is the product name “water A”, and the number included in the product is estimated to be “12”.

  Therefore, as shown in the total information TL41, the estimation device 100 totals 220 sales numbers of water A based on information such as the product information GD201 and the product information GD202. Specifically, as shown in the calculation formula CF41, the estimating apparatus 100 multiplies the number of sales “100” of the product information GD201 by the number “1” of the product information GD201 and the number of sales “10” of the product information GD202. ”Is multiplied by the number“ 12 ”of the product information GD202 to add up the total sales number of water A to 220.

  For example, when the product information GD201 and the product information GD202 are managed as different products, the number of sales as the water A as described above cannot be calculated. For example, when the sales numbers of the product information GD201 and the product information GD202 are simply added together, the value is “110”, and the meaning of the value is unknown. On the other hand, the estimation apparatus 100 manages the product information GD201 and the product information GD202 as products that target the products having the same product name, and calculates the number of sales based on the number of sales and the number of sales. The sales situation in a desired unit can be grasped. For example, the estimation apparatus 100 can generate a ranking that reflects the sales situation more accurately by generating a ranking of products based on the number of sales totaled in this way.

  In the example of FIG. 11, the estimation apparatus 100 manages the product information GD211 and the product information GD212 having different sizes as products with the same product name “skin lotion C”, and calculates the number of sales (60 = 50 + 10). As a result, it is possible to grasp the sales status of the product name “skin lotion C” in terms of the number of sales. In addition, the estimation apparatus 100 manages the product information GD211 and the product information GD212 as products that target products of the same product name “skin lotion C”, and totals the total capacity based on the number of sales and the capacity. May be. In the example of FIG. 11, the estimation apparatus 100 manages the product information GD221 and the product information GD222 having different colors as products with the same product name “wallet E”, and calculates the number of sales (6 = 2 + 4). Thus, it is possible to grasp the sales status by the number of sales of the product name “Wallet E”.

[8. effect〕
As described above, the estimation apparatus 100 according to the embodiment includes the acquisition unit 131 and the estimation unit 133. The acquisition unit 131 acquires information related to an exhibition in an electronic commerce service that enables purchase of products provided by a plurality of vendors. In addition, the estimation unit 133 estimates information related to an exhibition target based on information related to the exhibition acquired by the acquisition unit 131.

  As described above, the estimation device 100 according to the embodiment can improve the accuracy of information in the electronic commerce service by estimating information related to the subject of the exhibition based on the information related to the exhibition.

  Moreover, in the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates the attribute information of the exhibition target as information related to the exhibition target.

  As described above, the estimation apparatus 100 according to the embodiment can improve the accuracy of information in the electronic commerce service by estimating the attribute information of the exhibition target as the information related to the exhibition target.

  Moreover, in the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates the listing target as information related to the listing target.

  As described above, the estimation apparatus 100 according to the embodiment can improve the accuracy of information in the electronic commerce service by estimating the listing target as information related to the listing target.

  Further, in the estimation apparatus 100 according to the embodiment, the estimation unit 133 is a model used for estimation of an object, and estimates information related to an exhibition target based on an output of a model to which information related to exhibition is input.

  As described above, the estimation apparatus 100 according to the embodiment is a model that is used for estimation of an object, and based on the output of the model to which information related to exhibition is input, It is possible to improve the accuracy of information in a commercial transaction service.

  Moreover, in the estimation apparatus 100 according to the embodiment, the acquisition unit 131 acquires a plurality of types of information related to exhibition. The estimation unit 133 estimates information related to the subject of the exhibition based on the output of the model in which a plurality of types of information related to the exhibition are input.

  As described above, the estimation apparatus 100 according to the embodiment estimates the information related to the subject of the exhibition based on the output of the model in which a plurality of types of information related to the exhibition is input, thereby improving the accuracy of the information in the electronic commerce service. Can be improved.

  Moreover, in the estimation apparatus 100 according to the embodiment, the acquisition unit 131 acquires image information and character information related to the exhibition. The estimation unit 133 estimates information related to an exhibition target based on the output of the model to which image information and character information are input.

  As described above, the estimation apparatus 100 according to the embodiment improves information accuracy in the electronic commerce service by estimating information related to the subject of the exhibition based on the output of the model in which the image information and the character information are input. Can be made possible.

  Moreover, in the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates information related to an exhibition target using a model including a neural network generated based on predetermined information.

  As described above, the estimation apparatus 100 according to the embodiment uses the model including the neural network (CNN) generated based on the predetermined information to estimate the information related to the subject of the exhibition, thereby accurately correcting the information in the electronic commerce service. It is possible to improve the performance.

  Moreover, in the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates information related to an exhibition target using a model including a neural network that performs convolution processing and pooling processing.

  As described above, the estimation apparatus 100 according to the embodiment improves information accuracy in the electronic commerce service by estimating information related to an exhibition target using a model including a neural network that performs convolution processing and pooling processing. Can be made possible.

  In addition, the estimation apparatus 100 according to the embodiment includes a generation unit 132. The generation unit 132 generates a model.

