JP6401121B2 - RECOMMENDATION DEVICE, RECOMMENDATION METHOD, PROGRAM, AND RECORDING MEDIUM - Google Patents

Description

  The present invention relates to a recommendation device, a recommendation method, a program, and a recording medium for recommending information related to a second user to a first user who publishes an image via a network.

  Currently, an image publishing system that publishes an image via a network, such as an SNS (Social Networking Service) and an image sharing service, is used by many people. In the image publishing system, for example, a certain user uploads an image from his / her terminal device to a server and publishes (shares), and other users are interested in viewing it on his / her terminal device. The system is designed to find out a target, for example, a person who wants to be a friend or a favorite image.

  On the other hand, as disclosed in Patent Documents 1 to 3, the system recommends to a user information on friends who match the user's interests, hobbies, human relationships, and the like.

  In Patent Literature 1, event information of a first terminal, for example, transmission / reception message information is analyzed to generate first interest keyword information, and second interest keyword information of the second terminal is received. There is described a friend recommendation method for comparing information on the keyword of the subject and the keyword information of the second interest and displaying information for recommending the user of the second terminal as a friend based on the comparison result.

  In Patent Literature 2, a plurality of users stored in the SNS database are registered as being associated with each other as friends, and are not registered as friends of the target user, but from the point where the target user is located. A communication karaoke system that detects other users performing a karaoke performance using a karaoke device within a predetermined range and presents the other users as candidates for friend registration corresponding to the target user is described. Has been.

  Patent Document 3 includes a virtual community management system that includes a means for providing a friend service and manages a virtual community. For each second user having a relationship with the first user, for example, a best friend type Or, the type of relationship such as the standard friend type is stored, and the second user's information is displayed on the terminal device of the first user according to the type of relationship of the second user with the first user. It is described.

JP 2014-044721 A JP 2012-185382 A JP 2012-212456 A

In the image publishing system, the publishing of images is the center, and there are cases where there is no exchange of text data (text messages). Therefore, as in Patent Document 1, it may be difficult to know the user's interests only with the user's transmission / reception message.
Since Patent Literature 2 is a communication karaoke system, it can know a common hobby of users called karaoke. In the case of an image disclosure system, for example, it is necessary to detect a common hobby of users from an image. is there.
Patent Document 3 uses the type of relationship between users, but the user himself / herself has to specify the type of relationship between users, which is troublesome.

  An object of the present invention is to solve the problems of the prior art and accurately detect a second user who is close to the first user in an image publishing system that publishes an image via a network. To provide a recommendation device, a recommendation method, a program, and a recording medium capable of recommending related information to a first user.

In order to achieve the above object, the present invention provides a recommendation device that performs a display for recommending related information related to a second user to a terminal device of a first user who publishes an image via a network.
An image analysis unit that performs image analysis of the public image and the private image of the first user, and performs image analysis of the public image of the second user;
An image classifying unit that classifies the first user's public image and private image for each first theme based on the result of the image analysis, and classifies the second user's public image for each second theme;
Based on the result of the image analysis, a similar theme is detected which is a theme common to the first theme and the second theme, and whose first user interest level and second user interest level are equal to or higher than the first threshold. A similar theme detection unit,
A closeness calculating unit for calculating closeness between the first user and the second user based on the similar theme;
Provided is a recommendation device including a recommendation unit that performs a display for recommending related information related to the second user to the first user when the familiarity is equal to or greater than the second threshold. Is.

  Here, the image classification unit preferably uses preset themes as the first theme and the second theme.

  Moreover, it is preferable that an image classification part determines a 1st theme and a 2nd theme based on the result of an image analysis.

  It is preferable that the image classification unit further determines the first theme and the second theme based on text data posted on the first user's public image and the second user's public image.

  Further, the degree of interest of the first user with respect to each first theme is determined by the number of images classified into each first theme, and the degree of interest of the second user with respect to each second theme is the second degree of interest. It is determined by the number of images classified as a theme, and the first threshold is preferably a preset number of images.

  Further, the degree of interest of the first user with respect to each first theme is determined by the ratio of the images classified into each first theme, and the degree of interest of the second user with respect to each second theme is the second degree of interest. It is determined based on the ratio of images classified as a theme, and the first threshold is preferably a preset ratio of images.

  In addition, the degree of interest of the first user for each first theme is determined by the number of images classified into each first theme and the ratio of the images, and the degree of interest of the second user for each second theme is The first threshold value is a function value obtained by weighting the preset number of images and the preset percentage of images. It is preferable that

  The closeness is determined by the number of similar themes, and the second threshold is preferably a preset number.

  The familiarity is determined by adding a score to each similar theme and determining the total score of all similar themes, and the second threshold is preferably a preset score.

  In addition, the closeness is determined by the shooting duration from the shooting start date and time to the shooting end date and time of an image included in at least one similar theme, and the second threshold is preferably a preset time.

The similar theme detection unit calculates the short-term interest level of the first user based on the first user's public image and private image over a preset first period before the current time, and A short-term interest degree of the second user is calculated based on the public image of the second user over one period, and the short-term interest degree of the first user and the short-term interest degree of the second user are equal to or greater than the first threshold. Detects similar themes in
The intimacy calculation unit calculates the short-term intimacy between the first user and the second user based on the short-term similar theme,
It is preferable that a recommendation part performs the display which recommends, when short-term closeness is more than a 2nd threshold value.

In addition, the similar theme detection unit determines the long-term interest degree of the first user based on the first user's public image and private image over a preset second period that is longer than the first period before the current time. The long-term interest degree of the second user is calculated based on the calculated second user's public image over the second period, and the long-term interest degree of the first user and the long-term interest degree of the second user are the first. Calculate a similar long-term theme that is greater than or equal to one threshold,
The intimacy calculation unit calculates long-term intimacy between the first user and the second user based on the long-term similar theme,
It is preferable that a recommendation part performs the display which recommends, when long-term closeness is more than a 2nd threshold value.

