JP2013161116A - Recommendation system, recommendation method, and recommendation program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a recommendation system capable of recommending a commodity or service in view of person information of a user.SOLUTION: A recommendation system includes: a content collection part 1 for collecting a content containing at least one of an image and a moving image; an object recognition part 2 for collating a feature quantity of article information contained in the content and a feature quantity of an image of a commodity previously stored, and associating the identification information of the content with the identification information of the commodity when the difference therebetween is equal to or smaller than a predetermined value; a person information assessing part 3 for acquiring person information, including an age and a sexuality, contained in the content, associating the identification information of the content with the person information; and a recommendation calculating part 4 for associating identification information of a commodity, associated with the identification information of the same content, with person information, classifying person information on the basis of a predetermined section so as to calculate a point indicative of a recommendation priority for each commodity and section, and generating recommendation information including the point.

Description

  The present invention relates to a recommendation system that generates recommendation information for recommending a product or service to a user.

  In online shops, for example, a system that recommends recommended products and services to users by using data such as user browsing history, purchase history, product evaluation, and possessed products is generally used. ing.

  Further, in Patent Document 1, an image is input from a user's terminal, an article classification related to an article included in the image is recognized, and identification information of the article classification and information related to the product are stored in association with each other. There is disclosed a technique capable of recommending a product that is not shown in an image by using a storage unit.

Japanese Patent No. 4733236

  Among online shops, particularly in interior online shops, personal information such as age and gender of users who use products and services, usage conditions such as time and place of use are important information for recommending products. . However, with the above-described general system and the technique described in Patent Document 1, no recommendation is made in consideration of the user's personal information and usage status.

  An object of this invention is to provide the recommendation system which can recommend the goods and service which considered the user's personal information.

  A recommendation system according to the present invention includes a content collection unit that collects content including at least one of an image and a moving image, a feature amount of article information included in the content, and a feature amount of a product image stored in advance. If the difference is equal to or less than a predetermined value, the object recognition unit that associates the identification information of the content with the identification information of the product, and personal information including age and gender included in the content is acquired, A personal information determination unit that associates identification information with the personal information, and the product information associated with the same content identification information with the personal information, and the personal information is based on a predetermined category. Separately, a score indicating the priority of the recommendation is calculated for each of the product and the category, and recommendation information including the score is generated. Characterized by comprising a Komendo calculator.

  The recommendation method according to the present invention collects content including at least one of an image and a moving image, collates the feature amount of the article information included in the content with the feature amount of the image of the product stored in advance, and the difference If the content identification information is equal to or less than a predetermined value, the identification information of the content is associated with the identification information of the product, personal information including age and gender included in the content is acquired, and the identification information of the content and the personal information are Priorities of recommendation for each product and each category by associating the identification information of the product associated with the identification information of the same content and the person information, and dividing the personal information based on a predetermined category Is calculated, and recommendation information including the score is generated.

  The recommendation program according to the present invention includes a content collection process for collecting content including at least one of an image and a moving image in a computer, a feature amount of article information included in the content, and a feature of a product image stored in advance. The amount is collated, and when the difference is equal to or smaller than a predetermined value, the object recognition process for associating the identification information of the content with the identification information of the product, and personal information including age and gender included in the content are acquired. A classification in which the personal information is determined by associating the personal information determination process associating the identification information of the content with the personal information and the identification information of the product and the personal information associated with the same identification information of the content. The score indicating the priority of the recommendation is calculated for each product and each category, and the score is included. Characterized in that to execute the recommendation calculation process of generating Komendo information.

  ADVANTAGE OF THE INVENTION According to this invention, the goods and service which considered the user's personal information can be recommended.

It is a block diagram which shows the structure of embodiment of the recommendation system by this invention. It is a 1st sequence diagram which shows operation | movement of embodiment of the recommendation system by this invention. It is a 2nd sequence diagram which shows operation | movement of embodiment of the recommendation system by this invention. It is a 3rd sequence diagram which shows operation | movement of embodiment of the recommendation system by this invention. It is explanatory drawing which shows the information which image and moving image DB memorize | store. It is explanatory drawing which shows the information which product DB memorize | stores. It is explanatory drawing which shows the information which object DB memorize | stores. It is explanatory drawing which shows the information which age and sex DB memorize | store. It is explanatory drawing which shows the information which tag DB memorize | stores. It is explanatory drawing which shows the information which recommendation DB memorize | stores. It is a block diagram which shows the principal part of the recommendation system by this invention.

