KR100996997B1 - User ordered blogger analysis system and method - Google Patents

Abstract

PURPOSE: A system and a method for analyzing a user customized blogger are provided to extract feature information of a blogger community, thereby providing customized blogger analysis to a user by the feature information. CONSTITUTION: A server(100) classifies blogger information. The server extracts a blogger community. The server analyzes a user feature by comparing user information with feature information of the blogger community. The server analyzes features of the user and the community according to flow of an event based on time. A browser terminal includes a service UI(User Interface)(10). An analyzed result of the server is provided to the user through the service UI.

Description

User ordered blogger analysis system and method}

The present invention relates to a system and method for user-customized blogger analysis. In particular, the present invention analyzes the temporal / public analysis using characteristic information and user information of the blogger community found through blogger information stored in the ontology in an internet service such as a blog and a mini homepage. A system and method for performing custom bulllogger analysis is provided.

Recently, due to the popularization of blogs and mini homepages, users can easily share their contents (UCC) with others. This phenomenon improves the accessibility of users' information by making it easier for Internet users to obtain various information not obtained from a specific press or portal. To get this information, existing systems use keyword-based searches. Existing search-based systems like this are the only way for users to access information unless the user checks every piece of content.

Content generated and shared by a user may not only provide useful information by itself, but also provide information on a specific user who is a content creator by gathering such contents. For example, content about “iphone” posted by a particular blogger provides useful information about “iphone”. And content like this can provide information such as what areas the blogger is interested in and what products she likes.

However, in existing search-based systems, there is no way to find such information. In view of the above example, "iphone" can easily obtain information through a keyword, but information such as "area of interest by bloggers who hold" iphone "is not available. Therefore, using the existing search-based system, the user can not access or utilize all the information of the various user-generated content existing on the Internet.

In order to solve the above problems, the present invention extracts the feature information of the blogger community and provides the user with customized analysis using the extracted feature information so that the user can access more information on the Internet, To provide a customized blogger analysis system and method to facilitate the use of information.

In addition, the present invention is to find a community that meets the characteristics of the information required by the user and to show the content generated by these communities to extract the feature information of the extracted community and user information and maximize the usage of the extracted community and feature information It provides a customized blogger analysis system and method to make it easy for users to find the differences and commonalities through public comparison.

The customized blogger analysis system for achieving the objects of the present invention, extracts blogger community and feature information from the blogger information of the retrieved bloggers, and performs custom blogger analysis using the extracted blogger community and feature information and the user information. Performing server; And a browser having a service UI that is connected to the server through the Internet and receives the analyzed result from the server and displays the analyzed result to the user.

In addition, a custom blogger analysis method in a custom blogger analysis system for achieving the objects of the present invention, extracting blogger information of the retrieved bloggers; Searching for a blogger community and extracting feature information of the found blogger community from the extracted blogger information; Performing a customized blogger analysis using the extracted blogger community feature information and user information; And displaying a result of the customized blogger analysis to the user.

The present invention provides a personalized blogger analysis through a synchronic / public analysis method using feature information extracted by extracting feature information of the blogger community, so that the user can access more information on the Internet to obtain high quality information. It is easy to use, and the user can select more bloggers with the desired characteristics among the bloggers, thereby obtaining more useful information from the generated content.

In addition, since the present invention can present personalized community information that is generally difficult to obtain through Internet search, the user can easily find an item such as an information blog or a book / destination / movie that is more useful to them.

1 is a view showing the structure of a custom blogger analysis system according to an embodiment of the present invention;
2 illustrates a method for analyzing a custom blogger in a custom blogger analysis system according to an embodiment of the present invention;
3 illustrates a detailed process of clustering for custom blogger analysis according to an embodiment of the present invention.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. In describing the present invention, if it is determined that detailed descriptions of related known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted.

If the blog users can be analyzed and searched for their characteristics, the user can select bloggers with the desired characteristics among the bloggers and obtain more useful information from the contents generated by them. For example, by analyzing “Bloggers living in Seoul in their 20s”, the analyzed bloggers have four communities, among them are those who enjoy “movie” and like the “adventure, action” genre. Able to know. With this information, if a user wants to get information about a specific product, it is not provided by specific bloggers in all bloggers, but provided by the specific users they are interested in or selected (“It is provided by adventure action enthusiasts living in Seoul in 20s. Information ”(“ iphone ”) can be easily obtained. This analysis of blogger information not only increases the amount of information accessible to users, but also improves the quality of the search for information access.

