KR100871873B1 - Apparatus and method for managing user preference informations of the custom broadcasting system - Google Patents

Abstract

A preference information management apparatus of a customized broadcast system and a method thereof are provided to expand collected user preference information sufficiently for program recommendation, thereby providing a customized service of high reliability and satisfaction based on the expanded user preference information. A preference information analyzing unit analyzes preference information of a user based on a viewing pattern of a user. A preference information expanding unit conceptualizes user preference information outputted from the preference information analyzing unit by ontology mapping(S240). The preference information expanding unit expands the user preference information by using keyword information of the conceptualized user preference information(S250). A preference contents recommendation unit recommends contents whose user preference degree is high based on a frequency of the expanded preference information by the preference information expanding unit(S260).

Description

{APPARATUS AND METHOD FOR MANAGING USER PREFERENCE INFORMATIONS OF THE CUSTOM BROADCASTING SYSTEM}

The present invention relates to a customized broadcasting system, and more particularly, to an apparatus and method for managing preference information of a customized broadcasting system capable of extending user preference information collected and managed by the customized broadcasting system.

As the digital television environment is popularized, the development of user-tailored broadcasting service using metadata is becoming prominent. The customized broadcasting service is a service for providing customized broadcasting contents suitable for a usage environment including user preferences, terminal performance, network characteristics, and surrounding environment (eg, time, place, user mood, etc.). The customized broadcasting service allows viewers to participate in the broadcasting service directly or receive and watch a desired program at a desired time, instead of passive viewing, which receives the broadcasting program with limited channel selection.

The TV-Anytime Forum, a private standards organization, established standards for targeting services based on individual user preferences. The targeting service is to provide content suitable for an individual's current situation (eg, TV program, advertisement, information, etc.) based on additional detailed information (eg, user mood, preference, personal consumption information, etc.) to user preference information. It has the same meaning as a customized broadcast service.

The user-related metadata standard suggested by the TV-Anytime for a customized broadcasting service includes user history and user preference information. The user history records and stores a command, a target program, and a time that each user performs while watching for each user. The user preference information stores information useful for recommending a user's favorite program such as a title of a preferred program, a subject or genre of a preferred program, a favorite channel, a preferred viewing time zone, and the like.

Preferred programs when watching a broadcast vary depending on the user's life pattern and taste. For example, some users prefer to watch dramas and movies while others prefer to watch current events or sports. Since the user's preferences often change depending on the viewing time or season, the user's preference pattern is complicated and diverse. Therefore, there is a need for a method of increasing the viewing satisfaction by automatically identifying and managing patterns related to user-specific preference programs, and first recommending and providing a preference program suitable for each user.

However, due to the difficulty in accumulating and acquiring user viewing data, the conventional customized broadcasting systems did not extract various user preference information sufficient for program recommendation, and thus, the reliability and satisfaction in providing customized services were deteriorated.

Accordingly, the present invention has been devised to solve the above problems, and an object of the present invention is to provide an apparatus and method for managing preference information of a customized broadcasting system which can expand user preference information collected to be sufficient for program recommendation.

Another object of the present invention is to provide an apparatus and method for managing preference information of a customized broadcasting system capable of providing a customized service having high reliability and satisfaction based on extended user preference information.

In order to achieve the above object, the preference information management apparatus of the customized broadcasting system according to an aspect of the present invention, preference information analysis means for analyzing the user's preference information based on the user's viewing pattern; Preference information expansion means for extending user preference information output from the preference information analysis means through ontology mapping; It may be configured to include a preferred content recommending means for recommending content with a high preference of the user based on the preference information extended by the preference information expansion means.

Preferably, the preference information analyzing means comprises a pattern collecting means for collecting a viewing pattern for each user; Characteristic prediction means for determining characteristics of a preference pattern and a non-preferred pattern based on the viewing pattern collected by the pattern collecting means; It may be configured to include a preference prediction means for quantifying the characteristics determined by the characteristic prediction means to determine the rank of the preference information.

Preferably, the pattern collecting means extracts and collects a viewing pattern for each user from signals input from user devices.