  In addition, intimacy is determined by the total score of all short-term similar themes and all long-term similar themes, with a higher score for each long-term similar theme than for each short-term similar theme. The 2 threshold value is preferably a preset score.

The image classification unit further classifies the images classified into each first theme for each first lower theme included in each first theme, and images classified into each second theme, Furthermore, it classifies for each second lower theme included in each second theme,
The similar theme detection unit further includes a similar lower theme that is a common theme of the first lower theme and the second lower theme, and the degree of interest of the first user and the degree of interest of the second user are equal to or greater than the first threshold. Detect
It is preferable that the closeness calculation unit calculates closeness between the first user and the second user based on the similar lower theme.

  The image classification unit preferably uses preset themes as the first lower theme and the second lower theme.

  The image classification unit preferably determines the first lower theme and the second lower theme based on the result of the image analysis.

  It is preferable that the image classification unit further determines the first lower theme and the second lower theme based on text data posted on the first user's public image and the second user's public image.

  The degree of interest of the first user for each first lower theme is determined by the number of images classified into each first lower theme, and the degree of interest of the second user for each second lower theme is The first threshold is preferably a preset number of images, which is determined by the number of images classified as the second lower theme.

  The degree of interest of the first user for each first lower theme is determined by the ratio of the images classified into each first lower theme with respect to the first user's public image and private image, and each second lower theme. The degree of interest of the second user with respect to the theme is determined by the ratio of the images classified into the respective second lower themes with respect to the public image of the second user, and the first threshold is a preset ratio of the image. preferable.

  Further, the degree of interest of the first user with respect to each first lower theme is classified into the number of images classified into each first lower theme and the first lower theme with respect to the first user's public image and private image. The degree of interest of the second user for each second lower theme is determined by the ratio of images, and the number of images classified into each second lower theme and the second lower theme for the second user's public image are classified. The first threshold is preferably a function value obtained by weighting the preset number of images and the preset image ratio.

  The closeness is determined by the number of similar lower themes, and the second threshold is preferably a preset number.

  In addition, the closeness is given a higher score than the similar theme for the similar lower theme, and is determined by the total score of all the similar themes and all the similar lower themes, and the second threshold is a preset score. Preferably there is.

  In addition, the closeness is determined by the shooting duration from the shooting start date and time to the shooting end date and time of an image included in at least one similar lower theme, and the second threshold is preferably a preset time.

The similar theme detection unit calculates the short-term interest level of the first user based on the first user's public image and private image over a preset first period before the current time, and A short-term interest degree of the second user is calculated based on the public image of the second user over one period, and the short-term interest degree of the first user and the short-term interest degree of the second user are equal to or greater than the first threshold. Detect similar sub-themes,
The closeness calculation unit calculates short-term closeness between the first user and the second user based on the short-term similar lower theme,
It is preferable that a recommendation part performs the display which recommends, when short-term closeness is more than a 2nd threshold value.

In addition, the similar theme detection unit determines the long-term interest degree of the first user based on the first user's public image and private image over a preset second period that is longer than the first period before the current time. The long-term interest degree of the second user is calculated based on the calculated second user's public image over the second period, and the long-term interest degree of the first user and the long-term interest degree of the second user are the first. Calculate a similar long-term theme that is more than one threshold,
The intimacy calculating unit calculates long-term intimacy between the first user and the second user based on the long-term similar lower theme,
It is preferable that a recommendation part performs the display which recommends, when long-term closeness is more than a 2nd threshold value.

  Also, the intimacy is higher for each long-term similar sub-theme than for each short-term similar sub-theme, and is based on the total score of all short-term similar sub-themes and all long-term similar sub-themes It is determined and the second threshold value is preferably a preset score.

  Moreover, it is preferable that a recommendation part performs the display which recommends a 2nd user to a 1st user as a friend candidate with respect to a 1st user's terminal device.

  Moreover, when a 1st user adds a 2nd user as a friend, it is preferable that a recommendation part performs the display which recommends a 1st user to a 2nd user as a friend candidate with respect to a 2nd user's terminal device. .

Furthermore, based on the result of the image analysis, a recommendation information extraction unit that extracts recommendation information recommended to the first user from the public image of the second user included in each similar theme is provided.
It is preferable that a recommendation part performs the display which recommends recommendation information to a 1st user with respect to a 1st user's terminal device.

The present invention is also a recommendation method for performing a display for recommending related information related to the second user to the terminal device of the first user who publishes an image via a network,
An image analysis unit performing image analysis of the public image and the private image of the first user and performing image analysis of the public image of the second user;
A step of classifying the first user's public image and non-public image for each first theme and classifying the second user's public image for each second theme based on the result of the image analysis; ,
Based on the result of image analysis, the similar theme detection unit is a theme common to the first theme and the second theme, and the interest level of the first user and the interest level of the second user are equal to or higher than the first threshold. Detecting a similar theme that is,
A step of calculating an intimacy between the first user and the second user based on the similar theme;
A recommendation unit including a step of, when the intimacy is equal to or greater than a second threshold value, displaying, on the first user terminal device, a display for recommending related information related to the second user to the first user. I will provide a.

Here, the image classification unit further classifies the images classified into the respective first themes for each of the first lower themes included in the respective first themes, and classified into the respective second themes. The image is further classified for each second lower theme included in each second theme,
The similar theme detection unit further includes a similar lower theme that is a common theme of the first lower theme and the second lower theme, and the degree of interest of the first user and the degree of interest of the second user are equal to or greater than the first threshold. Detect
It is preferable that the closeness calculation unit calculates closeness between the first user and the second user based on the similar lower theme.

  The present invention also provides a program for causing a computer to execute each step of the recommendation method described above.

  The present invention also provides a computer-readable recording medium on which a program for causing a computer to execute each step of the recommendation method described above is recorded.

