JP7539650B2 - USER SELECTION DEVICE, USER SELECTION METHOD, AND PROGRAM - Google Patents

Description

The present invention relates to a technology for providing a specific user with information that is likely to interest the specific user.

Conventionally, a method for providing information that is likely to interest a particular user has been to infer the user's interests and preferences from the user's past browsing history, for example, and provide highly relevant information. It is also self-evident that the higher the quality of the information itself, the more likely it is to interest the user. From this perspective, as an example, Patent Literature 1 discloses a method for improving the quality of evaluations.

JP 2020-035021 A

It is generally known that people are more likely to be interested in information sent by users with whom they have a strong relationship. Here, a user with a strong relationship could be, for example, a real acquaintance or a user who is registered as a favorite.

Conventional technology that estimates a user's interests and preferences from their past browsing history and provides highly relevant information only takes into account the user's interests and preferences. Therefore, it is believed that by having information sent by users with strong relationships as described above, the likelihood of attracting greater interest is increased.

However, taking SNS as an example, it is common for users to register multiple users as favorites, making it difficult to identify which of those favorite users is most likely to provide information that will attract interest.

The present invention has been made in consideration of the above points, and aims to provide a technique for selecting users who are likely to have an influence on a particular user.

According to the disclosed technology, there is provided a user selection device that selects users who are likely to have an influence on a specific user by posting a message that spreads information, the user selection device comprising:
a user influence model construction unit that constructs a user influence model that models the conditions under which a third user interacts with a post by referring to a post database that accumulates posts via SNS, based on data on whether a third user interacts with a post by a first user when the post is spread and posted by a second user; and
There is provided a user selection device comprising: an inquiry reception/result output unit that receives as input a user and topic that one wishes to influence, and outputs users who are likely to influence the user that one wishes to influence by posting a spreading message based on the user influence model.

The disclosed technology makes it possible to select users who are likely to have an influence on a particular user.

1 is a diagram showing a configuration of an apparatus according to an embodiment of the present invention; 10 is a flowchart illustrating an operation of the user selection device. FIG. 13 is a diagram illustrating an example of a posting database. FIG. 11 is a diagram illustrating an example of an intermediate generation database. FIG. 2 illustrates an example of a feedforward loop structure. FIG. 2 illustrates an example of a hardware configuration of the apparatus.

The following describes an embodiment of the present invention (the present embodiment) with reference to the drawings. The embodiment described below is merely an example, and the embodiment to which the present invention is applicable is not limited to the following embodiment.

(Overview of the embodiment)
First, an outline of this embodiment will be described. In this embodiment, information posted by a certain user A is diffused and posted by a user B based on a three-node feedforward loop structure consisting of a node indicating a user A, a node indicating a user B, and a node indicating a user C. In other words, the information posted by the user A is diffused as it is or in accordance with the impressions of the user B, and is viewed by a target user C who is desired to attract the interest of the user. In other words, the feedforward loop structure corresponds to the minimum structure indicating that almost the same information has been transmitted to the target user C from multiple users.

Here, in a typical SNS (social network service), information about users who are registered as favorites (users who are followed) is displayed on the user's home screen, so assuming that User C has registered User A and User B as favorites, it is highly likely that User C will view the information in the manner described above.

Under the circumstances described above, based on data on whether user C has interacted with the post in question, for example, on data on whether he or she has performed an action such as giving a like rating flag, a user influence model is constructed to calculate the degree to which user C will interact under specific conditions, in other words, whether the post has piqued his or her interest. Using this model, users who can exert influence on user C by spreading posts are selected.

The above method makes it possible to instruct a user equivalent to user B, who is likely to have an influence on target user C, to make a spreading post in order to influence user C. The configuration and operation of the device that realizes the above operation will be described in detail below.

(Device configuration example)
1 shows an example of the configuration of a user selection device 100 according to the present embodiment. The user selection device 100 is connected to a network 200 that connects a large number of terminals 10. The network 200 is, for example, the Internet. Note that the "terminals" may also be called "users."

