KR100720762B1 - Method for calculating similarity and searching content using context information of user - Google Patents

Abstract

The present invention relates to a method for searching content, and more particularly, to a method for automatically searching for and recommending a content most suitable for a user's current situation based on the user's context information determined from the user's static information and dynamic information. It is about.

The content search method according to the present invention may search for content suitable for the user's current situation using the context information determined from the static and dynamic information of the user and recommend it to the user.

Content, search terms, customized, ontology

Description

Method for calculating similarity and searching content using context information of user}

1 schematically illustrates a system for searching content according to an embodiment of the present invention.

2 is a flowchart illustrating a method of searching for content according to an embodiment of the present invention.

3 illustrates an example of static and dynamic information of a user.

4 is a flowchart illustrating a content searching method according to another embodiment of the present invention.

Description of the main parts of the drawing

1: user terminal 2: external ontology server

3: wired / wireless network 10: search server

11: client part 12: situation analysis part

13: search engine unit 14: storage unit

15: information collector 16: server controller

The present invention relates to a method for searching content, and more particularly, to a method for automatically searching for and recommending a content most suitable for a user's current situation based on the user's context information determined from the user's static information and dynamic information. It is about.

Currently, the Internet is spread to the public around the world, and users obtain and use various information from the Internet. But too much information is stored on the Internet. A user uses a search engine provided by an Internet portal site to easily obtain the information he needs from a large amount of information.

In general, in a search engine provided by an Internet portal site, a user searches by directly inputting one or more search terms. In addition, the searched content searches only content that includes a search word input directly by a user in the body or meta information. Therefore, in order to obtain content suitable for the current situation among a large amount of content, the user must select a search word correctly.

However, for users who are unfamiliar with the Internet or unfamiliar with search engine content, using search engines and selecting search terms is not an easy task, so users spend a lot of time and money trying to get the right content for their current situation. It's a waste of effort.

In order to overcome the problems of the conventional content retrieval method, the user's behavior pattern is analyzed and analyzed based on the user's behavior history such as searching for and recommending content suitable for the user based on the user's static information, or searching and purchasing the user's content. By searching for content based on a behavior pattern and recommending it to the user, the user is assisted with the content search.

However, the method of searching and recommending content from the user static information is searched only if the keyword given as the user static information exists at the same value in the body or meta information of the content to be searched, and the user's current location, place, search intent, etc. The problem is that the dynamic information of the user is not considered at all. In addition, the method of recommending content according to the user's history of behavior is also provided to all users who purchased content of 'A' according to the association rule such as "User who purchases 'A' content always buys 'B' content". By searching and recommending content in a way of recommending content, there is a problem in that users search for and recommend content uniformly without considering different situations at all.

As a result, the conventional content retrieval method has no limitation in searching / recommending customized content suitable for the user by not considering the dynamic current situation information of the user at all.

Accordingly, an object of the present invention is to provide a method for searching for customized content suitable for a user according to the current situation information of the user determined from the static and dynamic information of the user.

An object of the present invention is to provide an intelligent content retrieval method that can automatically search a variety of content by extending the search range to the semantic information that can be generated from the current situation information of the user using the information ontology.

The content search method for achieving the object of the present invention comprises the steps of extracting the user's static information and dynamic information, determining the user's context information based on the extracted user's static information and dynamic information and the user's context information Calculating the similarity of the content corresponding to and selecting the content to be recommended to the user based on the calculated similarity of the content.

Hereinafter, a method of searching for content according to the present invention will be described in detail with reference to the accompanying drawings.

1 schematically illustrates a system for searching content according to an embodiment of the present invention.

Referring to FIG. 1, a user terminal 1 and an external ontology server 2 are connected to a wired / wireless internet network 3. The user terminal 1 may access the search server 10 through a wired / wireless internet network 3 such as a desktop personal computer, a mobile phone, a portable terminal (PDA), etc. and receive content provided by the search server 10. To use the terminal generically. On the other hand, the external ontology server 2 refers to a server that stores the ontology information and provides the ontology information according to the request of the search server 10.