  According to the present invention, by using not only the first user's public image and the second user's public image, but also the first user's private image, the second user who is close to the first user can be accurately identified. Can be detected.

  Further, in the present invention, a similar theme that is a theme common to the theme of the image of the first user and the theme of the image of the second user, and is a theme that is highly interested in the first user and the second user. And the closeness between the first user and the second user is detected based on the similar theme. Thereby, according to this invention, a 2nd user with high closeness with a 1st user can be detected correctly, and various related information relevant to a 2nd user can be recommended to a 1st user. .

It is a block diagram of one embodiment showing composition of an image public system to which a recommendation device of the present invention is applied. It is a block diagram of one Embodiment showing the structure of the server shown in FIG. It is a conceptual diagram of one Embodiment showing the mode of the image of the 1st user uploaded to the server and a 2nd user. It is a flowchart of one Embodiment showing operation | movement of the image disclosure system in the case of recommending a friend candidate. It is a flowchart of one Embodiment showing operation | movement of the image disclosure system in recommending recommendation information.

  DESCRIPTION OF EMBODIMENTS Hereinafter, a recommendation device, a recommendation method, a program, and a recording medium of the present invention will be described in detail based on preferred embodiments shown in the accompanying drawings.

  FIG. 1 is a block diagram of an embodiment showing a configuration of an image disclosure system to which a recommendation device of the present invention is applied. An image disclosure system 10 shown in FIG. 1 includes a server 12 and two or more clients 14 connected to the server 12 via a network 16 such as the Internet. The image publishing system 10 publishes an image uploaded from each client 14 to the server 12 via the network 16 and published to each client 14 via the network 16.

  FIG. 2 is a block diagram of an embodiment showing the configuration of the server shown in FIG. A server 12 shown in the figure is an example of a recommendation device according to the present invention, and includes an image analysis unit 18, an image classification unit 20, a similar theme detection unit 22, a closeness calculation unit 24, and a recommendation information extraction unit 26. And a recommendation unit 28. The server 12 performs a display for recommending related information related to the second user to the first user on the client (terminal device) 14 of the first user who publishes an image via the network 16.

  In the server 12, the image analysis unit 18 performs image analysis of the first user's public image and private image. The image analysis unit 18 performs image analysis of the second user's public image.

  In the present embodiment, as shown in FIG. 3, the first user's image A has already been uploaded from the first user's client 14 to the server 12 via the network 16, and similarly, the second user's image B is It is assumed that the second user's client 14 has already been uploaded to the server 12 via the network 16. The first user's image A includes a public image A-2 and a private image A-1, and the second user's image B includes a public image B-2 and a private image B-1.

  As the image analysis, the image analysis unit 18 performs processing for acquiring information on the shooting date and time and the shooting location (GPS: Global Positioning System) from the supplementary information of the image such as Exif (Exchangeable Image File Format) information. Also, for example, the image content is analyzed, such as annotation analysis processing (scene analysis processing) for analyzing an image scene (outdoor, mountain, sea, night view, etc.), similar image determination processing for determining a similar image, and the like. Process.

  The image includes at least one of a still image and a moving image. The image analysis unit 18 performs the image analysis on the still image. In the case of a moving image, KFE (Key Frame Extraction) technology is used to extract a frame image (still image) that is a key such as a scene change from the moving image, and the extracted frame image is extracted from the moving image. Perform the above image analysis.

  In the present embodiment, since the result of image analysis of the image of the first user is used to recommend to the first user, the image analysis unit 18 not only for the first user's public image but also for the private image. Perform image analysis.

  On the other hand, since the result of the image analysis of the image of the second user is used for recommendation to the first user, the image analysis unit 18 performs only the image analysis of the public image of the second user from the viewpoint of privacy, Normally, image analysis of private images is not performed. However, if the parties are allowed to use a private image, or if it is a safe image, such as a general hobby image May be used, etc., the private image may be analyzed in addition to the public image of the second user.

  Subsequently, the image classification unit 20 classifies the first user's public image and private image for each first theme based on the result of the image analysis performed by the image analysis unit 18. The image classification unit 20 classifies the second user's public image for each second theme.

  The image theme (first theme, second theme) is a keyword representing the atmosphere or impression that the user feels when viewing the image, and examples include travel, baby, children, weddings, pets, and the like. Can do.

  The image classification unit 20 may use preset themes as the first theme and the second theme. Alternatively, based on the result of image analysis, and further based on text data (text message) posted on the public image of the first user and the public image of the second user, the first theme and the second theme It can also be determined. For example, when text data “I want to eat a cake” is posted to a public image by another user, it can be estimated that the public image is an image of a cake.

  Subsequently, the similar theme detection unit 22 is a common theme for the first theme and the second theme based on the result of the image analysis, and has a high degree of interest of the first user and a high degree of interest of the second user ( A similar theme that is a theme (which is equal to or greater than a first threshold value) is detected.

  The degree of interest represents the degree of interest that the first user and the second user have for each theme.

  The similar theme detection unit 22 can determine the degree of interest of the first user for each first theme based on the number of images classified into each first theme. For example, as the first threshold value, the number of images serving as a criterion for determining interest is set in advance, and the number of images classified into each first theme is equal to or greater than the number of images serving as a criterion for determining interest. If there is, it determines with the 1st user's interest degree with respect to each 1st theme being high. Similarly, the degree of interest of the second user with respect to each second theme can be determined based on the number of images classified into each second theme.

  Moreover, the interest degree of the 1st user with respect to each 1st theme can be determined with the ratio of the image classified into each 1st theme. For example, as the first threshold value, the ratio of images that serve as criteria for determining the degree of interest is set in advance, and the ratio of images classified as the first theme with respect to the first user's public image and private image is set as the first threshold. If it is equal to or greater than the ratio of the image that is the criterion for determining the degree of interest, it is determined that the degree of interest of the second user for each second theme is high. Similarly, the degree of interest of the second user for each second theme can be determined based on the ratio of the images classified into each second theme.

