KR20110075259A - An online music service device and method for generating a music playlist reflecting user preferences and ratings - Google Patents

Abstract

The present invention relates to an online music service apparatus and method. To this end, the present invention calculates the score for the song to reflect the user preferences and ratings, and stores the score in the user library, selects the top N of the score value based on the score value when creating a playlist, collaborating filtering (CF: collaborating) After extracting the similar songs related to the favorite songs by using the filtering), the similar songs are created as a playlist and recommended to the user. At this time, when creating a playlist, the playlist is configured by reflecting the user's evaluation of the song in real time. Accordingly, the present invention does not require additional metadata and recommends songs not previously thought by the user since the playlist is generated by reflecting the user's viewing history information about the song when the playlist is preferred by the user. I can receive it. In addition, it is possible to recommend a more accurate playlist by reflecting the user rating of the song in the playlist in real time.

Description

ONLINE MUSIC SERVICE APPARATUS FOR GENERATING MUSIC PLAYLIST CONSIDERING USER'S PREFERENCES AND RATINGS AND METHOD THEREOF}

The present invention relates to an online music service apparatus and method for generating a music playlist reflecting user preferences and ratings, and more particularly, to selecting a user's favorite song based on user's listening history information and collaborating filtering (CF). An online music service apparatus and method for generating a music playlist reflecting user preferences and ratings that can generate a music playlist by extracting similar songs related to a selected favorite song using a filtering technique.

The online music service is one of various content services, and subscribers (for example, mobile terminals, PCs, etc.) of various types of music, for example, songs, pop songs, jazz, classics, and various melodies, can be accessed from a music service server through a network network. It is a service provided by.

The subscriber terminal accesses a specific site through the Internet to download desired music or to receive a streaming service.

On the other hand, the online music service provides a service that recommends to the user by creating a playlist of the user's favorite songs. As a method of providing such a song recommendation service, three methods have been proposed as follows.

The first method is to identify the user's favorite songs in the music service server and create a playlist from them. The number of regenerations and the presence of skips are used to calculate the song preference. When a playlist is created and recommended in this way, the user is recommended based on songs that are already familiar. In other words, the user cannot feel a new feeling for the recommended song.

In the second method, when a user inputs a seed song, the user extracts songs similar to the seed song by using metadata and creates a playlist from the extracted songs. In this case, the metadata may include chords, melodies, musical instrument configurations, singers, albums, and genres. However, the second method has a problem that it takes a lot of time and money to have a sufficient amount of metadata necessary for similar song extraction.

In the third method, when a user inputs a seed song, similar songs are extracted using audio signal information, and a playlist is created from the extracted songs. The audio signal information includes spectrum, fluctuation patterns, and the like. This third method has a problem that the similarity of the audio signal does not match that of the user.

An object of the present invention is to provide an online music service apparatus and method that can create a music playlist that can create a new and fresh feeling for a user who encounters a recommendation song.

Another object of the present invention is to provide an online music service apparatus and method for generating a music playlist without using metadata when generating a music playlist for recommendation.

Another object of the present invention is to provide an online music service apparatus and method for accurately extracting songs having similarities to a user's favorite song when creating a music playlist for recommendation.

In order to achieve the above, the online music service apparatus of the present invention includes a user library for managing user viewing history information about songs that have been streamed, downloaded, or evaluated, and a score calculation for calculating a score representing a user's preference for the song. And a preference song extracting unit for extracting the top N scores based on a score value from the user library as a favorite song, and extracting a recommendation song for extracting similar songs associated with the extracted favorite songs using a collaborative filtering (CF) method. And a playlist generator for generating the extracted similar songs as a playlist.

In addition, the online music service method of the present invention comprises the process of having a user library for managing the listening history information for the songs streamed, downloaded or evaluated by the user, the recent score of the song, the evaluation of the user's songs, songs A score is calculated using the number of times of play and the percentage of songs played, a process of extracting favorite songs from the user library using the calculated score value, and the extracted preference using a collaborative filtering (CF) method. And extracting the similar songs associated with the songs, and generating the extracted similar songs as a playlist.

