KR20200043687A - Providing Method for music based on personalization and service device supporting the same - Google Patents

Abstract

The present invention provides an electronic device capable of improving quality of a listening log used for music recommendation. The electronic device comprises: a communication circuit forming a communication channel with a terminal device; a memory storing at least one music item; and a processor functionally connected to the communication circuit and the memory. Moreover, the processor may be set to collect a music item usage history of the terminal device when the terminal device and the communication channel are formed, generate a music recommendation list related to the terminal device based on the music item usage history, classify music items included in the music recommendation list into a plurality of groups, name group information using music tag data for the classified groups, and store groups in which the group information is named as a personalized music recommendation list.

Description

Providing Method for music based on personalization and service device supporting the same}

The present invention relates to a music recommendation, and more particularly, to a personalized music recommendation method and apparatus that allows the characteristics of individual music recommendation lists to be consistent.

With the generalization of mobile electronic devices, listening to music becomes easier. Accordingly, many users are using the music service. In this regard, operators operating sound source services are developing various music recommendation services. Here, the music recommendation is based on predicting music similar to the music the user has previously heard. This method calculates the probability that the user will listen to the music based on the similarity of content or similarity of the listening history pattern. In general, although the accuracy of the algorithm has the greatest influence on the quality of the recommendation result, in the music recommendation, the manner in which the recommendation result is provided greatly affects the user's perceived quality.

Meanwhile, the user listens to several pieces of music at once. In the sound source service, the user listens to at least 5 to 10 songs at a time. This phenomenon reflects that the user selects and listens to several songs from the recommended music list. In addition, users listen to music in various contexts. There are contexts such as before going to bed, going to work, and studying. Basically, you listen to music that suits your taste, but you listen to different types of music depending on the context. This phenomenon reflects the mix of various types of music in the user's listening log. Therefore, the recommendation result based on the listening log is also composed of various types of music.

The conventional method of providing a recommendation result is a concept of 'Music you may like', and provides music selected in a recommendation algorithm at once. However, this method does not take into account the consumption method of the above-mentioned music contents, and the heterogeneous musics with different atmospheres are provided as one playlist. Since the user listens to several pieces of music at once using functions such as full play, the likelihood of listening to inconsistent music increases. In other words, the perceived quality deterioration of the recommended result occurs separately from the algorithm accuracy.

In order to overcome the above-mentioned problems, research on a context-based recommendation algorithm has been conducted, but directly extracting recommendation results for each context from a noisy listening log does not guarantee quality in a real service. In addition, there is an existing context-specific music curation method in which a music expert who specializes in music selects music suitable for each situation and provides it as a playlist, but this method ensures consistency and explanation of recommendation results, but works manually. It is impossible to provide personalization to numerous users. Therefore, in order to obtain the results desired by the user, a separate search process is required.

Korean Patent Publication No. 10-2018-0081856 (Name: Content analysis music recommendation system, Publication date: 2018.07.18.)

In order to solve the above-mentioned problems, the present invention provides the user's favorite music obtained through the recommendation algorithm in multiple playlists, by dividing them into multiple playlists, so as to maintain the consistency of the recommendation results and improve the user's perceived quality of experience. It provides a personalized music recommendation method and a service device that supports the personalized music recommendation method so that the list can have a consistent theme.

In addition, the present invention provides a divided playlist with a consistent theme, applies labeling to the divided playlist, so that the user can easily recognize its characteristics, and recommends music to each user based on an automated method A personalized music recommendation method capable of providing a service and a service device supporting the same are provided.

However, the object of the present invention is not limited to the above object, and other objects not mentioned will be clearly understood from the following description.

The service device of the present invention for achieving the above object includes a communication circuit forming a communication channel with the terminal device, a memory storing at least one music item, a communication circuit and a processor functionally connected to the memory. You can. When a communication channel is established with the terminal device, the processor collects a history of music item use of the terminal device, generates a music recommendation list related to the terminal device based on the music item use history, and includes the music recommendation list in the music recommendation list It may be configured to classify the classified music items into a plurality of clusters, name cluster information using music tag data for the classified clusters, and store the clusters named with the cluster information as a personalized music recommendation list. .

