JP2019061525A - Content recommendation system, content recommendation method, and program - Google Patents

Abstract

To provide a content recommendation system, a content recommendation method, and a program capable of making recommendations by reflecting content in accordance with a preference of a customer on a screen of a design in accordance with the preference of the customer.SOLUTION: A content recommendation system comprises: an acquisition unit that acquires customer attribute information indicating an attribute of a customer and customer behavior history information indicating a behavior history of the customer; a recommendation information generation unit that generates, based on the customer attribute information acquired by the acquisition unit and the customer behavior history information acquired by the acquisition unit, content information that is information related to content recommended to customers and template information indicating a template including an area for displaying information included in the content information using learning type artificial intelligence; and an image generation unit that generates image data in which the content information generated by the recommendation information generation unit is associated with the area included in the template indicated by the template information generated by the recommendation information generation unit.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a content recommendation system, a content recommendation method, and a program.

  In e-commerce, it is practiced to recommend a product to a customer based on information such as the attribute of the customer and the purchase history. In recent years, when recommending a product, it is practiced to analyze information such as customer attributes and purchase history using artificial intelligence and machine learning, and recommend the product based on the analysis result.

  For example, there is known an information processing apparatus for recommending a product that matches the preference of a customer based on product information including the image of the product and the content of the description of the product (Patent Document 1). The information processing apparatus described in Patent Document 1 learns the feature amount of product information and the feature amount of information of a customer who purchases a product by machine learning, and creates a model in which the correlation between both feature amounts is learned. . In the information processing apparatus described in Patent Document 1, information on a customer who has accessed a site for purchasing a product and a feature quantity of the product information are acquired, machine learning based on the created model is performed, and the degree of conformity is determined. calculate. In the information processing apparatus described in Patent Document 1, the order of recommendation of a product is determined based on the calculated degree of matching.

JP, 2016-181196, A

  However, in the information processing apparatus such as the information processing apparatus described in Patent Document 1, the product is calculated based on the feature amount of the product information and the feature amount of the information of the customer who purchases the product. It is only to decide the order of recommendation of Here, in the information processing apparatus such as the information processing apparatus described in Patent Document 1, when recommending a product to a customer, the design of the screen on which the product is displayed is not considered.

  Generally, in electronic commerce, a customer selects an item while viewing an image of the item displayed on the screen of the shopping site. Although there are various screen designs for such shopping sites, preference for screen design is considered to be different for each customer. Therefore, if the design of the screen matches the preference of the customer, it is likely that the product will be easily understood, and the customer will be motivated to purchase. Therefore, it is desirable that not only the product recommended to the customer but also the design of the screen of the shopping site also match the customer's preference in order to sufficiently motivate the customer's purchase intention.

  The present invention has been made in view of the above points, and a content recommending system, a product recommending method, and a program capable of reflecting the content according to the preference of the customer on the screen of the design according to the preference of the customer provide. Here, the content includes a product image and a review about the product.

  (1) The present invention has been made to solve the above-described problems, and one aspect of the present invention acquires customer attribute information indicating an attribute of a customer and customer behavior history information indicating a customer behavior history. Content information, which is information related to content to be recommended to the customer, based on the acquisition unit, the customer attribute information acquired by the acquisition unit, and the customer behavior history information acquired by the acquisition unit, and the content information A recommendation information generation unit that generates template information indicating a region including the information to be included by using learning artificial intelligence, the content information generated by the recommendation information generation unit, and the recommendation information generation unit An image for generating image data associated with the area included in the template indicated by the template information generated by A generating unit, a content recommendation system comprising a.

  (2) Further, according to one aspect of the present invention, there is provided an acquisition process of acquiring customer attribute information indicating an attribute of a customer and customer behavior history information indicating an activity history of the customer, and the customer attribute information acquired in the acquisition process. And a template including an area for displaying information included in the content information, the content information being information on content recommended for the customer based on the customer behavior history information acquired in the acquisition process. The recommendation information generation process of generating template information to be shown by learning type artificial intelligence, the content information generated in the recommendation information generation process, and the template indicated by the template information generated in the recommendation information generation process An image generation unit that generates image data associated with the region including And process, a content recommendation method comprising.

  (3) Further, according to one aspect of the present invention, there is provided an acquiring step of acquiring, on a computer, customer attribute information indicating an attribute of a customer and customer action history information indicating an action history of the customer; An area for displaying content information that is information related to content recommended for the customer based on customer attribute information and the customer behavior history information acquired in the acquisition step, and information included in the content information A recommendation information generation step of generating template information indicating a template to be included by the learning type artificial intelligence, the content information generated in the recommendation information generation step, and the template information generated in the recommendation information generation step The area included in the template is An image generation step of generating a response correlated obtained image data, a program including.

  According to the present invention, it is possible to select the content according to the preference of the customer and to select the template of the design according to the preference of the customer from a plurality of types of templates. Therefore, according to the present invention, it is possible to recommend the content according to the preference of the customer to be reflected on the screen of the design according to the preference of the customer.

It is a figure showing an example of an outline of a contents recommendation system concerning an embodiment of the present invention. It is a figure showing an example of composition of contents recommendation system S1 concerning an embodiment of the present invention. It is a figure which shows an example of the Kansei learning data which concern on embodiment of this invention. It is a figure which shows an example of the content learning data which concern on embodiment of this invention. It is a figure which shows an example of a learning process of the image display production | generation apparatus which concerns on embodiment of this invention. It is a figure which shows an example of the image display production | generation process which concerns on embodiment of this invention. It is a figure which shows an example of the image display information which concerns on embodiment of this invention. It is a figure which shows the 1st example of the template which concerns on embodiment of this invention, and an image display. It is a figure which shows the template and the 2nd example of an image display which concern on embodiment of this invention.

(Embodiment)
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. FIG. 1 is a diagram showing an example of an outline of a content recommendation system S1 according to the present embodiment. In the content recommendation system S1, when the content according to the preference of the customer is recommended to the customer in the internet shopping or direct mail, the design of the screen or the sheet on which the content is displayed is also selected according to the preference of the customer. In the content recommendation system S1, learning artificial intelligence is used to select content and design. Here, the content is information on the product such as an image of the product or information on a review on the product.