In addition, the online music service method of the present invention includes the process of providing a user library for managing the viewing history information for the songs streamed, downloaded or evaluated by the user, the recent score of the song, the evaluation of the user's song, playback of the song Calculating scores using the number of times of recovery and song reproduction, extracting a genre associated with a genre input by a user by using a collaborative filtering (CF) method, and scores among songs belonging to the extracted related genre It characterized in that it comprises a process of generating a high N songs as a playlist.

In addition, the online music service method of the present invention includes the process of providing a user library for managing the viewing history information for the songs streamed, downloaded or evaluated by the user, the recent score of the song, the evaluation of the user's song, playback of the song A process of calculating scores using the number of times and the percentage of song reproduction, extracting an artist associated with an artist input by a user by using a collaborative filtering method, and a score among songs belonging to the extracted related artists It characterized in that it comprises a process of generating a high N songs as a playlist.

According to the present invention, as songs composed of playlists are composed of similar songs extracted using CF techniques based on the preference songs reflecting user's listening history information, recommending songs that the user may like but did not know before There is an effect to receive.

In addition, when creating a user's favorite music playlist, since the playlist is generated by reflecting the user's listening history information on the song, it is not necessary to build metadata for extracting similar songs since no additional metadata is required.

In addition, by reflecting the user rating of the song in the music playlist in real time, it is effective to recommend the accurate music playlist in real time.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings. Specific details described below are provided to aid the overall understanding of the present invention. In describing the present invention, detailed descriptions of related known functions or configurations will be omitted when it is determined that the gist of the present invention may be unnecessarily obscured.

The present invention selects the user's favorite songs by using the user's music listening history information, and extracts similar songs similar to the favorite songs by using a cooperative filtering (CF) technique to generate a playlist.

Here, the cooperative filtering technique refers to a technique used for a service that identifies customers with similar preferences or tastes and recommends related products among customers with similar preferences or tastes.

That is, in the present invention, in order to extract similar songs using the cooperative filtering technique, the listening history information of each user is stored and based on the listening history information of another user who prefers the given song when the user's favorite song is given. Similar songs will be extracted. In this way, a playlist is generated from the extracted similar songs.

In this way, since the music playlist is generated through the user's listening history information, no additional metadata is required, and new and fresh songs can be provided when compared to similar songs recognized by the user.

1 is a block diagram of a system including a music service server for generating a music playlist according to the present invention.

Referring to FIG. 1, the mobile terminal 10 is connected to the music service server 120 through the mobile communication network 100 and the Internet network 110, and the PC 20 is a music service server through the Internet network 110. A music service may be provided by accessing 120.

The music service server 120 according to the present invention generates a music playlist by a preference based playlist generation method, a genre based playlist generation method, and an artist based playlist generation method. Here, the preference based playlist generation method is to create a playlist using the user's past viewing history information, and the genre and artist based method receives a seed genre and artist from the user together with the viewing history information and the playlist. To generate.

The internal configuration of the music service server 120 may be illustrated as shown in FIG. 2.

Referring to FIG. 2, the music service server 120 according to the present invention may include a score calculator 200, a user library 201, a favorite song extractor 202, a recommended song extractor 203, The playlist generating unit 204 and the recommendation song providing unit 205 are included.

The user library 201 manages song listening history information for each user. The listening history information is generated in response to the song streamed, downloaded or evaluated by the user, and the listening history information for each song includes various items as shown in FIG. 3.

Referring to Figure 3, it shows the viewing history information for any one user.

Listening history information includes song ID, score, genre to which the song belongs, last access time for the song, user rating, song count, and recent listening. Contains the percentage playback percentage item.

Here, the score is a score of a song, which indicates a user's preference for the song, and is expressed as a number between 0 and 100. The score is recalculated every time the user plays streaming or downloads the song. Such a score calculation method is explained in full detail below.

The rating of a user's song is expressed as a number between 0-10 and the score that the user rated for the song is indicated. At this time, the larger the evaluation value, the more the user likes the song. If the song is a user's favorite song for a given song, the user may set the song as the top favorite song, and if the song is set as the top favorite song, the evaluation score may be 10. In addition, a ban for banning the playlist may be set, and when the ban is set, the evaluation score may be zero.

