JP5973927B2 - Feature estimation device and feature estimation method - Google Patents

Description

  The present invention relates to a feature estimation apparatus and a feature estimation method for estimating a feature of a user.

  Conventionally, in user feature estimation using information on the Internet, a method based on data (text information or the like) of a person who is connected to the person to be estimated and a person such as a friend of the person concerned (text information below) is used. Patent Document 1).

JP 2006-309660 A

  However, the above method has the following problems. As information indicating user characteristics, for example, proper nouns such as person names and place names appearing in text information and their appearance frequency data can be used. Such user characteristics are used in a method of recommending (recommending) news and products. As such a user feature extraction source, for example, it is conceivable to use text posted by a user or information indicating a connection between users in a microblog called Twitter (registered trademark).

  However, the text posted to such microblogs is the frequency of appearance of proper nouns that can be used for user feature estimation compared to normal text documents, such as raining, train delays, etc. Is often low text. That is, as described above, the microblog information may have little or no data that can be used for user feature estimation in the data of the person and friends directly connected to the person. In that case, there is a problem that the user feature estimation with sufficient accuracy cannot be performed.

  The present invention has been made in view of the above-described problems, and sufficiently usable data for user feature estimation can be obtained from data relating to a user to be a feature estimation target and a user directly connected to the user. It is an object of the present invention to provide a feature estimation device and a feature estimation method that can perform user feature estimation with sufficient accuracy even when there is no such feature.

  In order to achieve the above object, a feature estimation apparatus according to the present invention is a feature estimation apparatus that performs user feature estimation, and includes link information acquisition means for acquiring link information indicating a link relationship between a plurality of users. Similarity calculation for calculating the similarity between one or more preset first users and one or more preset second users based on the link information acquired by the link information acquisition means Based on the similarity between the second user calculated by the means and the similarity calculating means, and the first user having a link relationship with the estimation target user, the feature estimation of the estimation target user is performed. A specifying unit that specifies a second user to be used, a feature estimating unit that performs feature estimation of a user to be estimated using data related to the second user specified by the specifying unit, and a feature estimating unit And an output means for outputting information indicating characteristics estimation of the user made Te.

  In the text extracting device according to the present invention, based on the degree of similarity between the second user and the first user having a link relationship with the estimation target user, the second text extraction device is used for feature estimation of the estimation target user. Two users are identified, and user feature estimation is performed from data relating to the second user. Here, the second user may not have a direct link relationship with the estimation target user. Therefore, according to the text extracting device of the present invention, if the second user is a user who has data that can be used for user feature estimation, the user who is the target of feature estimation and the user who is directly connected to the user are identified. Even if sufficient data that can be used for user feature estimation cannot be obtained from such data, it is possible to perform user feature estimation with sufficient accuracy.

  The similarity calculation means may calculate the similarity based on the degree of coincidence of the user having a link relationship with the first user and the user having a link relationship with the second user. According to this configuration, the similarity between users can be calculated by a calculation with a small amount of calculation.

  The similarity calculation means uses a smaller denominator of both the number of users linked to the first user and the number of users linked to the second user, both the first user and the second user. The number of users who have a link relationship with the numerator may be used as the degree of coincidence. According to this configuration, even when the number of users having a link relationship with either the first user or the second user is extremely small, the similarity between users can be accurately calculated.

  The feature estimation means may perform feature estimation of the estimation target user using a word included in the text relating to the second user as data relating to the second user. According to this configuration, it is possible to appropriately and reliably perform user feature estimation.

  The feature estimation means may perform feature estimation of the estimation target user using a word included in the text related to the estimation target user. According to this structure, a user's characteristic estimation can be performed more appropriately based on a user's own information.

  The link information acquisition means may acquire link information that also indicates directions of link relationships between a plurality of users. According to this configuration, it is possible to appropriately perform user feature estimation based on the link direction.

  By the way, the present invention can be described as an invention of a feature estimation method as described above, and can also be described as an invention of a feature estimation method as follows. This is substantially the same invention only in different categories, and has the same operations and effects.

