KR101678779B1

KR101678779B1 - Method for recommending contents using metadata and apparatus for performing the method

Info

Publication number: KR101678779B1
Application number: KR1020150060206A
Authority: KR
Inventors: 유환조; 오진오; 김성철
Original assignee: 포항공과대학교 산학협력단
Priority date: 2015-04-29
Filing date: 2015-04-29
Publication date: 2016-12-06
Also published as: KR20160128591A

Abstract

A content recommendation method and apparatus are disclosed. According to an aspect of the present invention, there is provided a content recommendation method comprising: constructing a heterogeneous graph indicating a relationship between a plurality of contents based on a similarity degree between metadata extracted from each of a plurality of contents; Extracting a pattern of a content preferred by the user using a content usage history of the user, and extracting a content to be recommended to the user from the heterogeneous graph using a pattern of a content preferred by the user. Accordingly, it is possible to improve the convenience of the user and the effectiveness of the content recommendation by providing the customized content recommendation service reflecting the content preference of the user.

Description

[0001] METHOD FOR RECOMMENDING CONTENTS [0002] METHOD AND APPARATUS FOR PERFORMING THE METHOD [

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a content recommendation technique, and more particularly, to a method of recommending a content to reflect a user's content preference using a heterogeneous graph constructed based on metadata and a content usage history, will be.

2. Description of the Related Art [0002] With the popularization of user terminals such as smart phones, tablet PCs, wearable devices, etc., users have access to a search engine system through a user terminal to use desired contents anytime and anywhere It was. In particular, with the advent of the Big Data era, the number of content has grown exponentially, allowing users to have a wide choice of content.

However, on the contrary, users have inconvenienced time and effort to search for a desired content among a large amount of contents. Accordingly, research on a content recommendation technology that helps a user to easily select a content in a search engine system in which a large amount of contents are produced and circulated is actively being studied.

In the past, a technique of recommending a content that is frequently used by users on a search engine system or recommending a content that is a current issue has been used.

The above-described conventional content recommendation technique can reduce the time and effort of a user who checks a large number of contents in order to search for a desired content, but can not explain that the recommended content satisfies the user.

Thus, in many search engine systems, a content recommendation technique is provided that provides users' evaluation information on the content so that the user can easily select the content. For example, a user selects a desired content by checking a rating, a rating, a ranking rate, or a viewing rank assigned to the movie by a user who has watched a movie in the past in the search engine.

However, such prior art also lacks an explanation as to why the recommended content is related to an individual user, and recommends the content based on the predicted rating simply. Therefore, the reliability and validity of the content recommendation Of the population.

Therefore, it is necessary to develop a content recommendation technique that can improve the reliability and validity of the content recommendation by confirming the content preference of each user from the history of the content usage history of the user, recommending the corresponding content, and providing the reason why the content is recommended There is a need.

In order to solve the above problems, an object of the present invention is to provide a personalized content recommendation service to a user by extracting content preference information in the form of a metapath from a heterogeneous graph constructed based on metadata and a history of content usage history of a user And to provide a method for recommending contents that can be used.

Another object of the present invention is to provide a content recommendation apparatus capable of increasing the utilization possibility of a content recommendation technique by additionally providing content preference information extracted from a content usage history of a user while providing a customized content recommendation service to a user have.

According to an aspect of the present invention, there is provided a content recommendation method for content recommendation, the method comprising: displaying a heterogeneous graph representing a relationship between a plurality of contents based on a degree of similarity between metadata extracted from each of a plurality of contents; Extracting a pattern of a content preferred by a user using a content usage history of the user in response to receiving a content recommendation request from a user, and extracting a pattern of a content preferred by a user from a heterogeneous graph And extracting contents to be edited.

The step of constructing the heterogeneous graph may include extracting metadata of a predefined type from each of the plurality of contents, comparing the similarity between the extracted metadata with each of the predefined types, Can create at least one metapath by concatenating a plurality of contents including high metadata.

Here, the heterogeneous graph may include at least one metaphor expressing a link representing a connection between a node expressing high similarity meta data and a plurality of contents including metadata having high similarity with high similarity meta data .

Here, the user's content usage history may be stored by mapping at least one content used by the user and evaluation scores of at least one content used by the user.