The search server 10 includes a client unit 11, a situation analysis unit 12, a search engine unit 13, a storage unit 14, an information collection unit 15, and a server control unit 16. The client unit 11 provides a user interface to the user terminal 1, manages user information input through the user interface, and modifies the searched content according to the type of the user terminal 1 to provide it to the user terminal 1. . When the user joins the search server 10, the user inputs static information of the user through the user interface, and the input static information of the user is stored in the storage unit 14.

The information collecting unit 15 collects various information ontology or content information from the external ontology server 2 and stores the collected information ontology or content information in the storage unit 14. The situation analysis unit 12 analyzes / determines the current situation information of the user by using the user static information and dynamic information received from the user terminal 1 and the information ontology stored in the storage unit 14. The search engine unit 13 calculates the similarity of the content corresponding to the user's current situation information based on the current situation information of the user, and the server controller 16 selects the content to be recommended to the user based on the calculated similarity. The search engine unit 13 calculates the similarity of the content to be recommended to the user based on the similarity of the content matching the user's static information, the similarity of the content matching the user's context information, and the user's preferred content similar to the user. . The server controller 16 provides a policy necessary for calculating the similarity, and the search engine 13 calculates the similarity based on the provided policy.

2 is a flowchart illustrating a method of searching for content according to an embodiment of the present invention.

Referring to FIG. 2, static information of a user is extracted (step 1). The user's static information is the user's ID, name, age, occupation, address / work place, preferred content classification, etc. The user's static information is stored in the storage unit 14. Other kinds of static information may be used depending on the field to which the present invention is applied, which is within the scope of the present invention.

Extract dynamic information of the user (step 2). The user's dynamic information can include the time the user requires a search for the content, the user's current location, the user's device type, the status of the last content available (e.g. music listening status for music recommendation services, a list of music content, etc.) to be. Different kinds of dynamic information may be used depending on the field to which the present invention is applied, which is within the scope of the present invention. The dynamic information of the user may be received from the user terminal 1 to the client unit 11 or extracted from the storage unit 14 when the user requests a search for content.

The information ontology is used to determine the user's current situation information from the static and dynamic information of the user (step 3). An information ontology is an ontology used to infer a user's current situation from user static information and dynamic information, or used to calculate similarity of content that matches the inferred user's current situation, and includes a public ontology and a domain ontology. .

A public ontology is an information ontology that is referred to to define the concepts and relationships of each word constituting a user's static and dynamic information. For example, time-related ontology such as day / night, am / pm, day of the week, month, season, year, vacation, holiday, holiday, etc., place-related ontology such as country, city, mountain, beach, amusement park, travel, work Situational ontologies such as business trips, business trips, honeymoons, and dating are used as public ontology.

On the other hand, domain ontology is an information ontology that is referred to to define the relationship of various information related to the content and to extend the semantic information inferred from the content information. For example, in the case of music content, a music title, singer, composer, genre, album, authorization chart, preferred listening time, preferred listening region, and preferred listening situation are used as domain ontology.

The current situation information of the user is determined from the relation between each word or words constituting the user static information and the dynamic information using the common ontology. 3 illustrates an example of static and dynamic information of a user.

Referring to FIG. 3, when the user wants to search for the content, the static information of the user's address (per minute), gender (male), age (36 years old), marital status (unmarried), preferred content classification (ballad) and the user Dynamic information of the current location of (Jeju Island) and the current time (11 pm on August 3, 2006) is extracted. Preferably, the extracted static and dynamic information of the user may be generated by a user ontology. The concept of summer and day can be extracted from the user's current time by using time-related ontology among public ontology, and the concept of beach, mountain, and island can be extracted by using the place-related ontology called Jeju Island. Bundang), current location (Jeju Island) and marital status (unmarried) can be extracted the concept of travel, vacation, honeymoon. From the extracted concepts, it is possible to determine the current situation information of the user such as summer vacation and summer travel.

When determining the current situation information of the user, different weights may be applied to each element constituting the user static information and dynamic information or to each concept constituting the common ontology.

The similarity of the content corresponding to the user's context information determined from the user's static information and the dynamic information is calculated (step 4).

The similarity of the content corresponding to the user's situation information is calculated from the following equation (1).

