JP5892839B2 - Profile generation apparatus and program - Google Patents

Description

  The present invention relates to a profile generation apparatus and a program for generating a user profile.

  In recent years, as one of information provision, a technique has been proposed in which a program that matches a user's preference is searched in advance and recommended to the user before the program starts. For example, in the program recommendation device described in Patent Literature 1, an EPG of a program that has been viewed in the past is analyzed based on a user's viewing history, a program that the user likes is grasped, and a program that meets the user's preference is selected. A method of recommending has been proposed (see, for example, Patent Document 1).

  Moreover, in the program recommendation device, in addition to using the viewing history, for example, the user's preference is inferred based on information such as keywords registered in advance by the user or blog writing managed by the user, A configuration is also possible in which a viewing profile is created and a program that matches the viewing profile is recommended.

JP 2008-199406 A

  However, when the user refuses to provide the viewing history from the viewpoint of privacy, such a program recommendation device cannot acquire the viewing history and cannot recommend a program that suits the user's preference.

  Also, registering a favorite keyword for program recommendation by the user himself / herself or writing on a blog may be burdensome and may not be used. In such a case, the program recommendation device cannot infer user preferences and cannot recommend programs that match the user preferences.

  In addition to the program, the user has an event that interests him / her (for example, watching a movie, watching a sport, etc.), and therefore wants to receive a recommendation for an event that suits his / her preference.

  Therefore, in order to solve the above problem, the present invention provides a profile generation device and a program capable of generating a profile for estimating a user's favorite event without depending on the user's own information. One of the purposes.

  The profile generation device according to claim 1 has a first account of its own in a first social networking service that provides a service for posting comments and communicating with other users, and The second social networking service is different from the first social networking service, and is a second social networking service that provides a service for posting a comment together with location information of an actually visited place to communicate with other users. A registration unit for registering a user who wants to recommend predetermined information together with the data of the first account and the data of the second account, and the first registered in the registration unit 1 data of the account is acquired and the predetermined information is recommended. A first obtaining unit that obtains data of accounts of other users who are followed by a user who desires, and obtains data of accounts of all users who are provided with the first social networking service. A second acquisition unit that acquires, for each user, the number of users indicating whether or not the user has been followed, and the number of utterances for the predetermined information, and the other user acquired by the first acquisition unit Based on the account data, a user who satisfies the first condition is extracted from the other users, and the second number is obtained based on the number of users and the number of utterances acquired by the second acquisition unit. A user who satisfies the above condition is extracted from all the users, the first related user specifying unit for specifying the extracted user as a first related user, and the registration unit Another user who obtains the recorded data of the second account, analyzes the behavioral form of the user who wants to recommend the predetermined information based on the posted location information, and is similar to the behavioral form Is extracted from the second social networking service, the data of the account extracted by the extraction unit that extracts the data of the account of the other user, and the account linked to the account A determination unit that determines whether or not the first social networking service exists in the first social networking service, and when the determination unit determines that the account to be linked exists in the first social networking service, A second related user specifying unit for specifying as a second related user, the first related user and the previous The information related to the second related user is analyzed from the data of each account provided by the first social networking service, and a user profile of the user who wants to recommend the predetermined information is created. The configuration includes a user profile creation unit.

  According to such a configuration, the profile generation apparatus of the present invention can generate a profile for estimating a user's favorite event without depending on the user's own information.

  According to a second aspect of the present invention, in the profile generation device according to the first aspect, the event profile creation unit that obtains the transmitted event information and creates an event profile indicating the characteristics of the event information; The user profile created by the user profile creation unit and the event profile created by the event profile creation unit are compared to calculate the similarity between both profiles, and the similarity calculation unit calculates On the basis of the result, a transmission unit is provided that transmits event information corresponding to the event profile to a user who wants to recommend the predetermined information registered in the registration unit.

  According to such a configuration, the profile generation device of the present invention can notify the user of event information that the user is likely to be interested in from various events based on the created user profile.

  According to a third aspect of the present invention, in the profile generation device according to the first aspect, the first related user specifying unit is based on user account data acquired by the first acquisition unit. As a first condition, a user who makes a speech of a predetermined number or more with respect to the predetermined information is extracted, and based on the number of users acquired by the second acquisition unit and the number of times of the speech, the second As a condition, a configuration is adopted in which users who are followed by a predetermined number or more of users and who make a statement of a predetermined number or more are extracted.

  According to such a configuration, the profile generation device of the present invention specifies the first related user based on the number of users following a page held by another user, and therefore desires to recommend predetermined information. A user who is connected to the user on the SNS or a popular representative user can be the first related user.

  The invention described in claim 4 is the profile generation device according to claim 1, wherein an account provided by the first social networking service is registered in a page provided by the second social networking service. An account registration function is provided, and the determination unit analyzes the account data extracted by the extraction unit and determines whether or not a page linked with the page exists in the first social networking service. It was set as the structure to do.

  According to such a configuration, the profile generation device of the present invention determines whether or not the account of the page provided by the first social networking service is registered in the page provided by the second social networking service. It is possible to easily determine whether or not the page is linked to the page provided by the first social networking service.

  According to a fifth aspect of the present invention, in the profile generation device according to the second aspect, the user profile creation unit is configured to use the first related user and the information related to the second related user as the first information. Analyzing from the data of each account provided by the social networking service of all, subjecting all writing to the information, and calculating the weight for a predetermined word appearing in the writing of interest, the event profile creation unit, The content vector constituting the generated event profile is shortened based on the word weight calculation, and the user profile creation unit creates the user profile based on the shortened content vector, and the similarity The calculation unit is the event profile creation unit. And event profiles shortened more content vectors, by comparing the user profile created by the user profile generator, and configured to calculate the similarity of the two profiles.

  According to such a configuration, the profile generation device of the present invention targets writing of all users of the first social networking service when there is a certain number of related users for users who want to recommend predetermined information. Instead, the content vector constituting the event profile is shortened only for writing by related users, so that an event profile that matches the preference of the user who wants to recommend more predetermined information can be created.

  According to a sixth aspect of the present invention, in the profile generation device according to the fifth aspect, the event profile creation unit uses a weight calculated with respect to the predetermined word for a content vector constituting the event profile. The content vector is set to 0 in the content vector with respect to an element having a weight lower than a specific threshold or an element other than a specific number of elements from a high weight according to the computing environment. It was set as the structure which shortens.

  According to such a configuration, the profile generation device of the present invention sets the weight in the content vector to 0 for an element having a weight lower than a specific threshold or an element other than a specific number of elements from a high weight element. Since the content vector is shortened, it is possible to create an event profile that matches the preference of the user who desires the recommendation of predetermined information.