In addition, the number of playbacks of the song is indicated by the number of times that has been played so far, and may be displayed by a predetermined number of times of playback in the case of download or free download. For example, a download may be regarded as 10 playbacks, and a free download (downloading a part of a song for free) may be regarded as 8 playbacks. Of course, the number of playbacks to be considered at the time of download or pre-download can be set differently.

Percentage of song playback is expressed as a number between 0 and 100, and refers to the percentage of actual playback of the total length of the song.

The user library 201 manages only songs that the user has listened to or rated. When a user streams, downloads, or evaluates a song that is not in the user library 201, the song is added to the user library 201, and listening history information on the added song is also generated. If the user does not respond to the predetermined song during the preset period, the song is removed from the user library 201.

The music service server 120 may limit the music managed per user in consideration of the storage or management cost of the user library 201. When the number of songs to be managed is limited to 1000, the existing song information is deleted using the Least Score First (LSF) replacement technique. In other words, the score values are deleted in descending order.

The method for calculating the score may be calculated in the order of Equations 1 to 3 below.

score1 = (5 * rating) + (prevscore / 2)

score2 = score1 + (100-score1) * (1-Math.exp (-playcount / 100))

score3 = score2 * (0.9 + 0.1 * percentage / 100)

Referring to Equation 1 to Equation 3, a score calculation method is described first. A score 1 is calculated using a user rating and a prevscore.

Next, the score 2 is calculated using Equation 2 by reflecting the playcount of the song. The larger the number of song regenerations, the larger the score, but the upper limit of the score is set to 100. Therefore, an exponential decay function is used to limit the increase of the score value. As the number of song reproduction increases, that is, the listening to the song increases, score 2 of the song gradually increases from score 1 to 100.

Lastly, score 3 is calculated by using Equation 3 to reflect the percentage of play during recent viewing. Here, the score 3 becomes a score finally stored in the viewing history information. According to Equation 3, the score 3 cannot be greater than the score 2, and the reproduction percentage is set so that only 10% of the score 2 (score2) is affected. That is, when the percentage of reproduction is 100%, the score 3 is equal to the value of score 2, and when the percentage of reproduction falls below 100%, it becomes smaller than the value of score 2. Here, the ratio of the effect of the reproduction percentage on the value of the score 2 can be changed.

Next, a description will be given of a method in which the music service server 120 generates a playlist while the listening history information of each subscriber is managed in the user library 201 as described above.

Hereinafter, a method of generating a preference based playlist, a method of generating a genre based playlist, and a method of generating an artist based playlist will be described with reference to FIGS. 4 to 6.

First, a method of generating a playlist based on preferences will be described.

The preference-based playlist generation method does not need to set predetermined information separately because the playlist is generated by using the user's viewing history information. That is, the preference-based playlist generation operation selects a user's favorite song based on the viewing history information and creates a playlist with songs associated with the favorite song. At this time, the songs associated with the preferred song are extracted through a cooperative filtering (CF) technique.

This will be described in detail with reference to FIGS. 2 and 4.

The favorite song extractor 202 extracts the preferred songs PrefSongs based on the listening history information of the user library 201 (S400).

At this time, the top N pieces are selected based on the score value as the preferred song.

Thereafter, the recommended song extractor 203 extracts songs associated with the selected preferred songs using the CF technique (S402). That is, related songs may be extracted from the user library 201 of another user who prefers the selected favorite song based on the viewing history information of the other user.

The playlist generation unit 204 generates the extracted recommended songs (CandiSongs) as a playlist. At this time, the playlist generation unit 204 reflects the user's song rating in the playlist generation in real time (S404). To this end, the playlist generation unit 204 updates the playlist whenever an evaluation occurs.

If the user sets the best love for a given song, the similar songs associated with the song are extracted by a cooperative filtering technique, and the extracted songs are inserted at the front of the playlist. By doing this, the user can immediately listen to a song similar to his or her favorite song.

At this time, if the songs extracted by the cooperative filtering are set by the user as a ban (ban) and the songs associated with them, they are excluded from the playlist.

For this purpose, a ban set can be managed separately. The prohibited song set is composed of songs selected by the user as prohibited songs and songs associated with them. The addition of the song to the forbidden song set is performed when the user enters an evaluation score of 0 for the predetermined song.

