KR20090033750A - Method and device for recommending content playlist - Google Patents

Abstract

The present invention relates to a method and apparatus for recommending a content playlist applied to a terminal storing content. The present invention provides a plurality of recommended keyword lists through clustering of a sound source listened to by a user. By suggesting a method of automatically generating a playlist in real time through the field search according to the user, the user can be recommended a playlist according to his or her taste without having to create a playlist himself.

Music Content, Playlists, Recommendations, Medley, Music Channels

Description

Method and apparatus for recommending playlists of content {Method and apparatus for recommending playlist of contents}

The present invention relates to a terminal storing content, and more particularly, to a method and apparatus for recommending a content playlist applied to a terminal storing content.

Recently, music contents playback through various mobile devices such as mobile phones, MP3 players, personal digital assistants (PDAs), portable multimedia players (PMPs), and the like, have become popular.

The user may select and play music content himself or herself, but may be recommended music content to play. Here, the method of recommending the music content to be played is generally a form in which a music list that a user may prefer is provided from a large-capacity content server mainly for sound source sales.

However, as the storage space of a personal terminal including a computer and various mobile terminals increases, it is possible to hold many sound sources in the personal terminal instead of a large capacity server. Therefore, in the future, playlist recommendation function for music contents stored in personal terminal is required.

Accordingly, an object of the present invention is to provide a method and apparatus for efficiently recommending a playlist in a terminal storing content.

In order to achieve the above object, the present invention provides a plurality of recommended keyword lists through clustering of a sound source listened to by a user, and automatically generates a playlist in real time through a field search according to a keyword inputted by a user with reference to the same. Suggest how to generate.

More specifically, the present invention provides a method for recommending a playlist for a content stored in a terminal, the method comprising: extracting an attribute from a sound source listened to by a user; And analyzing user tastes through clustering of the extracted attributes and providing a recommendation keyword list based on the analysis result.

In the extracting of the attribute, it is preferable to extract at least one attribute from the sound source listened to by the user using tag information or content-based audio analysis information.

Providing the recommendation keyword list may generate a recommendation keyword list that may be mapped to a user's taste through symbolic data clustering for the extracted attribute, and the symbolic data clustering may use an ontology database. Preferably to be carried out.

The providing of the recommendation keyword list may include configuring a basic cluster having the same value with respect to the extracted attribute; A primary clustering step of hierarchically grouping clusters in a parent-child relationship using an ontology database with respect to the primary cluster; A second clustering step of grouping similar clusters by calculating cluster similarity between extracted attributes; And generating a cluster name for the secondary clustering result as a recommendation keyword list, wherein the cluster name is preferably generated using attribute values of a higher concept in the grouped cluster.

Meanwhile, according to another aspect of the present invention, there is provided a method for recommending a playlist for a content stored in a terminal, the method comprising: providing a recommendation keyword list based on a user taste analysis result; And performing a search for the content according to the user's keyword input using the recommended keyword list, and generating a playlist in real time using the search result.

Here, in the step of providing the recommendation keyword list, it is preferable to display at least one recommendation keyword list based on the user taste analysis result to the user using the ontology database.

The generating of the playlist may include receiving a keyword from a user using the displayed at least one recommendation keyword list; Performing field search based on indexing information of content stored in the terminal using the input keyword; And generating a playlist for the content stored in the terminal in real time using the field search result.

Meanwhile, according to another aspect of the present invention, an object of the present invention is to provide a playlist for a content stored in a terminal, comprising: an attribute extractor extracting at least one attribute from a sound source listened to by a user; A user taste analyzer for analyzing user tastes through clustering of the extracted attributes and providing a list of recommended keywords based on the analysis result; And a search unit that searches for content stored in the terminal according to a user's keyword input using the recommended keyword list, and generates a playlist in real time using the search result. Is achieved.

As described above, according to the present invention, a method and apparatus for efficiently recommending a playlist in a terminal storing content are provided. That is, a list of recommended keywords for a plurality of music channels is provided according to a user's taste with respect to content stored in the terminal. In addition, a playlist for the channel selected by the user using the search engine is dynamically generated in real time. Accordingly, the user can be recommended a playlist that matches his or her taste without having to create the playlist directly. That is, the music DJ or medley function can be provided simply.