Preferably, the characteristic predicting means calculates Mutual Information (MI) value of the viewing pattern collected by the pattern collecting means, based on the calculated MI value and the average Mutual Information (MI _avg ) value of the entire pattern. As a result, the characteristics can be classified and managed into a preference pattern and a non-preferred pattern.

Preferably, the characteristic predicting means may determine the pattern having the MI _avg and the MI value greater than 0 as the preferred pattern and the pattern having the MI value less than 0 as the non-preferred pattern.

Preferably, the MI value can be calculated by the following formula.

, 여기에서

는 패턴 A의 빈도수이고,

는 사용자의 패턴 풀에서 패턴 A가 나타날 확률이며,

는 사용자가 선호 프로그램을 시청하는 동안에 패턴 A가 나타날 확률을 의미한다.

, From here

Is the frequency of pattern A,

Is the probability that pattern A appears in your pattern pool,

Denotes the probability that the pattern A appears while the user watches the preferred program.

Preferably, the preference predicting means scores the preference of the preference pattern determined by the characteristic predicting means, and ranks the preference information for each user by reflecting the non-preferability of the non-preferred pattern.

Preferably, the preference information expansion means includes ontology mapping means for conceptualizing the user-specific preference information output from the preference information analyzing means through ontology mapping; And a concept vector extracting means for extracting a concept vector by expanding keyword information of the conceptualized user preference information output from the ontology mapping means.

Preferably, the concept vector extracting means may manage keyword information of user preference information by dividing the keyword information of the user preference information into an original keyword vector (okv) and a concept keyword vector (ckv).

Preferably, the basic keyword vector and the concept keyword vector can be defined by the following formula.

ckv _i = {c ₁ , c ₂ ,, c _k } k = # of concept

okv _i = {w ₁ , w ₂ ,, w _{k '} } k' = # of uniq word, where i is user, c is frequency of appearance of each concept k in preferences of conceptualized user i, w is user It means the frequency of appearance of each basic keyword k 'in the preference information of i.

Preferably, the preferred content recommending means scores and ranks a similarity (similarity) between the preference information extended by the preference information expansion means and the metadata information of each content to obtain contents predicted to be preferred by the user. I can recommend it.

Preferably, the preferred content recommending means measures a vector similarity by using a vector of elements of the preference information extended by the preference information expansion means and a vector of corresponding metadata information of each content by cosine similarity calculation of the following equation. The average similarity can be scored.

Preferably, the scoring may be performed by the score calculation of the following formula.

, 여기에서

,

는 각 요소벡터 유사도의 가중치, N은 선호정보에서 벡터화하고 있는 요소의 개수를 의미한다.

, From here

,

Is the weight of each element vector similarity, and N is the number of elements vectorized in the preference information.

In order to achieve the above object, there is provided a method of managing preference information of a customized broadcasting system according to another aspect of the present invention, comprising: analyzing preference information of a user based on a viewing pattern of the user; Extending the analyzed user preference information through ontology mapping; And recommending a content having high user preference based on the expanded conceptual information.

Preferably, the analyzing step includes collecting a viewing pattern for each user; Determining characteristics of a preference pattern and a non-preferred pattern based on the collected viewing pattern; And determining the rank of the preference information by quantifying the determined characteristic.

The determining of the characteristics may include obtaining a mutual information (MI) value of the collected viewing pattern and classifying the characteristics into a preferred pattern and a non-preferred pattern based on the average mutual information (MI _avg ) value of the entire pattern. Can be managed.

Preferably, the pattern having the MI _avg and the MI greater than 0 may be determined as the preferred pattern, and the pattern having the MI less than 0 may be determined as the non-preferred pattern.

Preferably, the MI value can be calculated by the following formula.

, 여기에서

는 패턴 A의 빈도수이고,

는 사용자의 패턴 풀에서 패턴 A가 나타날 확률이며,

는 사용자가 선호 프로그램을 시청하는 동안에 패턴 A가 나타날 확률을 의미한다.

, From here

Is the frequency of pattern A,

Is the probability that pattern A appears in your pattern pool,

Denotes the probability that the pattern A appears while the user watches the preferred program.

Preferably, in the ranking step, the preference of the determined preference pattern may be scored, and the preference information for each user may be ranked by reflecting the non-preferability of the non-preferred pattern.