  Alternatively, the degree of interest of the first user for each first theme can be determined based on the number of images classified into each first theme and the ratio of the images. For example, as the first threshold value, a function value obtained by weighting the number of images and the ratio of images, which is a criterion for determining the degree of interest, is set in advance, and the images classified into the respective first themes are set. If the number of images and the ratio of the images are equal to or greater than the value of a function that is a criterion for determining the degree of interest, it is determined that the degree of interest of the first user for each first theme is high. Similarly, the degree of interest of the second user for each second theme can be determined from the number of images classified into each second theme and the ratio of the images.

  Subsequently, the familiarity calculation unit 24 calculates the familiarity between the first user and the second user based on the similar theme detected by the similar theme detection unit 22.

  Intimacy between users represents the degree to which the hobbies, preferences, behaviors, and the like are similar between users, in the case of this embodiment, between the first user and the second user. For example, the intimacy can be determined based on the number of similar themes. In this case, as the second threshold value, the number of similar themes used as the criteria for determining the familiarity is set in advance, and if the number of similar themes is equal to or greater than the number of similar themes used as the criteria for determining the familiarity, the familiarity Is determined to be high.

  Moreover, a score can be given to each similar theme, and a closeness degree can be determined by the total score of all the similar themes. For example, as the second threshold, a total score that is a criterion for determining the familiarity is set in advance, and if the total score for all similar themes is equal to or greater than the total score that is the criterion for determining the familiarity, It is determined that the density is high. For example, when the number of points per similar theme is 5 and the number of similar themes is 3, the total score is 15. In this case, if the total score that is the criterion for determining the intimacy is 10, the intimacy is determined to be high.

  Furthermore, the closeness can be determined based on the shooting duration from the shooting start date / time to the shooting end date / time of the image included in at least one similar theme. For example, as the second threshold, a shooting duration that is a criterion for determining the intimacy is set in advance, and a shooting duration of an image included in at least one similar theme is equal to or longer than the shooting duration that is a criterion for determining the familiarity. If so, it is determined that the intimacy is high.

  Then, the recommendation information extraction part 26 extracts the recommendation information recommended to the 1st user from the public image of the 2nd user contained in each similar theme based on the result of image analysis.

  The recommendation information is information that has a high degree of interest of the second user (is greater than or equal to the third threshold). The similar theme is a theme common to the first theme and the second theme, and is a theme having a high degree of interest of the first user and a high degree of interest of the second user. It can be assumed that the recommendation information extracted from the public image is information that is highly interesting to the first user.

For example, when it is detected that the second user has traveled to Hawaii from the sea image of the second user based on the contents of the image, GPS information, etc., for example, the travel to Hawaii should be used as recommendation information. Can do.
If it is detected from the pasta image that the second user has gone to an Italian restaurant, for example, going to the Italian restaurant can be used as recommendation information.
When it is detected from the temple image that the second user has visited the temple, for example, a trip around the temple can be used as recommendation information.
When the cosmetic used by the second user is detected from the image of the lipstick, for example, the cosmetic or the manufacturer that manufactured the cosmetic can be used as the recommendation information.

  Subsequently, the recommendation unit 28 displays a display for recommending related information related to the second user to the first user when the familiarity calculated by the familiarity calculation unit 24 is high (greater than or equal to the second threshold). This is for one user's client 14.

  The related information includes various information (recommendation information) related to the second user obtained from the result of image analysis of the public image of the second user in addition to the information of the second user.

  As the related information, the recommendation unit 28 can display, for example, the second user as a friend candidate and recommend the first user to the client 14 of the first user. Moreover, the display which recommends it to a 2nd user by making a 1st user into a friend candidate is performed with respect to the client 14 of a 2nd user. Moreover, the recommendation part 28 can also perform the display which recommends the recommendation information extracted from the image of a 2nd user to a 1st user as related information with respect to the client 14 of a 1st user.

  On the other hand, the client 14 is an example of a terminal device owned by each user, such as a PC (personal computer), a tablet PC, or a smartphone. The client 14 includes, for example, an input unit such as a keyboard, a mouse, and a touch panel, a display unit such as a liquid crystal display and a touch panel, a process in which the input unit acquires an instruction input from a user, and a process in which the display unit displays various information. A control unit and the like for controlling are provided.

  Each of the server 12 and the client 14 includes a transmission / reception unit that transmits and receives various data via the network 16. The server 12 further includes a storage unit that stores an image, and disclosure and non-disclosure of the image. However, in this embodiment, descriptions of the transmission / reception unit, the storage unit, the control unit, and the like are omitted in order to prevent complication.

  Next, the operation of the image disclosure system 10 when recommending friend candidates will be described with reference to the flowchart shown in FIG.

  In the server 12, first, the image analysis unit 18 performs image analysis of the public image A-2 and the private image A-1 of the first user shown in FIG. Further, the image analysis unit 18 performs image analysis of the public image B-2 of the second user (Step S1).

  Subsequently, the image classification unit 20 classifies the first user's public image A-2 and private image A-1 for each first theme based on the result of the image analysis. The image classification unit 20 classifies the second user's public image B-2 for each second theme (step S2).

  Subsequently, the similar theme detection unit 22 detects whether there is a common theme between the first theme and the second theme based on the result of the image analysis (step S3).

As a result, if there is no common theme (No in step S3), the process ends. In this case, the closeness between the first user and the second user is low.
On the other hand, if there is a common theme (Yes in step S3), the similar theme detection unit 22 determines whether there is a theme with high interest of the first user and the second user based on the result of the image analysis. Is detected (step S4).

As a result, if there is no theme with a high degree of interest (No in step S4), the process ends. Also in this case, the closeness between the first user and the second user is low.
On the other hand, when there is a theme with a high degree of interest (Yes in step S4), the familiarity calculation unit 24 is a common theme for the first theme and the second theme, and the interest degree of the first user and the second user The closeness between the first user and the second user is calculated based on the similar theme that is a theme with a high degree of interest (step S5).