Here, the processor may be configured to generate an item vector corresponding to a numeric vector based on the history of using the music item, and to generate the music recommendation list based on the similarity with the item vector.

Also, the processor may be configured to classify the music items into a plurality of clusters based on a designated clustering algorithm.

Meanwhile, the processor may be configured to collect music tag data for the music items from an external server device.

In addition, the processor may be configured to score music tag data for the clusters based on a text mining technique and to write music tag data within a designated rank as the cluster information.

On the other hand, if the history of using the music item does not satisfy a specified condition, the processor may recommend a music recommendation list prepared based on a preset rule or a music recommendation list prepared based on a history of using a plurality of users' music items to the terminal device. It can be set to provide.

Alternatively, when the personalized music recommendation list is stored in a memory by satisfying a condition in which a history of use of a music item of the terminal device satisfies a specified condition, the processor sets the personalized music recommendation list to the terminal device when the terminal device is accessed Can be.

In the personalized music recommendation method according to an embodiment of the present invention, a service device establishes a communication channel with a terminal device, and checks whether a personalized music recommendation list generated based on a history of using music items of the terminal device exists. Step, if the personalized music recommendation list does not exist, checking whether the music item usage history satisfies a specified condition, and when the music item usage history satisfies a specified condition, based on the music item usage history Generating a music recommendation list related to the terminal device, classifying music items included in the music recommendation list into a plurality of clusters, and naming cluster information using music tag data for the classified clusters , Personalized music recommendation list of clusters named as the cluster information It may include the step of providing to the terminal device.

In addition, the personalized music recommendation method may include a music recommendation list prepared based on a preset rule or a music recommendation list prepared based on a history of using a plurality of users' music items when the history of using the music item does not satisfy a specified condition. It may further include the step of providing to the terminal device.

In particular, the step of naming the cluster information includes collecting music tag data for the music items from an external server device, scoring music tag data for the clusters based on a text mining technique, and music within a specified rank And writing tag data as the cluster information.

According to the present invention, the present invention provides a plurality of consistent playlists as personalized recommendation results, thereby reducing inconvenient user experience such as turning to the next song while listening to music, and playing suitable for listening time among candidate groups for preferred music You can select a list to help you listen to music with similar attributes in succession.

In addition, the present invention can improve the quality of the listening log used for music recommendation by removing a significant amount of noise existing in the conventional listening log and removing the inconvenience of manipulation in listening to music.

In addition, the present invention can support the construction of a context-aware music recommendation service that automatically provides a recommendation result suitable for a user's context through a context-aware recommendation algorithm by providing an improved quality listening log.

In addition, various effects other than the above-described effects may be directly or implicitly disclosed in a detailed description according to an embodiment of the present invention to be described later.

1 is a view showing an example of a service system configuration for performing personalized music recommendation according to an embodiment of the present invention.
2 is a view showing an example of a configuration of a service device according to an embodiment of the present invention.
3 is a diagram illustrating an example of a processor configuration among service device configurations according to an embodiment of the present invention.
4 is a diagram illustrating an example of a personalized music recommendation list according to an embodiment of the present invention.
5 is a view showing a personalized music recommendation method according to an embodiment of the present invention.
6 is a diagram illustrating an example of a method of generating a personalized music recommendation list according to an embodiment of the present invention.

In order to clarify the features and advantages of the problem solving means of the present invention, the present invention will be described in more detail with reference to specific embodiments of the present invention shown in the accompanying drawings.

However, in the following description and accompanying drawings, detailed descriptions of well-known functions or configurations that may obscure the subject matter of the present invention are omitted. In addition, it should be noted that the same components throughout the drawings are denoted by the same reference numerals as much as possible.

The terms and words used in the following description and drawings should not be interpreted as being limited to ordinary or dictionary meanings, and the inventor can appropriately define the concept of terms for explaining his or her invention in the best way. Based on the principle that it should be interpreted as meanings and concepts consistent with the technical spirit of the present invention. Therefore, the configuration shown in the embodiments and drawings described in this specification is only one of the most preferred embodiments of the present invention, and does not represent all of the technical spirit of the present invention, and can replace them at the time of this application. It should be understood that there may be equivalents and variations.