  The program according to claim 5 has a first account of itself in a first social networking service that provides a service for posting comments and communicating with other users, and the first The second account of the second social networking service, which is different from the social networking service of the second social networking service and provides a service for communicating with other users by posting a comment together with location information of the actually visited place A registration unit for registering a user who wants to recommend predetermined information together with the data of the first account and the data of the second account, and the first registered in the registration unit Obtain account data and request recommendation of the specified information A first acquisition unit that acquires data of accounts of other users who are followed by the user, and acquires data of accounts of all users who are provided with the first social networking service. A second acquisition unit that acquires, for each user, the number of users indicating whether the user is following, and the number of remarks about the predetermined information; and the account of the other user acquired by the first acquisition unit Based on the data, a user who satisfies the first condition is extracted from the other users, and the second condition is determined based on the number of users and the number of utterances acquired by the second acquisition unit. First, a user is extracted from all the users, the first related user specifying unit that specifies the extracted user as a first related user, and registered in the registration unit. The second account data is acquired, and based on the posted position information, the behavioral form of the user who wants to recommend the predetermined information is analyzed, and another user similar to the behavioral form is analyzed. The social networking service identified from the two social networking services, and extracting the account data of the other user, analyzing the account data extracted by the extraction unit, the account linked to the account is the first A determination unit that determines whether or not the social networking service exists in the social networking service; and if the determination unit determines that the account to be linked exists in the first social networking service, A second related user specifying unit for specifying as a related user, the first related user and the second related A user profile creation unit that analyzes information that a continuous user is talking about from the data of each account provided by the first social networking service, and creates a user profile of a user who desires recommendation of the predetermined information The computer having the function is configured to function as a profile generation device.

  According to this configuration, the program of the present invention can generate a profile for estimating a user's favorite event without depending on the user's own information.

  According to the present invention, it is possible to generate a profile for estimating a user's favorite event without depending on the user's own information.

It is a functional block diagram which shows the structure of a profile production | generation apparatus. It is a figure which shows typically the structure of a 1st SNS page. It is a figure which shows typically the structure of a 2nd SNS page. It is a figure which shows typically the registration table registered into the registration part. It is a figure used for description about a series of operation | movement which extracts a charismatic user by a profile production | generation apparatus, determines a related user, estimates the program which the related user watched, and determines a recommended program. It is a figure where it uses for description about the production method of the viewing profile of a related user. It is a figure with which it uses for description about the method of recommending a program based on a viewing profile. It is a figure showing the viewing frequency for every program of a related user. It is a flowchart with which it uses for description about the flow of a process when shortening a content vector.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
As shown in FIG. 1, the profile generation device 1 includes a registration unit 10, an acquisition unit 11, a first related user specifying unit 12, an extracting unit 13, a determining unit 14, and a second related user specifying unit. 15 and a user profile creation unit 16. In addition, the profile generation device 1 is a first social networking service (hereinafter referred to as SNS1) that is a service for posting a comment with a plurality of user terminals 200 via the network 100 to communicate with other users. The SNS 1 is different from the Web server 300 that provides the service, and is a second service that is a so-called location-based service that posts a comment together with position information of a place actually visited and communicates with other users. Web servers 301 that provide a social networking service (hereinafter referred to as SNS2) are connected to each other.

  The registration unit 10 owns its own first account in SNS1 and owns its second account in SNS2, and desires to recommend predetermined information (hereinafter referred to as user U1). .) Together with the data of the first account and the data of the second account.

For example, the user U1 can acquire data of the first account by performing a predetermined registration process in the SNS1. In addition, the user U1 can access his home page created in the SNS 1 (hereinafter referred to as the first SNS page) by using the acquired data of the first account.
Similarly, the user U1 can acquire data of the second account by performing a predetermined registration process in the SNS2. Further, the user U1 can access his homepage (hereinafter referred to as a second SNS page) created in the SNS2 by using the acquired data of the second account.
Next, the main configuration of the first SNS page (hereinafter referred to as page P1) will be described with reference to FIG. The page P1 includes an area A1 for inputting the user U1's own message, a message area A2 in which the user U1's message and the user's following message are displayed in a new order, and a follow area A3 indicating the user who is following the user U1. Etc. Note that the user icon shown in the follow area A3 has a function (hyperlink function) that allows a transition (jump) to the user's page. Further, in SNS1, a unique URL is assigned to the posted message, and it is sequentially updated as a timeline in the message area A2 on the page of the user being followed.

  Next, the main configuration of the second SNS page (hereinafter referred to as page P2) will be described with reference to FIG. The page P2 includes a check-in area B1 indicating the number of places (check-in) where the user U1 has visited, a link information area B2 indicating a link to the page P1 in the SNS1, and the number of times the user U1 has visited in the check-in area. It is composed of a Mayer area B3 indicating the number of titles (Meyer) given when there are more than users, a posting area B4 for displaying posted comments, and the like.

  Specifically, the SNS 2 links the mobile terminal with GPS function and the page P 2 owned by the user U 1 by a predetermined registration work, and visits the place where the user U 1 possessing the mobile terminal is located and checks When the in operation is performed, a process of increasing the number of check-in areas B1 of the page P2 is performed. In SNS2, a title (Meyer) is assigned to the user who has checked in the same place most frequently, and a new mark is registered in the Mayer area B3 of page P2.

The acquisition unit 11 includes a first acquisition unit 11a and a second acquisition unit 11b.
The 1st acquisition part 11a acquires the data of the 1st account registered into registration part 10, and based on follow (following) information, the account of other users who user U1 is following (following) Get the data.
Moreover, the 2nd acquisition part 11b acquires the data of all the users' accounts in SNS1, and acquires the number of users which shows how many users are followed, and the frequency | count of remarks about predetermined information for every user. To do.

Moreover, the 1st related user specific | specification part 12 extracts the user which satisfy | fills 1st conditions from other users based on the data of the user's account acquired by the 1st acquisition part 11a. Moreover, the 1st related user specific | specification part 12 extracts the user who satisfy | fills 2nd conditions from all the users based on the number of users acquired by the 2nd acquisition part 11b, and the frequency | count of utterance. In addition, the user extracted from all the users means the charismatic user mentioned later in a present Example.
The 1st related user specific | specification part 12 specifies the user extracted by the 1st acquisition part 11a and the 2nd acquisition part 11b as a 1st related user.

  In addition, the extraction unit 13 acquires the data of the second account registered in the registration unit 10, analyzes the behavior form of the user U1 based on the posted position information, and is similar to the behavior form Are identified from SNS2 and the account data of the other users are extracted.

  The determination unit 14 analyzes the data of the account extracted by the extraction unit 13 and determines whether an account that cooperates with the account exists in the SNS 1.