Here, the step of extracting a pattern of a content preferred by the user may include extracting a predefined type of metadata from each of at least one content stored in the content usage history of the user, and defining the similarity between the extracted metadata At least one meta pass for the user's content usage history can be generated by connecting the at least one content including the meta data having high similarity and the meta data having high similarity.

Here, the step of extracting a pattern of a content preferred by the user may extract a pattern of a content preferred by the user based on the frequency with which each of the at least one metapass for the content usage history of the user is generated.

Here, the step of extracting a content to be recommended to the user may include extracting a content having an evaluation score of at least one content out of at least one content stored in the content usage history of the user equal to or greater than a predetermined reference score, A plurality of contents linked to a pattern of a user's preferred content are extracted from the content recommendation candidates starting from the content candidate and extracted from the content recommendation candidates as content to be recommended to the user have.

Here, the content recommendation method may further include providing the user with a content to be recommended to the user.

According to another aspect of the present invention, there is provided a content recommendation apparatus for generating a heterogeneous graph for establishing a heterogeneous graph representing a relationship between a plurality of contents based on a similarity between metadata extracted from each of a plurality of contents, A preferred content pattern extracting unit for extracting a pattern of a content preferred by a user by using a content usage history of a user upon receipt of a content recommendation request from a user, And a recommended content extracting unit for extracting a content to be recommended.

Here, the content recommendation apparatus may further include a content providing unit for providing the user with a content to be recommended to the user.

According to the content recommendation method and the apparatus for performing the content recommendation according to the embodiment of the present invention as described above, it is possible to provide a customized content recommendation service reflecting the content preference of the user.

In addition, by providing content preference information extracted from the user's content usage history while providing a customized content recommendation service to the user, the possibility of utilizing the content recommendation technique can be increased and the reliability and validity of the content recommendation can be improved.

1 is a flowchart illustrating a content recommendation method according to an embodiment of the present invention.
2 is an exemplary diagram illustrating the construction of a heterogeneous graph for a movie according to an embodiment of the present invention.
3 is an exemplary diagram illustrating a heterogeneous graph for a movie according to an embodiment of the present invention.
FIG. 4 is a diagram illustrating an example of extracting a pattern of a content preferred by a user according to an embodiment of the present invention. Referring to FIG.
5 is a diagram illustrating an example of extracting contents to be recommended to a user in a heterogeneous graph according to an embodiment of the present invention.
6 is a block diagram illustrating a content recommendation apparatus according to an embodiment of the present invention.

While the invention is susceptible to various modifications and alternative forms, specific embodiments thereof are shown by way of example in the drawings and will herein be described in detail. It is to be understood, however, that the invention is not to be limited to the specific embodiments, but includes all modifications, equivalents, and alternatives falling within the spirit and scope of the invention. Like reference numerals are used for like elements in describing each drawing.

The terms first, second, A, B, etc. may be used to describe various elements, but the elements should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as a second component, and similarly, the second component may also be referred to as a first component. And / or < / RTI > includes any combination of a plurality of related listed items or any of a plurality of related listed items.

It is to be understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, . On the other hand, when an element is referred to as being "directly connected" or "directly connected" to another element, it should be understood that there are no other elements in between.

The terminology used in this application is used only to describe a specific embodiment and is not intended to limit the invention. The singular expressions include plural expressions unless the context clearly dictates otherwise. In the present application, the terms "comprises" or "having" and the like are used to specify that there is a feature, a number, a step, an operation, an element, a component or a combination thereof described in the specification, But do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in commonly used dictionaries are to be interpreted as having a meaning consistent with the contextual meaning of the related art and are to be interpreted as either ideal or overly formal in the sense of the present application Do not.

Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 is a flowchart for explaining a content recommendation method according to an embodiment of the present invention. FIG. 2 is an exemplary diagram illustrating construction of a heterogeneous graph for a movie according to an embodiment of the present invention. Fig. 8 is an exemplary diagram illustrating a heterogeneous graph for a movie according to an example. Fig.

4 is a diagram illustrating an example of extracting a pattern of a content preferred by a user according to an embodiment of the present invention. FIG. 5 is a flowchart illustrating a method of extracting a content to be recommended to a user from a heterogeneous graph according to an exemplary embodiment of the present invention Fig.