In addition, terms including ordinal numbers such as first and second are used to describe various components, and are used only to distinguish one component from other components, and to limit the components It is not used. For example, the second component may be referred to as a first component without departing from the scope of the present invention, and similarly, the first component may also be referred to as a second component.

In addition, the terms used herein are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In addition, terms such as "comprises" or "have" described herein are intended to designate the existence of features, numbers, steps, operations, components, parts, or combinations thereof described in the specification, or one or more thereof. It should be understood that the above or other features or numbers, steps, operations, components, parts or combinations thereof are not excluded in advance.

In addition, terms such as “part”, “group”, and “module” described in the specification mean a unit that processes at least one function or operation, which may be implemented by hardware or software or a combination of hardware and software. In addition, "a (a or an)", "one (one)," "the (the)" and similar related terms in the context of describing the present invention (especially in the context of the following claims) are different herein. It can be used in a sense including both singular and plural unless indicated or clearly contradicted by context.

In addition to the above-mentioned terms, specific terms used in the following description are provided to help understanding of the present invention, and the use of these specific terms may be changed into other forms without departing from the technical spirit of the present invention.

In addition, embodiments within the scope of the present invention include computer-readable media having or having computer-executable instructions or data structures stored on computer-readable media. Such computer-readable media can be any available media that can be accessed by a general purpose or special purpose computer system. By way of example, such computer readable media may be in the form of RAM, ROM, EPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage, or computer executable instructions, computer readable instructions or data structures. Physical storage media, such as any other media that may be used to store or transfer certain program code means in, and can be accessed by, general purpose or special purpose computer systems. .

For example, the present invention provides a computer readable recording medium stored in a memory and stored in at least one instruction executable by a processor, wherein the computer readable recording medium recommends music related to a terminal device based on a history of using music items A computer program including at least one instruction set to perform an operation of generating a list and classifying music items included in the music recommendation list into a plurality of clusters and providing the personalized music recommendation list to the terminal device is stored. It can be a computer-readable recording medium.

Additionally, the computer-readable recording medium in which the computer program is stored may be provided based on a music recommendation list prepared based on a preset rule or a history of using music items of a plurality of users when the history of using the music item does not satisfy a specified condition. At least one command set to perform an operation of providing a music recommendation list to the terminal device may be stored.

In addition, the computer-readable recording medium in which the computer program is stored may include at least one instruction set to perform an operation of naming cluster information using music tag data for the classified clusters.

1 is a view showing an example of a service system configuration for performing personalized music recommendation according to an embodiment of the present invention.

1, the personalized music recommendation service system 10 according to an embodiment of the present invention may include a service device 200 and a terminal device 100, and the service device 200 and the terminal The device 100 is interlocked through the network 150.

Then, with reference to Figure 1 will be described with respect to each component.

First, the network 150 serves to transmit data for data transmission and reception between the terminal device 100 and the service device 200, and according to a system implementation method, Ethernet, xDSL (ADSL, VDSL), HFC ( Wired communication methods such as Hybrid Fiber Coaxial Cable (FTTC), Fiber to The Curb (FTTC), and Fiber To The Home (FTTH) can be used, and Wireless LAN (WLAN), Wi-Fi, Wibro, WiMAX (Wimax), HSDPA (High Speed Downlink Packet Access), LTE (Long Term Evolution), LTE-A (Long Term Evolution Advanced), and other wireless communication methods may be used. It can include any form of communication. The network 150 forms a communication channel between the terminal device 100 and the service device 200, and transmits various data related to a music search and purchase function provided by the service device 200 to the terminal device 100 , A message related to a search request or purchase request of the terminal device 100 may be transmitted to the service device 200.

The terminal device 100 refers to a user device capable of transmitting and receiving various data to and from the service device 200 through the network 150 according to the user's manipulation. The terminal device 100 may perform voice or data communication through the network 150, a memory storing programs and protocols for transmitting and receiving data, and a microprocessor for calculating and controlling various programs. And the like. In addition, the terminal device 100 of the present invention may be implemented in various forms. For example, the terminal device 100 described in the present specification is a mobile phone such as a smart phone, a tablet PC, a personal digital assistants (PDA), a portable multimedia player (PMP), or an MP3 player. Of course, a fixed terminal such as a smart TV, a laptop computer, a desktop computer, etc. may be used, and as long as it is a device capable of transmitting and receiving data according to the present invention, any device in the embodiment of the present invention It can be used as a terminal device 100 according to.