  The second related user specifying unit 15 specifies the owner of the account as the second related user when the determining unit 14 determines that the linked account exists in the SNS 1.

  The user profile creation unit 16 analyzes the information that the first related user and the second related user are interested in (hereinafter referred to as the topic information) from the data of each account in the SNS 1, Create a user profile for user U1. The user profile created in this way shows the characteristics of information that the user U1 is likely to be interested in. The user profile is created based on the content vector created by the event profile creation unit 17.

  Since it is configured in this way, the profile generation device 1 does not depend on the information of the user U1 itself (for example, writing to the page P1 in the SNS1, etc.), the user connected to the user U1 and the SNS1, and the popularity A profile for estimating user U1's favorite information (event) can be generated based on the writing of a page in SNS1 owned by a representative user or a user whose behavioral form is similar to that of user U1.

  Although the configuration of the profile generation device 1 has been described above, the configuration is not limited to this, and a computer that includes each component and generates a profile for estimating a favorite event of the user U1 is caused to function as the profile generation device. It may be configured as a program.

  Further, as shown in FIG. 1, the profile generation device 1 includes an event profile creation unit 17, a similarity calculation unit 18, and a transmission unit 19.

  The event profile creation unit 17 acquires event information transmitted from the event information transmission server 400 and creates an event profile indicating the characteristics of the event information. Here, when the event information is information of a television program scheduled to be broadcast, the event profile creation unit 17 performs morphological analysis on the EPG transmitted from the event information transmission server 400 and generates an event profile indicating the characteristics of the program contents. Create for each broadcast program. In addition, when the event information is information on a movie that is being released or is scheduled to be released, the event profile creation unit 17 is based on movie information described on a homepage of a distribution company as the event information transmission server 400, for example. An event profile showing the characteristics of the movie content is created for each movie. The event is not limited to a television program or a movie, but may be a radio program, a drama, sports, news, or the like.

  In addition, the similarity calculation unit 18 compares the user profile created by the user profile creation unit 16 with the event profile created by the event profile creation unit 17 to calculate the similarity between both profiles.

  Based on the result calculated by the similarity calculation unit 18, the transmission unit 19 transmits the event information corresponding to the event profile to the user U 1 registered in the registration unit 10 using a message. For example, the message is received by the page P1 and displayed in the speech area A2.

  With this configuration, the profile generation device 1 can notify the user of event information that the user is likely to be interested in from various events based on the created user profile.

Further, the first related user specifying unit 12 has a condition (first condition) that a predetermined number of statements have been made on the event based on the user account data acquired by the first acquiring unit 11a. As a user is extracted.
Moreover, the 1st related user specific | specification part 12 is followed by the predetermined number or more users based on the number of users and the frequency | count of utterance acquired by the 2nd acquisition part 11b, and utters more than a predetermined number of utterances. As a condition (second condition), the user is extracted.
The 1st related user specific | specification part 12 may be the structure which specifies the user extracted by the 1st acquisition part 11a and the 2nd acquisition part 11b as a 1st related user.

  Since it is configured in this manner, the profile generation device 1 specifies the first related user based on the number of users following pages owned by other users, and therefore desires to recommend predetermined information. A user who is connected to the user U1 and the SNS1 or a popular representative user can be specified as the first related user.

  In addition, the 1st related user specific | specification part 12 may specify only a user with the largest number of users acquired by the 2nd acquisition part 11b as a 1st related user.

  The page provided by SNS2 is provided with an account registration function for registering an account provided by SNS1 (see link information area B2 in FIG. 3).

  In such a configuration, the determination unit 14 analyzes the account data extracted by the extraction unit 13, refers to the account registration function of the page, and a page linked to the page exists in the SNS 1. Judge whether to do.

  Therefore, the profile generation device 1 can easily determine whether or not the page provided by SNS1 is linked to the page provided by SNS1 based on whether or not the account of the page in SNS1 is registered in the page in SNS2.

  Here, an embodiment of the profile generation device 1 as a program recommendation device that recommends a television program scheduled to be broadcast will be described below.

  As shown in FIG. 1, the profile generation device 1 includes a plurality of user terminals 200, a Web server 300 that provides SNS 1, a server 301 that provides SNS 2, It is connected to an event information transmission server 400 having electronic program information (EPG, Electronic Program Guide). The EPG includes broadcast time information, performer information, program summary information, and the like of all broadcast programs.

  The registration unit 10 has its own page P1 in SNS1 and has its own page P2 in SNS2, and a user who wants to recommend a broadcast program (hereinafter referred to as user U1) is referred to as page P1. It is registered together with information (URL etc.) of page P2. For example, the registration unit 10 has a registration table as shown in FIG. 4. The ID of a user who desires recommendation in the registration table, the user name, the account ID (first account ID) in SNS 1, The account ID (second account ID) in SNS2 is associated and registered.

Here, before registering by the registration unit 10, the user U1 needs to have his / her own page in SNS1 and SNS2.
The Web server 300 is unique to a registered user (for example, the user name is “abcd”), such as Twitter or Facebook, for example, http://SNS1.com/ abcd), and a page displayed by this URL provides SNS1 that allows users to communicate with each other by following each other's remarks.

  Note that the SNS 1 may be any SNS as long as the user's “now” is written and the relationship with other users is clearly indicated, such as Twitter. The present invention can be realized by using.

  Further, the Web server 301 sends a unique URL (for example, http://SNS2.com/abcd) to a registered user (for example, the user name is “abcd”), for example, like Foursquare. The SNS 2 is provided that communicates with other users by posting a comment (Tip) together with location information of the actually visited place on the page displayed by the URL.

  In addition, SNS2 is a form in which a check-in to a user's location is performed, a comment regarding the user's location is written together with check-in, and the relationship with other users is clearly shown, as in Foursquare. Any SNS may be used, and the present invention can be realized by using this.

The acquisition unit 11 browses the page P1 registered in the registration unit 10, refers to the follow area A3 of the page P1, changes to the page of another user that is followed by the user U1, and displays the page. Analyze and obtain the number of users following the page.
The first related user specifying unit 12 specifies the first related user from among other users based on the number of users acquired by the acquiring unit 11.

  In addition, the extraction unit 13 acquires, for example, account data registered in the registration unit 10, and the user who has checked in the same place, the user who is a Mayer at the place, or the posting area B 4 On the other hand, a user or the like who makes a posting is identified as another user similar to the behavior form of the user U1, and the page of the other user is extracted.

  The determination unit 14 analyzes the page extracted by the extraction unit 13 and determines whether or not a page linked to the page exists in the SNS 1.

  The second related user specifying unit 15 specifies the owner of the page as the second related user when the determining unit 14 determines that the linked page exists in the SNS 1.