The learning type artificial intelligence used in the content recommendation system S1 selects the template for displaying the content and the content to be recommended for each customer when acquiring the customer attribute information and the customer behavior history information (a) ((b) , (C)). Here, the customer attribute information is information indicating an attribute of a customer. The attributes of the customer include the gender and age of the customer. The customer behavior history information is information indicating a customer behavior history. The customer's behavior history is, for example, the history of the customer's purchase history or the history of the response to the customer's direct mail.
The template for displaying content includes design information such as layout patterns and font patterns.
When the template and the content are selected, the selected content is embedded in the selected template (d), and the embedded image is generated as screen display data or print layout data for recommendation (see FIG. e). Here, the screen display data is data of an image defined by HTML (HyperText Markup Language) by the WEB, data of an image displayed by a browser, or data of an image obtained by converting the image as a raster image. The print layout data is data for printing a screen display on a paper medium. As described above, in the content recommendation system S1, screen display data may be generated or print layout data may be generated according to the output destination for each application.

  Here, the learning type artificial intelligence used in the content recommendation system S1 matches the reaction with the image with the customer attribute information and the customer behavior history information, and learns for each customer serving as a learning model. This learning type artificial intelligence classifies the customer who serves as a learning model for each preference based on the learning result. The learning type artificial intelligence used in the content recommendation system S1 matches the customer attribute information and the customer behavior history information with the preference classified in advance, and as a design for displaying the content, it matches the customer's preference Choose a template.

  In the content recommendation system S1, a response to an image is associated with customer attribute information and customer behavior history information, and learning is performed for each customer serving as a learning model. Therefore, in the content recommendation system S1, even if the amount or type of data used to select a template is different from the amount or type of data such as purchase history, the reaction to the image, the customer attribute, and the like As compared with the case where the information and the customer behavior history information are not associated, the preference of the customer can be classified efficiently.

  FIG. 2 is a diagram showing an example of the configuration of the content recommendation system S1 according to the present embodiment. The content recommendation system S1 includes a learning database 1, an image display generation device 2, a customer database 3, and a recommendation database 4.

Kansei learning data 10 and content learning data 11 are stored in the learning database 1.
The sensitivity learning data 10 is data in which a customer ID, customer attribute information, information indicating a customer's sensitivity, and a classification label are associated with one another. Here, the affective learning data 10 will be described with reference to FIG. FIG. 3 is a diagram showing an example of the affective learning data 10 according to the present embodiment. In the example illustrated in FIG. 3, in the present embodiment, a case where questionnaire response information and one or more image feature amount response information are used is illustrated as the information indicating the customer's sensibility.

  In the sensitivity learning data 10, customer attribute information, questionnaire response information, image feature amount response information, and classification labels are associated with each other for each customer ID. However, although only one set of the image feature response information and the classification label is shown for each customer ID in FIG. 3, the image feature response information and the classification label correspond to each of the customer IDs. One or more pairs correspond to each other. That is, a plurality of sets of the reaction shown by one customer for each of a plurality of images and the classification of those images correspond to the customer ID of this customer. How many pairs of image feature amount reaction information and classification labels correspond to customer IDs may differ for each customer ID.

  The customer attribute information includes, for example, gender, age, residence area and family structure. The customer attribute information may include, in addition to these items shown in FIG. 3, a distinction between married singles, occupation, final academic background, payment method, and the like.

The questionnaire response information is information indicating the customer's response to the questionnaire question.
Questionnaire questions whose answers are indicated by the questionnaire response information, “I want to select good performance”, “I want to select good design”, “I want to select new product”, “select as low price as possible It includes questions such as “I want to”, “I want to choose popular items that are selling now”, and “I want to choose manufacturers and brands that can be reassured”. Questionnaire response information, in addition to these questions shown in Figure 3, "I want to select the one with many functions", "I want to select the best deals with discounts and coupons", and "in the net and word of mouth It may include an answer to a question such as “I want to select a good reputation”.

The image feature amount response information is information indicating a response to the customer's image feature amount. Here, the image feature amount is the lightness difference in the image, the color difference in the image, the amount of written information included in the image, and the like. However, the image feature amount may include information indicating features of various images in addition to the lightness difference, the color difference, and the amount of the described information. Here, the reaction to the customer's image feature value indicates whether the customer who visually recognizes an image has some impression and that the customer has a change in feeling or the like.
A classification label shows the result of having classified the image feature-value which image feature-value reaction information shows. This classification is performed in advance using a known clustering method such as the K-means method. The classification label is associated with the customer's preference for the design. The customer's preference for design includes, for example, preference for good design, preference for cute design, and preference for high quality design.

  In the sensitivity learning data 10, an input (customer attribute information, questionnaire response information, image feature amount response information) and an output (classification label) for the input are associated in advance and stored in the learning database 1. Kansei learning data 10 is used to generate a customer model. The customer model is information indicating the classification of a customer specified by a combination of customer attribute information, questionnaire response information, and image feature amount response information.

Returning to FIG. 2, the description of the configuration of the content recommendation system S1 will be continued.
The content learning data 11 is data in which a customer ID, customer attribute information, one or more customer purchase history information, and one or more direct mail reaction history information are associated with each other.

  Here, a specific example of the content learning data 11 will be described with reference to FIG. FIG. 4 is a diagram showing an example of content learning data 11 according to the present embodiment. In the content learning data 11, customer attribute information, one or more customer behavior history information, and one or more direct mail reaction history information are associated with each customer. In FIG. 4, one customer behavior history information is shown among one or more customer behavior history information, and one direct mail reaction history information is shown among one or more direct mail reaction history information. ing. The number of customer behavior history information corresponding to the customer ID and the number of direct mail response history information corresponding to the customer ID may be different for each customer ID.

The customer purchase history information is information indicating the history of the customer's purchase behavior.
The customer behavior history information includes, for example, a purchased product, a single purchase price, a purchase date, a discount rate, a web page browsing history, a favorite product registered on the web page, and a media reaction rate. The media reaction rate is, for example, a numerical value indicating the rate at which the customer reacted when advertising the customer with respect to the purchased product using direct mail or mail magazine.

  The direct mail reaction history information is information indicating the design of direct mail sent to the customer and the history of the customer reaction to the direct mail. The history of the reaction of the customer to the direct mail mentioned here is an action of the customer with respect to the guidance information for the customer from the store side such as the direct mail.

The direct mail reaction history information includes, for example, the design and layout of the direct mail, the ID of the product introduced by the direct mail, the date when the direct mail was sent, the reaction to the direct mail, and the like. The response to direct mail is whether this direct mail has been opened, whether the product link introduced by this direct mail has been clicked, whether the product introduced by this direct mail has been purchased, etc. is there.
Although the direct mail reaction history information shown in FIG. 4 indicates various types of information on direct mail, various types of information on a mail magazine may be included. The direct mail reaction history information may include information indicating whether the customer received direct mail or a mail magazine and visited an actual store.

  The content learning data 11 is associated with product information (not shown) through the information indicating the purchased product included in the customer behavior history information or the information of the ID of the product introduced by direct mail included in the direct mail reaction history information. There is. The product information includes the ID of the purchased product, the minimum keeping unit (SKU), the product image, the price, the release date, the product category, the product description, functions, performance, sales, etc. It is information.