Referring to FIG. 1, a content recommendation method may be performed by a content recommendation apparatus. Here, the content recommendation apparatus may be implemented by a search engine system. The search engine system may be a server such as a cloud server or a web server, but is not limited to, a system for quickly searching for and providing content in a plurality of databases according to a user's requirement.

As the age of big data has arrived and the number of contents produced and distributed through search engine system has surged, a search engine system must implement a content recommendation technology so that a user can easily select contents from a large number of contents.

Therefore, the search engine system currently operated by Naver, Daum, Google, and the like recommends the content that is frequently used by the users on the search engine system or the content that is the current issue to the user, The content recommendation technology provides users with evaluation information on the content so that the user can select the content.

However, the conventional content recommendation technique described above has a problem in that the reliability and validity of the recommendation of content is degraded in that it can not provide a customized content recommendation service reflecting the individual content preference of the user.

Accordingly, the present invention proposes a content recommendation method that can improve the convenience of users and the effectiveness of content recommendation by recommending content suitable for each user's content preference using a heterogeneous graph constructed based on metadata and a history of content usage do.

A content recommendation method according to the present invention includes a step of constructing a heterogeneous graph indicating a relationship among a plurality of contents (S100), extracting a pattern of a content preferred by a user using a content usage history (S200) A step S300 of extracting a content to be recommended to the user from the heterogeneous graph using a pattern of the content, and a step S400 of providing the extracted content to the user.

A heterogeneous graph indicating a relationship between a plurality of contents can be constructed based on the similarity between the metadata extracted from each of the plurality of contents for the content recommendation (S100).

The contents may be media produced by mixing information such as voice, text, images, and images, and may be a web page, a movie, a television program, a computer game, or the like, including various information, but is not limited thereto. At this time, the content may be stored in a content database inside the search engine system or may be stored in an external storage device connected to the search engine system via a wire / wireless network.

In order to construct a heterogeneous graph, metadata of a predefined type can be extracted from each of a plurality of contents stored in a content database in the search engine system or an external storage device. In this case, the type is a category for classifying the metadata describing the characteristics of each content, and is defined differently depending on the type of contents.

As the metadata is extracted from each of a plurality of contents, it is possible to compute metadata having a high degree of similarity by comparing the similarities among the metadata with respect to each of the predefined types. In this case, in order to calculate metadata having high similarity, the metadata extracted from each of a plurality of contents may be clustered, and the distance between the metadata may be compared. However, the present invention is not limited to this, High meta data can be calculated.

At least one metapath can be generated by connecting a plurality of contents including metadata having a high degree of similarity and metadata having a high degree of similarity. Thus, the meta data having high similarity in at least one meta pass is expressed as a node, and the connection between a plurality of contents including metadata having high similarity and metadata having high similarity is expressed as a link, Can be constructed.

Explain more specifically how heterogeneous graphs are constructed using movies as examples of various types of contents.

In order to construct a heterogeneous graph for a movie, metadata of a predefined type can be extracted from each of a plurality of movies. At this time, the metadata extracted from each of the plurality of movies are classified into the types of 'director', 'actor', 'genre', 'release country', and 'release year' . &Lt; / RTI >

For example, metadata of 'James Cameron', 'Leonardo DiCaprio, Kate Winslet', 'Melo', 'USA' and '1997' corresponding to each of the predefined types in movie A are extracted, Meta data for 'Martin Scorsese', 'Leonardo DiCaprio, Cameron Diaz', 'Action', 'Germany' and '2002' for each of the predefined types can be extracted. In addition, metadata of 'Lee Jung Bum', 'Won Bin, Kim Saraon', 'Action', 'Korea' and '2010' corresponding to each of the predefined types in the movie C can be extracted.

The metadata 'Leonardo DiCaprio' and 'action', which are determined to have high similarity, can be calculated by comparing the similarities between the metadata extracted from the movies A, B, and C by type.

Thus, as shown in FIG. 2 (b), it is possible to connect the movie A and the movie B including the 'actor' type metadata 'Leonardo DiCaprio' and 'Leonardo DiCaprio' The meta-path can be generated by connecting the movie B and the movie C including the metadata 'action' based on the metadata 'action' of the type.