Hereinafter, with reference to the accompanying drawings will be described a preferred embodiment of the present invention; In the following description of the present invention, if it is determined that detailed descriptions of related well-known functions or configurations may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. In addition, terms to be described below are terms defined in consideration of functions in the present invention, which may vary according to intention or custom of a user or an operator. Therefore, the definition should be made based on the contents throughout the specification.

1 is an example of the configuration of a content playlist recommendation device according to the present invention.

Referring to FIG. 1, the content playlist recommendation apparatus according to the present invention includes an attribute extractor 10, a user taste analyzer 20, and a searcher 30. In addition, the ontology database 50 and the index database 60 may be further provided.

The feature extraction module 10 extracts a feature from the sound source listened to by the user. The attribute extracting unit 10 uses mood 3, theme, artist, genre, album, and the like by using ID 3 tag information or content-based audio analysis information from a sound source listened to by a user. Attributes according to classification criteria such as album, year, etc. are extracted.

The user preference analysis module 20 analyzes the user's taste through clustering of the attributes extracted by the attribute extraction unit 10 and recommended the keyword list based on the analysis result. keyword-list). That is, the keyword extractor 10 generates a recommendation keyword list that can be mapped to a user's taste through symbolic data clustering for the attribute extracted by the attribute extractor 10. In this case, it is preferable to use an ontology database 50 having a hierarchical structure for efficient clustering.

In addition, the recommendation keyword list indicates a playlist title list including at least one attribute reflecting a user's taste. For example, it may be provided in the form of " classic / tranquil music " combining the user's favorite genre and mood attributes, or " Lee Hyo Lee / Dance music " combining the user's favorite artist and genre attributes.

The retrieval module 30 searches for the content stored in the terminal using the input keyword when the user inputs a keyword using the recommendation keyword list provided by the user taste analyzer 20. The playlist is used to generate a playlist in real time. In this case, it is preferable that the inverted index information of the metadata of the music content stored in the terminal is pre-built in the indexing database 60 for the real-time search.

2 is a flowchart illustrating an example of a content playlist recommendation method according to the present invention.

Referring to FIG. 2, first, attributes according to classification criteria such as mood, theme, artist, genre, album, year, etc., are read from a sound source listened to by a user. Extract (S500). Clustering the extracted attributes using the ontology database 50 is performed to analyze the user's taste and generate a recommendation keyword list based on the analysis result and provide it to the user (S600). Now, when a user inputs a keyword using the provided recommended keyword list, the user searches for the content stored in the terminal using the indexing database, and generates a playlist in real time using the search result (S700).

Hereinafter, each step of the content playlist recommendation method according to the present invention shown in FIG. 2 will be described in detail.

3 is a flowchart illustrating an example of a method S500 of extracting an attribute from a sound source.

Referring to FIG. 3, the attribute extractor 10 receives a music list listened to by a user as an input and extracts music attributes to be used for user taste analysis. In the present invention, not only ID3 tag information is used (S510), but also mood and theme classification information is further used (S520) as a content-based audio analysis method.

At this time, the mood classification is classified at high speed by using timbre and tempo information in the compressed region through MP3 partial decoding. For a specific mood classification method, refer to "High speed music mood classification method and apparatus" of Korean Patent Application No. 2005-107957 filed and registered by the applicant of November 11, 2005.

In addition, theme classification extracts ID3 tag information, and then extracts and classifies attributes necessary for subject classification from the title and lyrics of the song. For a specific method for subject classification, refer to Korean Patent Application No. 2006-17768 filed by the present applicant on February 23, 2006 and published on August 28, 2006, entitled "Method and apparatus for classifying music subject using music title". In addition, a plurality of music-related features are used, including an artist, a genre, an album, a year, and the like obtained from ID3 tag extraction. The illustrated attributes are merely exemplary representative attributes highly related to user taste, but are not limited thereto. Various attributes may be further used.

On the other hand, an object of the present invention is to automatically generate a playlist in accordance with the user's taste for the music content in the terminal possessed by the user. To this end, a process of analyzing a user's taste is necessary.

4 is a flowchart illustrating an example of a taste analysis method for user listening music. Referring to FIG. 4, a detailed process of generating, by the user taste analyzer 20, a recommended keyword list that can be mapped to user taste through symbolic data clustering for the extracted attribute is illustrated. The user taste analysis may be specifically composed of the following four steps.