The content learning data 11 is used to predict a customer's purchasing tendency and a response to the customer's product review.
The customer ID included in the affective learning data 10 and the customer ID included in the content learning data 11 may or may not correspond to each other.

  Customer's sensitivity learning data 10 and content learning data 11 are collected in advance in a predetermined number or more and stored in the learning database 1. Kansei learning data 10 and content learning data 11 may be supplied from the EC site 5.

Returning to FIG. 2, the description of the configuration of the content recommendation system S1 will be continued.
The customer database 3 stores customer attribute information 30 and customer behavior history information 31. Here, the customer attribute information 30 indicates the attribute of the customer. The customer behavior history information 31 is information in which customer purchase history information and direct mail reaction history information are associated with each other. The customer attribute information 30 and the customer behavior history information 31 are used by the image display generating device 2 to generate image display data that matches the preference of the customer. The customer attribute information 30 and the customer behavior history information 31 are, for example, data of a customer who uses the EC site 5. The customer attribute information 30 and the customer behavior history information 31 may be data in which data of customers other than the customer using the EC site 5 is collected.

  Content information is stored in the recommendation database 4. Content information is information indicating candidates for content to be recommended to the customer. The content information includes, for example, template information 40, product information 41, greeting information 42, and word-of-mouth information 43. The template information 40 is information indicating a template including an area for displaying information included in the content information. In the template information 40, the preference for the customer's design is associated with the type of template.

The template includes layout patterns and font patterns. The layout pattern is information in which one or more areas for displaying content such as a product image, a number indicating a discount, a greeting sentence, and a word of mouth are provided, and the layout of the area is defined. The number of areas for displaying these contents is different for each template.
The font pattern is information for specifying a font for displaying text characters when text data is embedded in the layout area.
The template may include at least a layout pattern.

  The product information 41 is information indicating a product to be sold or recommended to the customer at the EC site 5. That is, the product information 41 is information on a product to be recommended to the customer. The product information 41 includes information of a product image to be embedded in the template indicated by the template information 40. The greeting information 42 is information indicating a greeting sentence to be embedded in the template indicated by the template information 40. There are types of greeting sentences indicated by the greeting information 42 for each customer preference. Word-of-mouth information 43 is information which shows the text of word-of-mouth information embedded in the template which template information 40 shows.

  The image display generation device 2 generates image display data for recommending to the customer. The image display generation device 2 includes an emotional learning data acquisition unit 20, a content learning data acquisition unit 21, a recommendation information generation unit 22, a customer information acquisition unit 23, a template information acquisition unit 24, a content information acquisition unit 25, and the like. , And an image display generation unit 26.

The sensitivity learning data acquisition unit 20 acquires sensitivity learning data 10 from the learning database 1. The sensitivity learning data acquisition unit 20 supplies the acquired sensitivity learning data 10 to the recommendation information generation unit 22.
The content learning data acquisition unit 21 acquires content learning data 11 from the learning database 1. The content learning data acquisition unit 21 supplies the acquired content learning data 11 to the recommendation information generation unit 22.

The customer information acquisition unit 23 acquires customer attribute information 30 and customer behavior history information 31 from the customer database 3. The customer information acquisition unit 23 supplies the acquired customer attribute information 30 and the customer behavior history information 31 to the recommendation information generation unit 22.
The template information acquisition unit 24 acquires template information 40 from the recommendation database 4. The template information acquisition unit 24 supplies the acquired template information 40 to the recommendation information generation unit 22.

The content information acquisition unit 25 acquires the product information 41 from the recommendation database 4. The content information acquisition unit 25 supplies the acquired product information 41 to the recommendation information generation unit 22.
The content information acquisition unit 25 acquires the greeting information 42 from the recommendation database 4. The content information acquisition unit 25 supplies the acquired greeting information 42 to the recommendation information generation unit 22.
The content information acquisition unit 25 acquires word-of-mouth information 43 from the recommendation database 4. The content information acquisition unit 25 supplies the acquired word-of-mouth information 43 to the recommendation information generation unit 22.

The recommendation information generation unit 22 is based on the customer attribute information 30 acquired by the customer information acquisition unit 23 and the customer behavior history information 31 acquired by the customer information acquisition unit 23 (content information 41, greeting information 42, Content information selected from the word-of-mouth information 43) and template information selected from the template information 40 are generated by the learning type artificial intelligence (the sensitivity learning unit 220 and the content learning unit 221). The recommendation information generation unit 22 generates content information and template information by, for example, selecting content that matches the customer's preference and a template that matches the customer's preference.
The recommendation information generation unit 22 includes a sensitivity learning unit 220, a content learning unit 221, a customer model classification unit 222, a template selection unit 223, and a content selection unit 224.

  The sensitivity learning unit 220 learns the preference for the customer's design for each customer serving as a model based on the sensitivity learning data 10 acquired by the sensitivity learning data acquisition unit 20. The sensitivity learning unit 220 learns the customer's sensitivity for each customer by learning the preference of the customer for the design. The customer that the sensitivity learning unit 220 models is modeled by the customer model. That is, the customer as the model of the affective learning unit 220 is modeled by the customer attribute information, the questionnaire response information, and the image feature amount response information included in the affective learning data 10.

  The sensitivity learning unit 220 includes an individual sensitivity learning unit 220-i (i = 1, 2,..., N: N is the number of customers serving as a model). Each individual sensitivity learning unit 220-i is assigned to each customer serving as a model. Each of the individual sensitivity learning units 220-i learns a classification label to be output when questionnaire response information and image feature amount response information are given as input for each customer serving as a model. Here, the classification label is information indicating the preference of the customer for the design. That is, the individual sensitivity learning unit 220-i learns the sensitivity for each customer serving as the assigned model, and models the customer's sensitivity.

  The content learning unit 221 learns the behavior history of the customer for each customer based on the content learning data 11 acquired by the content learning data acquisition unit 21 and the product information 41 acquired by the content information acquisition unit 25. The customer's behavior history is the history of the customer's purchase history and the response to the customer's direct mail. The customer as a model of the content learning unit 221 is classified based on customer attribute information, customer purchase history information, and direct mail reaction history information included in the content learning data 11.

  The content learning unit 221 includes a personal content learning unit 221-i (i = 1, 2,..., N: N is the number of customers serving as a model). Each of the personal content learning units 221-i is assigned to each customer serving as a model. Each of the personal content learning units 221-i learns, for each customer serving as an assigned model, a customer's purchasing tendency serving as the model and a reaction to a review of a product. The content learning unit 221 calculates, based on the learning result, a function (for example, a predicted product score function or a predicted word score function) for predicting the purchase tendency of the customer and the reaction of the customer for the word of mouth. Do.