Two technologies are required for the analysis of blogger information. The first is to find a community from bloggers' information (individual information and generated content) and to extract the features of the found community and to customize the found communities. It is a technique to analyze.

Hereinafter, in the embodiment of the present invention using the above two techniques to extract the information of the bloggers from the content generated by them, the blogger community and its features, and the user to show through the method of perceptual / public analysis from the user's point of view A custom blogger analysis system and method will be described in detail.

First, a custom blogger analysis system according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

 1 is a view showing the structure of a user-customized blogger analysis system according to an embodiment of the present invention.

The customized blogger analysis system may be largely classified into a service user interface (hereinafter, referred to as a service UI) 10 and a server 100 of a browser. In addition, the custom blogger analysis system basically uses a web service of the Internet, connects the service UI 10 and the server 20 to communicate through the Internet, and Ajax (Asynchronous JavaScript and XML) for real time updating. Use a protocol. Here, the customized blogger analysis system may be used in any environment as well as the web, if there is an environment in which the blogger is provided.

The service UI 10 includes an analysis UI 11 and a search UI 12, and refers to a user interface portion of an Internet content service (user information, blog, etc.) to which a customized blogger analysis is applied. In addition, the service UI 10 transmits user information to the server part by login in order to perform user-specific analysis. The main service line of the Internet content service is not shown in FIG. 1 for convenience of description, and for example, a part of a result after a search and analysis and an event in each UI are omitted.

The search UI 11 is a portion of a user interface through which a user searches for a blogger through a browser, and serves to show a search result to a user. In the embodiment of the present invention, a general search operation such as a query input or an input event used in the existing system will be omitted.

The analysis UI 12 is a part of the user interface through which a user views a comparison result between himself and the person who performed the search through a browser. The analysis UI 12 allows the user to examine the analysis result obtained through the synchronic / public analysis on the server side. It serves to show.

The server 100 may include a blogger information extraction unit 110, a blogger community and feature extraction unit 120, an analysis unit 130, a user information unit 140, a blogger search unit 150, and a blog database (DB) 160. ), The tag search unit 170 and the tag database (DB) 180 can be configured.

The blogger information extracting unit 110 extracts blogger information such as member information of the bloggers, content of the blogger, and postings from the blogger search unit 150 and the tag search unit 170, and extracts the extracted blogger information from the blog community and features. Transfer it to the extraction unit. The blogger information (build information) can be divided into static information and dynamic information. Static information is the member information (age, name, gender, etc.) to be created when creating a blog, and has been used in existing search / analysis methods. The dynamic information can be viewed as to what product the blogger has created and what area of interest they have. Such dynamic information can be easily obtained when the blogger tags the content. According to an embodiment of the present invention, the semantic tag that imposes various information is imposed on the content generated by the blogger, and the blogger extracting unit 110 extracts the blogger information by extracting the dynamic and static information.

The blogger community and feature extractor 120 extracts the community and feature of the bloggers searched from the blogger information extracted by the blogger information extractor 110 and delivers the extracted blogger community and feature information to the analyzer 130. In addition, the blogger community and feature extractor 120 classifies the set of blogger information into a preset type according to the characteristics of the extracted blogger information, calculates the distance value for each set of blogger information, and integrates the calculated distance value. Blogger community and feature information is extracted from the final calculated distance value. The blogger community is divided into a public community and an implicit community. For example, bloggers belonging to a specific club may belong to a public community. An implicit community here refers to a group of bloggers that are not publicly designated by the blogger. Since bloggers obtained through search are very unlikely to belong to some public community, finding implicit communities and extracting their characteristics is very important in terms of actual service. Therefore, in the embodiment of the present invention, the implicit community is searched through the static and dynamic information of the bloggers, and defined as information that has importance in users belonging to the community as their characteristics.

The analysis unit 130 is a preset analysis method based on the user information transmitted from the user information unit 140 and the blog community and feature information transmitted from the blog community and feature extraction unit 120 (temporal / public analysis method). ) Performs a customized blogger analysis, and transmits the analyzed result to the analysis UI 11 of the user through the Internet and the AJAX protocol.

The user information unit 140 manages user information required by the analysis unit 130.

The blogger search unit 150 searches for bloggers through the blog DB 160, that is, retrieves blogger information and related posts from a specific blog, and delivers the searched bloggers to the blog information extraction unit 110.