1) Basic cluster configuration steps ( S610 )

First, the user taste analyzer 20 configures a basic cluster having the same value for each attribute extracted by the attribute extractor 10. Referring to FIG. 5, sound sources having the same attribute value of Kim Gwang-seok in attribute 1 (artist), that is, songs 1, 2, and 3 may be configured into one basic cluster, and sound sources having the same attribute value of Lee Hyo-ri. , Namely, the song names 3, ..., m can be configured as one basic cluster. In the case of attribute 2 (genre), sound sources having the same attribute value such as dance, ballad, blues, etc. may be configured as basic clusters, respectively.

2) First clustering step ( S620 )

Next, the user taste analyzer 20 hierarchically configures clusters by grouping clusters having upper and lower relations using an ontology for the basic cluster. At this time, in the integrated cluster, an attribute value of a higher concept is determined as a centroid of the cluster, and is used in a cluster naming step to be described later.

It is preferable that the cluster composed of the sound sources MP3 below the threshold for the cluster in the finally generated attribute is deleted. In this case, the reference value may be set as a reference that can be viewed as an abnormal distribution rather than an average distribution based on the average number of documents for the attribute.

For example, if the total number of grains is n and the number of clusters in the current attribute is m, the average cluster size is n / m. Therefore, if the size of the cluster is larger than the reference value n / m, it can be interpreted that the user prefers the frequency of occurrence above the average distribution. If this is expressed as an expression, it is as follows.

size of Cluster> weight * (n / m)

Here, the weight is greater than 1 and can be determined experimentally depending on the application.

As a simple example, suppose you have 100 sound sources (MP3), and the genre distribution includes dance, ballad, jazz, and rock, and consists of 12, 19, 51, and 18 songs, respectively. In this case, the average occurrence frequency is 25 pieces, which is 100/4. Therefore, it can be judged that the jazz which is 25 or more pieces is a genre which a user especially likes.

3) Second clustering stage ( S630 )

Next, the user taste analyzing unit 20 groups similar clusters by calculating cluster similarity between attributes. At this time, the similarity between the groups is calculated using the distribution of sound sources (MP3 files) constituting the cluster. In other words, the more documents that are duplicated through the AND operation, the higher the similarity between the two groups. If this is expressed as an expression, it is as follows.

Cluster A = [dA1, dA2, ..., dAn], Cluster B = [dB1, dB2, ..., dBn] Similarity = | A∩B | / | A∪B |

Clusters A and B may be represented by a vector of sound sources included in each cluster. dA1 means whether a specific sound source d1 is included in the cluster A. Similarity (Similarity) has a value between 0 and 1, and the closer the similarity value is to 1, the higher the similarity between the two clusters. Therefore, if the similarity is higher than a certain threshold, the two clusters will be merged. In this case, the reference value may be determined experimentally depending on the application.

For example, 25 songs such as {1, 3, 20, 46, 60, ...} are included in the "Ballard" cluster among genre attributes, and {1, 3, 20, 46, 60, ...} are included in the "Lee Moon Se" cluster among artist properties. , 3, 46, 70, ...}. Suppose that a common song (| A∩B |) is 10 songs. In this case, the similarity is 10/30 = 0.33. If the obtained similarity value is larger than the reference value, the two clusters are combined. In the cluster naming step described later, a cluster name of the form “Lee Mun Se / Ballard” is generated.

4) cluster Naming ( naming ) step( S640 )

Finally, the user taste analyzer 20 generates a name for the cluster by using a central word (Centroid) for the generated cluster. The generated name corresponds to the name of the recommended playlist (also referred to as 'recommended keyword list'). The recommended keyword list is provided to the user in the form of a music channel or a medley.

Referring to FIG. 5, a process of obtaining a recommendation keyword list (playlist title) of “Lee Moon Se / Ballard” or “Kim Kwang Seok / Sad Music” through a first clustering process within an attribute and a second clustering process between the attributes is illustrated. It is.

6 is a flowchart illustrating an example of a method for automatically generating a playlist by keyword search.

Referring to FIG. 6, when the user selects one of a plurality of recommended keyword lists (playlist titles), the search unit 30 uses the selected recommended keyword list to search for a sound source (MP3 file) stored in the terminal. Perform a field search. For example, if the selected recommendation keyword list is "Lee Moon-se / Balad", the field search for "Artist: Lee Moon-se" and "Genre: Ballard" is performed, and then an AND operation on the search result is used to determine the playlist.