  Subsequently, the familiarity calculating unit 24 determines whether the familiarity between the first user and the second user is high (step S6).

As a result, if the intimacy is low (No in step S6), the process ends.
On the other hand, when the intimacy is high (Yes in step S6), the recommendation unit 28 displays on the client 14 of the first user a recommendation for recommending the second user to the first user as a friend candidate. Moreover, the recommendation part 28 performs the display which recommends a 1st user to a 2nd user as a friend candidate with respect to the 2nd user's client 14 (step S7).

  Subsequently, the control unit of the server 12 determines whether or not the first user has accepted the second user as a friend. Moreover, the control part of the server 12 determines whether the 2nd user received the 1st user as a friend (step S8).

As a result, when both the first user and the second user have not accepted the second user and the first user as friends (No in step S8), the process is terminated. In this case, the first user and the second user are not recognized as friends.
On the other hand, when both the first user and the second user accept the second user and the first user as friends, respectively (Yes in step S8), the control unit of the server 12 performs the first user and the second user. The user is identified as a friend (step S9), and the process is terminated.

  In addition, in the said embodiment, although a 1st user is made into a friend candidate and recommended also to a 2nd user, you may recommend only a 1st user by making a 2nd user a friend candidate. In this case, for example, when the first user does not accept the second user as a friend, the process ends. On the other hand, when the first user accepts the second user as a friend, the second user is notified that a request to become a friend is coming from the first user.

  Next, the operation of the image disclosure system 10 when recommending recommendation information will be described with reference to the flowchart shown in FIG.

  In the present embodiment, it is assumed that the first user and the second user are certified as friends as described above. That is, between the first theme and the second theme, there is a theme that is a common theme and has a high degree of interest of the first user and a high degree of interest of the second user.

  In the server 12, first, the image analysis unit 18 performs image analysis of the public image B-2 of the second user (step S10).

  Subsequently, the image classification unit 20 classifies the public image B-2 of the second user for each second theme based on the result of image analysis (step S11).

  Subsequently, the similar theme detection unit 22 detects a theme having a high degree of interest of the second user based on the result of the image analysis (step S12). When the 1st user and the 2nd user are authorized as a friend, the theme with a high degree of interest of the 2nd user is a common theme of the 1st theme and the 2nd theme, and the degree of interest of the 1st user And a theme with a high degree of interest of the second user, that is, a similar theme.

  Subsequently, the recommendation information extraction unit 26 extracts recommendation information to be recommended to the first user from the public image B-2 of the second user included in each similar theme based on the result of the image analysis (step S13). ).

  Subsequently, the recommendation information extraction unit 26 determines whether there is recommendation information extracted from the similar theme (step S14).

As a result, if there is no recommendation information (No in step S14), the process is terminated.
On the other hand, when there is recommendation information (Yes in step S14), the recommendation unit 28 displays the recommendation information to the first user on the client 14 of the first user (step S15), and performs processing. finish.

  As described above, in the recommendation device, not only the first user's public image A-2 and the second user's public image B-2 but also the first user's private image A-1 is used, so that It is possible to accurately detect the second user who has high intimacy with the user.

  Further, in the recommendation apparatus, a similar theme that is a theme common to the theme of the image of the first user and the theme of the image of the second user, and that is a theme that is highly interested in the first user and the second user. And the closeness between the first user and the second user is detected based on the similar theme. Thereby, a 2nd user with high closeness with a 1st user can be detected correctly, and various related information (recommendation information) relevant to a 2nd user can be recommended to a 1st user.

  In the above embodiment, when the first user and the second user are recognized as friends, the recommendation information extracted from the public image B-2 of the second user is recommended to the first user. However, the present invention is not limited to this. That is, even if the first user and the second user are not recognized as friends, for example, “A user with a hobby close to you went on a trip. Would you like to go on a trip?” Etc., it is possible to recommend the recommended information to the first user.

  In this case, for example, by executing steps S1 to S6 instead of steps S10 to S12, the process is terminated when the intimacy between the first user and the second user is low, and when the intimacy is high, steps S13 to S15 are performed. Execute.

  The similar theme detection unit 22 also displays the first user's public image and non-disclosure over a first period set before the current time, for example, three months, six months, one year, etc. before the current time. The short-term interest level of the first user may be calculated based on the image. Further, the short-term interest level of the second user is calculated based on the public image of the second user over the first period. As a result, it is possible to detect short-term similar themes in which the first user's short-term interest level and the second user's short-term interest level are high.

  In this case, the closeness calculation unit 24 calculates short-term closeness between the first user and the second user based on the short-term similar theme, and the recommendation unit 28 recommends when the short-term closeness is high. Display.

  As described above, by calculating the short-term closeness based on the short-term similar theme, it is possible to find a second user who has an interest that matches the current interest of the first user.

  On the other hand, the similar theme detection unit 22 applies the first user's public image and private image over a preset second period, for example, five years, ten years, etc., which is longer than the first period before the present. Based on this, the long-term interest level of the first user may be calculated. Further, the long-term interest level of the second user is calculated based on the public image of the second user over the second period. Thereby, a long-term similar theme with a long-term interest degree of a 1st user and a long-term interest degree of a 2nd user is computable.

  In this case, the closeness calculation unit 24 calculates the long-term closeness between the first user and the second user based on the long-term similar theme, and the recommendation unit 28 recommends when the long-term closeness is high. Display.

  In this way, by calculating the long-term intimacy based on the long-term similar theme, it is possible to find a second user who has a hobby etc. that matches the hobby etc. that the first user has continued for a long time.