  That is, the feature estimation method according to the present invention is a feature estimation method for estimating a feature of a user, and is acquired in a link information acquisition step for acquiring link information indicating a link relationship between a plurality of users, and a link information acquisition step. A similarity calculation step for calculating a similarity between one or more preset first users and one or more preset second users based on the link information, and a similarity calculation step Based on the similarity between the second user calculated in step 1 and the first user having a link relationship with the estimation target user, the second user to be used for feature estimation of the estimation target user is A specifying step for specifying, a feature estimating step for performing feature estimation of the user to be estimated using data relating to the second user specified in the specifying step, and a feature estimating step Tsu comprising an output step of outputting information indicating characteristics estimation of a user performed in up, the.

  According to the present invention, even when data that can be used for user feature estimation cannot be sufficiently obtained from data related to a user whose feature is to be estimated and a user who is directly connected to the user, sufficient accuracy can be obtained. User feature estimation can be performed.

It is a figure which shows the function structure of the feature estimation apparatus which concerns on embodiment of this invention. It is a figure which shows the follow relationship between a user and another user. It is a figure which shows the follow relationship between a celebrity and users other than a celebrity, and the follow relationship between a presumed assistant and users other than a presumed assistant. It is a figure which shows the follow relationship between a celebrity and users other than a celebrity, and the follow relationship between a presumed assistant and users other than a presumed assistant. It is a figure which shows the hardware constitutions of the feature estimation apparatus which concerns on embodiment of this invention. It is a flowchart which shows the feature estimation method which is a process performed with the feature estimation apparatus which concerns on embodiment of this invention.

  Hereinafter, a feature estimation apparatus and a feature estimation method according to the present invention will be described in detail with reference to the drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted.

  FIG. 1 shows a feature estimation apparatus 10 according to the present embodiment. The feature estimation device 10 is a device that performs user feature estimation. The estimated user characteristics are used, for example, when content is distributed. Specifically, the feature estimation device 10 is a server device connected to a network such as the Internet N, for example. In this embodiment, the user's feature estimation is performed using microblog data such as Twitter, for example. In this embodiment, Twitter will be described as an example.

  Twitter is a system in which each user posts a text (tweet). The posted text is referred to by other users in the Twitter system. In the present embodiment, the text is used as user data for user feature estimation. Note that the user data, which is text, may include not only tweets but also user self-introduction sentences registered by the user and referable by other users. Each user (account) in Twitter is given a user ID, which is information for identifying the user.

  Twitter has the function of registering another user and displaying posts from the other user. This registration of one user by another user is called follow. As shown in FIG. 2, for example, a certain user 30 follows a general user 31 that is a friend A, a friend B, a friend n, and a celebrity 32. Thus, in Twitter, a link is provided between users. The follow is a link relationship in the direction from the following side to the following side. In this embodiment, the link relationship is used for user feature estimation. In addition, in the above, a celebrity is a user with many followers, such as an entertainer, a politician, and a talent.

  As described above, an ordinary user's post in Twitter often does not include information for performing user feature estimation. On the other hand, there are users who frequently post texts that can be used for user feature estimation. For example, a spokesperson for a sports newspaper or a talent group. Since these user posts are information about news and entertainers, they contain many proper nouns and can be used to estimate user characteristics. In this embodiment, these users are called estimation assistants.

  When there is a direct link relationship between the estimation target user and the estimation assistant (for example, when the estimation target user follows the estimation assistant), the posting of the estimation assistant is used for user feature estimation. be able to. However, when there is no direct link relationship, it is usually impossible to use the post of the estimation assistant for user feature estimation. In the present embodiment, even if there is no direct link relationship between the estimation target user and the estimation assistant, the posting of the estimation assistant can be used for user feature estimation. As shown in FIG. 3, the celebrity 32 and the estimation assistant 33 are followed by many general users 31 (for example, users other than the celebrity 32 and the estimation assistant 33). In FIG. 3, a general user 31 indicated by a hatched circle follows both a celebrity 32 and an estimated assistant 33.