Meanwhile, the terminal device 100 according to the present invention includes a configuration of a communication circuit, a display, an input device, an audio device, a processor, a memory, etc., and a service device through the network 150 using a communication circuit according to user manipulation. (200) Connection can be performed. For example, the terminal device 100 executes a web browser or an application corresponding to the user input, and accesses the service device 200 based on pre-entered Internet address information or address information input by the user. You can. The terminal device 100 may receive a designated web page from the service device 200 and output it on the display. For example, the web page may include a virtual page through which music items can be searched or purchased. The terminal device 100 may request login information input through a display in response to user authentication required by the service device 200. The terminal device 100 may transmit log-in information to the service device 200 in response to a user input, receive the logged-in web page in response to the approval completion, and output it to the display. In this operation, the terminal device 100 may transmit user information (eg, age, gender, residential area, occupation, education, weight, height, hobby, etc.) to the service device 200 in response to a user input.

The terminal device 100 may generate a music search request message in response to a user input and transmit it to the service device 200. The terminal device 100 may receive a list including at least one music item in response to a music search request from the service device 200. The terminal device 100 may generate a specific music item selection play request message or a specific music item purchase request message in response to a user input, and transmit it to the service device 200. The terminal device 100 may receive and output playback data of a specific music item from the service device 200 in a streaming or download method.

The terminal device 100 may receive a personalized music recommendation list from the service device 200 when logging in while connected to the service device 200. The personalized music recommendation list may be a list generated by cluster and cluster classification and cluster naming generated based on the predicted preference using the user information and search history, listening history, or purchase history of the terminal device 100. have. The terminal device 100 may store the personalized music recommendation list in a memory or output it to a display. The personalized music recommendation list provided to the terminal device 100 may be updated corresponding to the music search history, music listening history, or music purchase history of the terminal device 100.

The service device 200 may provide a designated service to the terminal device 100 accessed through the network 150. For example, the service device 200 may provide a web page designated to the connected terminal device 100. Alternatively, the service device 200 may provide the logged in web page to the terminal device 100 upon completion of the login of the terminal device 100. When the service device 200 receives a request message for searching for a music item from the terminal device 100, the service device 200 parses a search term within the message provided by the terminal device 100 and performs a music item search based on the parsed search term can do. The service device 200 may provide a search result to the terminal device 100. Alternatively, the service device 200 may receive a music purchase request message from the terminal device 100. In this case, the service device 200 may identify a music item to be purchased in the music purchase request message, and provide the identified music item to the terminal device 100.

In particular, the service device 200 may provide a personalized music recommendation list to the terminal device 100. In this regard, the service device 200 may store and manage music items, and classify music items according to specified criteria. The service device 200 may perform clustering and naming of music items based on the user's music item usage history information and the music item recommendation list in connection with creating a personalized music recommendation list. When the personalized music recommendation list is generated through the above-described process, the personalized music recommendation list may be provided to the terminal device 100.

As described above, the service apparatus 200 of the present invention performs clustering and naming operations on a user's music item use history and music recommendation list, thereby providing music suitable for a user's taste or preference for each individual, It can improve the quality or reliability of music recommendations they feel.

2 is a view showing an example of a configuration of a service device according to an embodiment of the present invention.

First, referring to FIG. 2, the service device 200 according to an embodiment of the present invention may include a communication circuit 210, a memory 240, and a processor 250.

The communication circuit 210 may support a communication function of the service device 200. The communication circuit 210 may support the formation of a communication channel for connecting at least one terminal device 100 through the network 150. The communication circuit 210 may support at least one communication method among various communication methods. For example, the communication circuit 210 may support various generations of mobile communication channels and Internet communication channels. When the communication circuit 210 receives at least one music item from an external server device in response to the control of the processor 250, the communication circuit 210 may transfer the music item to the memory 240. When the terminal device 100 is connected to the communication circuit 210, in response to the control of the processor 250, the designated web page may be delivered to the terminal device 100. In particular, the communication circuit 210 may transmit a personalized music recommendation list to the terminal device 100 in response to the control of the processor 250. When the communication circuit 210 receives a specific music item purchase request message from the terminal device 100, the communication circuit 210 transmits it to the processor 250 and, in response to the processor 250 control, transmits the requested music item to the terminal device ( 100).