The tag search unit 170 searches for a tag for extracting blogger information from the tag database 180.

The tag database 180 is a record that holds all records of which tags form a set of tags in any combination.

Next, a method for analyzing the custom blogger in a custom / blogger analysis system having such a structure will be described in detail with reference to the accompanying drawings.

In the existing system, the extracted community and feature information cannot be accessed. Therefore, in order to utilize the information, the user must find a community that matches the characteristics of the information required by the user and show the contents generated by the community. Therefore, in the embodiment of the present invention, the user can easily find the difference and commonality through the user's information and the chronological / public comparison of the extracted community information to maximize the utilization of the extracted community and feature information. We will explain how to analyze custom bloggers.

The customized blogger analysis method can be largely divided into the implicit community and feature extraction process of the bloggers and the customized poetic / public analysis process. Here, the implicit community refers to a group of people who have similar information based on bloggers' member information and event information.

The implicit community and feature extraction process of bloggers categorizes the blogger's information sets into four types and measures the distance values using different calculation methods for each set. The four distance values thus calculated are added together to obtain a single distance value. Then, based on the obtained distance values, bloggers with small distance values are assigned to the same cluster. Detailed description thereof will be described below with reference to FIG. 3.

The user-defined poetic / public analysis process analyzes the characteristics of the user to be analyzed using the community composed of multiple bloggers and the features extracted from the community from the perspective of the poetic / public.

The following describes these processes in more detail.

2 is a diagram illustrating a method for analyzing a custom blogger in a custom blogger analysis system according to an embodiment of the present invention.

Referring to FIG. 2, in step 210, the server 100 of the user-customized blogger analysis system searches for the bloggers and the tags of the bloggers in the blog DB 160 through the blogger search unit 150, and the bloggers from the searched bloggers. Extract the information. The information of the blogger basically includes member information such as age, name, gender, etc. input when the blog is first created, and event information about the user's experience extracted from content registered by the user using the blog. Member information is static information that is fixed regardless of the flow or user's experience in time, and event information is dynamic information in which more information is added or modified over time. Such dynamic information includes, for example, information about the genre of the book, the author, the publisher, the genre of the film, the director, the actor, the type of the IT device purchased, and the manufacturer, which can be extracted from the content registered by the user. In an embodiment of the present invention, such static and dynamic information is constructed as an ontology, and the constructed ontology is represented and used in a vector form to facilitate information extraction and calculation. When expressed as a vector, the information is divided into three levels of depth. Here, Depth 1 is member information which is static information, Depth 2 is the type of each event and the date of the event, and Depth 3 is the object of each event.

As described above, one user generates and expresses a vector having each extracted information as a feature, and generates and expresses an object module for each event separately.

In operation 220, the server 100 finds a blogger community from blogger information extracted through the blogger community and the feature extractor 110, and extracts feature information of the found blogger communities. A detailed process of extracting such blogger community and community feature information will be described below with reference to FIG. 3.

Next, in step 230, the server 100 analyzes the community composed of a plurality of bloggers and the features extracted from the community by the blogger community and the feature extraction unit 110 through the analysis unit 130. Analyze the characteristics of the users from the point of view. Here, the meanings of the syntactic analysis and the syntactic analysis will be described in detail below.

In operation 240, the server 100 transmits the analysis result to the analysis UI 11 of the service UI 10. Accordingly, the analysis UI 11 of the service UI 10 displays the analysis result received so that the user can grasp the analysis result.

The public analysis and the historical analysis described in step 230 will be described in detail.

First, public analysis analyzes blogger community and feature information and user information as one data regardless of time, and compares blogger community and feature information and user information to analyze user characteristics, that is, the distance between two data types. It is a way to analyze the characteristics between them based on the difference. For example, there are three types of analysis methods: analysis based on the location between the user and each community, user characteristic analysis according to the user's event frequency, and social network.

The analysis method based on the location between the user and each community is based on the fact that the user's characteristics will be similar to the characteristics of the community that are close to the user. This is how you analyze. This not only makes it easy to analyze the characteristics of the user, but also makes it easy to determine whether the user belongs to the types of people to be compared. This can be easily seen by arranging each community group and user's position in a two-dimensional plane in proportion to the distance. In this case, the distance between the user and each community is calculated based on the events of each community and my event information, and the types of data are as Depth 1, 2, and 3 above.