  In the present embodiment, Twitter is described as an example, but the present invention can be implemented using any system and information as long as the user data and the link relationship described above can be used. For example, user data in SNS (Social Networking Service) and a link relationship between users can be used.

  Subsequently, the function of the feature estimation apparatus 10 according to the present embodiment will be described in detail. The feature estimation apparatus 10 is connected to a server that provides a Twitter service via a network such as the Internet N so that data used for user feature estimation can be obtained (received). As shown in FIG. 1, the feature estimation apparatus 10 includes a celebrity user ID storage unit 11, an estimation assistant user ID storage unit 12, a data acquisition unit 13, a link information storage unit 14, and a user data storage unit 15. The similarity calculation unit 16, the specifying unit 17, the user feature generation unit 18, and the user feature storage unit 19 are configured.

  The celebrity user ID storage unit 11 is means for storing the user ID of an account of a celebrity as the first user in the present invention. As described above, the celebrity is a user who is frequently followed by many general users. Which user is a celebrity is set in advance by the administrator of the feature estimation device 10 or the like, and the user ID of the celebrity is input to the feature estimation device 10 by the administrator or the like. The celebrity user ID storage unit 11 stores one or more celebrity user IDs, for example, in a list. In this embodiment, the celebrity is the first user. However, the first user is not necessarily famous, and any user can be selected as long as the user satisfies the qualification as the first user in the present invention. It is good also as a 1st user.

  The estimated assistant user ID storage unit 12 is a means for storing the user ID of the account of the estimated assistant as the second user in the present invention. As described above, the estimation assistant is a user who makes a post or the like that can be utilized for user feature estimation, and is a user who is frequently followed by many general users. Which user is the estimation assistant is set in advance by the administrator or the like of the feature estimation apparatus 10, and the user ID of the estimation assistant is input to the feature estimation apparatus 10 by the administrator or the like. The estimated assistant user ID storage unit 12 stores, for example, a list (list) of user IDs of one or more estimated assistants.

  The data acquisition unit 13 is means for acquiring data necessary for user feature estimation. For example, the data acquisition unit 13 acquires (receives) the data by requesting the data from the Twitter server via the Internet N. Specifically, the data acquisition unit 13 is a link information acquisition unit that acquires link information indicating a link relationship between a plurality of users. The link information is information indicating the follow relationship between users described above.

  The data acquisition unit 13 inputs the user ID of the estimation target user. This input is performed, for example, by receiving a user ID transmitted from a terminal connected to the feature estimation apparatus 10 by an operation of the terminal by an administrator of the feature estimation apparatus 10 or the like. The data acquisition unit 13 acquires, as link information, a list of user IDs of the users that the user is following from the Twitter server using the user ID as a key. The link information indicates a link in the direction from the estimation target user to another user. Note that the data acquisition unit 13 refers to the celebrity user ID stored in the celebrity user ID storage unit 11 when acquiring link information related to the estimation target user, so that the estimation target user and the celebrity It is good also as acquiring only the link information which concerns on the link relationship (follow relationship from the said user to a celebrity).

  In addition, the data acquisition unit 13 reads the celebrity user ID from the celebrity user ID storage unit 11, and uses the celebrity user ID as a key to follow the celebrity from the Twitter server (the user is the celebrity. A list (list) of user IDs of “following” is acquired as link information. This link information indicates a link in the direction from the user other than the celebrity to the celebrity.

  In addition, the data acquisition unit 13 reads out the user ID of the estimated assistant from the estimated assistant user ID storage unit 12, and follows the estimated assistant from the Twitter server using the user ID of each estimated assistant as a key. A list (list) of user IDs of existing users (the user is following the estimated assistant) is acquired as link information. This link information indicates a link in the direction from a user other than the estimated assistant to the estimated assistant. The data acquisition unit 13 inputs link information indicating the follow relationship between users acquired as described above to the link information storage unit 14.