Upon acquiring the customer attribute information 30 and the customer behavior history information 31, the customer model classification unit 222 classifies the customer into a customer model for each customer indicated by the customer attribute information 30 and the customer behavior history information 31. The customer model classification unit 222 supplies the classified result to the template selection unit 223 and the content selection unit 224.
The customer model classification unit 222 associates the customer model of the personal affective learning unit 220-i with the customer model of the personal content learning unit 221-i based on the correlated result. Here, the customer model of the personal content learning unit 221-i is information in which a combination of customer attribute information, customer purchase history information, and direct mail reaction history information is associated with the customer model.

The template selection unit 223 selects, based on the customer model acquired from the customer model classification unit 222, the individual sensitivity learning unit 220-i corresponding to the customer model from among the plurality of individual sensitivity learning units 220-i. The template selection unit 223 acquires the learning result of the individual affective learning unit 220-i from the selected individual affective learning unit 220-i.
The template selection unit 223 selects a template according to the preference for each customer based on the customer attribute information 30 and the customer behavior history information 31 acquired by the customer information acquisition unit 23 and the acquired learning result. The template selection unit 223 supplies the image display generation unit 26 with information indicating the selected template.

  The content selection unit 224 selects, based on the customer model acquired from the customer model classification unit 222, the personal content learning unit 221-i corresponding to the customer model from among the plurality of personal content learning units 221-i. The content selection unit 224 acquires the learning result of the personal content learning unit 221-i from the selected personal content learning unit 221-i.

  Among the products indicated by the product information 41 acquired by the content information acquisition unit 25, the content selection unit 224 is based on the customer attribute information 30 and the customer behavior history information 31 acquired by the customer information acquisition unit 23 and the acquired learning result. Then, one or more products are selected according to the preference of the customer for each customer. The content selection unit 224 supplies the image display generation unit 26 with information indicating the selected product.

  The content selection unit 224 selects one of the greetings indicated by the greeting information 42 by the content information acquisition unit 25 based on the customer attribute information 30 and the customer behavior history information 31 acquired by the customer information acquisition unit 23 and the acquired learning result. Select one or more greetings that match the customer's preferences for each customer. The content selection unit 224 supplies the selected greeting sentence to the image display generation unit 26.

  The content selection unit 224 uses the customer attribute information 30 and the customer behavior history information 31 acquired by the customer information acquisition unit 23 and the text of the review indicated by the word-of-mouth information 43 acquired by the content information acquisition unit 25 based on the acquired learning result. From among, select one or more word-of-mouth texts that match the customer's preference for each customer. The content selection unit 224 supplies the text of the selected review to the image display generation unit 26.

  The content selection unit 224 calculates a discount rate or discount price that matches the preference of the customer for each customer based on the selected product and the customer behavior history information 31 acquired by the customer information acquisition unit 23. The content selection unit 224 supplies the calculated discount rate or discount price to the image display generation unit 26.

The image display generation unit 26 associates the content (product, greeting, word, word, discount rate or discount price) selected by the recommendation information generation unit 22 with the area included in the template selected by the recommendation information generation unit 22. Generate image display data. Here, the image display data may be, for example, data indicating an image in which the content is embedded in the area included in the template. Further, the image data may be, for example, data including data indicating a content, data indicating a template, and an instruction indicating embedding of the content in an area included in the template.
The image display generation unit 26 supplies the generated image display data to the EC site 5. The image display generation unit 26 may supply the generated image display data to the database of the business operator as data for printing.

The EC site 5 supplies the image display acquired from the image display generation device 2 to the customer terminal 6. The EC site 5 creates, for example, a direct mail or a mail magazine based on the image display data acquired from the image display generation device 2 and supplies it to the customer terminal 6. Further, for example, upon receiving an access from the customer terminal 6, the EC site 5 supplies the image display data acquired from the image display generation device 2 to the customer terminal 6.
The EC site 5 collects information for generating the customer attribute information 30 and the customer behavior history information 31 based on the information received from the customer terminal 6. The EC site 5 supplies the collected information to the customer database 3. The EC site 5 may supply the collected information to the learning database 1 and update the content learning data 11.

  The learning process of the image display generation device 2 will be described with reference to FIG. FIG. 5 is a diagram showing an example of learning processing of the image display generation device 2 according to the present embodiment. The learning process shown in FIG. 5 is executed in advance before the content recommendation system S1 is operated.

  The sensitivity learning data acquisition unit 20 acquires the sensitivity learning data 10 from the learning database 1 (step S10). The sensitivity learning data acquisition unit 20 supplies the acquired sensitivity learning data 10 to the sensitivity learning unit 220.

  The sensitivity learning unit 220 assigns the individual sensitivity learning unit 220-i to each customer ID indicated by the sensitivity learning data 10 acquired from the sensitivity learning data acquisition unit 20. The sensitivity learning unit 220 supplies the acquired sensitivity learning data 10 to the individual sensitivity learning unit 220-i assigned to the customer ID indicated by the sensitivity learning data 10.

  The individual affective learning unit 220-i executes learning based on the supplied affective learning data 10 (step S20). The personal affection learning unit 220-i receives the customer attribute indicated by the customer attribute information corresponding to the assigned customer ID, the questionnaire response indicated by the questionnaire response information, and the image feature amount reaction information as an input of the reaction to the image feature amount. The learning is performed so as to output the classification label corresponding to the assigned customer ID when it is received. However, since there are a plurality of sets of image feature quantity reaction information and classification labels for each customer ID, the number of sets of image feature quantity reaction information and classification labels is the data used by the individual affectability learning unit 220-i for learning. It becomes a number.

  The individual affective learning unit 220-i executes learning using a known method of machine learning. The individual affective learning unit 220-i executes learning using, for example, a random forest or a neural network.

  The content learning data acquisition unit 21 acquires content learning data 11 from the learning database 1 (step S30). The content learning data acquisition unit 21 supplies the acquired content learning data 11 to the content learning unit 221.

  The content information acquisition unit 25 acquires the product information 41 and the word-of-mouth information 43 from the recommendation database 4 (step S40). The content information acquisition unit 25 supplies the acquired product information 41 and the word-of-mouth information 43 to the content learning unit 221.

  The personal content learning unit 221-i (i = 1, 2,..., N: N is the number of customers serving as a model) includes the content learning data 11 acquired by the content learning data acquiring unit 21 and the content information acquiring unit Based on the product information 41 acquired by the information processing unit 25, the customer's purchasing tendency and the response to the customer's product review are learned for each customer serving as a model (step S 50).