On the network 200, an SNS service is provided, and the SNS service is used by each terminal 10. The SNS service is a well-known general SNS service that allows users to distribute posts, rate posts (such as "likes"), and spread posts. Retweet (registered trademark) is a specific example of a spread post. Note that a spread post may be abbreviated as RT.

As shown in FIG. 1, the user selection device 100 includes a post collection unit 110, a post database 120, a user influence model construction unit 130, a user influence model storage unit 140, and an inquiry reception and result output unit 150. The operation of each unit will be described later.

The user selection device 100 may be implemented as a single computer, or as multiple computers connected via a network. The user selection device 100 may also be a virtual machine on the cloud.

In addition, the post collection unit 110, the post database 120, and the user influence model storage unit 140 may all be provided outside the user selection device 100.

(Example of Operation of User Selection Device 100)
An example of the operation of the user selection device 100 will be described below in accordance with the procedure shown in the flowchart of FIG.

<S101: Collection of posts>
In S<b>101 , the post collection unit 110 collects posts from SNS services on the network 200 , organizes the posts in accordance with the format of the post database 120 , and accumulates the posted contents in the post database 120 .

An example of the format of the post database 120 is shown in FIG. 3. As shown in FIG. 3, the post database 120 has the following items: post ID, posting user, number of followers of the posting user, post content, RT flag (flag indicating whether the post is a spread post), RT original post ID (ID of the original post that was spread), following list of the posting user, and list of users who "liked" the post.

<S102: Generate intermediate database>
In S102, the user influence model construction unit 130 extracts data used for constructing the user influence model from the post database 120, and stores the data in a separate database (called an intermediate generation database). An example of the format of the intermediate generation database is shown in Fig. 4. Specifically, the intermediate generation database is generated as follows.

For example, the user influence model construction unit 130 performs batch processing once a day late at night, and when it is determined from the post database 120 that information posted by a certain user A has been diffused by another user B, and that user C, who has both user A and user B registered as favorite users, has given the diffused post a high rating, the user influence model construction unit 130 fills each of the values shown in FIG. 4 in the post database 200 and stores them as a row.

An example will be described with reference to FIG. 5 in addition to FIG. 4. FIG. 5 shows an example of a feedforward loop structure having three nodes, User A, User B, and User C. Here, the first line of the record in FIG. 4 will be used as an example. The numbers 1 to 5 in FIG. 5 correspond to the numbers 1 to 5 in FIG. 4.

Let us assume that the user with the username aa in the first line of Figure 4 is User C in Figure 5. In this case, the "RT user" in Figure 4 (the user who posted the spread post) is User B in Figure 5, and the "RT source user" (the user who posted the spread post) is User A.

In the example of the first line in Figure 4, information posted by user A is spread by another user B, and user C (user name aa), who has both A and B registered as favorite users, gives the post a high rating.

In this case, since there is an interaction in the form of a high rating, "presence or absence of interaction" becomes "TRUE." Then, the number of followers of user B (RT user) shown in 3, the number of interactions between user C and user B (RT user) shown in 4, the number of followers of user A (RT user) shown in 1, the number of interactions between user C and user A (RT user) shown in 2, and the number of likes for the shared post shown in 5 are extracted from the post database 120 and stored in the intermediate generation database.

The number of interactions between user A and user C shown in 2 is, for example, the number of times user C liked a post by user A in the last n days. The same is true for the number of interactions between user B and user C shown in 4, and is, for example, the number of times user C liked a post by user B in the last n days.

After all data that meets the above conditions is recorded in the intermediate generation database from the posting database 120, proceed to S103.

<S103: Building a user influence model>
In S103, the user influence model construction unit 130 uses the data recorded in the intermediate database (FIG. 4) to set the values of each of the items 1 to 5 in the intermediate database as explanatory variables, and sets the presence or absence of interaction as a response variable with TRUE being 1 and FALSE being 0, and applies logistic regression to obtain weight parameters for each item by maximum likelihood estimation. This user influence model is calculated for each user.

Specifically, the user influence model for user aa is expressed by the following formula:

上記の数式において、各変数の意味は下記のとおりである。

In the above formula, the meaning of each variable is as follows:

y_aa...degree of influence on user aa a...number of followers of RT user b...number of interactions with RT user c...number of followers of RT original user d...number of interactions with RT original user e...number of likes on retweeted post k_n...parameters obtained as a result of maximum likelihood estimation The user influence model as described above is stored in the user influence model storage unit 140 for each user.