Preferably, the expanding step includes conceptualizing the analyzed user-specific preference information through ontology mapping; And extracting a concept vector by expanding keyword information of the conceptualized preference information for each user.

Preferably, the concept vector extracting step may further include dividing and managing keyword information of user preference information into an original keyword vector (okv) and a concept keyword vector (ckv).

Preferably, the basic keyword vector and the concept keyword vector can be defined by the following formula.

ckv _i = {c ₁ , c ₂ ,, c _k } k = # of concept

okv _i = {w ₁ , w ₂ ,, w _{k '} } k' = # of uniq word, where i is user, c is frequency of appearance of each concept k in preferences of conceptualized user i, w is user i The frequency of occurrence of each basic keyword k 'in the preference information of.

Preferably, the recommendation step may recommend contents that are predicted to be preferred by the user by scoring and ranking a similarity (sim) between the extended preference information and metadata information of each content.

Preferably, the recommendation step is performed by measuring the vector similarity of the vector of the elements of the extended preference information and the corresponding metadata information of each content by calculating the cosine similarity of the following equation and then scoring the average similarity. Can be.

Preferably, the scoring may be performed by the score calculation of the following formula.

, 여기에서

,

는 각 요소벡터 유사도의 가중치, N은 선호정보에서 벡터화하고 있는 요소의 개수를 의미한다.

, From here

,

Is the weight of each element vector similarity, and N is the number of elements vectorized in the preference information.

The apparatus and method for managing preference information of a customized broadcasting system according to the present invention extends the collected user preference information enough to recommend a program, thereby providing a customized service having high reliability and satisfaction based on the expanded user preference information. It works.

Ontology is the definition or specification of a vocabulary or concept, and in the field of information systems, it refers to the components (concepts) corresponding to the contents handled by the system. Ontologies are made up of definitions of the basic concepts of a particular field and specifications of their relationships. They are not simply enumerated, but are structured and concrete enough to be processed by a computer. The basics of ontology are the concepts that exist in the realm. For example, books, computers, desks, chairs, purchases, bids. Each concept again has attributes that describe the concept, for example, a book might have attributes such as author, publisher, page, and price, and a bid might have attributes such as object, date, method, and condition. Concepts can also be related to each other, the most basic of which is up-and-down relationship. For example, fairy tales become sub-concepts included in books.

The present invention proposes a method for alleviating the data shortage problem by conceptualizing and expanding user preference information extracted through pattern analysis using such ontology. For example, if a user's preference information has a basic keyword 'baseball', this keyword is conceptualized as an upper concept of 'sports' or 'ball game' through ontology mapping, Keywords such as 'volleyball', 'soccer' and 'basketball' may be used as the user's preference information.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be noted that the same elements in the figures are represented by the same numerals wherever possible. In addition, detailed descriptions of well-known functions and configurations that may unnecessarily obscure the subject matter of the present invention will be omitted.

1 is a block diagram showing the configuration of a preference information management apparatus of a customized broadcasting system according to an embodiment of the present invention.

As shown in FIG. 1, the preference information managing apparatus 100 of the customized broadcasting system includes a preference information analyzer 110 analyzing the preference information of the user based on a viewing pattern of the user; A preference information extension unit 120 for expanding user preference information output from the preference information analysis unit 110 through ontology mapping; The preference information extension unit 120 may include a preferred content recommendation unit 130 that recommends content with high user preference based on the preference information extended by the preference information extension unit 120.

The preference information analyzing unit 110 includes a pattern collecting unit 112 for collecting a viewing pattern for each user; A characteristic predictor 114 that determines characteristics of a preference pattern and a non-preferred pattern based on the viewing pattern collected by the pattern collector 112; It may be configured to include a preference predictor 116 to determine the rank of the preference information by quantifying the characteristics determined by the feature predictor 114.

The pattern collector 112 extracts and collects a viewing pattern for each user from signals input from user devices.

The characteristic predictor 114 obtains Mutual Information (MI) values of the viewing patterns collected by the pattern collector 112 and then prefers them based on the average mutual information (MI _avg ) values of the entire patterns. Manage characteristics by classifying them into patterns and non-preferred patterns. In this case, the characteristic predictor 114 determines a pattern having a MI _avg and a MI greater than 0 as a preferred pattern and a pattern having a MI smaller than 0 as a non-preferred pattern. The MI value may be calculated by Equation 1 below.