The content learning unit 221 calculates a product feature amount vector for each product indicated by the product information 41 from the acquired product information 41. Here, the product feature amount vector corresponds to, for example, the representative color included in the product image, the ratio of the representative color in the product image, the brand, the product category, the price, the discount rate, and the like, respectively. , It is a vector which makes those numerical values a component.
The content learning unit 221 calculates word-of-mouth feature amount vectors that are feature amount vectors indicating various features of the text of the word of mouth from the obtained word-of-mouth information 43.
The content learning unit 221 supplies the calculated product feature amount vector and the calculated word feature amount vector to each of the personal content learning units 221-i.

The content learning unit 221 assigns the customer ID included in the content learning data 11 to each of the personal content learning units 221-i. Thus, the content learning unit 221 assigns a customer serving as a model to each of the personal content learning units 221-i.
The personal content learning unit 221-i extracts customer purchase history information corresponding to the assigned customer ID from the content learning data 11.

The personal content learning unit 221-i calculates a product score for each product indicated by the acquired product information 41 based on the extracted customer purchase history information. Here, the product score is a numerical value indicating a customer's evaluation of the product.
The personal content learning unit 221-i derives a predicted product score function based on the calculated product score and the product feature quantity vector of the product corresponding to the product score for the assigned customer ID. Here, the predicted product score function is a function that outputs a product score when a product feature amount vector is input. The personal content learning unit 221-i derives a predicted product score function using a known machine learning method.

The personal content learning unit 221-i calculates a word-sword score for each word shown by the acquired product information 41 based on the extracted customer purchase history information. Here, the word-of-mouth score is a numerical value indicating the customer's evaluation for the word-of-mouth.
The personal content learning unit 221-i derives a predicted word-of-mouth score function based on the calculated word-of-mouth score and the word-of-mouth feature amount vector of word-of-mouth corresponding to the word-of-mouth score. Here, the predicted word-of-mouth score function is a function that outputs a word-of-mouth score when a word-of-mouth feature amount vector is input. The personal content learning unit 221-i derives a predicted word score function using a known machine learning method.

  In the present embodiment, the personal content learning unit 221-i has described the form in which the predicted product score function and the predicted word-of-mouth score function are derived based on only the information of each individual. -The form which i derives a prediction goods score function and a prediction word-of-mouth score function is not restricted to this. The personal content learning unit 221-i is configured to predict each score using information (customer behavior history information 31) of other customers using distance information between customers specified using the sensitivity learning data. It may be Here, the distance information between customers specified using the affective learning data is an index indicating how similar the feature amounts expressed using the affective learning data are among the customers.

  The customer model classification unit 222 associates the customer model of the individual sensitivity learning unit 220-i with the customer model of the individual content learning unit 221-i (step S60). The customer model classification unit 222 associates each individual sensitivity learning unit 220-i with each individual content learning unit 221-i based on the sensitivity learning data 10 and the content learning data 11.

  The customer model classification unit 222 associates the customer purchase history included in the content learning data 11 used for learning by the personal content learning unit 221-i with the questionnaire response information included in the sensitivity learning data 10. The customer model classification unit 222 associates direct mail reaction history information included in the content learning data 11 used for learning by the personal content learning unit 221-i with the image feature amount reaction information included in the sensitivity learning data 10.

  The customer model classification unit 222 uses a predetermined algorithm from the customer purchase history included in the content learning data 11 for the item for which the response is not obtained among the questionnaire response information included in the sensitivity learning data 10 Generate In this predetermined algorithm, the questionnaire response indicated by the questionnaire response information is uniquely determined from the customer purchase history. The customer model classification unit 222 associates the customer purchase history with the questionnaire response information based on the generated questionnaire response.

For example, the customer model classification unit 222 refers to the information on the function and performance among the product information corresponding to the purchased product in the customer purchase history information, and for the questionnaire “I want to select one with good performance” in the questionnaire response information. Generate an answer.
For example, the customer model classification unit 222 refers to the product image information among the product information associated with the purchased product in the customer purchase history information, and answers the questionnaire “I want to select a good design” in the questionnaire response information. Generate
For example, the customer model classification unit 222 refers to the information on the release date among the product information associated with the purchased product in the customer purchase history information or the information on the purchase date in the customer purchase history information, and Generate an answer to the “I want to select a new product” questionnaire.
The customer model classification unit 222, for example, refers to the discount rate in the customer purchase history, and generates an answer to the questionnaire “I would like to select the one with the lowest price possible” in the questionnaire response information.
The customer model classification unit 222 refers to, for example, sales information among the product information corresponding to the purchased product of the customer purchase history information, and for the questionnaire response information, “I want to select a popular item sold now” Generate answers to the questionnaire.
The customer model classification unit 222, for example, the information of the ID of the purchased product among the product information associated with the purchased product of the customer purchase history information and the information of the descriptive text about the product, and the browsing history of the Web among the customer purchase history. Refer to the information in and generate an answer to the questionnaire “I would like to select a maker or brand that can be reassured” from the questionnaire answer information.

  The customer model classification unit 222 generates an answer to the questionnaire for each customer purchase history because one or more customer purchase histories correspond to the customer ID in the content learning data 11. For each questionnaire indicated by the questionnaire response information, the customer model classification unit 222 sets, for example, the more “Yes” or “No” of the responses to the questionnaire generated for each customer purchase history as the response to this questionnaire.

The customer model classification unit 222 generates image feature amount reaction information included in the affective learning data 10 from the direct mail reaction history information included in the acquired content learning data 11 using a predetermined algorithm. In this predetermined algorithm, the image feature amount reaction information is uniquely determined from the direct mail reaction history information.
The customer model classification unit 222, for example, from the information on the product image among the product information associated with the direct mail design and layout information included in the direct mail reaction history information, the product information associated with the presented product ID, and the reaction to the direct mail Image feature amount response information is generated.

  The customer model classification unit 222 generates image feature amount reaction information for each direct mail reaction history information because the content learning data 11 corresponds to one or more direct mail reaction history information with respect to the customer ID. The customer model classification unit 222, for each response to the image feature amount indicated by the image feature amount reaction information, for example, information indicating a reaction indicated by the image feature amount reaction information generated for each direct mail reaction history information, and a reaction The larger one of the information indicating that the image is not to be generated is used as this image feature amount response information.