The memory 240 may store at least one application or program in connection with the operation of the service device 200. In addition, the memory 240 may store a recommendation list for a plurality of music items and stored music items in relation to the personalized music recommendation function according to an embodiment of the present invention. The recommendation list may be determined according to a policy defined in the service device 200. The recommended list may include, for example, lists classified into various genres or various singers, various periods, and various age groups. The memory 240 may store a cluster application or various algorithms for clustering a recommendation list.

The processor 250 may perform data processing or data transfer related to the operation of the service device 200. For example, the processor 250 may wait to access the terminal device 100 through the network 150 according to an administrator request. The processor 250 may store and manage user information according to membership of the terminal devices 100 in the memory 240. The processor 250 collects and stores music item use history (eg, music item search history, music item listening history, or music item purchase history) while the terminal device 100 accesses and performs the use of music items. You can. When the processor 250 receives a new music item from the outside, the processor 250 may store and manage it in the memory 240. In this operation, the processor 250 may classify the received music items or music items stored in the memory 240 in response to a preset policy. For example, the processor 250 may generate a recommendation list for music items according to criteria such as a genre, a singer, an age, and an age. The processor 250 may generate a personalized music recommendation list based on the recommendation list and user information of the terminal device 100, and provide at least one of them to the terminal device 100. In this regard, the processor 250 may generate a music recommendation list and perform cluster generation based on the generated music recommendation list. The processor 250 may classify the generated clusters and perform naming for the classified music clusters. The processor 250 may include a configuration as shown in FIG. 3.

3 is a diagram illustrating an example of a processor configuration among service device configurations according to an embodiment of the present invention.

Referring to FIG. 3, the processor 250 according to an embodiment of the present invention includes a preference prediction unit 251, a recommendation music providing unit 252, a recommendation music cluster processing unit 253, and a cluster naming processing unit 255. It can contain.

The preference predicting unit 251 may generate a music recommendation model recommending music based on data obtained according to the use of music services of various users, and based on this, generate a music recommendation list. In this regard, the preference predicting unit 251 may generate and operate a music recommendation model based on machine learning. Alternatively, the preference prediction unit 251 may generate a music recommendation model according to a specified rule base. In relation to the creation of a machine learning-based music recommendation model, the preference predicting unit 251 may express each music item as a numeric vector for score calculation. That is, the preference predicting unit 251 may generate each item vector (Item vector: a numeric vector for an item) as an intermediate product derivation process of creating a music recommendation list, and generate a music recommendation list based on the item vector.

The item vector is a numerical vector of a predetermined dimension. The preference prediction unit 251 may generate an item vector using at least one of various algorithms. In this process, the preference predicting unit 251 may provide a result represented by a numeric vector as input data of the recommended music cluster processing unit 253 regardless of the algorithm logic. That is, the preference predicting unit 251 of the present invention can use the result value of the item vector regardless of the type of the recommendation algorithm using the item vector.

In relation to item vector generation, the preference prediction unit 251 may use item vector generation based on a matrix factorization algorithm. Matrix Factorization can decompose a user-music item matrix into a product of two matrices as shown in Equation 1 below.

Here, each of the matrices A, X, and Y is defined as in Equation 2 below.

Here, A represents the user's music item usage history (eg, music item search history, music item purchase history, or music listening history) as a matrix, and a_ij means whether the i-th user used the j-th music item do. If a music item is used, a_ij = 1, and if a music item is not used, a_ij = 0. The m * n-dimensional matrix A can be approximated by the product of the m * k matrix X and the k * n matrix Y through matrix decomposition techniques such as SVD (singular value decomposition). Each row x_i of X is a k-dimensional vector summarizing the information of the i-th user, and each column y_j of Y is a k-dimensional vector summarizing the information of the j-th music item, each of which can be interpreted as a User vector, Item vector. . Also, the matrix X and Y can be interpreted as a user matrix and an item matrix, respectively.