는 패턴 A의 빈도수이고,

는 사용자의 패턴 풀에서 패턴 A가 나타날 확률이며,

는 사용자가 선호 프로그램 을 시청하는 동안에 패턴 A가 나타날 확률을 의미한다.In [Equation 1]

Is the frequency of pattern A,

Is the probability that pattern A appears in your pattern pool,

Denotes the probability that pattern A appears while the user is watching a favorite program.

The preference predictor 116 scores the preference of the preference pattern determined by the characteristic predictor 114, and ranks the preference information for each user by reflecting the non-preferability of the non-preferred pattern.

The preference information extension unit 120 includes an ontology mapping unit 122 for conceptualizing user-specific preference information output from the preference information analysis unit 110 through ontology mapping; It may include a concept vector extraction unit 124 for extracting a concept vector by expanding the keyword information of the conceptualized preference information for each user output from the ontology mapping unit 122.

The ontology mapping unit 122 conceptualizes user-specific preference information having a rank output from the preference information analysis unit 110 through ontology mapping. This conceptualized user preference information can be expected to expand the preference information by sharing terms belonging to the concept of the ontology.

The concept vector extracting unit 124 is a keyword information that can be expected positively the effect of expansion among the various elements (eg genre, title, keywords, etc.) of the conceptualized user preference information output from the ontology mapping unit 122 Expand to extract the concept vector. Accordingly, keyword information of user preference information may be managed by dividing it into an original keyword vector (okv) and a concept keyword vector (ckv). The equation associated with the basic keyword vector and the concept keyword vector is shown in Equation 2 below.

ckv _i = {c ₁ , c ₂ ,, c _k } k = # of concept

okv _i = {w ₁ , w ₂ ,, w _{k '} } k' = # of uniq word

In Equation 2, the c value of the concept keyword vector ckv _i of the preference information for the user i represents the frequency of appearance of each concept k after conceptualizing each keyword. The w value of the basic keyword vector okv _i represents the frequency of appearance of each basic keyword k '.

The preferred content recommendation unit 130 scores and ranks the similarity (similarity: sim) between the preference information extended by the preference information extension unit 120 and the metadata information of each content, and predicts that the user will prefer. Recommended content.

A score of each content may be calculated through the following [Equation 3] and [Equation 4].

,

는 각 요소벡터 유사도의 가중치를 의미한다. N은 선호정보에서 벡터화하고 있는 요소의 개수이다.[Equation 4] is to score the degree of similarity between the user preference information and the metadata of each program, the vector of each element of the preference information and the corresponding metadata information of each program information of Equation 3 The vector similarity is measured by cosine similarity calculation and then the average similarity is scored. In this case, the keyword vector is scored in consideration of both the concept keyword vector and the basic keyword vector extracted through the ontology mapping as described above. In [Equation 4] above

,

Denotes the weight of each element vector similarity. N is the number of elements vectorized in the preference information.

Units of the apparatus of the present invention as described above may be configured in the form of a software module to perform the same function.

2 is a flowchart illustrating a procedure of a method for managing preference information of a customized broadcasting system according to another embodiment of the present invention.

Referring to FIG. 2, in step S210, the preference information analyzing unit 110 collects a viewing pattern for each user.

In addition, in step S220, the preference information analysis unit 110 determines the characteristics of the preference pattern and the non-preferred pattern based on the collected viewing pattern. At this time, as described above, after obtaining the mutual information (MI) value of the collected viewing pattern, and classifies the characteristics into a preferred pattern and a non-preferred pattern based on the average mutual information (MI _avg ) value of the entire pattern do. Here, the pattern having MI _avg and MI greater than 0 is determined as a preferred pattern, and the pattern having MI smaller than 0 is determined as a non-preferred pattern, and the MI value may be calculated by Equation 1 above.

In step S230, the preference information analyzer 110 determines the rank of the preference information by quantifying the determined characteristic. At this time, the preference of the determined preference pattern is scored, and the preference information for each user is ranked by reflecting the non-preference of the non-preferred pattern.