In the present embodiment, although the case where the customer model classification unit 222 generates the image feature amount reaction information based on the direct mail reaction history information has been described, the customer model classification unit 222 performs the image feature amount reaction information. The method of generation is not limited to this. The customer model classification unit 222 generates image feature amount reaction information from, for example, reaction information to an image presented to the customer in advance in a questionnaire or the like, or feature amount information of image information of an item included in the customer purchase history information. It is also good.
As a method of obtaining image feature amount reaction information from customer purchase history information, for example, purchase history information for a product having an image feature amount similar to an image specified in advance in an image feature amount space as a product image A method of obtaining image feature amount reaction information is conceivable. As another example of a method for obtaining image feature amount reaction information from customer purchase history information, image feature amount reaction information is predicted from an image feature amount of a product image included in the purchase history and a reaction to a product image included in the purchase history It is conceivable to use a machine learning device.

The customer model classification unit 222 includes customer attribute information included in the content learning data 11 used for learning by the personal content learning unit 221-i and the content learning data 11 used for learning by the personal content learning unit 221-i A set of questionnaire response information associated with a customer purchase history and image feature response information associated with direct mail reaction history information included in the content learning data 11 used for learning by the personal content learning unit 221-i Determine the customer model that corresponds to
The customer model classification unit 222 determines the individual sensitivity learning unit 220-i based on the determined customer model. The customer model classification unit 222 associates the determined customer model of the individual sensitivity learning unit 220-i with the customer model of the individual content learning unit 221-i.

When there are a plurality of determined customer IDs, the customer model classification unit 222 matches the customer model of the personal content learning unit 221-i among the individual affectability learning units 220-i to which the determined customer IDs are assigned. The customer model of the individual sensitivity learning unit 220-i not corresponding to the customer model of the individual content learning unit 221-i is associated with each other.
When all of the individual affectability learning units 220-i to which the determined customer IDs are assigned already correspond to the customer models of the individual content learning units 221-i, the customer models of the individual content learning units 221-i are , And may not correspond to the customer model of the individual affective learning unit 220-i.

  The process of the image display generation device 2 will be described with reference to FIG. FIG. 6 is a diagram showing an example of the image display generation process according to the present embodiment. Here, a process of selecting the content and the template in accordance with the preference for the design of the customer using the learning result obtained by the learning process shown in FIG. 5 will be described.

  The customer information acquisition unit 23 acquires the customer attribute information 30 and the customer behavior history information 31 from the customer database 3 (step S100). The customer information acquisition unit 23 supplies the acquired customer attribute information 30 and the customer behavior history information 31 to the customer model classification unit 222. The customer information acquisition unit 23 supplies the acquired customer attribute information 30 and the customer behavior history information 31 to the template selection unit 223. The template selection unit 223 supplies the acquired customer attribute information 30 and the customer behavior history information 31 to the content selection unit 224.

  The template information acquisition unit 24 acquires template information 40 from the recommendation database 4 (step S101). The template information acquisition unit 24 supplies the acquired template information 40 to the template selection unit 223.

  The content information acquisition unit 25 acquires the product information 41, the greeting information 42, and the word-of-mouth information 43 as content information from the recommendation database 4 (step S102). The content information acquisition unit 25 supplies the acquired product information 41, greeting information 42, and word-of-mouth information 43 to the content selection unit 224.

  The customer model classification unit 222 acquires the customer attribute information 30 and the customer behavior history information 31 from the customer information acquisition unit 23. The customer model classification unit 222 classifies the customer by determining a customer model for each customer ID indicated by the acquired customer attribute information 30 and customer behavior history information 31 (step S103).

The customer model classification unit 222 generates questionnaire response information from customer purchase history information included in the acquired customer behavior history information 31. The customer model classification unit 222 generates image feature amount reaction information from the direct mail reaction history information included in the acquired customer behavior history information 31. The customer model classification unit 222 indicates the acquired customer attribute information 30 and the customer behavior history information 31 based on a set of the acquired customer attribute information 30, the generated questionnaire response information, and the generated image feature amount response information. A customer model is determined for each customer ID.
In the case where items of information included in the customer attribute information 30 and the customer action history information 31 are insufficient, the customer model classification unit 222 corresponds to the customer model with information indicating that these items are insufficient. Assign to an item
The customer model classification unit 222 supplies the set of the customer ID and the calculated customer model to the template selection unit 223. The customer model classification unit 222 supplies the combination of the customer ID and the calculated customer model to the content selection unit 224.

  The template selection unit 223 acquires template information from the template information acquisition unit 24. The template selection unit 223 acquires a set of the customer ID and the customer model from the customer model classification unit 222. The template selection unit 223 acquires the customer attribute information 30 and the customer behavior history information 31 from the customer information acquisition unit 23.

  The template selection unit 223 selects, for each of the acquired customer IDs, the individual sensitivity learning unit 220-i having a customer model forming a pair with the customer ID. If there are a plurality of individual sensitivity learning units 220-i having this customer model, template selecting unit 223 selects individual sensitivity learning units 220-i according to a predetermined order from among the plurality of individual sensitivity learning units 220-i. Do.

  The template selection unit 223 acquires the learning result from the selected individual sensitivity learning unit 220-i. Here, the result learned by the individual affective learning unit 220-i is information on a decision tree when the individual affective learning unit 220-i executes learning using, for example, a random forest. When the individual affective learning unit 220-i executes learning using, for example, a neural network, the result learned by the individual affective learning unit 220-i is information on the weight of the neural network.

The template selection unit 223 uses the customer attribute information 30 acquired from the customer information acquisition unit 23, the customer behavior history information 31 acquired from the customer information acquisition unit 23, and the customer indicated by the customer attribute information 30 based on the acquired learning result. Calculate the classification label for each ID.
The template selection unit 223 selects the type of template corresponding to the calculated classification label for each customer ID from the acquired template information 40 (step S104). The classification label and the type of template are associated in advance. The template selection unit 223 associates the selected template type with the customer ID, and supplies this to the image display generation unit 26.

  When the item of the customer model forming a pair with the acquired customer ID includes information indicating that the item is insufficient, the template selection unit 223 ignores the insufficient item and the individual The sensitivity learning unit 220-i may be selected. Even if the type of information included in the customer attribute information 30 and the customer behavior history information 31 is insufficient, the template selection unit 223 is a customer close to the customer's sensitivity indicated by the customer attribute information 30 and the customer behavior history information 31. The personal affective learning unit 220-i may be selected that has a customer model that indicates the classification of. Even if the types of information included in the customer attribute information 30 and the customer behavior history information 31 are insufficient, the template selection unit 223 selects a template based on the learning result of the selected individual sensitivity learning unit 220-i. It can be selected.

  The content selection unit 224 acquires the product information 41, the greeting information 42, and the word-of-mouth information 43 from the content information acquisition unit 25. The content selection unit 224 acquires a set of the customer ID and the customer model from the customer model classification unit 222. The content selection unit 224 acquires the customer behavior history information 31 from the customer information acquisition unit 23.