The heterogeneous graph for the movie can be constructed as shown in FIG. 3 by expressing a plurality of metapaths generated in this manner by a node and a link. In the heterogeneous graph for the movie shown in FIG. 3, a circle represents a movie, and a node indicated by a figure other than a circle represents metadata having a high degree of similarity among predefined types of metadata extracted from a plurality of movies. In addition, the link between the metadata having high similarity and the movie including the metadata can be expressed by a link to express the relationship between the movies.

Upon receipt of the content recommendation request from the user, the pattern of the content preferred by the user can be extracted using the content usage history of the user (S200).

Here, the content usage history may be stored by mapping the content used by the user and the content evaluation score given by the user, but the present invention is not limited to this, and usage information such as time and frequency using each content of the user may be additionally mapped .

In order to extract a pattern of a content preferred by a user, a metapass for a user's content usage history can be generated. At this time, the process of generating a metapath for a user's content usage history can be performed in the same manner as the generation process of a metapath for constructing a heterogeneous graph.

That is, metadata of a predefined type is extracted from each of at least one content stored in the content usage history of the user, and the similarity between the extracted metadata is compared with each of the predefined types to calculate metadata having high similarity , At least one meta pass for a user's content usage history can be generated by linking at least one content including meta data having high similarity and metadata having high similarity.

Thus, a pattern of a content preferred by the user can be extracted based on the frequency at which each of the at least one metapass for the content usage history of the user is generated.

More specifically, for example, as shown in FIG. 4A, evaluation scores '95', '35', and '50' assigned to the contents 'content_1', 'content_2', 'content_5' ',' 77 'may be mapped and stored. Based on the metadata of the 'actor' type as shown in FIG. 4 (b), 'meta-pass' connected between 'content_1' and 'content_6' At least one meta pass for a user's content usage history including a meta path connected between 'content_5' and 'content_5' may be generated. Among these, it can be known that a meta-pass A composed of "content-type" metadata of an actor-content "is generated with the highest frequency. Accordingly, the meta-pass A can be extracted as a pattern of content preferred by the user.

When a pattern of a user-preferred content is extracted, a content to be recommended to the user can be extracted from the heterogeneous graph using the extracted pattern (S300).

More specifically, it is possible to extract a content whose evaluation score for each of at least one content among at least one content stored in the content usage history of the user is equal to or higher than a predetermined reference score. Thus, it is possible to extract a plurality of contents linked to the pattern of the user's preferred content, starting from the content extracted from the heterogeneous graph, as a content recommendation candidate. A content that is not stored in the content usage history of the user among the content recommendation candidates can be extracted as a content to be extracted to the user.

For example, when the reference score is set to 70 points in order to extract a content to be recommended to the user, 'content_1' of 95 points and 'content_6' of 77 points in the content usage history of FIG. 4A can be extracted. Thus, in the heterogeneous graph constructed in the form of a heterogeneous graph as shown in FIG. 5, "content-type" metadata-content ", which is a pattern of the content preferred by the user, starting from" content_1 " It is possible to extract a plurality of following contents as content recommendation candidates.

That is, a plurality of contents 'content_3', 'content_4', and 'content_6' reaching through the 'actor' type metadata node linked by a link starting from a node indicating 'content_1' can be extracted as content recommendation candidates . Likewise, a plurality of contents 'content_1', 'content_7', and 'content_8' reaching through the metadata node of the 'actor' type can be extracted as content recommendation candidates with a node representing 'content_6' as a starting point.

In this way, it is confirmed whether the 'content_1', 'content_3', 'content_4', 'content_6', 'content_7', and 'content_8' extracted by the content recommendation candidate are stored in the content usage history of the user, 'Content_3', 'content_4', 'content_7', and 'content_8', except for 'content_1' and 'content_6' with content, can be extracted as contents to be recommended to the user.

In this case, the contents to be recommended are determined through the determination of whether or not the content recommendation candidate is included in the user's content usage history, but the present invention is not limited to this. However, after the content recommendation candidate is evaluated using the existing recommendation system, Recommended.

Thus, the finally extracted content is provided to the user (S400), and the personalized content recommendation service can be provided to each user.

In addition, a pattern of a content preferred by the user extracted from the content usage history of the user can be additionally provided so that the user can understand why the content is recommended.

6 is a block diagram illustrating a content recommendation apparatus according to an embodiment of the present invention.