[Equation 1]

Here, w1 means the similarity of the first candidate content that matches the user's static information, and w2 means the similarity of the second candidate content that matches the user's situation information. α and β are weights applied to the similarity w1 of the first candidate content and the similarity w2 of the second candidate content, respectively.

Based on the similarity (w) of the calculated content, content that exceeds a predetermined threshold value is selected and recommended to the user (step 5).

4 is a flowchart illustrating a content searching method according to another embodiment of the present invention. Referring to FIG. 4, steps 11 to 13 are the same as steps 1 to 3 described with reference to FIG. 2, and a description thereof will be omitted.

Determine a user with static information, contextual information or behavior history similar to the user (step 14). Considering not only the user's situation information determined from the user's static information and dynamic information, but also the user's preferred content similar to the user, the similarity w of the content corresponding to the user's situation information is calculated (step 15). The similarity of the content corresponding to the user's context information is calculated from the following equation (2).

[Equation 2]

Where w1 is the similarity of the first candidate content that matches the user's static information, w2 is the similarity of the second candidate content that matches the user's context information, and w3 is a third candidate to recommend to the user among the content recommended by the similar user. The similarity of the content. Α, β, and γ are weights applied to the similarity w1 of the first candidate content, the similarity w2 of the second candidate content, and the similarity w3 of the third candidate content, respectively.

Based on the similarity (w) of the calculated content, a content that exceeds a predetermined threshold is selected and recommended to the user (step 16). Preferably, the similarity w1 of the first candidate content, the similarity w2 of the second candidate content, and the similarity w3 of the third candidate content are applied with different weights to calculate the similarity w of the final content. .

Hereinafter, a method of calculating the similarity w1 of the first candidate content, the similarity w2 of the second candidate content, and the similarity w3 of the third candidate content will be described in more detail.

<Calculation of Similarity (w1) of First Candidate Content>

The similarity (w1) of the first candidate content corresponding to the user static information is 1) the content similarity (w11) based on the recommendation rule 2) the similarity (w12) based on the preferred content by the static information 3) based on the tag tagged in the content It is calculated by applying different weights for each type of similarity (w13).

The similarity w11 based on the recommendation rule is a case where the content to be recommended is defined as a rule by the expert or the operator of the search server 10 according to the static information of the user. For example, if you are a man in his 20s and like ballads, the recommendation rule for providing "A" content is to calculate the similarity of content that matches the user's static information. Similarity (w12) based on the preferred content for each user's static information is used when the user's static information is extracted from the history of the user's purchase or listening for the content. The calculated similarity. The similarity w13 based on the tag of the content is a similarity calculated based on the static information of the user who created the tag when the user generates the tag in the content. The weight may be separately defined according to the expertise of the user who created the tag or the number of users who created the same tag.

Similarity of the first candidate content is calculated according to Equation (3) below.

[Equation 3]

Α1, β1, and γ1 are weights applied to content similarity (w11) based on recommendation rules, similarity (w12) based on preferred content for each static information, and similarity (w13) based on tags tagged in the content.

<Calculation of Similarity (w2) of Second Candidate Content>

The similarity (w2) of the second candidate content that matches the user context information is 1) the content similarity (w21) based on the recommendation rule 2) the similarity (w22) based on the preferred content by situation information 3) based on the tag tagged in the content It is calculated by applying different weights for each type of similarity (w23).

Similarity (w21) based on the recommendation rule is a case where the content to be recommended is defined as a rule by the expert or the operator of the search server 10 as a rule, and the similarity (w22) based on the preferred content for each situation information of the user. Is the similarity calculated using the relevant information when the contextual information of the users who are interested in the content is extracted from the behavior history of the users who purchase or listen to the content, and the similarity (w23) based on the content tag is When users create a tag in the content, the similarity is calculated based on context information of the user who created the tag. The weight may be separately defined according to the expertise of the user who created the tag or the number of users who created the same tag.

The similarity degree of the second candidate content is calculated according to Equation (4) below.

[Equation 4]

Here, α2, β2, and γ2 are weights applied to the content similarity w21 based on the recommendation rule, the similarity w22 based on the preferred content for each contextual information, and the similarity w23 based on the tag tagged to the content.

<Calculation of Similarity (w3) of Third Candidate Content>

In order to calculate the similarity w3 of the third candidate content to be recommended to the user among the contents recommended by the similar user, first, the user similar to the user should be selected.