  Further, as described above, when calculating the short-term similar theme and the long-term similar theme, the intimacy calculating unit 24 gives a higher score to each long-term similar theme than each short-term similar theme, Intimacy can be determined by the total score of all short-term similar themes and all long-term similar themes. For example, as the second threshold value, the total score of similar themes used as the criteria for determining intimacy is set in advance, and the total scores of all short-term similar themes and all long-term similar themes are determined as the intimacy determination. If the score is equal to or higher than the total score of similar themes as a reference, it is determined that the closeness between the first user and the second user is high.

  Further, the image classification unit 20 may classify the images classified into the respective first themes for each first lower theme included in each of the first themes. Further, the image classification unit 20 further classifies the images classified into each second theme for each second lower theme included in each second theme. For example, if the theme (first theme, second theme) is “sea”, the lower themes (first lower theme, second lower theme) are “sand beach”, “shell”, “lighthouse”, etc. A theme related to “sea” can be exemplified.

  In this case, the similar theme detection unit 22 detects a similar lower theme that is a common theme of the first lower theme and the second lower theme and has a high degree of interest of the first user and a high degree of interest of the second user, The closeness calculation unit 24 calculates closeness between the first user and the second user based on the similar lower theme.

  As described above, the theme is hierarchized and further subdivided to more accurately detect the second user having a high degree of closeness with the first user and relate to the second user having a high closeness with the first user. Various related information can be recommended to the first user.

  In this case, the image classification unit 20 may use preset themes as the first lower theme and the second lower theme. Alternatively, the first lower theme and the second lower theme are determined based on the result of the image analysis, and further based on the text data posted on the first user's public image and the second user's public image. You can also.

  In this case, the similar theme detection unit 22 can determine the degree of interest of the first user for each first lower theme based on the number of images classified into each first lower theme. For example, as the first threshold value, the number of images serving as a criterion for determining interest is set in advance, and the number of images classified into each first lower theme is equal to or greater than the number of images serving as a criterion for determining interest. If so, it is determined that the degree of interest of the first user for each first lower theme is high. Similarly, the degree of interest of the second user for each second lower theme can be determined based on the number of images classified into each second lower theme.

  Further, the degree of interest of the first user for each first lower theme can be determined based on the ratio of the images classified as the first lower theme to the first user's public image and private image. For example, as a first threshold value, a ratio of an image serving as a criterion for determining the degree of interest is set in advance, and a ratio of an image classified into each first lower theme with respect to a public image and a private image of a first user is If it is more than the ratio of the image used as the criterion of interest degree, it will determine with each 1st user's interest degree being high. Similarly, the degree of interest of the second user for each second lower theme can be determined based on the ratio of the images classified into each second lower theme with respect to the public image of the second user.

  Alternatively, the first user for each first lower theme according to the number of images classified into each first lower theme and the ratio of the images classified as the first lower theme to the public images and private images of the first user. The degree of interest can be determined. For example, as the first threshold value, a function value obtained by weighting the number of images and the ratio of images, which is a criterion for determining the degree of interest, is set in advance, and the images classified into the respective first lower themes And the ratio of the images classified into the first lower theme with respect to the first user's public image and private image are equal to or greater than the value of the function that is a criterion for determining the degree of interest, It is determined that the interest level of one user is high. Similarly, the degree of interest of the second user with respect to each second lower theme according to the number of images classified into each second lower theme and the ratio of images classified into the second lower theme with respect to the second user's public images. Can be determined.

  Intimacy can be determined from the number of similar lower themes. For example, as the second threshold value, the number of similar lower themes that are the criteria for determining the familiarity is set in advance, and if the number of similar lower themes is equal to or greater than the number of similar lower themes that are the criteria for determining the familiarity, It is determined that the intimacy is high.

Further, the similar lower theme is a more detailed theme than the similar theme, and the closeness between the first user and the second user is higher when the similar lower theme matches the similar theme. it is conceivable that.
Therefore, a higher score than the similar theme is assigned to the similar lower theme, and the familiarity can be determined based on the total score of the similar theme and the similar lower theme. For example, as the second threshold value, a total score that is used as a criterion for determining intimacy is set in advance, and the total score of all similar themes and all similar lower-level themes is used as a criterion for determining the familiarity. If it is above, it is determined that the familiarity is high.

  Furthermore, the closeness can be determined based on the shooting duration from the shooting start date and time to the shooting end date and time of the image included in at least one similar lower theme. For example, the second threshold is used as a determination criterion of closeness. A shooting duration is set in advance, and if the shooting duration of an image included in at least one similar lower theme is equal to or longer than the shooting duration that is a criterion for determining the familiarity, it is determined that the familiarity is high.

  Further, the similar theme detection unit 22 may calculate the short-term interest degree of the first user based on the public image and the private image of the first user over a preset first period before the current time. . Similarly, the similar theme detection unit 22 calculates the short-term interest level of the second user based on the public image of the second user over the first period. As a result, it is possible to detect short-term similar lower themes in which the first user's short-term interest level and the second user's short-term interest level are high.

  In this case, the closeness calculation unit 24 calculates short-term closeness between the first user and the second user based on the short-term similar lower theme, and the recommendation unit 28 recommends when the short-term closeness is high. Display.

  On the other hand, the similar theme detection unit 22 determines the long-term interest level of the first user based on the first user's public and private images over a second period that is set longer than the first period before the current time. May be calculated. Similarly, the similar theme detection unit 22 calculates the long-term interest level of the second user based on the public image of the second user over the second period. Accordingly, it is possible to calculate a long-term similar lower theme having a high long-term interest degree of the first user and a long-term interest degree of the second user.

  In this case, the closeness calculation unit 24 calculates the long-term closeness between the first user and the second user based on the long-term similar lower theme, and the recommendation unit 28 recommends when the long-term closeness is high Display.