The content selection unit 224 selects a personal content learning unit 221-i having a customer model associated with the customer model of the individual affectability learning unit 220-i selected by the template selection unit 223. When there are a plurality of personal content learning units 221-i having a customer model associated with the customer model of the individual affectability learning unit 220-i selected by the template selection unit 223, the content selection unit 224 selects the plurality of individuals The personal content learning unit 221-i is selected from the content learning units 221-i in a predetermined order.
The content selection unit 224 acquires a predicted product score function and a predicted word-of-mouth score function from the selected personal content learning unit 221-i.

The content selection unit 224 selects a product from among the products indicated by the product information 41 based on the acquired predicted product score function (step S105). The content selection unit 224 selects a product for each customer ID included in the customer behavior history information 31.
The content selection unit 224 calculates, for each product indicated by the acquired product information 41, a product feature amount vector for the product. The content selection unit 224 calculates a product score for each product indicated by the acquired product information 41 based on the calculated product feature amount vector and the acquired predicted product score function. The content selection unit 224 selects a predetermined number of products from among the products included in the product information 41 in descending order of the calculated product scores. Here, the predetermined number may be equal to or larger than the maximum number of areas for displaying product images that may be included in the template. In the present embodiment, the predetermined number is ten as an example.

The content selection unit 224 selects a greeting from the greetings included in the greeting information 42 (step S106).
The content selection unit 224 uses the same product category as the product calculated as having the highest product score among the products selected from the product information 41 among the purchased products indicated by the customer purchase history information included in the customer behavior history information 31 Extract the corresponding purchased goods. The content selection unit 224 generates a questionnaire response indicated by the questionnaire response information from the customer purchase history information corresponding to the extracted purchased product, using a predetermined algorithm. The content selection unit 224 selects a greeting sentence from the greeting information 42 acquired from the content information acquisition unit 25 in accordance with the generated questionnaire response. The greeting texts indicated by the greeting information 42 are, for example, “To you who want to get deals!”, “To you who has purchased high quality goods”, and the like.
The content selection unit 224 may generate a greeting according to the extracted purchased item. For example, when the extracted purchased product is a watch, the content selection unit 224 generates a text of “Greet to buy a watch!” As a greeting.

The content selection unit 224 selects a review from the reviews indicated by the review information 43 based on the acquired predicted word-of-mouth score function (step S107). The content selection unit 224 selects a review for each customer ID included in the customer behavior history information 31.
The content selection unit 224 calculates, for each of the reviews indicated by the acquired review information 43, a word-of-mouth feature amount vector for this review. The content selection unit 224 calculates a word score for each word indicated by the acquired word-of-mouth information 43 based on the calculated word feature amount vector and the acquired predicted word-of-mouth score function. The content selection unit 224 selects a predetermined number of reviews from the text of the review included in the review information 43 in descending order of the calculated review score. Here, the predetermined number may be equal to or larger than the maximum number of areas for displaying reviews that may be included in the template. In the present embodiment, the predetermined number is four as an example.

The content selection unit 224 selects a discount rate (step S108). The content selection unit 224 uses the same product category as the product calculated as having the highest product score among the products selected from the product information 41 among the purchased products indicated by the customer purchase history information included in the customer behavior history information 31 Extract the corresponding purchased goods. The content selection unit 224 selects the discount rate of the extracted purchased product from the customer purchase history information.
For a customer whose number of pieces of customer purchase history information included in the customer behavior history information 31 is smaller than a predetermined number, for example, the content selection unit 224 may convert the selected discount rate into a discount price.

  The content selection unit 224 generates product information by associating the customer ID with the ID of a predetermined number of products. The content selection unit 224 supplies the generated image information, the ID of the greeting, the ID of a predetermined number of reviews, the discount rate or the discount price to the image display generation unit 26 in association with each customer ID. .

The image display generation unit 26 acquires the template and the customer ID from the template selection unit 223. The image display generation unit 26 acquires, from the content selection unit 224, a customer ID, an ID of a predetermined number of products, an ID of a greeting, an ID of a predetermined number of reviews, and a discount rate or discount price.
The image display generation unit 26 acquires the type of acquired template, the ID of the acquired product, the ID of the acquired greeting sentence, the ID of the acquired predetermined number of reviews, the acquired discount rate or discounted price, and the like. The image display information is generated in association with the selected customer ID.
The image display generation unit 26 generates image display data based on the generated image display information (step S109). The image display generation unit 26 supplies the generated image display data to the EC site 5.

  FIG. 7 is a diagram showing an example of image display information according to the present embodiment. In the image display information, the customer ID, the type of template, the ID of a predetermined number of products, the text of a greeting, the text of a predetermined number of reviews, and the discount rate or discount price are mutually associated. Information.

  FIG. 8 is a view showing a first example of a template and an image display according to the present embodiment. In the template TA, variable regions RA1 to RA7 are provided. The image display GA is an example of an image in which the contents CA1 to CA7 are embedded in the variable area RA1 to the variable area RA7 of the template TA.

  Content CA1 which is a goods picture of goods 1 and goods 2 among goods which image display information shows in variable field RA1 is embedded. Content CA2 which is a discount amount is embedded in variable region RA2. A greeting sentence is embedded in the variable area RA3. In the variable area RA4 to the variable area RA6, content CA4, content CA5 and content CA6 which are texts of word of mouth are embedded. In the variable area RA7, content CA7 which is a product image of the product 3 to the product 10 among the products indicated by the image display information is embedded.

FIG. 9 is a view showing a second example of the template and the image display according to the present embodiment. In the template TB, variable regions RB1 to RB9 are provided. The image display GB is an example of an image in which the content CB1 to the content CB9 are embedded in the variable region RB1 to the variable region RB9 of the template TB.
Template TB (FIG. 9) is different from template TA (FIG. 8) in the number of variable areas for displaying a product image of a product. Template TB (FIG. 9) is different from template TA (FIG. 8) in the number of variable areas for displaying word of mouth.

  Content CB1 which is a goods picture of goods 1 among goods which image display information shows is embedded in variable field RB1. Content CB2 which is a goods picture of goods 2 among goods which image display information shows is embedded in variable field RB2. Content CB3 which is a discount rate is embedded in variable region RB3. Content CB4 which is a greeting sentence is embedded in variable region RB4. In the variable region RB5 to the variable region RB8, content CB6 to content CB8 which are texts of word of mouth are embedded respectively. In the variable region RB9, content CB9 which is a product image of the product 3 to the product 6 among the products indicated by the image display information is embedded. In the template TB, the product images of the product 7 to the product 10 among the products indicated by the image display information are not displayed.