Referring to FIG. 6, the content recommendation apparatus 100 may be implemented by a search engine system.

Here, the search engine system may refer to a server such as a cloud server or a web server as a system for quickly searching for and providing content in a plurality of databases according to a user's requirement, no.

The search engine system can be connected to a user terminal operated by a user through a wired / wireless network such as local area wireless communication, Wi-Fi, 3G (3G), LTE (Long Term Evolution) At this time, the user terminal may mean an information communication terminal such as a smart phone, a tablet PC, a notebook, a computer, a smart home appliance, and a system robot, but is not limited thereto.

The content recommendation apparatus 100 according to the present invention may include a heterogeneous graph construction unit 110, a preferred content pattern extraction unit 120, a recommended content extraction unit 130, and a content provider 140.

The heterogeneous graph building unit 110 may construct a heterogeneous graph indicating a relationship between a plurality of contents based on the similarity between the metadata extracted from each of the plurality of contents.

The contents may be media produced by mixing information such as voice, text, images, and images, and may be a web page, a movie, a television program, a computer game, or the like, including various information, but is not limited thereto. The content may be stored in the content DB 150 of the search engine system or an external content storage device connected to the search engine system via a wired or wireless network, although not separately shown in FIG.

Thus, the heterogeneous graph building unit 110 can generate a heterogeneous graph from a content DB 150 of a search engine system storing a plurality of contents or a predefined type the metadata of the type can be extracted. In this case, the type is a category for classifying the metadata describing the characteristics of each content, and is defined differently depending on the type of contents.

The preferred content pattern extracting unit 120 may extract the content preferred by the user by using the content usage history of the user upon receipt of the content recommendation request from the user terminal operated by the user.

At this time, the preferred content pattern extracting unit 120 may be interlocked with the content usage history DB 160 in which the content usage history is identified and stored for each user. Here, the content usage history may be stored by mapping the content used by the user and the content evaluation score given by the user, but the present invention is not limited to this, and usage information such as time and frequency using each content of the user may be additionally mapped .

Thus, the preferred content pattern extracting unit 120 extracts predefined types of meta data from each of at least one content stored in the user's content usage history, compares the similarities between the extracted meta data on a predefined type basis It is possible to generate at least one meta pass for the user's content usage history by connecting the at least one content including the meta data having high similarity and the meta data having high similarity after calculating the meta data having high similarity.

It is possible to extract a pattern of the content preferred by the user based on the frequency of each of the at least one metapass for the content usage history of the user.

The recommended content extraction unit 130 can extract the content to be recommended to the user from the heterogeneous graph using the pattern of the content preferred by the user.

More specifically, the recommended-content extracting unit 130 can extract a content whose evaluation score for each of at least one content among the at least one content stored in the content usage history of the user is equal to or greater than a predetermined reference score. Thus, it is possible to extract a plurality of contents linked to the pattern of the user's preferred content, starting from the content extracted from the heterogeneous graph, as a content recommendation candidate. A content that is not stored in the content usage history of the user among the content recommendation candidates can be extracted as a content to be extracted to the user.

The content providing unit 140 can provide a personalized content recommendation service to each user by providing the extracted content to the user.

The configuration of the content recommendation apparatus according to the embodiment of the present invention is applied to the heterogeneous graph building unit 110, the preferred content pattern extracting unit 120, the recommended content extracting unit 130, and the content providing unit 140 The present invention is not limited to the above embodiments, and at least two of the components may be combined into one component, or one component may be divided into a plurality of components to perform the functions. And is included in the scope of the present invention unless it departs from its essence.

In addition, the operation of the content recommendation apparatus according to the embodiment of the present invention can be implemented as a computer-readable program or code on a computer-readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored. The computer-readable recording medium may also be distributed and distributed in a networked computer system so that a computer-readable program or code can be stored and executed in a distributed manner.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the present invention as defined by the following claims It can be understood that

100: content recommendation apparatus 110: heterogeneous graph building unit
120: preferred content pattern extracting unit 130: recommended content extracting unit
140: Content providing unit 150: Contents DB
160: Content usage history DB