The user profile creation unit 16 analyzes (for example, morphological analysis) pages in the SNS 1 of the first related user and the second related user (hereinafter simply referred to as related users), and is commented based on the result of the analysis. One or a plurality of broadcast programs viewed by related users are estimated with reference to the EPG from the current time and the comment content.
The user profile creation unit 16 notifies the event profile creation unit 17 of one or more estimated broadcast programs.

Then, the event profile creation unit 17 extracts the notified one or more broadcast program information from the broadcasted EPG, performs morphological analysis on the extracted one or more broadcast program information, and based on the analysis result, A program profile is created based on the characteristics of the broadcast program that the related user was viewing.
The event profile creation unit 17 notifies the user profile creation unit 16 of the content vector of the created program profile.
The user profile creation unit 16 creates a user profile for the user U1 based on the notified content vector.

  The event profile creation unit 17 acquires an unbroadcast EPG transmitted from the event information transmission server 400, performs morphological analysis, and generates an event profile (hereinafter referred to as a program profile) indicating the characteristics of the program content for each broadcast program. To create.

  The similarity calculation unit 18 compares the user profile of the user U1 created by the user profile creation unit 16 with the program profile for each broadcast program created by the event profile creation unit 17, and the program profile having a high similarity The broadcast program corresponding to is determined as a recommended program.

  The transmission unit 19 transmits the recommended program determined by the similarity calculation unit 18 to the user U1 registered in the registration unit 10 with a message. For example, the message transmitted by the transmission unit 19 is displayed on the page P1 of the user U1.

  In the present embodiment, the recommended program information transmitted by the transmission unit 19 is displayed in the speech area A2 in the page P1 managed by the user U1 (reply function), but is not limited to this form. For example, the transmission unit 19 may transmit information on a recommended program determined by a transmission unit that does not use SNS such as e-mail.

  In this way, the profile generation device 1 recommends a program that the user U1 is likely to like based on the remarks of other users or the like that are similar to the user U1's behavior, without depending on the information of the user U1 itself. Is possible. The profile generation device 1 can also recommend a broadcast program that the user U1 potentially wants, although it is not the content of the program that is normally viewed.

  Further, for example, the profile generation device 1 displays recommended program information (information such as a program title and broadcast time) at a predetermined location (for example, a speech area A2) on the page P1 managed by the user U1. Information of recommended programs can be provided on the SNS, and it is not necessary to use other applications separately, and can be realized efficiently and inexpensively.

  Moreover, since the profile generation apparatus 1 estimates the viewing history of the program from writing to the SNS 1, it is possible to recommend a program to the user U1 without having each user provide viewing history as personal information. Can do. Further, since the profile generation device 1 does not require keyword registration or personal profile setting as in the prior art, it is possible to recommend a program without imposing a burden on the user U1.

  The profile generation device 1 can also recommend a program from the user U1's own writing. However, even if the user U1 does not have a statement, the profile generation device 1 can perform a program recommendation based on the relevant user's statement. Can do.

  In addition, if the user U1 is following, the profile generating apparatus 1 recommends the program in consideration of the information, but the function such as following is not necessarily used. This is because the profile generation device 1 recommends a program based on the user's remarks if there is a user who is related on the SNS. This is because program recommendation can be performed based on a statement of a user (charisma user) who satisfies (for example, condition 1 and condition 2 described later). Further, the profile generation device 1 may make a program recommendation based on the remarks of a plurality of charismatic users. In this case, the profile generation device 1 can recommend a wide variety of information to the user U1.

  In addition, when a program recommendation is made based only on the remarks of the user U1, there is a high possibility that a program close to the program that the user U1 himself has viewed so far is recommended. On the other hand, since the profile generation device 1 recommends information based on the remarks of other users, for example, it is possible to recommend programs that the user U1 did not know, such as popular programs or programs that friends are watching. , Can be used to create opportunities to watch new programs.

  Here, the profile generation device 1 performs the following operations: 1. Extract charismatic users; 2. Determine related users (users who serve as criteria for recommendation to user U1); A series of operations for estimating a program viewed by a related user and determining a recommended program will be described with reference to FIG.

<1. Extraction of charismatic users>
Here, a method for extracting a charismatic user will be described.
The 1st related user specific | specification part 12 analyzes the utterance of all the users in SNS1, and the situation of following (the situation of being followed and the situation of being followed), and the user who satisfy | fills the following two conditions beforehand is analyzed. Extract as a charismatic user. However, this operation is not necessary when a charismatic user is not used.

Condition 1. The number of followers is a certain number or more.
“Following” means registering another user as a favorite. In SNS1, this favorite user's remark is displayed in the remark area A2 on the page of the user being followed.

  The 1st related user specific | specification part 12 analyzes the page of the related user registered into favorites, and when many users are followed, presumes that it is a popular user. In the example shown in FIG. 5, the 1st related user specific | specification part 12 estimates the user U2 followed by the some user as a popular user. In addition, the one-way | direction arrow in FIG. 5 shows the user whose start end follows, and the end shows the user who is followed by the user. In addition, the bidirectional arrows indicate that users at both ends are following each other.

Condition 2. Talking about the broadcast program more than a certain number.
The 1st related user specific | specification part 12 analyzes the content of an utterance, and when the hash tag regarding a broadcast program is added to the utterance, it considers that it was uttered about the program.

  In the present embodiment, conditions 1... Are selected from all users in SNS 1. And condition 2. The charismatic user satisfying the above condition is the related user. However, the present invention is not limited to this, and the user that the related user is following may be the related user.

<2. Determination of related users>
Next, a related user determination method will be described. The related user refers to a user who becomes a reference for recommendation to the user U1.
In principle, the related user is a user that the user U1 follows on the page P1 managed by the user U1 in the SNS1.
In addition, the related user checked in to the user who is the Meyer or the place where the user U1 checked in (or the place where the comment was written) on the page P2 managed by the user U1 in the SNS2. A user or a user who has entered a comment and has a page linked to SNS1 also corresponds.
In particular, it can be said that the greater the number of places checked-in to each other (or places where comments are written), the higher the degree of association between the user U1 and other users.

Here, an example of a method for calculating the degree of association will be shown. Based on the following equation (1), the extraction unit 13 calculates the behavioral form similarity A for each place where the user U1 has visited, and normalized the behavioral form similarity A calculated for each place to 0 to 1. Above, adding up, it is judged that the degree of association (similarity) is higher as the total value is larger.
A = [(a × user U1 Mayer flag) + (b × user U1 check-in count) + (c × user U1 post count)] × [(d × other user Meyer flag) + (e × Number of check-ins of other users) × (f × number of posts of other users)] ... (1)

  The Mayer flag is “1” if it is a Mayer, and “0” if it is not a Mayer. Further, a, b, c, d, e, and f are variables. The profile generation device 1 can set a large value for a, b, c, d, e, or f for important items. For example, when the profile generation device 1 determines that “the user U1 Meyer flag” is the most important item in calculating the similarity of the behavioral form, “a” is compared with other variables. Set a large value.