The pattern collection and characteristic determination may be managed by collecting new patterns periodically and optimally classifying them into a preferred pattern and a non-preferred pattern even when the user does not watch a broadcast. This automatic user behavior pattern observation is advantageous to manage the user's latest preference information.

Subsequently, in step S240, the preference information extension unit 120 conceptualizes the ranked user preference information through ontology mapping.

In operation S250, the conceptual vector may be extracted by expanding keyword information of the conceptualized preference information for each user. Accordingly, the keyword information of the user preference information may be divided into an original keyword vector (okv) and a concept keyword vector (ckv). The basic keyword vector and the concept keyword vector are defined by Equation 2 above.

Subsequently, in step S260, the preferred content recommendation unit 130 scores and ranks a similarity (similarity) between the expanded preference information and the metadata information of each content to predict the contents that the user is likely to prefer. I recommend you. In this case, a score of each content may be calculated through the above Equation 3 and Equation 4.

As described above, the methods of the present invention may be implemented as a program and stored in a recording medium (eg, CD-ROM, RAM, floppy disk, hard disk, magneto-optical disk, flash memory, etc.) in a computer-readable form.

In the above, specific preferred embodiments of the present invention have been illustrated and described. However, the present invention is not limited to the above-described embodiments, and various modifications can be made by any person having ordinary skill in the art without departing from the gist of the present invention attached to the claims. .

1 is a block diagram showing the configuration of a preference information management apparatus of a customized broadcasting system according to an embodiment of the present invention.

2 is a flowchart illustrating a procedure of a method for managing preference information of a customized broadcasting system according to another embodiment of the present invention.

Claims (26)

Preference information analyzing means for analyzing preference information of the user based on the viewing pattern of the user; Preference information expansion means for conceptualizing user preference information output from the preference information analyzing means through ontology mapping and extending the user preference information using keyword information of the conceptualized user preference information; And preference content recommendation means for recommending content with high user preference based on the frequency of the preference information extended by the preference information expansion means. The method of claim 1, wherein the preference information analyzing means Pattern collecting means for collecting a viewing pattern for each user; Characteristic prediction means for determining characteristics of a preference pattern and a non-preferred pattern based on the viewing pattern collected by the pattern collecting means; And preference preference means for quantifying the characteristics determined by the characteristic prediction means to determine the rank of the preference information. The method of claim 2, wherein the pattern collecting means The preference information management apparatus of the customized broadcasting system, characterized in that for extracting and collecting the viewing pattern for each user from the signals input from the user equipment. The method of claim 2, wherein the characteristic prediction means After calculating the Mutual Information (MI) value of the viewing pattern collected by the pattern collecting means, it is characterized by the preferred pattern and the non-preferred pattern based on the calculated MI value and the average mutual information (MI _avg ) value of the entire pattern. Device for preference information management of a customized broadcasting system, characterized in that to classify and manage. The method of claim 4, wherein the characteristic prediction means Preference information managing apparatus of the personalized broadcasting system, characterized in that the pattern having the large MI MI _avg and negative value as the preferred pattern, determining a pattern having a MI value less than 0 as a non-preferred pattern. The method of claim 4, wherein the MI value is Preferred information management apparatus for a customized broadcasting system, characterized in that it is calculated by Equation 5.

Is the frequency of pattern A,

Is the probability that pattern A appears in your pattern pool,

Denotes the probability that the pattern A appears while the user watches the preferred program. The method of claim 2, wherein the preference prediction means And scoring preferences of the preference patterns determined by the feature prediction means, and ranking preference information for each user by reflecting the non-favorites of the non-preferred patterns. According to claim 1, wherein said preference information expansion means Ontology mapping means for conceptualizing the user preference information output from the preference information analyzing means through the ontology mapping; And concept vector extraction means for extracting a concept vector by using keyword information of the conceptualized user preference information output from the ontology mapping means. The method of claim 8, wherein the concept vector extraction means A preference information management apparatus of a customized broadcasting system, characterized in that the keyword information of user preference information is divided into an original keyword vector (okv) and a concept keyword vector (ckv) through concept vector extraction. The method of claim 9, wherein the basic keyword vector and the concept keyword vector Device for preference information management of a customized broadcasting system, characterized in that defined by Equation 6. ckv _i = {c ₁ , c ₂ ,, c _k } k = # of concept okv _i = {w ₁ , w ₂ ,, w _{k '} } k' = # of uniq word Here, i denotes the frequency of appearance of each concept k in the preference information of the user i, c denotes the user, and c denotes the frequency of appearance of each basic keyword k 'in the preference information of the user i. The method of claim 1, wherein the preferred content recommending means Customized broadcasting system, characterized by recommending contents predicted by the user by scoring and ranking the similarity (similarity: sim) between the preference information extended by the preference information expansion means and the metadata information of each content. Preferred Information Management Device. The method of claim 11, wherein the preferred content recommending means The vector similarity of each element of the preference information extended by the preference information expansion means and the corresponding metadata information of each content is measured by cosine similarity calculation of Equation 7 and then the average similarity is scored. Device for preference information management of a customized broadcast system, characterized in that.