Users who are similar to users are: 1) users with similar user static information 2) users with similar user context information 3) users with similar tags on the same content 4) users with similar scores on the same content 5) views / views on similar content 6) Users who have performed the same actions, such as purchases. 6) Users who have been explicitly registered as similar users are classified by type.

Finds users with similar user static information and extracts the user's preferred content as recommended content, extracts content that users have viewed / purchased with similar situational information as recommended content, and generates the same or similar tags for the same content Extracts the user's preferred content as recommended content, finds users who have given similar ratings to the same content, extracts the user's preferred content as recommended content, and uses the same behavior for similar content (e.g. The user searches for a user who has made a purchase, listens to music, etc., and extracts the user's preferred content as the recommended content. If the user who requested the recommendation explicitly registers the similar user, the registered content of the registered similar user is extracted as the recommended content. The degree of similarity between tags determines the degree of similarity between tag concepts using a public ontology.

 The similarity w3 of the third candidate content to be recommended to the user among the recommended content extracted from the similar user is calculated according to the number of overlaps and the weight of the extracted recommended content of the similar user.

Meanwhile, the above-described embodiments of the present invention can be written as a program that can be executed in a computer, and can be implemented in a general-purpose digital computer that operates the program using a computer-readable recording medium.

The computer-readable recording medium may be a magnetic storage medium (for example, a ROM, a floppy disk, a hard disk, etc.), an optical reading medium (for example, a CD-ROM, DVD, etc.) and a carrier wave (for example, the Internet). Storage medium).

Although the present invention has been described with reference to the embodiments shown in the drawings, this is merely exemplary, and it will be understood by those skilled in the art that various modifications and equivalent other embodiments are possible. Therefore, the true technical protection scope of the present invention will be defined by the technical spirit of the appended claims.

The content search method according to the present invention may search for content suitable for the user's current situation using the context information determined from the static and dynamic information of the user and recommend it to the user.

Claims (9)

(a) extracting static and dynamic information of the user; (b) determining situation information of the user based on the extracted static and dynamic information of the user; And (c) calculating a similarity (w) of the content corresponding to the determined situation information of the user and selecting content to be recommended to the user based on the similarity of the content. The method of claim 1, wherein the user's context information is And a content retrieval method determined from static and dynamic information of the user by using an information ontology for determining context information. The method of claim 2, wherein the similarity (w) of the content corresponding to the user's context information is And a similarity (w1) of first candidate content that matches the static information of the user and a similarity (w2) of second candidate content that matches the situation information of the user. The method of claim 3, wherein the similarity (w) of the content is A method of searching for content, calculated by applying different weights to the similarity (w1) of the first candidate content and the similarity (w2) of the second candidate content. The method of claim 3, wherein the similarity (w1) of the first candidate content is The similarity (w11) of the content that matches the static information of the user based on the recommendation rule, the similarity (w12) of the content that matches the static information of the user based on the preferred content for each static information, and the tag tagged to the content. A content retrieval method based on the similarity (w13) of the content corresponding to the static information of the user. The method of claim 5, wherein the similarity (w) of the first candidate content is And searching for the similarity (w11, w12, w13) of the calculated content by applying different weights. The method of claim 5, wherein the similarity (w2) of the second candidate content is The similarity (w21) of the content that matches the context information of the user based on the recommendation rule, the similarity (w22) of the content that matches the context information of the user based on the preferred content for each context information, and the tag tagged to the content. And a content retrieval method based on the similarity (w23) of the content corresponding to the contextual information of the user. 8. The method of claim 7, wherein the similarity (w2) of the second candidate content is The search method of the content calculated by applying different weights to the similarity (w21, w22, w23) of the calculated content. 8. The method of claim 7, wherein the searching method of the content is performed after the step (b). Determining a user having static information, contextual information, or behavioral history similar to the user, The similarity (w) of the content is the similarity (w1) of the first candidate content that matches the static information of the user, the similarity (w2) of the second candidate content that matches the situation information of the user, and the determined similar user selected. A content retrieval method calculated from the similarity (w3) of third candidate content to be recommended to a user.

Images