When the User matrix and Item matrix (or User vector, Item vector) are used for recommendation, each element value of X * Y, which is an approximation of the user-music item matrix, is treated as a score for user i purchasing music item j. Or, after obtaining k-nearest neighbors with a similar user i and a user vector from User matrix X, they can recommend the items they purchased.

As another embodiment, in relation to item vector generation, the preference predicting unit 251 may use a Neural Item Embedding algorithm. Neural Item Embedding Algorithm refers to the use of Neural Word Embedding, which is used to express words as a numeric vector in a language model, to recommend music items. Neural Word Embedding is a method based on artificial neural networks, and the preference predicting unit 251 may use the Word2Vec algorithm.

The Word2Vec algorithm 1) determines a reference word in a sentence and predicts the existence of a reference word by words appearing before and after the reference word, and 2) words appearing before and after the reference word as reference words in the sentence ( Context).

With regard to obtaining the item vector, the word2Vec algorithm can use the appearance of a word as an input variable. Whether a word appears or not is indicated by 1 when a word appears or 0 when it does not appear, and has a dimension equal to the number of word types (n, usually hundreds of thousands) in the corpus. In the Word2Vec algorithm, a process of summarizing to a low dimension (k) is performed by receiving a high-dimensional input and passing it through a hidden layer of an artificial neural network. After learning is completed, a weight matrix of n * k dimensions can be obtained from the hidden layer, and each row is referred to as a word vector. In order to obtain an item vector for the above-described method, the preference predicting unit 251 1) determines a reference item (music here) in the music item usage history, and a reference item as items (context) appearing before and after the reference item Method for predicting the existence of the 2) As a reference item in the music item use history, at least one of the methods for predicting the existence of items (context) appearing before and after the reference item may be operated. The preference prediction unit 251 may apply the same method as the above-described word vector acquisition to the music item vector scheme and obtain an item vector for the music item as a learning result.

In relation to the item vector acquisition, the preference prediction unit 251 may use a deep learning algorithm. The deep learning algorithm is similar to the Neural Item Embedding based recommendation above in that it utilizes an artificial neural network. Normally, the item appearance is used as input data, as in Neural Item Embedding. However, if Neural Item Embedding is an approach that focuses on summarizing the distribution of data in a low dimension, the difference is that Deep learning algorithms use the appearance of items as input variables and free the objective function or artificial neural network structure design. There is.

The preference predicting unit 251 may obtain an item vector from a hidden layer summarizing the appearance of an item in a low dimension according to completion of learning using the deep learning algorithm. In Neural Item Embedding, the item vector is the main result, but in the Deep learning algorithm, it can be obtained additionally. Deep learning algorithms do not have a uniform method because of their high degree of freedom in design, so this document does not cover how it works. In the present invention, an item vector can be extracted and clustered for a case in which the appearance of an item is used as an input variable regardless of a specific structure and is summarized in a low dimension.

Since the item vector represented by the numerical vector through the above-described method is an intermediate product of the recommendation algorithm, there is no need to separately generate it in the process of the present invention. When using a special type of recommendation algorithm such as a graph algorithm, a numeric vector may not be generated. However, a graph embedding technique that summarizes the adjacency matrix of a graph into a low-dimensional matrix can be converted into a numeric vector.

The recommended music providing unit 252 may store and manage the music recommendation list generated by the preference prediction unit 251. The music recommendation list is a single list composed of a plurality of music items. According to the listening behavior assumed in the present invention, music items having different personalities may be mixed. The recommended music providing unit 252 may deliver a music recommendation list to the recommended music cluster processing unit 253 according to a user request or according to a policy. For example, the recommended music providing unit 252 may provide the terminal device 100 with a list of music recommendations derived according to machine learning according to the service usage level of the connected terminal device 100 when the terminal device 100 is connected. have. The recommended music providing unit 252, when the terminal device 100 has a history of using a music item over a specified number of times or a specified period of time, lists the music recommendations related to the terminal device 100 in the preference prediction unit 251 You can request creation. For example, when the music item search history (or purchase history) of a specific terminal device 100 is collected over a certain number of times or the music item search history (or purchase history) is specified for a specified period, the recommended music provider 252 When it is repeatedly performed, requesting the preference prediction unit 251 to generate a music recommendation list through machine learning, and receiving the music recommendation list, delivers the music recommendation list to the recommendation music cluster processing unit 253, while the User information can be delivered together.