  Further, as described above, when calculating the short-term similar lower theme and the long-term similar lower theme, the intimacy calculating unit 24 is higher for each long-term similar lower theme than for each short-term similar lower theme. The degree of intimacy can be determined by assigning a score and the total score of all short-term similar sub-themes and all long-term similar sub-themes. For example, as the second threshold value, a total score that is a criterion of intimacy is set in advance, and the total score of all short-term similar lower themes and all long-term similar sub-themes are determined as a criterion of intimacy If the total score is greater than or equal to, it is determined that the closeness between the first user and the second user is high.

  Further, although the image disclosure system 10 including the server 12 and two or more clients 14 connected to the server 12 via the network 16 has been described as an example, the present invention is not limited to this. Although the recommendation apparatus of this invention can be comprised as a function with which the server 12 is provided like this embodiment, for example, it is comprised also as a function with which the client 14 is provided, without requiring the server 12 and the network 16. be able to. Further, two or more clients 14 may be directly connected via the network 16 without requiring the server 12.

In the apparatus of the present invention, each component included in the apparatus may be configured by dedicated hardware, or each component may be configured by a programmed computer.
The method of the present invention can be implemented, for example, by a program for causing a computer to execute each step. It is also possible to provide a computer-readable recording medium in which this program is recorded.

  Although the present invention has been described in detail above, the present invention is not limited to the above-described embodiment, and it is needless to say that various improvements and modifications may be made without departing from the gist of the present invention.

DESCRIPTION OF SYMBOLS 10 Image publication system 12 Server 14 Client 16 Network 18 Image analysis part 20 Image classification part 22 Similar theme detection part 24 Intimacy calculation part 26 Recommendation information extraction part 28 Recommendation part

Claims (33)