  The data acquisition unit 13 acquires data related to an estimation assistant used for user feature estimation. Specifically, the data acquisition unit 13 uses the estimated assistant's user ID as a key, the text that is the posting (tweet) of the estimated assistant from the Twitter server, and the text that is a self-introduction sentence of the estimated assistant. To get. In addition, the data acquisition unit 13 uses the user ID of the estimation target user as a key, the text that is the posting (tweet) of the estimation target user from the Twitter server, and the text that is the self-introduction sentence of the estimation assistant. get. Moreover, it is good also as acquiring a celebrity's contribution and a self-introduction sentence similarly to the above.

  Note that the data acquisition unit 13 may acquire all the posts of each user, or may acquire only posts for a certain period such as the past one month. The data acquisition unit 13 inputs text related to each user acquired as described above to the user data storage unit 15.

  In the above example, each data is acquired via a network such as the Internet N (from an external Internet environment). For example, the feature estimation device 10 itself provides a microblog service. If it can be obtained from within the server, it may be obtained from there. With this configuration, the time taken to acquire each data is shortened, and the operation performance of the apparatus is improved.

  The link information storage unit 14 stores the link information input from the data acquisition unit 13. For example, the user ID of the follow source user and the user ID of the follow destination user are stored in association with each other.

  The user data storage unit 15 stores text relating to each user input from the data acquisition unit 13. For example, the user ID of the user and the text are stored in association with each other.

  The similarity calculation unit 16 is a similarity calculation unit that calculates the similarity between a celebrity and an estimation assistant based on link information stored in the link information storage unit 14. The similarity calculation unit 16 calculates the similarity for each combination of each celebrity and each estimation assistant. The similarity calculation unit 16 indicates the user who follows the celebrity (the user who has a link relationship with the celebrity) and the user who follows the estimation assistant (the user who has a link relationship with the estimation assistant) indicated by the link information. ) To calculate the similarity.

  Specifically, the similarity calculation unit 16 counts the number of users following the celebrity (number of user IDs) and the number of users following the estimated assistant (number of user IDs). In addition, the number of users (user IDs) who follow both the celebrity and the estimated assistant (users linked to both the celebrity and the estimated assistant) is counted. The similarity calculation unit 16 follows the denominator, both the celebrity and the estimated assistant, among the counted number of users following the celebrity and the number of users following the estimated assistant. A value (Simpson coefficient) with the number of users who are present as a numerator is calculated as the similarity.

  The degree of similarity indicates whether the user following the celebrity and the estimated assistant is similar. It can be said that this shows the similarities between the celebrity and the estimated assistant. The similarity calculation unit 16 may calculate the similarity only between each celebrity followed by the estimation target user and all estimation assistants. In addition, the above calculation of the similarity is an example, and any method can be used as long as the similarity between the celebrity and the estimated assistant is calculated based on the link relationship between the celebrity and the estimated assistant. Can be used.

  The similarity calculation unit 16 compares the calculated similarity with a threshold and determines whether or not the threshold is exceeded. The threshold value is set in advance by an administrator of the feature estimation apparatus 10, input to the feature estimation apparatus 10, and stored in the similarity calculation unit 16. The threshold value is, for example, a value at which it can be determined that a celebrity and an estimated assistant are similar. When the similarity calculation unit 16 determines that the similarity exceeds the threshold, the user ID of the celebrity related to the similarity, the user ID of the estimation assistant related to the similarity, and the similarity are associated with the specifying unit 17. Output. This information is, for example, information as shown in the table of FIG.

  The specifying unit 17 is a specifying unit that specifies an estimation assistant to be used for estimating the feature of the estimation target user based on the similarity calculated by the similarity calculating unit 16. First, the specifying unit 17 refers to the link information stored in the link information storing unit 14 and the celebrity user ID stored in the celebrity user ID storing unit 11, and the user to be estimated follows. A celebrity (a celebrity who has a link relationship with the user to be estimated) is specified.

  Subsequently, the specifying unit 17 specifies the estimation assistant associated with the celebrity as the estimation assistant used for estimating the feature of the estimation target user based on the information input from the similarity calculation unit 16. This estimation assistant is an estimation assistant having a degree of similarity exceeding a threshold with a celebrity that the user to be estimated follows.