  Further, the determination unit 14 may select a user exceeding a predetermined value as a candidate for a related user (second related user) based on the total value calculated as described above, or a user with the highest value. It is good also considering only a candidate of a related user (2nd related user). And the judgment part 14 judges whether the user who became a candidate has the page which SNS1 cooperates.

  The second related user specifying unit 15 specifies the owner of the page as the second related user when the determining unit 14 determines that the linked page exists in the SNS 1.

  The first related user specifying unit 12 further analyzes the relationship between users based on the follow status, and specifies the users that the following users are following, the users that are following each other interactively, etc. Users who satisfy the condition can also be included in related users.

<3. Method for estimating program viewed by related user and method for determining recommended program>
Next, a method for estimating a program that has been viewed by a related user performed by the profile generation device 1 will be described. In the following description, the first related user specifying unit 12 and the second related user specifying unit 15 have three first related users and second related users (hereinafter simply referred to as related users) (related users). Specific operations of the user profile creation unit 16, event profile creation unit 17, and similarity calculation unit 18 in the case where they are specified will be described below.

  As shown in FIG. 6, the user profile creation unit 16 determines the characteristics of the broadcast program that the related user was watching from the EPG based on the result of analyzing the page of the related user and the time and content of the utterance. A profile indicating the viewing status of each related user (hereinafter referred to as a viewing profile) is extracted and expressed as a vector.

  For example, in the user profile creation unit 16, the related user A sends “1/28 23:00 patent forefront” and “1/31 21:00 trademark transmission” based on the time and the content of the statement by the related user A. Determine that you were watching. Then, the user profile creation unit 16 analyzes the EPG program information of the broadcast program “patent front line” and the broadcast program “trademark transmission” from this determination result, calculates the feature amount for each word, and relates to the related user A's A viewing profile a is created.

In addition, the user profile creation unit 16 creates the viewing profile b of the related user B and the viewing profile c of the related user C, as described above.
Next, the user profile creation unit 16 adds the viewing profiles a, b and c of each related user to create a user profile (hereinafter referred to as user profile x) of the user U1 (hereinafter simply referred to as user U1). To do. Note that the user profile creation unit 16 may create a user profile by applying a high weight to the viewing profile of a user who has a high degree of relevance with the user U1 (for example, a second related user with a close behavior).

  Then, the similarity calculation unit 18 compares the user profile x of the user U1 with the program profile for each broadcast program created by the event profile creation unit 17, and recommends a broadcast program corresponding to the program profile having a high similarity. To decide.

  In this way, the profile generation device 1 creates the user profile of the user U1 based on the viewing profiles of the related users Mr. A, Mr. B, and Mr. C, and creates the user profile of the user U1 and each broadcast program. Since the program profiles are compared and the broadcast program corresponding to the program profile having a high degree of similarity is determined as the recommended program, the user U1 and the user who is connected on the SNS or the popular representative user (related users A and B) It is possible to recommend a broadcast program having the same characteristics as the broadcast program being viewed by Mr. and Mr. C).

Next, a method for creating a program profile and a method for calculating similarity will be described in detail.
First, a method for creating a viewing profile of a related user of the user q _u (user U1) who desires recommendation will be described. The user profile creation unit 16 performs morphological analysis on the program information included in the EPG of the broadcast program that has been viewed, according to the appearance frequency of characteristic words (for example, names of performers and names of places) and the type of part of speech. Assign an evaluation value to each word. The user profile creation unit 16 creates a profile vector Q _r of one related user q _r from the result of morphological analysis. For example, the profile vector Q _r can be expressed as shown in equation (2).
Q _r = <(k ₁ , Z _r1 ), ..., (k _n , Z _rn )> (2)
However, Z _rj represents the weight of the word k _j for the related user q _r. For example, the weight of the word k _j in each broadcast program viewed by the related user q _r is calculated by the tf-idf method. This is the sum of the weights of the words k _j in all broadcast programs viewed by the user _qr . In addition, the profile vector Q _r, in order to be something that shows the characteristics of the broadcast program, word about the performer's name and location, are at a higher weighting compared to the other word.
In this manner, viewing profiles of other related users are created sequentially.

ただし、Ｑ_ｋは、推薦を希望するユーザｑ_ｕの関連ユーザｑ_ｋのプロファイルベクトルを示しており、ａ_ｋは、関連ユーザｑ_ｋの関連パラメータを示している。関連パラメータとは、関連ユーザごとに付与される重み付けである。詳細は後述するように、ユーザプロファイル作成部１６は、関連ユーザの中でも重要度の高い関連ユーザ（カリスマユーザ）の視聴プロファイルを他の関連ユーザの視聴プロファイルよりも優先し、高い重み付けを付与する。 Next, the user profile creation unit 16 creates a user profile of the user q _u who desires recommendation based on the viewing profiles of all related users. For example, the user profile of the user q _u who desires recommendation can be expressed as a profile vector Q ′ _u as shown in the equation (3).

Here, Q _k indicates the profile vector of the related user q _k of the user q _u who desires recommendation, and a _k indicates the related parameter of the related user q _k . The related parameter is a weight given to each related user. As will be described in detail later, the user profile creation unit 16 gives priority to a viewing profile of a related user (charisma user) having a high degree of importance among related users over a viewing profile of another related user, and assigns a higher weight.

In addition, when there is only one related user of the user q _u who desires recommendation, the user profile creation unit 16 uses the viewing profile of this related user as the user profile of the user q _u who desires recommendation.

In addition, the event profile creation unit 17 acquires the EPG of the broadcast scheduled program from the event information transmission server 400, analyzes it, and creates a program profile for each broadcast program. For example, the program profile of the broadcast program d _i can be expressed as the content vector D _i as shown in equation (4).
D _i = <(k ₁ , W _i1 ),... (K _n , W _in )> (4)
However, W _ij represents the weight of the word k _j in the broadcast program d _i calculated by the tf-idf method, for example. For example, the content vector indicates the characteristics of a broadcast program, and words related to performer names and places are weighted higher than other words.

Next, the similarity calculation unit 18 determines a broadcast program having a high similarity between the profile vector Q ′ _u and the content vector D _i as a recommended program. Here, the similarity (SIMILARY) between the user q _u who desires recommendation and a certain broadcast program d _i can be expressed as shown in equation (5).