  As described above, the estimation apparatus 100 according to the embodiment can improve the accuracy of information in the electronic commerce service by generating a model and estimating information related to an exhibition target using the generated model. Can do.

  In addition, the estimation apparatus 100 according to the embodiment includes a providing unit 134. The providing unit 134 provides information related to the subject of the exhibition estimated by the estimating unit 133.

  As described above, the estimation apparatus 100 according to the embodiment can provide information with high accuracy in the electronic commerce service by providing information related to the estimated subject of the exhibition.

  In addition, in the estimation apparatus 100 according to the embodiment, the providing unit 134 provides information related to a product recommended to a predetermined user based on information related to an exhibition target estimated by the estimation unit 133.

  As described above, the estimation apparatus 100 according to the embodiment provides highly accurate information in the electronic commerce service by providing information related to the recommended product to a predetermined user based on the information related to the estimated exhibition target. Can be provided.

  Further, in the estimation apparatus 100 according to the embodiment, the providing unit 134 provides information for assisting information input in the exhibition based on the information regarding the subject of the exhibition estimated by the estimation unit 133.

  As described above, the estimation apparatus 100 according to the embodiment provides highly accurate information in the electronic commerce service by providing information that assists information input in the exhibition based on the information related to the estimated exhibition target. can do.

  Moreover, in the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates information related to an exhibition target using information in an electronic commerce service different from the electronic commerce service. Further, the providing unit 134 provides information for assisting information input in the electronic commerce service based on the information related to the exhibition target estimated by the estimating unit 133.

  As described above, the estimation apparatus 100 according to the embodiment uses information in an electronic commerce service (for example, an online shopping mall) that is different from the electronic commerce service (for example, an auction service), and based on information on an estimated exhibition target. By providing information for assisting information input in an electronic commerce service (for example, an auction service), it is possible to provide highly accurate information in an electronic commerce service (for example, an auction service).

  Moreover, in the estimation apparatus 100 according to the embodiment, the estimation unit 133 estimates information related to an exhibition target using a model generated based on information in an electronic commerce service different from the electronic commerce service. Further, the providing unit 134 provides information for assisting information input in the electronic commerce service based on the information related to the exhibition target estimated by the estimating unit 133.

  As described above, the estimation apparatus 100 according to the embodiment uses the model generated based on the information in the electronic commerce service (for example, an online shopping mall) different from the electronic commerce service (for example, the auction service) to estimate the exhibition. By providing information for assisting information input in an electronic commerce service (for example, an auction service) based on information on the target of the target, highly accurate information in the electronic commerce service (for example, an auction service) can be provided.

  Further, in the estimation apparatus 100 according to the embodiment, the providing unit 134 provides product ranking information based on information related to the subject of listing estimated by the estimation unit 133.

  As described above, the estimation apparatus 100 according to the embodiment can provide highly accurate information in the electronic commerce service by providing product ranking information based on the information related to the estimated exhibition target. .

[9. Hardware configuration)
The estimation apparatus 100 according to the embodiment described above is realized by a computer 1000 configured as shown in FIG. 12, for example. FIG. 12 is a hardware configuration diagram illustrating an example of a computer that realizes the function of the estimation device. The computer 1000 includes a CPU 1100, a RAM 1200, a ROM (Read Only Memory) 1300, an HDD (Hard Disk Drive) 1400, a communication interface (I / F) 1500, an input / output interface (I / F) 1600, and a media interface (I / F). ) 1700.

  The CPU 1100 operates based on a program stored in the ROM 1300 or the HDD 1400 and controls each unit. The ROM 1300 stores a boot program executed by the CPU 1100 when the computer 1000 is started up, a program depending on the hardware of the computer 1000, and the like.

  The HDD 1400 stores programs executed by the CPU 1100, data used by the programs, and the like. The communication interface 1500 receives data from other devices via the network N and sends the data to the CPU 1100, and transmits data generated by the CPU 1100 to other devices via the network N.

  The CPU 1100 controls an output device such as a display and a printer and an input device such as a keyboard and a mouse via the input / output interface 1600. The CPU 1100 acquires data from the input device via the input / output interface 1600. In addition, the CPU 1100 outputs the generated data to the output device via the input / output interface 1600.

  The media interface 1700 reads a program or data stored in the recording medium 1800 and provides it to the CPU 1100 via the RAM 1200. The CPU 1100 loads such a program from the recording medium 1800 onto the RAM 1200 via the media interface 1700, and executes the loaded program. The recording medium 1800 is, for example, an optical recording medium such as a DVD (Digital Versatile Disc) or PD (Phase change rewritable disk), a magneto-optical recording medium such as an MO (Magneto-Optical disk), a tape medium, a magnetic recording medium, or a semiconductor memory. Etc.

  For example, when the computer 1000 functions as the estimation apparatus 100 according to the embodiment, the CPU 1100 of the computer 1000 performs control by executing a program or data (for example, model M11 (model data MDT11)) loaded on the RAM 1200. The function of the unit 130 is realized. The CPU 1100 of the computer 1000 reads and executes these programs or data (for example, the model M11 (model data MDT11)) from the recording medium 1800. As another example, these programs or data are transmitted from other devices via the network N. May be obtained.