  FIG. 1 is a block diagram showing the configuration of an embodiment of a recommendation system according to the present invention. As shown in FIG. 1, the recommendation system of the present embodiment includes an image and video collection unit 101, an image and video DB (Data Base) 102, a product DB (Data Base) 103, an object recognition unit 201, an age And gender determination unit 202, tag information analysis unit 203, object DB (Data Base) 301, age and gender DB (Data Base) 302, tag DB (Data Base) 303, recommendation calculation unit 401, recommendation A DB (Data Base) 402.

  The image and moving image collection unit 101 collects content including at least one of images and moving images from the Web (World Wide Web) or a file server, and stores the collected content in the image and moving image DB 102. The product DB 103 stores images of products handled in the online shop.

  The object recognition unit 201 extracts each product image from the product DB 103 and stores the feature amount of the extracted product image. Then, the object recognizing unit 201 collates the feature amount of the product image extracted from the product DB 103 with the feature amount of the article information reflected in the image and the moving image stored in the image and moving image DB 102, and the difference is equal to or less than a predetermined value. Are associated and stored in the object DB 301.

  The age and gender determination unit 202 determines personal information (age and gender) of a person shown in the image and video stored in the image and video DB 102 and stores the determined age and gender in the age and gender DB 302. In the following description, only the age and gender are used for the person information. However, the information included in the person information is not limited to age and gender, and may be anything related to a person such as a physique.

  The tag information analysis unit 203 extracts and extracts location information (GPS (Global Positioning System) information) and time information (date of shooting) of the location from the image and movie tag information stored in the image and movie DB 102. The information is stored in the tag DB 303.

  The recommendation calculation unit 401 generates recommendation information based on the information stored in the object DB 301, the age and gender DB 302, and the tag DB 303, and stores the generated information in the recommendation DB 402.

  The operation of the recommendation system of this embodiment will be described. FIG. 2 is a first sequence diagram showing the operation of the embodiment of the recommendation system according to the present invention. The image and moving image collection unit 101 collects images and moving images (contents) from the Web (homepage, blog or SNS (Social Network Service)) or a file server (album storage service, document management system, etc.), and the collected images In addition, the content ID is associated with the moving image. Then, the image and moving image collection unit 101 stores the associated image and moving image and the content ID in the image and moving image DB 102 (step S1).

  FIG. 5 is an explanatory diagram showing information stored in the image and moving image DB 102. As shown in FIG. 5, the image and moving image DB 102 stores an image file and a moving image file associated with the content ID. FIG. 6 is an explanatory diagram showing information stored in the product DB 103. As illustrated in FIG. 6, the product DB 103 stores a product ID, a product name associated with the product ID, and an image of the product.

  The object recognition unit 201 accesses the product DB 103 storing product images and product IDs handled in the online shop, acquires product images and product IDs, and performs image analysis on the acquired product images to extract feature quantities. (Step S2). And the object recognition part 201 hold | maintains the extracted feature-value as a goods feature pattern matched with goods ID.

  Next, the object recognition unit 201 accesses the image and moving image DB 102, acquires the image or moving image and the content ID, performs image analysis on the acquired image or moving image, and obtains the feature amount of the article information included in the image or moving image. Extract (step S3). Then, the object recognition unit 201 collates the feature amount extracted in step S2 with the feature amount extracted in step 3. If there is a product whose difference is equal to or less than a predetermined value in the collation performed in step S4, the object recognizing unit 201 adds the product ID of the product and the object ID to the content ID associated with the image or video including the product. Are associated (step S4). Further, the object recognition unit 201 stores the associated content ID, product ID, and object ID in the object DB 301 (step S5).

  FIG. 7 is an explanatory diagram showing information stored in the object DB 301. As illustrated in FIG. 7, the object DB 301 stores the object ID, the content ID, and the product ID that are associated in step 5. There may be one or more product IDs associated with each object ID and content ID.

  FIG. 3 is a second sequence diagram showing the operation of the embodiment of the recommendation system according to the present invention. The age and gender determination unit 202 accesses the image and moving image DB 102 to acquire an image or moving image, extracts a part that seems to be a face from the acquired image or moving image, for example, and extracts personal information (age and sex) from the face information. ) Is determined (step S6). Then, the age and gender determination unit 202 associates the information regarding the determined age and gender and the age and gender ID with the content ID of the image or video. Then, the age and gender determination unit 202 stores the associated content ID, information about the age and gender, and the age and gender ID in the age and gender DB 302 (step S7).

  FIG. 8 is an explanatory diagram showing information stored in the age and gender DB 302. As shown in FIG. 8, the age and sex DB 302 stores the associated age and sex ID, content ID, and information about the age and sex. In addition, the information regarding the age and sex linked | related with one content ID may be one, and plural may be sufficient as it.