When the user sets a prohibited song for a predetermined song, similar songs associated with the song are extracted by the CF technique, and the similar songs are deleted from the playlist. In addition, the selected prohibited songs and similar songs are added to the prohibited set.

Thereafter, the recommended song reproducing unit 205 recommends songs in the order of the playlist. At this time, the recommendation order may be random, or may be recommended in order (S406). In this case, in order to prevent the same song from being recommended to the user continuously, the recommended song is not included in the recently played 10 songs.

Next, the genre-based playlist generation method will be described.

In the genre-based playlist generation method, when a user selects a favorite genre, a related genre is extracted by using a CF technique, and a playlist is created with songs belonging to the related genre. At this time, the genre is generally described by the sound source provider.

This will be described in detail with reference to FIG. 5.

When one or more genres preferred by the user are input, genres associated with these genres are extracted using CF techniques (S500 and S502).

In operation S504, N pieces of music having a high score are extracted among the pieces of music belonging to the related genres. At this time, songs belonging to the related genres are searched in the user library 201.

The playlists are generated by reflecting the song ratings of the extracted songs as described above (S505). Here, the operation of reflecting the user's song rating when generating the playlist is performed so that the user's song rating is reflected in the playlist in real time as described in operation S404 of FIG. 4.

Then, the songs are recommended in the order of the playlist. At this time, the recommendation order may be random, or may be recommended in order (S506). At this time, in step S506, the song is recommended to the user by the operation described with reference to S406 of FIG. 4.

Finally, we will look at how to create an artist-based playlist.

When the user selects his or her favorite artist, a CF technique is used to extract the associated artist and create a playlist of songs belonging to the associated artist.

This will be described in detail with reference to FIG. 6.

If a favorite artist is input by the user, the associated artist associated with the artist is extracted using the CF (S600 and S602).

Then, N pieces of songs having a high score are extracted among the songs belonging to the associated artist (S604). At this time, songs belonging to the associated artist are searched in the user library 201.

The playlists are generated by reflecting the song ratings of the extracted songs as described above (S605). Here, the operation of reflecting the user's song rating when generating the playlist is performed so that the user's song rating is reflected in the playlist in real time as described in operation S404 of FIG. 4.

Then, the songs are recommended in the order of the playlist. At this time, the recommendation order may be random, or may be recommended in sequence (S606). In step S606, the music is recommended to the user by the operation described with reference to S406 of FIG.

Although embodiments of the present invention have been described above with reference to the accompanying drawings, those skilled in the art to which the present invention pertains may implement the present invention in other specific forms without changing the technical spirit or essential features thereof. I can understand that.

Therefore, since the embodiments described above are provided to completely inform the scope of the invention to those skilled in the art, it should be understood that they are exemplary in all respects and not limited. The invention is only defined by the claims.

In the conventional online music service, the user may compose songs that the user prefers when composing the recommended play music list, or extract songs that are similar to the favorite songs by using metadata or audio signal information. By selecting the, and using the CF technique to extract the similar songs associated with the favorite songs and create a playlist consisting of the extracted similar songs to recommend the user to the user, the user is his favorite songs, but not more frustrated, Since it has the advantage of recommending songs that give a fresh feeling, it can be widely applied to the field of providing online music service.

1 is a system configuration diagram including a music service server according to the present invention;

2 is an internal configuration diagram of a music service server according to an embodiment of the present invention;

FIG. 3 is an exemplary diagram illustrating items representing a music listening history for each user managed by a user library of a music service server according to an exemplary embodiment of the present invention; FIG.

4 is a flowchart illustrating a process of creating a preference based playlist according to an embodiment of the present invention;

5 is a flowchart illustrating a process of generating a genre based playlist according to an embodiment of the present invention;

6 is a flowchart illustrating a process of creating an artist-based playlist according to an embodiment of the present invention.