The user influence model storage unit 140 has the function of storing the user influence model and outputting the user influence model of a specified user upon request from the inquiry reception and result output unit 150.

<S104: Inquiry Reception>
In S104, the inquiry reception and result output unit 150 receives, as input in the form of an inquiry, posts related to target users whose interest should be attracted and topics that should attract the interest of the users.

The target users whose interest should be attracted are those users whose interest in a certain topic is desired to be attracted to, and who are desired to be influenced by. For example, the target users in question are those users who are desired to "like" a post on that topic.

For example, if a person wants user C to "like" post X on a certain topic, that person sends a query to the user selection device 100, including "user C, post X," and the query reception and result output unit 150 of the user selection device 100 receives this query.

<S105: User selection, selection result output>
In S105, the inquiry reception/result output unit 150 queries the user influence model storage unit 140 to obtain a user influence model for the user related to the inquiry, and performs calculations using the user influence model for the user, thereby outputting users who are likely to influence the user.

Specifically, if the inquiry is "user C, post X," in the user influence model for user C (equation 1), c, d, and e can be obtained from the post database 120 as the number of followers of the poster of post X, the number of interactions between user C and the poster of post X, and the number of likes for post X, respectively.

Then, y_aa (degree of influence on user C) is calculated by calculating the formula using the number of followers (a in the model) and the number of interactions (b in the model) for each user as a "RT user", and the users are output as results in descending order of the calculated value.

For example, if the result of the calculation using the number of followers of user 1 and the number of interactions with user 1 is 0.9, and the result of the calculation using the number of followers of user 2 and the number of interactions with user 2 is 0.5, then since user 1 has a greater influence on user C than user 2, the results will be output in the order of user 1, user 2.

In the above example, an inquiry is input in the form of a post about a topic that is desired to interest the user, but this is just one example. It is also possible to accept input at the granularity of a topic and determine the post to be used in subsequent processing by searching the post content in the post database 120.

Note that since a "post" is a specific example of a "topic," the meaning of "topic" may include a "post."

(Hardware configuration example>
The user selection device 100 can be realized, for example, by causing a computer to execute a program. This computer may be a physical computer or a virtual machine on the cloud.

In other words, the user selection device 100 can be realized by executing a program corresponding to the processing performed by the user selection device 100 using hardware resources such as a CPU and memory built into a computer. The program can be recorded on a computer-readable recording medium (such as a portable memory) and stored or distributed. The program can also be provided via a network such as the Internet or email.

Figure 6 is a diagram showing an example of the hardware configuration of the computer. The computer in Figure 6 has a drive device 1000, an auxiliary storage device 1002, a memory device 1003, a CPU 1004, an interface device 1005, a display device 1006, an input device 1007, an output device 1008, etc., all of which are interconnected by a bus BS.

The program that realizes the processing on the computer is provided by a recording medium 1001, such as a CD-ROM or a memory card. When the recording medium 1001 storing the program is set in the drive device 1000, the program is installed from the recording medium 1001 via the drive device 1000 into the auxiliary storage device 1002. However, the program does not necessarily have to be installed from the recording medium 1001, but may be downloaded from another computer via a network. The auxiliary storage device 1002 stores the installed program as well as necessary files, data, etc.

When an instruction to start a program is received, the memory device 1003 reads out and stores the program from the auxiliary storage device 1002. The CPU 1004 realizes functions related to the user selection device 100 in accordance with the program stored in the memory device 1003. The interface device 1005 is used as an interface for connecting to a network. The display device 1006 displays a GUI (Graphical User Interface) or the like according to a program. The input device 1007 is composed of a keyboard and mouse, buttons, a touch panel, or the like, and is used to input various operational instructions. The output device 1008 outputs the results of calculations.

(Effects of the embodiment)
The technology according to the present embodiment makes it possible to select users who are likely to have an influence on a particular user, and thereby makes it possible to instruct the selected users to make a diffusion post that will have an influence on the particular user.