Meanwhile, the search unit 30 may generate a final playlist by ranking the searched sound sources MP3. The ranking process may be configured to rank the most popular music or the latest music according to the user's selection.

To do this, it must have indexing information about music stored in the terminal. In the present invention, the indexing of ID3 tag information, mood, and subject classification information is performed using a search engine, and then an indexing database 60 is constructed and used in an inverted index form. The construction of the indexing database is preferably performed offline.

Accordingly, according to the present invention, a method and apparatus for efficiently recommending a playlist in a terminal storing content are provided. That is, a list of recommended keywords for a plurality of music channels is provided according to a user's taste with respect to content stored in the terminal. In addition, a playlist for the channel selected by the user using the search engine is dynamically generated in real time. Accordingly, the user can be recommended a playlist that matches his or her taste without having to create the playlist directly. That is, the music DJ or medley function can be provided simply.

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

1 is an example of the configuration of a content playlist recommendation device according to the present invention;

2 is a flowchart illustrating an example of a content playlist recommendation method according to the present invention;

3 is a flowchart illustrating an example of a method of extracting an attribute from a sound source;

4 is a flowchart illustrating an example of a taste analysis method for user listening music;

5 is a reference diagram illustrating an example of a clustering method;

6 is a flowchart illustrating an example of a method for automatically generating a playlist by keyword search.

Claims (12)

In the method for recommending a playlist (playlist) for the content stored in the terminal, Extracting an attribute from a sound source listened to by a user; And Analyzing user taste through clustering of the extracted attributes and providing a recommendation keyword list based on the analysis result. The method of claim 1, wherein the extracting of the attribute comprises: And at least one attribute is extracted from the sound source listened to by the user using tag information or content-based audio analysis information. The method of claim 1, wherein providing the recommendation keyword list comprises: Generating a recommendation keyword list that can be mapped to a user's taste through symbolic data clustering for the extracted attribute; And the symbolic data clustering is performed using an ontology database. The method of claim 1, wherein providing the recommendation keyword list comprises: Constructing a basic cluster having the same value with respect to the extracted attribute; A primary clustering step of hierarchically grouping clusters having a parent-child relationship using an ontology database with respect to the primary cluster; A second clustering step of grouping similar clusters by calculating cluster similarity between the extracted attributes; And Generating a cluster name for the secondary clustering result and providing the recommendation keyword list as a recommendation keyword list. The method of claim 4, wherein the cluster name is Content playlist recommendation method characterized in that it is generated using the attribute value of the upper concept in the grouped cluster. In the method for recommending a playlist (playlist) for the content stored in the terminal, Providing a recommendation keyword list based on a user taste analysis result; And Performing a search for the content according to a user's keyword input using the recommended keyword list, and generating a playlist in real time using a search result. The method of claim 6, wherein providing the recommendation keyword list comprises: And displaying at least one recommendation keyword list based on the user taste analysis result to the user using an ontology database. The method of claim 7, wherein generating the playlist comprises: Receiving a keyword from a user using the displayed at least one recommended keyword list; Performing field search based on indexing information of content stored in the terminal using the input keyword; And Generating a playlist for the content stored in the terminal in real time using the field search result. In the device for recommending a playlist (playlist) for the content stored in the terminal, An attribute extraction unit for extracting at least one attribute from a sound source listened to by a user; A user taste analyzing unit configured to analyze user tastes through clustering of the extracted attributes and to provide a list of recommended keywords based on the analysis result; And And a search unit for searching for content stored in the terminal according to a user's keyword input using the recommended keyword list, and generating a playlist in real time using a search result. The method of claim 9, wherein the attribute extraction unit, And at least one attribute is extracted from the sound source listened to by the user using tag information or content-based audio analysis information. The method of claim 9, wherein the user taste analysis unit, Generating a recommendation keyword list that can be mapped to a user's taste through symbolic data clustering for the extracted attribute; And the symbolic data clustering is performed by using an ontology database. The method of claim 9, wherein the search unit, Receives a keyword from a user using the provided recommendation keyword list, performs a field search based on indexing information of the content stored in the terminal using the input keyword, and uses the field search result in real time. Content playlist recommendation apparatus for generating a playlist for the content stored in the terminal.

Images