  The presumed assistant identified in this way is considered to have similar characteristics to the celebrity that the user is following. Therefore, even if there is no direct link relationship between the user and the estimation assistant, it is possible to estimate the user's characteristics using data on the estimation assistant. For example, if a celebrity is a talent belonging to a talent group and the estimated assistant similar to the celebrity is a spokesperson for the talent group, even if there is no follow-up relationship between the user and the spokesperson The assistant's post can be used for user feature estimation.

  As described above, for example, in Twitter, each user rarely makes a post that can be used for feature estimation. The same applies to celebrities, for example. Therefore, even if the user is interested in a celebrity and is following, it is difficult to estimate the user's feature from the celebrity's post. However, special users such as the above-mentioned estimation assistants often post that can perform feature estimation. For example, a spokesperson for the talent group posts information related to the talent group (for example, a TV program to appear or the name of a released song).

  The specifying unit 17 outputs the specified information indicating the estimation assistant used for feature estimation (for example, the estimation assistant) to the user feature generation unit 18. Note that the specifying unit 17 may specify a plurality of estimation assistants that satisfy the above-described determination criteria for one estimation target user. Or it is good also as specifying an estimation assistant (for example, one estimation assistant) to a specific number in order of similarity.

  The user feature generation unit 18 is a feature estimation unit that performs feature estimation of a user to be estimated using data related to an estimation assistant indicated by information input from the specifying unit 17. The user feature generation unit 18 reads the text relating to the estimation assistant stored in the user data storage unit 15 as data relating to the estimation assistant. The user feature generation unit 18 performs feature estimation of the user to be estimated using words included in the text related to the estimation assistant. The user feature generation unit 18 generates numerical information associated with a feature word that is a preset word as information indicating the feature of the user. This information is a vector of dimensions of the number of feature words. The feature amount indicates that the larger the numerical value associated with the feature word is, the more the user has the feature indicated by the feature word. The feature words are registered in a dictionary prepared by an administrator of the feature estimation apparatus 10 and stored in the user feature generation unit 18.

  The user feature generation unit 18 counts the number of appearances of feature words in the text related to the estimation assistant for each feature word. The user feature generation unit 18 uses a value based on the number of appearances for each feature word as a feature amount. For example, the number of appearances for each feature word itself is information indicating the user's feature.

  Further, the user feature generation unit 18 may perform feature estimation of the estimation target user using not only the text related to the estimation assistant but also words included in the text related to the estimation target user. In that case, the user feature generation unit 18 reads the text related to the estimation target user stored in the user data storage unit 15. The user feature generation unit 18 counts the number of appearances of the feature word in the text related to the estimation target user for each feature word. The user feature generation unit 18 adds the number of appearances in the text related to the estimation assistant and the number of appearances in the text related to the estimation target user for each feature word, and uses the added value as information indicating the feature of the user. At the time of this addition, the number of appearances in the text related to the estimation assistant and the number of appearances in the text related to the estimation target user may be weighted (at a fixed ratio) and added together. Moreover, you may use the text which concerns on the celebrity who has a link relation with the estimation object user for the production | generation of the information which shows the said user's characteristic.

  Note that the information indicating the user's characteristics is not necessarily limited to the information described above, and any information may be used as long as the characteristics of the estimation target user are estimated using data on the estimation assistant. . The user feature generation unit 18 outputs information indicating the user features generated as described above to the user feature storage unit 19. In other words, the user feature generation unit 18 is also an output unit that outputs information indicating the user's feature estimation.

  The user feature storage unit 19 is a unit that stores information indicating the user feature input from the user feature generation unit 18. For example, information indicating user characteristics is stored in association with the user ID. Information indicating the characteristics of the user generated by the feature estimation device 10 is used, for example, for determining content to be distributed or recommended to the user. For example, a content including a feature word representing a user (for example, a feature word having a value equal to or greater than a certain value) in the text (description) of the content is associated with the user (determined as content to be distributed or recommended to the user) . By determining the content in this way, it is possible to distribute or recommend content highly relevant to the user. The functional configuration of the feature estimation apparatus 10 has been described above.