  The tag information analysis unit 203 accesses the image and moving image DB 102 to acquire an image or moving image, and extracts tag information (time information and position information) attached to the acquired image or moving image (step S8). Then, the tag information analysis unit 203 associates the time information, the position information, and the tag ID with the image or moving image content ID. The tag information analysis unit 203 stores the associated content ID, time information, position information, and tag ID in the tag DB 303 (step S9).

  FIG. 9 is an explanatory diagram showing information stored in the tag DB 303. As illustrated in FIG. 9, the tag DB 303 stores the tag ID, content ID, time information, and position information associated in step 9. The time information is, for example, the date when the image was taken. The position information is, for example, the longitude and latitude of the place where the image was taken.

  In addition, the execution order of the process of step S2-step S5, the process of step S6 and step S7, and the process of step S8 and step S9 is not restricted to said order.

  FIG. 4 is a third sequence diagram illustrating the operation of the embodiment of the recommendation system according to the present invention. The recommendation calculation unit 401 accesses the object DB 301 and acquires the object information associated in step S5 (step S10). Specifically, the recommendation calculation unit 401 acquires content ID and information related to a product.

  Also, the recommendation calculation unit 401 acquires content ID, information on age and sex from the age and sex DB 302 (step S11). Also, the recommendation calculation unit 401 accesses the tag DB 303 and acquires the content ID, time information, and position information (step S12). In addition, the execution order of the process of step S10, the process of step 11, and the process of step S12 is not restricted to said order.

  Next, the recommendation calculation unit 401 performs a calculation for obtaining recommendation information based on the acquired product, age and sex, time, and position information (step S13). For example, scores for each age and sex in the product A are calculated as follows. In addition, although the score of each age and sex in the following examples is calculated according to the age every 10 years, it is not restricted to such distinction.

Scores for each age of product A (for men in their 30s) =
Number of images of products A and men in their 30s x coefficient + number of images of products A and men in their teens x coefficient + number of images of products A and men in their 20s x coefficient + products A And 40s ...
+ ...
+ Number of images of product A and women in their 30s x coefficient + ...
+ Weighting factor

  The coefficient is a different value depending on the target age and gender. For example, when males in their 30s are targeted, the coefficient for males in their 30s is 1, the coefficient for males in their 20s and 40s is 0.5, the coefficient for males in their 10s is 0.3, and the coefficient for males in their 50s is 0. .4. If the gender is different, for example, a value obtained by reducing 0.3 from the count of each age is used. Further, the weight coefficient is, for example, the number of clicks / the number of purchases of the product A. The higher the score calculated, the higher the priority for making recommendations.

The time information of the product is obtained by, for example, calculating the average of the time information of the image showing the product using the time information of the tag information acquired from the tag DB 303. For example, the time information of the product A is calculated as follows.
Time information of product A = average time information of images and videos in which product A is shown

The position information of the product is obtained by calculating the average of the position information (for example, longitude and latitude) of the image showing the product using the position information of the tag information acquired from the tag DB 303, for example. For example, the position information of the product A is calculated as follows.
Location information of product A = average of location information of images and videos in which product A is shown

  The recommendation calculation unit 401 stores the calculated result as recommendation information in the recommendation DB 402 (step S14). FIG. 10 is an explanatory diagram showing information stored in the recommendation DB 402. As shown in FIG. 10, in the recommendation DB 402, the product ID is stored in association with each age and sex score, time information, and position information calculated as described above.

  And recommendation is performed based on the information memorize | stored in recommendation DB402. For example, in furniture named Product A, males in their 30s have high scores for each age and gender, location information indicates Tokyo, and time information is winter. Since the information that the recommended furniture for winter is “Product A” is obtained, a recommendation is made based on the information.

  The recommendation system of this embodiment uses not only product information but also user information (age and gender) and tag information (position information and time information) of a user shown in an image and a video, and recommends them in consideration. can do. Therefore, it is possible to make a recommendation with a high sales promotion effect.

  The recommendation system according to the present embodiment uses a product image stored in advance and videos and images already stored on the Web or file server to generate recommendation information, so that the system can be easily constructed. be able to.

  FIG. 11 is a block diagram showing a main part of the recommendation system according to the present invention. As shown in FIG. 11, the recommendation system according to the present invention includes, as main components, a content collection unit 1 that collects content including at least one of an image and a moving image, a feature amount of article information included in the content, The object recognition unit 2 that compares the feature amount of the image of the product stored in advance and associates the content identification information with the product identification information when the difference is equal to or less than a predetermined value, and the age and gender included in the content The personal information determination unit 3 that acquires personal information including the content information and associates the content identification information with the personal information, associates the product identification information and the personal information associated with the same content identification information, and determines the personal information in advance. A recommendation that generates a piece of recommendation information including the score by calculating the score indicating the priority of the recommendation for each product and each category. And a calculation unit 4.