  In addition, the user profile creation unit 16 searches for, for example, a statement in which a predetermined symbol indicating that it is a broadcast program is inserted, and in principle, only a statement in which the predetermined symbol is inserted is subject to analysis. To. Here, the predetermined symbol is a hash tag (a word prefixed with “#”).

For example, if “#nhk” is included in the utterance, the user profile creation unit 16 determines that the utterance is an utterance about the NHK broadcast program, and based on the time when the utterance is made, With reference to the EPG, the broadcast program that the related user was viewing is estimated, and information on the broadcast program is analyzed. In Twitter, while listening to a broadcast program, a style of speaking by adding a broadcast station name or a hash tag of the program name is becoming customary.
The user profile creation unit 16 extracts not only predetermined symbols but also keywords specific to a broadcast program such as a program name and a character name from an EPG in advance, and these keywords are included in the remarks. For example, it may be configured that the broadcast program is estimated and the information of the broadcast program is analyzed.
Further, if only the hash tag such as “#nhk” is included, the user profile creation unit 16 estimates that the program that was being broadcast at the time when the speech was made was viewed.
In addition, when a hash tag unique to a program is added (for example, “#shohuyouden”), the user profile creation unit 16 has the hash tag regardless of which program is being broadcast at the time of the speech. The broadcast program (for example, “trademark transmission”) may be estimated to have been viewed.
Note that the user profile creation unit 16 regards the past program as a speech when the time when the speech is made is somewhat away from the time zone during which the broadcast program with the hash tag was broadcast, and the hash tag It is not estimated that you are watching the broadcast program.

  In this way, since the profile generation apparatus 1 analyzes only messages that are closely related to the broadcast program, it is possible to exclude messages that are not related to the broadcast program, so that it is possible to efficiently analyze the messages. It has become.

  Next, a specific operation of the profile generation device 1 will be described. In the following, a case 1 in which the program that the related user A was watching is recommended as it is and a case 2 in which a performer of the program that the related user A has been viewing appears is recommended. . In the following description, it is assumed that the related user is only one related user Mr. A, and that there is only one broadcast program viewed by the related user A.

  In case 1, as shown in FIG. 7A, the user profile creation unit 16 makes a statement “2010/1/28 (Thu) 23:38 of the related user A on the page on the SNS1 of the user U1. I want to go. Analyze "#nhk". The user profile creation unit 16 presumes that the related user A was watching the NHK broadcast program “Patent Frontline”.

  Also, the user profile creation unit 16 analyzes the information of the broadcast program “patent front line” acquired from the EPG, creates a viewing profile of the related user A, and uses this for the user profile of the user U1.

  The similarity calculation unit 18 compares the user profile of the user U1 created by the user profile creation unit 16 with the program profile for each broadcast program created by the event profile creation unit 17, and from the comparison result, for example, in the next week The “patent frontline” to be broadcast is determined as the recommended program.

The transmission unit 19 transmits “patent front line” as a recommended program to the page of the user U1 at a predetermined timing (at least before the broadcast time).
For example, in the speech area A2 on the page of user U1, “Now, the“ patent front line ”will begin with NHK. Is displayed.

  Further, in case 2, as shown in FIG. 7B, the user profile creation unit 16 makes a statement “2010/1/31 (Sunday) 21:26 of the related user B on the page on the SNS1 of the user U1. Analyze “#nhk”. The user profile creation unit 16 estimates that the related user B was watching the NHK broadcast program “Trademark”.

  Further, the user profile creation unit 16 analyzes the information of the broadcast program “trademark transmission” acquired from the EPG, extracts performers of “trademark transmission” (Taro Patent), and creates a viewing profile of the related user B. This is used for the user profile of the user U1.

The similarity calculation unit 18 compares the user profile of the user U1 created by the user profile creation unit 16 with the program profile for each broadcast program created by the event profile creation unit 17. Based on the comparison result, the similarity calculation unit 18 determines, for example, another broadcast program “utilization of design” in which Taro Patent, who appeared in “Trademark Den”, appears as a recommended program.
The transmission unit 19 transmits “trademark transmission” and “utilization of design” as a recommended program to the page of the user U1 at a predetermined timing (at least before the broadcast time).

  For example, in the speech area A2 on the page of the user U1, “Now,“ use of design ”will begin in NHK. ”Or“ Soon, NHK will begin “Trademarks” ”.

  Further, when recommending a program, the profile generation device 1 may transmit the URL (link information) of the homepage of the recommended program and the previous synopsis together to the page of the user U1.

<Determining other recommended programs>
Further, the user profile creation unit 16 analyzes the related user's page, and extracts from the EPG a broadcast program that the related user is likely to have watched based on the time and content of the speech.
The similarity calculation unit 18 records the viewing frequency for each related user based on the broadcast program extracted by the user profile creation unit 16 (see FIG. 8). The similarity calculation unit 18 records the viewing frequency only when the extracted broadcast program is a series program such as a scheduled program or a series drama.

  The similarity calculation unit 18 determines a recommended broadcast program based on the recorded viewing frequency. In the example shown in FIG. 8, the similarity calculation unit 18 determines the recommended program as a program that is frequently viewed by related users who are at or above a certain level of the program 3.

  Here, <3. Recommended Program Determining Method Described in the Method for Estimating Programs Watched by Related Users and Determining Recommended Programs> (hereinafter referred to as <Method 1>) and <Method for Determining Other Recommended Programs> (hereinafter, <Method 2) >)) Will be described. <Method 1> is a recommendation method based on the program content itself because a viewing profile is created based on the EPG information of the viewed program. Therefore, in <Method 1>, when the viewing profile of the user who wants to recommend has the characteristics of a specific person (for example, an actor), in a single program or a series program, the time in which that person appears in particular. There is a merit that can only be recommended.

  Also, <Method 2> is a recommendation method based on a program title because a recommended program is determined based on whether or not a series program has been viewed. Therefore, there is an advantage that a program that is regularly viewed can be recommended.

  The profile generation apparatus 1 may be a method that recommends all the results of <Method 1> and <Method 2> described above, or may be a method that recommends only the results common to <Method 1> and <Method 2>.

  In this way, the profile generation device 1 creates a user profile based on the comments of the user U1 who wants to recommend predetermined information and a user who is connected on the SNS or a popular representative user, and the user An event profile similar to the profile can be extracted, and an event (for example, a broadcast program or news) corresponding to the event profile can be notified to U1.

  Further, since the profile generation device 1 uses a plurality of SNSs, in addition to the above-described effects, even if there is no direct relationship such as a follow relationship on the SNS1 with the user U1, the user U1 on the SNS2 Can be provided, and information that is estimated to be of interest to the user U1 based on the user's comments can be provided.