(Summary of the embodiment)
This specification discloses at least the following user selection device, user selection method, and program.
(Section 1)
A user selection device that selects a user who is likely to have an influence on a specific user by posting a spreading message,
a user influence model construction unit that constructs a user influence model that models the conditions under which a third user interacts with a post by referring to a post database that accumulates posts via SNS, based on data on whether a third user interacts with a post by a first user when the post is spread and posted by a second user; and
A user selection device comprising: an inquiry reception/result output unit that receives as input a user and topic that one wishes to influence, and outputs users who are likely to influence the user that one wishes to influence by posting a spreading message based on the user influence model.
(Section 2)
The user selection device described in paragraph 1, wherein the user influence model construction unit constructs the user influence model based on the number of followers of the first user, the number of interactions between the first user and the third user, the number of followers of the second user, the number of interactions between the second user and the third user, an evaluation value assigned to a post posted by the first user, and data on whether the third user has interacted with the post.
(Section 3)
3. The user selection device according to claim 1, wherein the user influence model construction unit constructs the user influence model using a logistic regression model.
(Section 4)
4. The user selection device according to any one of claims 1 to 3, wherein a feed-forward loop structure is formed by the second user spreading and posting the post of the first user.
(Section 5)
A user selection method executed by a user selection device that selects a user who is likely to have an influence on a specific user by posting a spreading message, comprising:
a user influence model construction step of constructing a user influence model that models the conditions under which a third user interacts with a post by referring to a post database that accumulates posts via SNS, based on data on whether a third user interacts with a post by a first user when the post is spread and posted by a second user;
A user selection method comprising: an inquiry reception/result output step of receiving as input a user and topic that one wishes to influence, and outputting users who are likely to influence the user one wishes to influence by posting a spreading message based on the user influence model.
(Section 6)
A program for causing a computer to function as each part of the user selection device according to any one of claims 1 to 4.

Although the present embodiment has been described above, the present invention is not limited to this specific embodiment, and various modifications and variations are possible within the scope of the gist of the present invention as described in the claims.

REFERENCE SIGNS LIST 10 Terminal 100 User selection device 110 Post collection unit 120 Post database 130 User influence model construction unit 140 User influence model storage unit 150 Inquiry reception and result output unit 200 Network 1000 Drive device 1001 Recording medium 1002 Auxiliary storage device 1003 Memory device 1004 CPU
1005 Interface device 1006 Display device 1007 Input device 1008 Output device

Claims (6)

A user selection device that selects a user who is likely to have an influence on a specific user by posting a spreading message,
a user influence model construction unit that constructs a user influence model that models the conditions under which a third user interacts with a post by referring to a post database that accumulates posts via SNS, based on data on whether a third user interacts with a post by a first user when the post is spread and posted by a second user; and
A user selection device comprising: an inquiry reception/result output unit that receives as input a user and topic that one wishes to influence, and outputs users who are likely to influence the user that one wishes to influence by posting a spreading message based on the user influence model. The user selection device of claim 1, wherein the user influence model construction unit constructs the user influence model based on the number of followers of the first user, the number of interactions between the first user and the third user, the number of followers of the second user, the number of interactions between the second user and the third user, an evaluation value assigned to a post posted by the first user, and data on whether the third user has interacted with the post. The user selection device according to claim 1 or 2, wherein the user influence model construction unit constructs the user influence model using a logistic regression model. The user selection device according to claim 1 , wherein a feed-forward loop structure is formed by the second user spreading and posting the post of the first user. A user selection method executed by a user selection device that selects a user who is likely to have an influence on a specific user by posting a spreading message, comprising:
a user influence model construction step of constructing a user influence model that models the conditions under which a third user interacts with a post by referring to a post database that accumulates posts via SNS, based on data on whether a third user interacts with a post by a first user when the post is spread and posted by a second user;
A user selection method comprising: an inquiry reception/result output step of receiving as input a user and topic that one wishes to influence, and outputting users who are likely to influence the user one wishes to influence by posting a spreading message based on the user influence model. A program for causing a computer to function as each part of a user selection device according to any one of claims 1 to 4.

Images