Claims

A method performed by a content recommendation apparatus,
Constructing a heterogeneous graph indicating a relationship between the plurality of contents based on similarity between metadata extracted from each of a plurality of contents;
Extracting a pattern of a content preferred by the user using a content usage history of the user upon receipt of a content recommendation request from the user; And
And extracting a content to be recommended to the user from the heterogeneous graph using a pattern of a content preferred by the user,
Wherein the step of extracting content to be recommended to the user comprises:
Extracting a content having an evaluation score for each of the at least one content from the at least one content stored in the content usage history of the user equal to or greater than a preset reference score,
Extracting a plurality of contents linked to a pattern of the user's preferred content as content recommendation candidates starting from the extracted content in the heterogeneous graph,
Wherein the content recommendation candidate is extracted from a content recommendation candidate that is not stored in the content usage history of the user as a content to be recommended to the user.

The method according to claim 1,
Wherein the step of constructing the heterogeneous graph comprises:
Extracting a predefined type of metadata from each of the plurality of contents,
And generating at least one metapath by connecting the plurality of contents including the highly similar metadata and the metadata having the high similarity by comparing the similarities between the extracted metadata with each of the predefined types, The content recommendation method comprising:

The method of claim 2,
In the heterogeneous graph,
A node representing the meta data having a high degree of similarity and a link representing a connection between a plurality of contents including the highly similar metadata and the metadata having the high similarity, Path of the content.

The method of claim 2,
The content usage history of the user,
Wherein at least one content used by the user and an evaluation score for each of at least one content used by the user are mapped and stored.

The method of claim 4,
Wherein the step of extracting a pattern of the user-
Extracting a predefined type of metadata from each of at least one content stored in the content usage history of the user,
Comparing the similarities between the extracted metadata with each other by the predefined types, and connecting at least one content including metadata having a high degree of similarity and metadata having a high degree of similarity, And generating one meta pass.

The method of claim 5,
Wherein the step of extracting a pattern of the user-
Wherein a pattern of the content preferred by the user is extracted based on a frequency at which each of the at least one metapass for the content usage history of the user is generated.

delete

The method according to claim 1,
And providing the user with a content to be recommended to the user.

A heterogeneous graph building unit for constructing a heterogeneous graph representing a relationship between the plurality of contents based on similarity between metadata extracted from each of a plurality of contents;
A preferred content pattern extracting unit for extracting a pattern of a content preferred by the user by using the content usage history of the user upon receipt of a content recommendation request from the user; And
And a recommended content extracting unit for extracting a content to be recommended to the user from the heterogeneous graph using a pattern of a content preferred by the user,
The recommended content extracting unit extracts,
Extracting a content having an evaluation score for each of the at least one content from the at least one content stored in the content usage history of the user equal to or greater than a preset reference score,
Extracting a plurality of contents linked to a pattern of the user's preferred content as content recommendation candidates starting from the extracted content in the heterogeneous graph,
And extracts, as the content to be recommended to the user, a content that is not stored in the content usage history of the user out of the content recommendation candidates.

The method of claim 9,
Wherein the heterogeneous graph building unit comprises:
Extracting a predefined type of metadata from each of the plurality of contents, comparing the similarity between the extracted metadata with each of the predefined types, and comparing metadata having high similarity with metadata having high similarity And at least one metapath is created by concatenating the plurality of contents including the contents.

The method of claim 10,
In the heterogeneous graph,
A node representing the meta data having a high degree of similarity and a link representing a connection between a plurality of contents including the highly similar metadata and the metadata having the high similarity, Wherein the content recommendation apparatus is constructed by expressing a path.

The method of claim 10,
The content usage history of the user,
Wherein at least one content used by the user and an evaluation score for each of at least one content used by the user are mapped and stored.

The method of claim 12,
The preferred content pattern extracting unit extracts,
Extracts a predefined type of metadata from each of at least one content stored in the content usage history of the user
Comparing the similarities between the extracted metadata with each other by the predefined types, and connecting at least one content including metadata having a high degree of similarity and metadata having a high degree of similarity, And generates one metapath.

The method of claim 12,
The preferred content pattern extracting unit extracts,
Wherein the content recommendation apparatus extracts a pattern of a content preferred by the user based on a frequency at which each of the at least one metapass for the content usage history of the user is generated.

delete

The method of claim 9,
The content recommendation apparatus includes:
And a content providing unit for providing the user with a content to be recommended to the user.