  In addition, simply using location information from GPS, which is a common way to acquire action history, where the user has visited a restaurant or supermarket around the location acquired from GPS, and what Because of this, it is difficult to identify even if you have visited there. Since the profile generation apparatus 1 according to the present embodiment is based on data with a clear meaning of a place such as information obtained by the SNS 2, an accurate and detailed profile can be created.

In the above embodiment, the weights must be calculated for all words appearing in the EPG. Then, as the number of programs increases, the total number of words that appear increases, and the profile vector Q _r defined by equation (2) and the content vector D _i defined by equation (4) both increase in vector length. As a result, the cost of the vector calculation of equation (5) also increases.

  In this embodiment, in order to realize a service for many programs and users, words that do not significantly affect the calculated similarity among words that are elements of vectors, that is, words that are not very important, weights, etc. By reducing the number of words that are not high, the vector length is reduced and the cost of vector calculation is reduced.

In the above-described embodiment, vectors are obtained from the text of the EPG, and words that frequently appear in the EPG or words that have a high degree of rarity (words that characterize the program) are evaluated with a higher weight by the tf-idf method.
On the other hand, the vector needs to accurately represent the user's preference for the program. A user who is watching a program may pay attention to a word different from a word highly evaluated by the tf-idf method in the same program.

  That is, the word that the broadcast station wants to emphasize in the EPG does not necessarily match the word that the user is paying attention to.

  Therefore, in this embodiment, not only the characteristic words in the EPG but also the characteristics of the words that the user pays attention to in the program are added to the vector representing the user's preference for the program.

  However, in order to obtain which word the user is paying attention to in the program, it is a burden for the user to input directly, and it is realistic to have the user input each time the program is viewed Not.

  Therefore, by adding the feature of the word written by the user in SNS1 to the vector, the feature of the word that the user is paying attention to in the program can be expressed. Or, if there is no writing of the user himself / herself, the feature of the word written by the general user is added to the vector to express the feature of the word that is likely to be noticed by the general viewer in the program. it can.

  Specifically, the user profile creation unit 16 analyzes the information that the first related user and the second related user are talking about from the data of each account provided by the SNS 1, The weight calculation is performed for a predetermined word that appears in the target writing, and the calculation result is supplied to the event profile creation unit 17.

Specifically, the user profile creation unit 16 targets all writings on information that the first related user and the second related user are talking about, and sets the total number of occurrences of all words appearing in the target writing to m. And the word weight calculation is performed with the appearance frequency of the word k _{j being} written for the topic d _i as m _{ij and} the weight of the word k _j as m _ij / m. Note that the weight of the word k _j may be the appearance frequency m _ij .

For example, the user profile creation unit 16 morphologically analyzes the message “I think I'm watching a trademark story, but it gets more interesting as I repeat the number of times. I was impressed by the current lines. Then, extract nouns and verbs “Trademarks I see. I think.
The user profile creation unit 16 calculates the appearance frequency m _ij as “1” when the total number m of the extracted words is “10” and the word k _j is “trademark transmission”. In the case of this embodiment, the weight of “trademark transmission” is “1/10” or “1”. The above-described word weight calculation is an example, and is not limited thereto.

  The event profile creation unit 17 performs an operation of shortening the content vector constituting the generated event profile based on the word weight calculation, and supplies the shortened content vector to the user profile creation unit 16.

The user profile creation unit 16 creates a user profile based on the shortened content vector.
The similarity calculation unit 18 compares the event profile whose content vector is shortened by the event profile creation unit 17 with the user profile created by the user profile creation unit 16 and calculates the similarity of both profiles.

  By being configured in this way, the profile generation device 1 can be used by all users of the SNS 1 when there are some related users for the user U 1 who wants to recommend predetermined information (for example, a television program). Since the content vector constituting the program profile is shortened only for related users' writing and not the target writing, the user profile that matches the preference of the user U1 is created based on the shortened content vector. be able to. Moreover, since the profile generation apparatus 1 determines a recommended program using this program profile, it is possible to recommend a program more suited to the preference of the user U1.

The event profile creation unit 17 performs a predetermined calculation on the content vector constituting the event profile using the weight calculated for the predetermined word supplied from the user profile creation unit 16, and performs a specific calculation according to the calculation environment. The content vector is shortened by setting the weight to 0 in the content vector for elements having a weight lower than the threshold or elements other than the specific number of elements from the high weight.
For example, the event profile creation unit 17 calculates the word weights as content vectors D _i = <(k ₁ , W _i1 ),..., (K _n , W _in )> constituting the event profile (program profile). D ′ _i = <(k ₁ , m _i1 / m),..., (K _n , m _in / m)> is normalized, and D _i + b · D ′ _i (b is defined as a specific value) Each element, or D _i and D ′ _i after normalization, D _i = <(k ₁ , A _i1 ),..., (K _n , A _in )>, D ′ _i = <(k _{1 , B i1), ···, (} k n, and a _{B in)>, <(k} 1, a i1 · B i1), ···, (k n, a in · B in) each of the elements of> For elements with a weight lower than a specific threshold or elements other than a specific number of elements from a higher weight depending on the computing environment The content vector is shortened by setting the weight to 0 in D _i .

With this configuration, the profile generation device 1 sets the weight in D _i to 0 for an element having a weight lower than a specific threshold or an element other than a specific number of elements from a high weight element. Thus, since the content vector is shortened, it is possible to create a user profile more suitable for the user U1 based on the shortened content vector. Moreover, since the profile generation apparatus 1 determines a recommended program using this program profile, it is possible to recommend a program more suited to the preference of the user U1.

Here, the flow of processing when the content vector is shortened by the profile generation device 1 will be described with reference to the flowchart shown in FIG.
In step ST1, the user profile creation unit 16 analyzes the information (program) that the related user is talking about from the data of each account provided by the SNS 1, and estimates the viewing program. Specifically, the user profile creation unit 16 analyzes the page of the related user's SNS 1 (for example, morphological analysis), and the related user views the comment based on the commented time and the comment content based on the result of the analysis. One or a plurality of broadcast programs that have been used are estimated with reference to the EPG.

  In step ST2, the user profile creation unit 16 extracts the estimated information of one or more broadcast programs from the broadcasted EPG, and performs morphological analysis on the extracted information of the one or more broadcast programs.

  In step ST3, the user profile creation unit 16 calculates the weight of words in the program based on the result of the morphological analysis in the process of step ST2. The user profile creation unit 16 supplies the calculated word weight to the event profile creation unit 17.

  In step ST <b> 11, the event profile creation unit 17 acquires an unbroadcast EPG transmitted from the event information transmission server 400 and performs morphological analysis.