  FIG. 5 shows a hardware configuration of the feature estimation apparatus 10 according to the present embodiment. As shown in FIG. 5, the feature estimation apparatus 10 includes a central processing unit (CPU) 101, a random access memory (RAM) 102 and a read only memory (ROM) 103 that are main storage devices, a communication module 104 for communication, and The computer is configured to include a computer including hardware such as an auxiliary storage device 105 such as a hard disk. The functions of the feature estimation device 10 described above are exhibited by the operation of these components by a program or the like. The above is the configuration of the feature estimation apparatus 10 according to the present embodiment.

  Subsequently, a feature estimation method, which is a process executed by the feature estimation apparatus 10 according to the present embodiment, will be described using the flowchart of FIG. In this process, the user ID of the estimation target user is input by the data acquisition unit 13 (S01). The user ID is input from, for example, a terminal operated by an administrator of the feature estimation apparatus 10 or the like. Subsequently, link information and user data are acquired by the data acquisition unit 13 (S02, link information acquisition step). As link information, at least the user ID of the user that the estimation target user is following, the user ID of the user who is following the celebrity whose user ID is stored in the celebrity user ID storage unit 11, and the estimated assistant user The user ID of the user who is following the estimated assistant whose user ID is stored in the ID storage unit 12 is acquired. In addition, as user data, at least text relating to the estimated assistant is acquired. The acquired link information is stored in the link information storage unit 14. The acquired user data is stored in the user data storage unit 15.

  Subsequently, based on the link information stored in the link information storage unit 14, the similarity calculation unit 16 calculates the similarity between each celebrity and each estimated assistant from the degree of coincidence of each link relationship. (S03, similarity calculation step). Subsequently, the similarity calculation unit 16 determines whether or not the calculated similarity exceeds a threshold value. Information indicating a combination of a celebrity and an estimation assistant related to the similarity exceeding the threshold is output from the similarity calculation unit 16 to the specifying unit 17.

  Subsequently, based on the similarity calculated by the similarity calculation unit 16, the specifying unit 17 specifies an estimation assistant to be used for estimating the characteristics of the estimation target user (S04, specifying step). Specifically, the specifying unit 17 refers to the link information stored in the link information storage unit 14 and the celebrity user ID storage unit 11 to store the celebrity user ID, and the estimation target user follows. Celebrities are identified. Subsequently, the information input from the similarity calculation unit 16 is referred to by the specifying unit 17, and the estimation assistant associated with the celebrity is specified as the estimation assistant used for estimating the characteristics of the estimation target user. The Information indicating the identified estimated assistant is output from the specifying unit 17 to the user feature generating unit 18.

  Subsequently, the user feature generation unit 18 obtains the text related to the estimation assistant indicated by the information input from the specifying unit 17 from the user data storage unit 15, and performs the feature estimation of the estimation target user from the text. (S05, feature estimation step). Information indicating the user's feature estimation is output from the user feature generation unit 18 to the user feature storage unit 19 and stored in the user feature storage unit 19 (S06, output step). The above is the feature estimation method that is a process executed by the feature estimation apparatus 10 according to the present embodiment.

  As described above, in the present embodiment, the estimation assistant used for estimating the feature of the estimation target user is identified based on the similarity between the estimation assistant and the celebrity having a link relationship with the estimation target user. Then, the feature of the user is estimated from the data related to the estimation assistant. As described above, an estimation assistant who has data that can be used for user feature estimation, that is, a text that can be used for user feature estimation has a direct link relationship with the estimation target user. It does not have to be. That is, according to the present embodiment, even if sufficient data that can be used for user feature estimation cannot be obtained from data related to a user whose feature is to be estimated and a user who is directly connected to the user. User feature estimation with high accuracy can be performed.

  In addition, as in the present embodiment, the degree of similarity between a celebrity and an estimated assistant can be calculated based on the degree of coincidence of users who are linked to those users. According to this configuration, the above-described similarity can be calculated by a calculation with a small calculation amount. More specifically, a Simpson coefficient can be used. By using the Simpson coefficient, even when there are extremely few users who have a link relationship with either a celebrity or an estimated assistant, the similarity between users can be accurately compared with the Jackard coefficient. Can be calculated. However, the similarity is not necessarily calculated by the above method, and may be calculated by any method as long as it is calculated based on the link relationship.