The template information 40 may include, for each template, information indicating a rule for embedding an image. The rule for embedding an image is a rule for designating which variable region the product, a greeting, a word of mouth, or a number indicating a discount is to be embedded in. The rules for embedding the image may be updated based on the customer purchase history information of the content learning data 11 and the direct mail reaction history.
Further, in the template information 40, information on the layout of the template and information on the font may be held separately.

As described above, the image display generation device 2 of the content recommendation system S1 according to the present embodiment includes the customer information acquisition unit 23, the recommendation information generation unit 22, and the image display generation unit 26.
The customer information acquisition unit 23 acquires customer attribute information 30 indicating an attribute of a customer and customer action history information 31 indicating an action history of the customer.
The recommendation information generation unit 22 is content information that is information on a product to be recommended to the customer based on the customer attribute information 30 acquired by the customer information acquisition unit 23 and the customer behavior history information 31 acquired by the customer information acquisition unit 23. And template information indicating a template including an area for displaying information included in the content information are generated by the learning type artificial intelligence (the affective learning unit 220 and the content learning unit 221).
The image display generation unit 26 generates image data in which the content information generated by the recommendation information generation unit 22 and the area included in the template indicated by the template information generated by the recommendation information generation unit 22 are associated with each other.

With this configuration, in the content recommendation system S1 according to the present embodiment, it is possible to select a product according to the preference of the customer, and to select a template of a design according to the preference of the customer from a plurality of types of templates. . Therefore, according to the present invention, it is possible to recommend the image of the product according to the preference of the customer to be reflected on the screen of the design according to the preference of the customer.
Further, in the content recommendation system S1 according to the present embodiment, the amount and type of data used for selecting a template are different from the amount and type of data such as a customer's purchase history used for selecting a product. Efficiency by supplementing the amount and type of data such as the customer's purchase history used for product selection with the data of customers of similar classification without sequentially matching the amount and type of data used for template selection You can often classify the client's preferences.

In addition, a part of the image display generation device 2 in the embodiment described above, for example, the sensitivity learning data acquisition unit 20, the content learning data acquisition unit 21, the recommendation information generation unit 22, the customer information acquisition unit 23, the template information acquisition unit 24, The content information acquisition unit 25 and the image display generation unit 26 may be realized by a computer. In that case, a program for realizing the control function may be recorded in a computer readable recording medium, and the program recorded in the recording medium may be read and executed by a computer system. Here, the “computer system” is a computer system built in the image display generation device 2 and includes an OS and hardware such as peripheral devices. The term "computer-readable recording medium" refers to a storage medium such as a flexible disk, a magneto-optical disk, a ROM, a portable medium such as a ROM or a CD-ROM, or a hard disk built in a computer system. Furthermore, the “computer-readable recording medium” is one that holds a program dynamically for a short time, like a communication line in the case of transmitting a program via a network such as the Internet or a communication line such as a telephone line. In this case, a volatile memory in a computer system serving as a server or a client in that case may also include one that holds a program for a certain period of time. The program may be for realizing a part of the functions described above, or may be realized in combination with the program already recorded in the computer system.
Further, part or all of the image display generation device 2 in the above-described embodiment may be realized as an integrated circuit such as a large scale integration (LSI). Each functional block of the image display generation device 2 may be individually processorized, or part or all may be integrated and processorized. Further, the method of circuit integration is not limited to LSI's, and implementation using dedicated circuitry or general purpose processors is also possible. In the case where an integrated circuit technology comes out to replace LSI's as a result of the advancement of semiconductor technology, integrated circuits based on such technology may also be used.

  As mentioned above, although one embodiment of this invention was described in detail with reference to drawings, a specific structure is not restricted to the above-mentioned thing, Various design changes etc. in the range which does not deviate from the summary of this invention It is possible to

S1: content recommendation system, 1: learning database, 10: sensitivity learning data, 11: content learning data, 2: image display generating device, 20: sensitivity learning data acquisition unit, 21: content learning data acquisition unit, 22: recommendation information Generation unit, 23: customer information acquisition unit, 24: template information acquisition unit, 25: content information acquisition unit, 26: image display generation unit, 220: sensitivity learning unit, 220-1, 220-2, 220-N: individual Kansei learning unit, 221: content learning unit, 221-1, 221-2, 221-N: personal content learning unit, 222: customer model classification unit, 223: template selection unit, 224: content selection unit, 3: customer database , 30 ... customer attribute information, 31 ... customer behavior history information, 4 ... recommendation database, 40 ... template Information, 41 ... commodity information, 42 ... greeting information, 43 ... word of mouth information, 5 ... EC site, 6 ... customer terminal, TA, TB ... template, GA, GB ... image display, RA1, RA2, RA3, RA4, RA5 , RA6, RA7, RB1, RB2, RB3, RB4, RB5, RB6, RB7, RB9, variable regions, CA1, CA2, CA3, CA4, CA5, CA6, CA7, CB1, CB2, CB3, CB4, CB5, CB5, CB6, CB7, CB8, CB9 ... content

Claims (3)

An acquisition unit that acquires customer attribute information indicating an attribute of a customer and customer action history information indicating an action history of the customer;
Content information that is information related to content to be recommended to the customer based on the customer attribute information acquired by the acquisition unit and the customer behavior history information acquired by the acquisition unit, and information included in the content information A recommendation information generation unit that generates template information indicating a template including a region to be displayed by learning artificial intelligence;
An image generation unit that generates image data in which the content information generated by the recommendation information generation unit and the area included in the template indicated by the template information generated by the recommendation information generation unit are associated;
Content recommendation system with An acquisition process of acquiring customer attribute information indicating a customer attribute and customer behavior history information indicating a customer behavior history;
Content information that is information related to content recommended for the customer based on the customer attribute information acquired in the acquisition process and the customer behavior history information acquired in the acquisition process, and the content information Recommending information generation process of generating template information indicating a template including an area for displaying information by learning artificial intelligence;
An image generation process of generating image data in which the content information generated in the recommendation information generation process and the area included in the template indicated by the template information generated in the recommendation information generation process are associated;
Content recommendation method with On the computer
An acquisition step of acquiring customer attribute information indicating an attribute of the customer and customer action history information indicating an action history of the customer;
Content information that is information related to content to be recommended to the customer based on the customer attribute information acquired in the acquisition step and the customer behavior history information acquired in the acquisition step, and the content information Recommending information generating step of generating template information indicating a template including an area for displaying information by learning artificial intelligence;
An image generation step of generating image data in which the content information generated in the recommendation information generation step and the area included in the template indicated by the template information generated in the recommendation information generation step are associated;
Program with

Images