  In step ST12, the event profile creation unit 17 performs word weight calculation based on tf-idf based on the result of the morphological analysis in the process of step ST11.

  In step ST13, the event profile creation unit 17 creates a program profile indicating the characteristics of the program content for each broadcast program based on the word weight calculation by tf-idf in the process of step ST12.

  In step ST14, the event profile creation unit 17 performs an operation for shortening the content vector constituting the program profile generated by the event profile creation unit 17 based on the word weight calculated in the process of step ST3. The event profile creation unit 17 supplies the shortened content vector to the user profile creation unit 16.

In step ST15, the user profile creation unit 16 creates a user profile based on the shortened content vector.
In step ST16, the similarity calculation unit 18 compares the program profile obtained by shortening the content vector with the user profile, and calculates the similarity between both profiles.

  By being configured in this way, the profile generation device 1 does not target writing of all users of the SNS 1 when there are some related users for the user U 1 who wants to recommend a television program. Since the content vector constituting the program profile is shortened only for writing by related users, a user profile more suitable for the user U1 can be created based on the shortened content vector. Moreover, since the profile generation apparatus 1 determines a recommended program using this program profile, it is possible to recommend a program more suited to the preference of the user U1.

DESCRIPTION OF SYMBOLS 1 Profile production | generation apparatus 10 Registration part 11 Acquisition part 11a 1st acquisition part 11b 2nd acquisition part 12 1st related user specific part 13 Extraction part 14 Judgment part 15 2nd related user specific part 16 User profile creation part 17 Event profile creation unit 18 Similarity calculation unit 19 Transmission unit

Claims (7)

The first social networking service that provides a service for posting a comment and communicating with other users has its own first account, and is different from the first social networking service. Owns a second account in the second social networking service that provides a service for communicating with other users by posting a comment together with location information of the actually visited place. A registration unit for registering a user who desires recommendation together with the data of the first account and the data of the second account;
A first acquisition unit that acquires data of the first account registered in the registration unit, and acquires data of an account of another user followed by a user who desires recommendation of the predetermined information; ,
Acquire data of accounts of all users who are provided with the first social networking service, and indicate the number of users indicating how many users are followed and the number of remarks about the predetermined information for each user. A second acquisition unit to acquire
Based on the account data of the other user acquired by the first acquisition unit, a user who satisfies the first condition is extracted from the other users and acquired by the second acquisition unit. Based on the number of users and the number of utterances, a first related user specifying unit that extracts users satisfying a second condition from all the users and specifies the extracted user as a first related user; ,
The second account data registered in the registration unit is obtained, and based on the posted position information, the behavior form of the user who desires the recommendation of the predetermined information is analyzed, and similar to the behavior form An extracting unit for identifying other users to be selected from the second social networking service and extracting data of the account of the other users;
Analyzing the data of the account extracted by the extraction unit, and determining whether or not an account linked to the account exists in the first social networking service;
A second related user specifying unit for specifying the owner of the account as a second related user when the determining unit determines that an account to be linked exists in the first social networking service;
The information related to the first related user and the second related user is analyzed from the data of each account provided by the first social networking service, and the predetermined information is recommended. A profile generation device comprising a user profile creation unit for creating a user profile of a user. An event profile creation unit that obtains event information being sent and creates an event profile indicating the characteristics of the event information;
A similarity calculation unit that compares the user profile created by the user profile creation unit with the event profile created by the event profile creation unit, and calculates the similarity of both profiles;
A transmission unit that transmits information on an event corresponding to the event profile to a user who wants to recommend the predetermined information registered in the registration unit, based on a result calculated by the similarity calculation unit. Item 1. The profile generation device according to Item 1.   The said 1st related user specific | specification part has made a statement more than the predetermined number regarding the said predetermined information as said 1st condition based on the data of the user's account acquired by the said 1st acquisition part. Based on the number of users extracted by the second acquisition unit and the number of utterances, the second condition is followed by a predetermined number or more users, and a predetermined number or more is extracted. The profile generation apparatus according to claim 1, wherein a user who is speaking is extracted. The page provided by the second social networking service is provided with an account registration function for registering an account provided by the first social networking service,
The profile generation apparatus according to claim 1, wherein the determination unit analyzes the account data extracted by the extraction unit and determines whether a page associated with the page exists in the first social networking service. . The user profile creation unit analyzes the information that the first related user and the second related user are talking about from the data of each account provided by the first social networking service, and the information To calculate the weights for a given word that appears in the target writing,
The event profile creation unit performs an operation to shorten the content vector constituting the generated event profile based on the word weight calculation,
The user profile creation unit creates the user profile based on the shortened content vector,
The similarity calculation unit calculates a similarity between both profiles by comparing an event profile whose content vector is shortened by the event profile creation unit with a user profile created by the user profile creation unit. Profile generator.   The event profile creation unit performs a predetermined calculation on the content vector constituting the event profile using a weight calculated for the predetermined word, and an element having a weight lower than a specific threshold according to a calculation environment 6. The profile generation apparatus according to claim 5, wherein the content vector is shortened by setting the weight in the content vector to 0 for elements other than a specific number of elements from elements having a high weight. The first social networking service that provides a service for posting a comment and communicating with other users has its own first account, and is different from the first social networking service. Owns a second account in the second social networking service that provides a service for communicating with other users by posting a comment together with location information of the actually visited place. A registration unit for registering a user who desires recommendation together with the data of the first account and the data of the second account;
A first acquisition unit that acquires data of the first account registered in the registration unit, and acquires data of an account of another user followed by a user who desires recommendation of the predetermined information; ,
Acquire data of accounts of all users who are provided with the first social networking service, and indicate the number of users indicating how many users are followed and the number of remarks about the predetermined information for each user. A second acquisition unit to acquire
Based on the account data of the other user acquired by the first acquisition unit, a user who satisfies the first condition is extracted from the other users and acquired by the second acquisition unit. Based on the number of users and the number of utterances, a first related user specifying unit that extracts users satisfying a second condition from all the users and specifies the extracted user as a first related user; ,
The second account data registered in the registration unit is obtained, and based on the posted position information, the behavior form of the user who desires the recommendation of the predetermined information is analyzed, and similar to the behavior form An extracting unit for identifying other users to be selected from the second social networking service and extracting data of the account of the other users;
Analyzing the data of the account extracted by the extraction unit, and determining whether or not an account linked to the account exists in the first social networking service;
A second related user specifying unit for specifying the owner of the account as a second related user when the determining unit determines that an account to be linked exists in the first social networking service;
The information related to the first related user and the second related user is analyzed from the data of each account provided by the first social networking service, and the predetermined information is recommended. A program that causes a computer including a user profile creation unit that creates a user profile of a user to function as a profile generation device.

Images