The user's feature analysis method according to the user's event frequency is a feature analysis according to the frequency of the event, and the method is analyzed by comparing the frequency according to the same event type. The user's feature analysis method according to the frequency of the user's event can measure the interest of the object by the overall frequency of the event or object, and through this, based on the difference between the interest of the community in a specific area and the interest of me Analyze features. The difference between users and communities can be easily identified by listing the frequency of events in each community and the frequency of users in a bar graph. Based on this, it is possible to determine whether they have similar interests in general or have completely different interests.

The social network method builds a social network based on various events or objects in relation to members and users belonging to a community. The relationship can be a variety of domains, such as travel, books, IT devices, or a specific product. Based on these detailed social networks, we can grasp the relationship with each member of the community. Each member and user on a two-dimensional plane means a node, and these nodes can be represented by connecting them by a relationship line.

Next, the temporal analysis is a method of analyzing the characteristics of the user and the community based on the flow of events over time by viewing the information of the community and the user as continuous data that changes over time. For example, there is an analysis method from the perspective of trend.

Trends show how interest changes over time for a particular object or event between the user and the community. This allows you to compare and analyze what areas each community member is interested in recently and how users are following the flow of their interests. Trend analyzes the similarity of each other based on the ratio of community and user's time-based events, and finds out how much the interest flows between users and the community are similar and uses them in the analysis. The similarity used here is the same as the method for calculating the distance value for the following four types.

Through such a trend, the analysis unit 130 finds the most similar pairs of data of each section of the user and each section of the community, finds matching months, and analyzes whether the user follows or precedes. This can be expressed by drawing and drawing two line graphs on one plane at the same time, which makes it easy to see how similar two event flows are.

A detailed process of extracting the blogger community and the feature information of the community in step 220 in the process of analyzing the user-customized blogger as described above will be described with reference to FIG. 3.

In step 310, the blogger community and feature extractor 120 classifies the set of extracted blogger information into types that are preset according to characteristics. Here, the information of the blogger is not only different in the form in which the value is expressed but also in the characteristics of the user to reflect, so in the embodiment of the present invention, the numeric type, the category type, the string type, the string type, Four types, such as a Gaussian type, may be classified. In addition to these types, other types may be applied according to characteristics of blogger information.

Numeric type represents data with numeric value (ex> age, key), and Category type represents data with one of several selectable items (ex> gender, region ). String type represents data (ex> E-mail, homepage) that has a value in string, and Gaussian type represents data having distribution with multiple sets of one dimension. Book author, device manufacturer).

The information of the bloggers classified into the four types is different from each other in the form of the values represented and the characteristics thereof, so that the distance value cannot be calculated based on one criterion. Therefore, in step 320, the blogger communicator and the feature extractor 120 calculate a distance value for each classification and integrate the calculated values as follows. The distance value calculated here means the distance between individual bloggers.

For the character type, the difference between two values is obtained using Equation 1 below. Where is the value of the blogger attribute, is the i th cluster, and k is the number of bloggers in the cluster.

[Equation 1]

The category types are compared by using the following Equation 2 for equality. Where denotes an indicator function.

[Equation 2]

For the string type, the difference between two string values is compared using Equation 3 below.

[Equation 3]

The Gaussian type calculates the degree to which the two values are equal by considering the set size of the corresponding dimension using Equation 4 below. Where is the number of dimensions of the Gaussian Type set currently being compared in the cluster.

&Quot; (4) "

If the distance value calculated for each classified set is used as it is, the number of information for each classified set is different and the size to be considered is different, so that accurate information cannot be extracted. For example, numeric type information is numerical information such as age, height, and weight, and the number is not so large. On the other hand, Gaussian-type information is numerous because all information, genre, actor, director, and writer can be represented as a single data. Therefore, if each type is given the same weight, the number of types, that is, the number type, is greatly affected.

Accordingly, in step 330, the blogger community and the feature extractor 120 calculate the final distance values by applying weights to the distance values, respectively, as calculated by the distance values for each set, as shown in Equation 5 below. Integrate. Here, denotes weights for numbers, categories, strings, and Gaussians, respectively.

[Equation 5]

When the calculated final distance value is calculated, in step 340, the blogger community and the feature extracting unit 120 perform the clustering of bloggers who assign bloggers having the small distance value to the same cluster based on the final distance value. That is, find the blogger's community (implicit community).

Then, in step 350, the blogger community and feature extractor 120 extracts feature information of the found blogger community.

The clustering of the bloggers as described above will be described in more detail with reference to FIG. 4.