The recommended music providing unit 252 may check user information of the terminal device 100 when the terminal device 100 is connected. In this regard, the recommended music provider 252 checks the login information of the terminal device 100, and when the login authentication is successfully completed, the personalized music recommendation list associated with the terminal device 100 is stored in the memory 240 To see if it exists. The recommended music providing unit 252 may provide the personalized music recommendation list to the terminal device 100 when a personalized music recommendation list related to the terminal device 100 exists in the memory 240. When the personalized music recommendation list does not exist in the memory 240, the recommended music providing unit 252 may check the history of using the music item of the terminal device 100 and confirm whether the specified condition is satisfied. When the recommended music providing unit 252 does not satisfy the specified condition, the list of music recommendations generated based on the general music item usage history of various users is provided to the terminal device 100, or music generated according to the specified rules The recommendation list may be provided to the terminal device 100.

The recommended music cluster processing unit 253 may receive an item vector and a music recommendation list as input data. Also, the recommended music cluster processing unit 253 may receive user information related to the music recommendation list. The recommended music cluster processing unit 253 may cluster individual music recommendation results using an item vector obtained as a result of various algorithms. The recommended music cluster processing unit 253 may use a clustering technique to divide music items in a music recommendation list into several clusters. Since music items are represented by a k-dimensional numeric vector, a group of users who listen can form clusters of similar music by applying a general clustering technique defined in Euclidean space. For example, the recommended music cluster processing unit 253 uses K-average clustering. K-means clustering is a method of determining a cluster's reference point for a given number of clusters, and assigning each point to the cluster to which the nearest reference point belongs. The quality of clustering can be different in determining K, so choose K that maximizes the silhouette coefficient used as the clustering quality index. In this case, K is greater than or equal to 1 and less than or equal to the number of music in the music recommendation list. In actual application, the search range can be reduced by setting the maximum number of contexts that are supposed to exist as an upper limit. Based on this method, the recommended music cluster processing unit 253 may convert a music recommendation list (a single list composed of multiple pieces of music) into K music clusters (a plurality of lists composed of multiple pieces of music). The union of K music clusters is the same as the music recommendation list. Since similar music is assigned to the same cluster, the music in the product has a consistent theme.

The cluster naming processing unit 255 receives K music clusters to be labeled and music tag data, which is source data for labeling, as input data. Also, the cluster naming processing unit 255 may receive user information from the recommended music cluster processing unit 253. The K music clusters are received from the recommended music cluster processing unit 253. The music tag data is a tag attached to music by users, and mainly relates to appreciation of music, atmosphere of the music, and genre. The tag is data generated by users to easily search for information, and the cluster naming processor 255 may obtain information from various social network services. Specifically, the cluster naming processing unit 255 may obtain information summarizing the frequency of tags for music through the API of a social network service, or collect tags existing in various sound source service servers. Additionally, the cluster naming processing unit 255 may artificially generate a tag through natural language processing from music-related texts existing in a social network. In the present invention, in order to name a cluster by utilizing the fact that the relationship between music and tags is similar to the relationship between a document and a word in the text mining field, the cluster naming processing unit 255 may use a text mining tf-idf scoring method You can. Since the data to be used is a collection of tags placed by a plurality of users, each cluster can be represented by a word-document matrix as shown in Equation 3 below.

Here, C_k denotes the k th cluster, and w_ij denotes the j th tag appearance frequency of the i th recommended song.

Labeling used in the present invention is performed based on tf-idf used to quantify the importance of words in a document. In detail, first, the cluster naming processor 255 converts all elements of the word-document matrix into tf-idf values, as in Equation 4 below.

Here, D is a set of entire music appearing in the music tag data. The label score of tag j within cluster k is defined as in Equation 5 below.