  The user's feature estimation can be performed using a feature word included in the text related to the estimation assistant. Thereby, a user's feature estimation can be performed appropriately and reliably. In addition, user characteristics can be expressed in a concise manner. Further, as described above, text related to the estimation target user may be used. As described above, for example, there are many cases where information that can be used for feature estimation is not sufficiently included in the post of the estimation target user himself such as Twitter. However, if the user's own post to be estimated includes information that can be used for feature estimation, the above-described configuration enables the user's feature estimation to be performed more appropriately based on the user's own information. .

  Further, the direction of the link relationship between users may be considered so that the link relationship shown in the present embodiment is a follow relationship. For example, in the above-described embodiment, the follow link relation to the celebrity or the estimation assistant is used, but the follow link relation from the celebrity or the estimation assistant is not necessarily used. According to this configuration, it is possible to appropriately perform user feature estimation based on the link direction. However, it is not always necessary to consider the direction of the link relationship, and it is possible to implement the present invention based on the link relationship that is not oriented.

  DESCRIPTION OF SYMBOLS 10 ... Feature estimation apparatus, 11 ... Celebrity user ID storage part, 12 ... Estimation assistant user ID storage part, 13 ... Data acquisition part, 14 ... Link information storage part, 15 ... User data storage part, 16 ... Similarity calculation part , 17 ... identification unit, 18 ... user feature generation unit, 19 ... user feature storage unit, 101 ... CPU, 102 ... RAM, 103 ... ROM, 104 ... communication module, 105 ... auxiliary storage device.

Claims (7)

A feature estimation device that performs feature estimation of a user,
Link information acquisition means for acquiring link information indicating a link relationship between a plurality of users;
Similarity calculation for calculating the similarity between one or more preset first users and one or more preset second users based on the link information acquired by the link information acquisition means Means,
Based on the similarity between the second user calculated by the similarity calculation means and the first user who has a link relationship with the estimation target user, it is used for feature estimation of the estimation target user. Identifying means for identifying a second user;
Feature estimation means for performing feature estimation of the estimation target user using data related to the second user specified by the specification means;
Output means for outputting information indicating the feature estimation of the user performed by the feature estimation means;
A feature estimation apparatus comprising:   2. The similarity calculation unit according to claim 1, wherein the similarity calculation unit calculates the similarity based on a degree of coincidence between a user having a link relationship with the first user and a user having a link relationship with the second user. Feature estimation device.   The similarity calculating means calculates a smaller denominator of the number of users having a link relationship with the first user and the number of users having a link relationship with the second user, the first user and the first user. The feature estimation apparatus according to claim 2, wherein the number of users having a link relationship with both of the two users is a numerator.   The said feature estimation means performs the feature estimation of the said estimation object user using the word contained in the text which concerns on the said 2nd user as data regarding the said 2nd user, The any one of Claims 1-3 The feature estimation device according to claim 1.   The feature estimation apparatus according to claim 4, wherein the feature estimation unit performs feature estimation of the estimation target user using a word included in text related to the estimation target user.   The said link information acquisition means is a feature estimation apparatus as described in any one of Claims 1-5 which acquires the link information which also shows the direction of the link relationship between these users. A feature estimation method for performing user feature estimation,
A link information acquisition step of acquiring link information indicating a link relationship between a plurality of users;
Similarity calculation for calculating the similarity between one or more preset first users and one or more preset second users based on the link information acquired in the link information acquisition step Steps,
Based on the similarity between the second user calculated in the similarity calculation step and the first user who has a link relationship with the estimation target user, it is used for feature estimation of the estimation target user. A specific step of identifying a second user;
A feature estimation step of performing feature estimation of the estimation target user using data related to the second user identified in the identification step;
An output step of outputting information indicating the user's feature estimation performed in the feature estimation step;
A feature estimation method including:

Images