13. The method of claim 12, wherein said scoring Device for preference information management of a customized broadcasting system, characterized in that performed by the score (Score) of the equation (8).

From here

,

Is the weight of each element vector similarity, and N is the number of elements vectorized in the preference information. Analyzing preference information of the user based on the viewing pattern of the user; Conceptualizing the analyzed user preference information through ontology mapping; Expanding the user preference information by using keyword information of conceptualized user preference information; And recommending content with high user preference based on the extended frequency of the preference information. The method of claim 14, wherein said analyzing step Collecting a viewing pattern for each user; Determining characteristics of a preference pattern and a non-preferred pattern based on the collected viewing pattern; And determining the rank of the preference information by quantifying the determined characteristic. 16. The method of claim 15, wherein determining the characteristic Obtaining mutual information (MI) values of the collected viewing patterns, and classifying and managing characteristics into preference patterns and non-preferred patterns based on average mutual information (MI _avg ) values of all patterns. How to manage preference information of customized broadcasting system. 17. The method of claim 16, wherein the pattern having MI _avg and MI greater than 0 is determined as a preference pattern, and the pattern having MI less than 0 is determined as a non-preferred pattern. The method of claim 16, wherein the MI value is Method for managing preference information of a customized broadcasting system, characterized in that calculated by Equation (9).

Is the frequency of pattern A,

Is the probability that pattern A appears in your pattern pool,

Denotes the probability that the pattern A appears while the user watches the preferred program. The method of claim 15, wherein the ranking step Scoring preferences of the determined preference patterns, and ranking preference information for each user by reflecting the non-preferability of the non-preferred pattern. 15. The method of claim 14, wherein said expanding step And extracting a concept vector by using the keyword information of the conceptualized preference information for each user. The method of claim 20, wherein the concept vector extraction step And managing the keyword information of the user preference information into an original keyword vector (okv) and a concept keyword vector (ckv). . The method of claim 21, wherein the basic keyword vector and the concept keyword vector Method for managing preference information of a customized broadcasting system, characterized in that defined by the following [Equation 10]. ckv _i = {c ₁ , c ₂ ,, c _k } k = # of concept okv _i = {w ₁ , w ₂ ,, w _{k '} } k' = # of uniq word Here, i denotes the frequency of appearance of each concept k in the preference information of the user i, c denotes the user, and c denotes the frequency of appearance of each basic keyword k 'in the preference information of the user i. 15. The method of claim 14, wherein said recommendation step And scoring and ranking the similarity (similarity: sim) between the extended preference information and metadata information of each content to recommend contents predicted by the user. The method of claim 23, wherein the recommendation step The vector similarity of each element of the extended preference information and the corresponding metadata information of each content is measured by cosine similarity calculation of Equation 11, and then performed by scoring the average similarity. Preferred information management method of a customized broadcasting system.

The method of claim 24, wherein the scoring is Method for managing preference information of a customized broadcasting system, characterized in that performed by the score (Score) of [Equation 12].

From here

,

Is the weight of each element vector similarity, and N is the number of elements vectorized in the preference information. Analyzing preference information of the user based on the viewing pattern of the user; Conceptualizing the analyzed user preference information through ontology mapping; Expanding the user preference information by using keyword information of conceptualized user preference information; A computer-readable recording medium having recorded thereon a program for realizing a step of recommending content with high user preference based on the frequency of the expanded preference information.

Images