A recommendation device that performs a display for recommending related information related to a second user to a terminal device of a first user who publishes an image via a network,
An image analysis unit that performs image analysis of the public image and the private image of the first user, and performs image analysis of the public image of the second user;
An image classification unit that classifies the first user's public image and private image for each first theme based on the image analysis result, and classifies the second user's public image for each second theme; ,
Based on the result of the image analysis, the theme is a common theme for the first theme and the second theme, and the interest of the first user and the interest of the second user are equal to or greater than a first threshold. A similar theme detection unit for detecting a similar theme;
A closeness calculating unit for calculating closeness between the first user and the second user based on the similar theme;
A recommendation unit comprising: a recommendation unit configured to display, on the terminal device of the first user, related information related to the second user when the closeness is equal to or greater than a second threshold. apparatus.   The recommendation device according to claim 1, wherein the image classification unit uses a preset theme as the first theme and the second theme.   The recommendation device according to claim 1, wherein the image classification unit determines the first theme and the second theme based on a result of the image analysis.   The said image classification | category part further determines the said 1st theme and the said 2nd theme based on the text data contributed with respect to the said 1st user's public image and the said 2nd user's public image. The recommendation apparatus as described in.   The degree of interest of the first user for each of the first themes is determined by the number of images classified into each of the first themes, and the degree of interest of the second user for each of the second themes is The recommendation apparatus according to claim 1, wherein the first threshold is a preset number of images. 5. The recommendation device according to claim 1, wherein the first threshold is a preset number of images.   The degree of interest of the first user for each of the first themes is determined by the ratio of images classified into each of the first themes, and the degree of interest of the second user for each of the second themes is The recommendation device according to claim 1, wherein the first threshold is a preset image ratio. 5. The recommendation apparatus according to claim 1, wherein the first threshold is a preset image ratio.   The degree of interest of the first user with respect to each of the first themes is determined by the number of images classified into each of the first themes and the ratio of images, and the interest of the second user with respect to each of the second themes. The degree is determined by the number of images classified into each of the second themes and the ratio of the images, and the first threshold is obtained by weighting a preset number of images and a preset ratio of images. It is the value of the function obtained, The recommendation apparatus as described in any one of Claims 1-4.   The recommendation device according to claim 1, wherein the closeness is determined based on a number of the similar themes, and the second threshold value is a preset number.   The degree of familiarity is determined by adding a score to each of the similar themes, and is determined by a total score of all the similar themes, and the second threshold value is a preset score. The recommendation apparatus as described in a term.   The closeness is determined by a shooting duration from a shooting start date and time to a shooting end date and time of an image included in at least one of the similar themes, and the second threshold is a preset time. The recommendation apparatus as described in any one of. The similar theme detection unit calculates a short-term interest degree of the first user based on the public image and the private image of the first user over a preset first period before the current time; and The short-term interest degree of the second user is calculated based on the public image of the second user over the first period, and the short-term interest degree of the first user and the short-term interest degree of the second user are Detect short-term similar themes that are above the first threshold,
The closeness calculation unit calculates short-term closeness between the first user and the second user based on the short-term similar theme,
The recommendation device according to any one of claims 1 to 7, wherein the recommendation unit displays the recommendation when the short-term closeness is equal to or more than the second threshold value. The similar theme detection unit is configured to obtain a long-term interest of the first user based on the public image and the private image of the first user over a preset second period longer than the first period before the current time. Calculating the degree of long-term interest of the second user based on the public image of the second user over the second period, and calculating the degree of long-term interest of the first user and the second user A long-term similar theme whose long-term interest is equal to or greater than the first threshold,
The intimacy calculating unit calculates long-term intimacy between the first user and the second user based on the long-term similar theme,
The recommendation device according to claim 11, wherein the recommendation unit displays the recommendation when the long-term closeness is equal to or greater than the second threshold.   The intimacy is scored higher for each of the long-term similar themes than for each of the short-term similar themes, and determined by the total score of all the short-term similar themes and all the long-term similar themes The recommendation device according to claim 12, wherein the second threshold value is a preset score. The image classification unit further classifies the images classified into each of the first themes for each first lower theme included in each of the first themes, and further classifies the images into the respective second themes. The images are further classified for each second lower theme included in each of the second themes,
The similar theme detection unit is a common theme for the first lower theme and the second lower theme, and the degree of interest of the first user and the degree of interest of the second user are equal to or higher than the first threshold. Detect similar sub-themes
The recommendation device according to claim 1, wherein the closeness calculating unit calculates closeness between the first user and the second user based on the similar lower theme.   The recommendation device according to claim 14, wherein the image classification unit uses preset themes as the first lower theme and the second lower theme.   The recommendation device according to claim 14, wherein the image classification unit determines the first lower theme and the second lower theme based on a result of the image analysis.   The image classification unit further determines the first lower theme and the second lower theme based on text data posted on the public image of the first user and the public image of the second user. Item 17. The recommendation device according to Item 16.   The degree of interest of the first user for each of the first lower themes is determined by the number of images classified into each of the first lower themes, and the degree of interest of the second user for each of the second lower themes The recommendation device according to any one of claims 14 to 17, wherein the determination is made based on the number of images classified into each of the second lower themes, and the first threshold is a preset number of images. .   The degree of interest of the first user with respect to each of the first lower themes is determined by a ratio of images classified into the first lower theme with respect to the first user's public image and private image, and The degree of interest of the second user with respect to the second lower theme is determined by a ratio of the images classified into the second lower theme with respect to the public image of the second user, and the first threshold is set in advance. The recommendation device according to any one of claims 14 to 17, which is a ratio of an image.   The degree of interest of the first user with respect to each of the first lower themes is the number of images classified into each of the first lower themes and the first lower theme with respect to the first user's public and private images. The degree of interest of the second user with respect to each of the second lower themes is determined by the ratio of the classified images, and the degree of interest of the second user with respect to each of the second lower themes The determination is based on a ratio of images classified as the second lower theme, and the first threshold is a function value obtained by weighting a preset number of images and a preset ratio of images. The recommendation apparatus as described in any one of claim | item 14-17.   21. The recommendation device according to any one of claims 14 to 20, wherein the closeness is determined based on a number of the similar lower themes, and the second threshold value is a preset number.   The intimacy is given a higher score than the similar theme for the similar lower theme, and is determined by a total score of all the similar themes and all the similar lower themes, and the second threshold is set in advance The recommendation device according to any one of claims 14 to 20, wherein the recommendation device is a score obtained.   The closeness is determined by a shooting duration from a shooting start date and time to a shooting end date and time of an image included in at least one of the similar lower themes, and the second threshold is a preset time. The recommendation device according to any one of 20. The similar theme detection unit calculates a short-term interest degree of the first user based on the public image and the private image of the first user over a preset first period before the current time; and The short-term interest degree of the second user is calculated based on the public image of the second user over the first period, and the short-term interest degree of the first user and the short-term interest degree of the second user are Detect short-term similar sub-themes that are above the first threshold,
The intimacy calculation unit calculates a short-term intimacy between the first user and the second user based on the short-term similar lower theme,
The recommendation device according to any one of claims 14 to 20, wherein the recommendation unit performs the recommendation display when the short-term closeness is equal to or greater than the second threshold value. The similar theme detection unit is configured to obtain a long-term interest of the first user based on the public image and the private image of the first user over a preset second period longer than the first period before the current time. Calculating the degree of long-term interest of the second user based on the public image of the second user over the second period, and calculating the degree of long-term interest of the first user and the second user A long-term similar sub-theme whose long-term interest degree is equal to or greater than the first threshold value,
The intimacy calculating unit calculates long-term intimacy between the first user and the second user based on the long-term similar lower theme.
The recommendation device according to claim 24, wherein the recommendation unit displays the recommendation when the long-term closeness is greater than or equal to the second threshold value.   The intimacy is scored higher for each of the long-term similar sub-themes than for each of the short-term similar sub-themes, and the total of all the short-term similar sub-themes and all of the long-term similar sub-themes The recommendation device according to claim 25, wherein the second threshold is a preset score.   The said recommendation part is a recommendation apparatus as described in any one of Claims 1-26 which performs the display which recommends it to the said 1st user by making the said 2nd user into a friend candidate with respect to the said 1st user's terminal device.   When the first user adds the second user as a friend, the recommendation unit displays a display for recommending the second user to the second user as a friend candidate for the terminal device of the second user. The recommendation device according to claim 27. Furthermore, based on the result of the image analysis, a recommendation information extraction unit that extracts recommendation information recommended to the first user from a public image of the second user included in each of the similar themes,
The said recommendation part is a recommendation apparatus as described in any one of Claims 1-23 which performs the display which recommends the said recommendation information to the said 1st user with respect to the said 1st user's terminal device. A recommendation method for performing a display for recommending related information related to a second user to a terminal device of a first user who publishes an image via a network,
An image analysis unit performing image analysis of the public image and the private image of the first user, and performing image analysis of the public image of the second user;
An image classification unit classifies the first user's public image and private image for each first theme based on the result of the image analysis, and classifies the second user's public image for each second theme. And steps to
Based on the result of the image analysis, the similar theme detection unit is a common theme of the first theme and the second theme, and the degree of interest of the first user and the degree of interest of the second user are first. Detecting a similar theme that is one or more threshold themes;
A step of calculating an intimacy between the first user and the second user based on the similar theme;
A recommendation unit, when the closeness is equal to or greater than a second threshold, performing a display for recommending related information related to the second user to the first user on the terminal device of the first user; Recommendation method including. The image classification unit further classifies the images classified into each of the first themes for each first lower theme included in each of the first themes, and further classifies the images into the respective second themes. The images are further classified for each second lower theme included in each of the second themes,
The similar theme detection unit is a common theme for the first lower theme and the second lower theme, and the degree of interest of the first user and the degree of interest of the second user are equal to or higher than the first threshold. Detect similar sub-themes
The recommendation method according to claim 30, wherein the closeness calculating unit calculates closeness between the first user and the second user based on the similar lower theme.   The program for making a computer perform each step of the recommendation method of Claim 30 or 31.   32. A computer-readable recording medium on which a program for causing a computer to execute each step of the recommendation method according to claim 30 or 31 is recorded.

Images