In general, the number of information used to cluster bloggers is very large. There are thousands of book genres, and tens of thousands of book authors. When clustering all of the information, the number of information to be considered is large, and the actual data is small, resulting in data sparsity.

In order to solve this problem, instead of finding distance values for all the information given when clustering, find the information that shows characteristics of each cluster well and obtain the distance value using only the corresponding information. That is, the distance value between the found blogger community and the user is obtained. This is a solution to the problem of data sparsity, but it is significant in that it finds more meaningful information for each cluster.

Referring to FIG. 4, in step 410, the blogger community and feature extractor 110 converts the blogger information into a vector to perform clustering. The method of converting into a vector puts the values of each information into data of each dimension. For example, if a blogger's email is aa@aa.com, it means that aa@aa.com is a value for email.

Next, in step 420, the blogger communicator and the feature extracting unit 110 generate a desired number (K * A) clusters using a first algorithm (K-means Clustering algorithm). Where A is a user-defined constant that allows you to determine how many smaller clusters to create initially. The first algorithm initially serves to group the bloggers into small clusters, and calculates distance values in consideration of all generated vector dimensions.

In step 430, the blogger communicator and feature extractor 110 searches for M vector dimensions having a low variance and a high mean using the mean and the variance for each generated cluster. It is assumed here that the dimension with low variance and high mean is a meaningful dimension.

In step 440, the blogger community and feature extractor 110 merges the data into one cluster using a second algorithm (Hierarchical Clustering Algorithm) in consideration of only the searched dimension. Here, the second algorithm finds similar clusters among clusters of small sizes bound by the first algorithm and binds the clusters with the clusters. In the second algorithm, clustering is performed by finding a meaningful vector dimension for each cluster, calculating a distance value using only the selected dimension.

In step 450, the blogger communicator and feature extractor 110 finds M vector dimensions having a low variance and a high mean for each cluster, and reassigns bloggers for each cluster in consideration of only the dimensions found in step 460.

Thereafter, in step 470, the blogger communicator and the feature extracting unit 110 check whether K clusters remain and terminate the operation for clustering if K clusters remain, otherwise return to step 440 afterwards. Repeat the steps.

As a result of such clustering, K clusters may be obtained in an embodiment of the present invention, and each cluster may have a blogger included in the cluster, an average of the entire cluster, a variance, M selected dimensions, and an average and variance of each selected dimension. Can be. At this time, the selected M dimensions are low in variance and high in average for each cluster, so that the bloggers in the cluster have similar values with respect to the selected dimension.

The clusters thus obtained contain a lot of information that cannot be found in the existing search, and you can see the preferred relationship or trends over time between specific events.

On the other hand, in the detailed description of the present invention has been described with respect to specific embodiments, various modifications are of course possible without departing from the scope of the invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined not only by the scope of the following claims, but also by the equivalents of the claims.

10: Service UI 11: Analysis UI
12: Search UI 100: Server
110: blogger information extraction unit 120: blogger community and feature extraction unit
130: analysis unit 140: user information unit
150: blogger search unit 160: blog DB
170: tag search unit 180: tag DB

Claims (19)