  Further, in the above embodiment, recommendation systems as shown in the following (1) to (3) are also disclosed.

(1) A content collection unit (for example, an image and video collection unit 101) that collects content including at least one of an image and a moving image, a feature amount of product information included in the content, and a prestored product image The object recognition unit (for example, the object recognition unit 201) that associates the content identification information with the product identification information when the difference is equal to or smaller than a predetermined value, and the age and gender included in the content. A personal information determination unit (for example, age and gender determination unit 202) that acquires personal information including content and associates the content identification information with the personal information, and product identification information and personal information associated with the same content identification information And calculate the score indicating the priority of the recommendation for each product and category by dividing the person information based on the predetermined category, and the record including the score Recommendation calculation unit for generating a command information (e.g., the recommendation calculation unit 401) recommendation system characterized in that a.

(2) The recommendation system includes a time information analysis unit (for example, tag information analysis unit 203) that acquires time information included in the content and associates the content with the time information, and the recommendation calculation unit associates with the same content. It may be configured to generate recommendation information including information in which the received product is associated with the time information. Therefore, according to the recommendation system of the present embodiment, it is possible to generate recommendation information in consideration of the time (for example, date or season) in which the user uses the recommended product.

(3) The recommendation system includes a position information analysis unit (for example, tag information analysis unit 203) that acquires position information included in the content and associates the identification information of the content with the position information. You may be comprised so that the recommendation information containing the information which linked | related the identification information and the positional information on the goods linked | related with the content may be produced | generated. Therefore, according to the recommendation system of the present embodiment, it is possible to generate recommendation information in consideration of a place (position information) where the user uses the recommended product.

  The present invention can be applied to online shops such as interiors.

DESCRIPTION OF SYMBOLS 1 Content collection part 2 Object recognition part 3 Person determination part 4 Recommendation calculation part 101 Movie collection part 102 Movie DB
103 Product DB
DESCRIPTION OF SYMBOLS 201 Object recognition part 202 Age and sex determination part 203 Tag information analysis part 301 Object DB
302 Age and Gender DB
303 Tag DB
401 recommendation calculation unit 402 recommendation DB

Claims (5)

A content collection unit that collects content including at least one of an image and a video;
An object for associating the feature amount of the product information included in the content with the feature amount of the product image stored in advance and associating the content identification information with the product identification information when the difference is equal to or less than a predetermined value A recognition unit;
A personal information determination unit that acquires personal information including age and gender included in the content, and associates the identification information of the content with the personal information;
The identification information of the product associated with the identification information of the same content is associated with the person information, and the personal information is divided based on a predetermined category to indicate the priority of the recommendation for each product and the category. A recommendation system comprising: a recommendation calculation unit that calculates a score and generates recommendation information including the score. A time information analysis unit that acquires time information when the content is generated and associates the content with the time information;
The recommendation calculator is
2. The recommendation system according to claim 1, further comprising: generating recommendation information including information relating the product associated with the same content and the time information. A position information analysis unit that acquires position information indicating a place where the content is generated and associates the identification information of the content with the position information;
The recommendation calculator is
3. The recommendation system according to claim 1, wherein the recommendation information includes information that associates identification information of a product associated with the same content and the position information. Collecting content that includes at least one of images and videos,
The feature amount of the article information included in the content is compared with the feature amount of the product image stored in advance, and when the difference is equal to or less than a predetermined value, the content identification information and the product identification information are associated with each other.
Obtaining personal information including age and gender included in the content, associating the identification information of the content with the personal information,
The identification information of the product associated with the identification information of the same content is associated with the person information, and the personal information is divided based on a predetermined category to indicate the priority of the recommendation for each product and the category. A recommendation method characterized by calculating a score and generating recommendation information including the score. On the computer,
Content collection processing for collecting content including at least one of images and videos;
An object for associating the feature amount of the product information included in the content with the feature amount of the product image stored in advance and associating the content identification information with the product identification information when the difference is equal to or less than a predetermined value Recognition processing,
Obtaining personal information including age and gender included in the content and associating the identification information of the content with the personal information;
The identification information of the product associated with the identification information of the same content is associated with the person information, and the personal information is divided based on a predetermined category to indicate the priority of the recommendation for each product and the category. A recommendation program for executing a recommendation calculation process for calculating a score and generating recommendation information including the score.

Images