<Explanation of symbols for the main parts of the drawings>

10: mobile terminal 20: PC

100: mobile communication network 110: Internet network

120: music service server 200: score calculation unit

201: user library 202: favorite song extraction unit

203: recommended song extractor 204: playlist generation unit

205: recommended song provider

Claims (18)

A user library that manages user history information for songs that have been streamed, downloaded or rated; A score calculator for calculating a score indicating a user's preference for the song; A preference song extracting unit configured to extract the top N scores based on score values from the user library as a preference song; A recommendation song extraction unit for extracting similar songs associated with the extracted preferred songs using a collaborative filtering (CF) method; And a playlist generator for generating the extracted similar songs as a playlist. The method of claim 1, On-line music service device further comprises a recommendation song providing unit for recommending to the user in the order or randomly similar songs included in the generated playlist. The method of claim 1, wherein the user library, Songid, score, genre to which the song belongs, last access time to the song, user ratings, and playcount And managing at least one item of a percentage of playing at a last listening time. The method of claim 1 or 3, wherein the score calculation unit, An online music service apparatus, comprising: calculating a score using a recent score of a song, a user rating of the song, a number of times of playing the song, and a percent of the song playing. The method of claim 1, wherein the playlist generation unit, On-line music service apparatus, characterized in that the user rating for the song to the playlist in real time to generate it. The method of claim 1, wherein the playlist generation unit, If the user sets the best love for a given song, the associated similar songs of the song are extracted by cooperative filtering, and the extracted songs are inserted into the front of the playlist. Online music service device. The method of claim 1, wherein the playlist generation unit, When the user sets a ban for a predetermined song, the similar songs associated with the songs set as the banned songs are extracted by cooperative filtering, and the songs set as the banned songs and the similar songs are excluded from the playlist. An online music service device, characterized in that. Providing a user library for managing viewing history information about songs streamed, downloaded or evaluated by the user; Calculating a score using the latest score of the song, the user's rating of the song, the number of songs played, and the percentage of the song played; Extracting favorite songs from the user library using the calculated score value; Extracting similar songs associated with the extracted preferred songs using a collaborative filtering (CF) method; On-line music service method comprising the step of generating the extracted similar songs as a playlist. The method of claim 8, And recommending to the user in order or randomly the songs included in the generated playlist. The method of claim 8, wherein the playlist creation process comprises: Online music service method characterized in that the process of generating by reflecting the user rating for the song in real time when generating the playlist. The method of claim 8, wherein the playlist creation process comprises: If the user sets the best love for a given song, it is the process of extracting the similar songs associated with the song by cooperative filtering and inserting the extracted songs to the front of the playlist. An online music service method characterized by the above. The method of claim 8, wherein the playlist creation process comprises: When the user sets a ban for a predetermined song, the similar songs associated with the songs set as the banned songs are extracted by cooperative filtering, and the songs set as the banned songs and the similar songs are excluded from the playlist. Online music service method characterized in that the process. Providing a user library for managing viewing history information about songs streamed, downloaded or evaluated by the user; Calculating a score using the latest score of the song, the user's rating of the song, the number of songs played, and the percentage of the song played; Extracting a genre associated with a genre input by a user using a collaborative filtering method; And generating a playlist having N scores having a high score among the songs belonging to the extracted related genres. Providing a user library for managing viewing history information about songs streamed, downloaded or evaluated by the user; Calculating a score using the latest score of the song, the user's rating of the song, the number of songs played, and the percentage of the song played; Extracting an artist associated with an artist input by a user using a collaborative filtering method; And generating N playlists having a high score among the songs belonging to the extracted associated artist as a playlist. The method according to claim 13 or 14, And recommending to the user in order or randomly the songs included in the generated playlist. 15. The method of claim 13 or 14, wherein the playlist creation process, Online music service method characterized in that the process of generating by reflecting the user rating for the song in real time when generating the playlist. 15. The method of claim 13 or 14, wherein the playlist creation process, If the user sets the best love for a given song, it is a process of extracting the related similar songs of the song by cooperative filtering and inserting the extracted songs to the front of the playlist. An online music service method characterized by the above. 15. The method of claim 13 or 14, wherein the playlist creation process, When the user sets a ban for a predetermined song, the similar songs associated with the songs set as the banned songs are extracted by cooperative filtering, and the songs set as the banned songs and the similar songs are excluded from the playlist. Online music service method characterized in that the process.

Images