Classify blogger information of searched bloggers into a set type of preset numeric type, category type, string type, and Gaussian type, calculate distance values for each category of the blogger information, and integrate the calculated distance values for each classification. After calculating the final distance value, and grouping the bloggers based on the calculated final distance value to extract the blogger community, using the extracted blogger community and feature information, the user information and the user information regardless of time And a public analysis method or the user information and characteristics of the blogger community, which report the feature information of the blogger community as one piece of data, and analyze the feature of the user by comparing the user information with the feature information of the blogger community. When reporting information as continuous data that changes over time Based on the flow of events through a diachronic analysis of the method of analyzing the characteristics of the users and the community server to perform customized analysis bloggers; And
And a browser terminal connected to the server through the Internet and having a service UI which receives the analyzed result from the server and displays the analyzed result to the user.
The method of claim 1, wherein the server,
A blogger information extracting unit extracting blogger information of the retrieved bloggers;
Custom blogger analysis system comprising a.
The server according to claim 2,
A blogger searching unit for searching the blogs from the blogs;
A tag search unit for searching a tag necessary for extracting the blogger information; And
Custom blogger analysis system, characterized in that it further comprises a user information for managing user information required for custom blogger analysis.
The method of claim 1,
The server is
A weight is applied to each of the calculated distance values for each classification, and the final distance value is obtained by summing each weighted distance value.
[Equation]
d (m, X _j ) = w _n Nd (m, X _j ) + w _c Cd (m, X _j ) + W _s Sd (m, X _j ) + W _g Gd (m, X _j )
(W _n denotes a weight of the distance function between attributes having a numeric type, w _c denotes a weight of the distance function between attributes having a category type, and w _s denotes a weight of the distance function between attributes having a string type) Where w _g represents a weight of the distance function between attributes having a Gaussian type, and Nd (m, Xj) represents a function for obtaining a distance between attributes having a numeric type, and Cd (m, Xj) represents a category. Represents a function for calculating the distance between attributes, wherein Sd (m, Xj) represents a function for calculating the distance between attributes having a string value, and Gd (m, Xj) represents a function for calculating the distance between attributes having a distribution. .)
Custom blogger analysis system, characterized in that the calculation through the above equation.
delete delete delete The method according to claim 1 or 2,
The blogger information is composed of static information, which is member information input when creating a blog, and dynamic information, which is event information about a user's experience extracted from content registered by the user while using the blog. system. Extracting blogger information of bloggers retrieved by the server;
The server classifies the extracted blogger information into a set type of a predetermined number type, a category type, a string type, and a Gaussian type, calculates a distance value for each category of the blogger information, and calculates the calculated distance values for each category. Integrating bloggers based on the calculated final distance value and extracting blogger community and feature information to search for blogger community and feature information;
By using the feature information of the blogger community extracted by the server and the user information irrespective of time, the user information and the feature information of the blogger community are viewed as one data, respectively, the user information and the feature of the blogger community. A public analysis method for analyzing user characteristics by comparing information or reporting the user information and the characteristic information of the blogger community as continuous data that changes with time, based on the flow of events over time. Performing a user-specific blogger analysis through a historical analysis method of analyzing a feature of the; And
Displaying, by the browser terminal, a user customized blogger analysis result;
Custom blogger analysis method in a custom blogger analysis system comprising a.
delete delete 10. The method of claim 9,
The calculating of the final distance value may include applying a weight to each of the calculated distance values and adding the weighted distance values together to obtain the final distance value.
[Equation]
d (m, X _j ) = w _n Nd (m, X _j ) + w _c Cd (m, X _j ) + W _s Sd (m, X _j ) + W _g Gd (m, X _j )
(W _n denotes a weight of the distance function between attributes having a numeric type, w _c denotes a weight of the distance function between attributes having a category type, and w _s denotes a weight of the distance function between attributes having a string type) Where w _g represents a weight of the distance function between attributes having a Gaussian type, and Nd (m, Xj) represents a function for obtaining a distance between attributes having a numeric type, and Cd (m, Xj) represents a category. Represents a function for calculating the distance between attributes, wherein Sd (m, Xj) represents a function for calculating the distance between attributes having a string value, and Gd (m, Xj) represents a function for calculating the distance between attributes having a distribution. .)
Custom blogger analysis method characterized in that the calculation through the above equation.
The method of claim 9, wherein performing clustering of the bloggers comprises:
Converting the blogger information into a vector;
Generating a desired number of clusters by a user using a first algorithm;
Searching for a vector dimension having a low variance and a high mean using the mean and the variance for each generated cluster;
Merging clusters generated by the second algorithm into one cluster, considering only the retrieved vector dimension;
Rescanning the vector dimension having a low variance and a high mean for each of the generated clusters; And
And reassigning bloggers for each of the clusters using only the re-searched vector dimensions.
The method of claim 13,
The clustering of the bloggers may be performed by repeatedly merging the bloggers into one cluster by using the second algorithm until the predetermined number of clusters remain, and reassigning the bloggers for each cluster. A custom blogger analysis method in a custom blogger analysis system.
The method of claim 13,
The first algorithm is a custom blogger analysis method in a custom blogger analysis system, characterized in that the algorithm that serves to initially group the bloggers in a small cluster size.
The method of claim 13,
The second algorithm is a user-specific blogger analysis method in a user-defined blogger analysis system, characterized in that the algorithm that serves to find a cluster similar to each other among the clusters of the small size grouped by the first algorithm.
delete delete 10. The method of claim 9,
The blogger information is composed of static information, which is member information input when creating a blog, and dynamic information, which is event information about a user's experience extracted from content registered by the user while using the blog. How to analyze custom bloggers in the system.

Images