  As described above, some of the embodiments and modifications of the present application have been described in detail with reference to the drawings. It is possible to carry out the present invention in other forms that have been modified and improved.

[10. Others]
In addition, among the processes described in the above-described embodiments and modifications, all or a part of the processes described as being automatically performed can be manually performed, or are described as being performed manually. All or part of the processing 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 and modifications can be combined as appropriate within a range that does not contradict processing contents.

  In addition, the “section (module, unit)” described above can be read as “means” or “circuit”. For example, the acquisition unit can be read as acquisition means or an acquisition circuit.

DESCRIPTION OF SYMBOLS 1 Estimation system 100 Estimation apparatus 121 Learning data storage part 122 Model information storage part 123 Product information storage part 130 Control part 131 Acquisition part 132 Generation part 133 Estimation part 134 Providing part 10 Administrator apparatus 20 Sales origin apparatus N Network

Claims (19)

An acquisition unit that acquires information related to listings in an electronic commerce service that can purchase products provided by a plurality of vendors;
Based on information related to the listing acquired by the acquisition unit, an estimation unit that estimates information related to the subject of the listing;
An estimation apparatus comprising: The estimation unit includes
The estimation apparatus according to claim 1, wherein the attribute information of the exhibition target is estimated as information related to the exhibition target. The estimation unit includes
The estimation apparatus according to claim 1, wherein the exhibition target is estimated as information related to the exhibition target. The estimation unit includes
The model used for estimation of an object, wherein the information related to the object of the exhibition is estimated based on an output of the model to which the information related to the exhibition is input. The estimation apparatus described in 1. The acquisition unit
Obtaining multiple types of information about the listing,
The estimation unit includes
The estimation apparatus according to claim 4, wherein information relating to the subject of the exhibition is estimated based on an output of the model to which a plurality of types of information relating to the exhibition are input. The acquisition unit
Obtain image information and text information related to the exhibition,
The estimation unit includes
The estimation apparatus according to claim 5, wherein information related to the exhibition target is estimated based on an output of the model to which the image information and the character information are input. The estimation unit includes
The estimation apparatus according to any one of claims 4 to 6, wherein information relating to the exhibition target is estimated using the model including a neural network generated based on predetermined information. The estimation unit includes
The estimation apparatus according to claim 7, wherein information relating to the exhibition target is estimated using the model including the neural network that performs convolution processing and pooling processing. A generating unit for generating the model;
The estimation apparatus according to claim 4, further comprising: A providing unit that provides a service based on information related to the subject of the listing estimated by the estimating unit;
The estimation apparatus according to claim 1, further comprising: The providing unit includes:
The estimation apparatus according to claim 10, wherein information related to the subject of the exhibition estimated by the estimation unit is provided. The providing unit includes:
The estimation device according to claim 10 or 11, wherein information on a product recommended to a predetermined user is provided based on information on the target of the exhibition estimated by the estimation unit. The providing unit includes:
The estimation apparatus according to any one of claims 10 to 12, wherein information for assisting information input in the exhibition is provided based on information on the target of the exhibition estimated by the estimation unit. The estimation unit includes
Using information in an electronic commerce service different from the electronic commerce service, estimating information related to the subject of the exhibition,
The providing unit includes:
14. The estimation according to claim 10, wherein information for assisting information input in the electronic commerce service is provided on the basis of information on the subject of the exhibition estimated by the estimation unit. apparatus. The estimation unit includes
Using a model generated based on information in an electronic commerce service different from the electronic commerce service, estimating information on the subject of the listing,
The providing unit includes:
The estimation according to any one of claims 10 to 14, wherein information for assisting information input in the electronic commerce service is provided based on information related to the subject of the exhibition estimated by the estimation unit. apparatus. The providing unit includes:
The estimation apparatus according to any one of claims 10 to 15, wherein product ranking information is provided on the basis of information related to the exhibition target estimated by the estimation unit. An estimation method performed by a computer,
An acquisition process for acquiring information related to listings in an electronic commerce service that can purchase products provided by a plurality of vendors;
Based on information related to the listing acquired in the acquiring step, an estimation step for estimating information related to the subject of the listing;
The estimation method characterized by including. An acquisition procedure for acquiring information related to listings in an electronic commerce service where products provided by multiple vendors can be purchased;
An estimation procedure for estimating information related to the subject of the exhibition based on information related to the exhibition acquired by the acquisition procedure;
An estimation program for causing a computer to execute. An input layer for inputting image information and character information related to a predetermined object;
The output layer,
A first element belonging to any layer from the input layer to the output layer other than the output layer;
A second element whose value is calculated based on the first element and a weight of the first element;
An operation based on the first element and the weight of the first element is performed on the image information and the character information input to the input layer, with each element belonging to each layer other than the output layer as the first element. In order to output information on the value obtained by quantifying the estimation of the predetermined target from the output layer,
A model for making a computer function.

Images