The cluster naming processing unit 255 may name clusters by using a tag having the highest predetermined number (for example, three) based on a label score within each cluster. The generated K music clusters to which cluster naming is applied by the cluster naming processing unit 255 is a personalized music recommendation list, and may have a form as shown in FIG. 4. The cluster naming processing unit 255 may map a personalized music recommendation list together with user information and store and manage it in the memory 240.

4 is a diagram illustrating an example of a personalized music recommendation list according to an embodiment of the present invention.

Referring to FIG. 4, the personalized music recommendation list has a consistent theme and may be easy for a user to understand its characteristics. As a result, the user can easily select a music item based on a personalized music recommendation list, and after selecting, only the music that matches the intention of the selection can be purchased or listened.

As described above, the service device 200 for generating and providing the personalized music recommendation list of the present invention is personalized while compensating for the problem of the result of the existing personalized recommendation providing method that is inconsistent and cannot easily recognize features. It can provide different music curation and simplified user search effects.

5 is a view showing a personalized music recommendation method according to an embodiment of the present invention.

Referring to FIG. 5, in connection with the personalized music recommendation method according to an embodiment of the present invention, in step 501, the processor 250 of the service device 200 may check whether the terminal device 100 is connected. If there is no connection to the terminal device 100, the processor 250 may process the performance of the designated function in step 503. For example, the processor 250 may collect music items from an external server device or music tag data for music items.

When the terminal device 100 is connected, the processor 250 may check user information in operation 505. In this regard, the processor 250 may request login information from the terminal device 100. When receiving the login information from the terminal device 100, the processor 250 may determine whether to authenticate the login by comparing with the previously stored login information.

Upon checking the user information, the processor 250 may check whether the user information satisfies the specified condition in step 507. The specified condition may include a condition in which the history of using the music item of the terminal device 100 exceeds a specified number of times or a specified time. For example, the specified condition may include a condition in which at least one of the music item search history, listening history, and purchase history of the terminal device 100 exceeds a specified number of times or a specified period.

If the specified condition is satisfied, the processor 250 may provide a personalized music recommendation list in step 509. When the specified condition is satisfied, the processor 250 may check whether a personalized music recommendation list corresponding to the connected terminal device 100 exists in the memory 240. When the personalized music recommendation list exists in the memory 240, the processor 250 provides it to the terminal device 100, and when the personalized music recommendation list does not exist in the memory 240, the personalized music recommendation list is provided. You can create a list.

If the specified condition is not satisfied, the processor 250 may provide a general music recommendation list in step 511. For example, the processor 250 may provide the terminal device 100 with a music recommendation list set according to a specified rule of the service device 200 or a music recommendation list generated based on the history of using a plurality of users' music items. .

6 is a diagram illustrating an example of a method of generating a personalized music recommendation list according to an embodiment of the present invention.

Referring to FIG. 6, in connection with the method for generating a personalized music recommendation list of the present invention, the processor 250 of the service device 200 may collect a history of using music items in step 601. In this regard, when the processor 250 searches for, listens to, or purchases music items while the terminal device 100 is connected, the processor 250 may store a corresponding history. In this operation, the processor 250 may map and store each history in user information of the terminal device 100.

In operation 603, the processor 250 may generate an item vector for each music item and a music recommendation list through preference prediction. The item vector is a numerical vector of music as an intermediate product according to preference prediction. If the similarity of the item vector is high, it may be similar music. When the item vector is generated, the processor 250 may collect music items based on the similarity with the item vector, and generate a music recommendation list.

In operation 605, the processor 250 may perform cluster processing on the music recommendation list. That is, the processor 250 may divide music items in the music recommendation list into multiple clusters. In this operation, the processor 250 may separate the music within the cluster so that the similarity between the clusters is high and the similarity between the clusters is low. Through the above-described steps, the processor 250 may acquire music lists against K music clusters.

In step 607, the processor 250 may perform cluster naming on the clustered music recommendation list. In this operation, the processor 250 may perform a task of labeling the characteristics of music items so that the user can distinguish the separated music lists (or clusters). The label may be determined by a person's perception or mood for a music item, and in this connection, users are attached to respective music data from an external server device (eg, a social network service server device or a sound source service providing server device). Music tag data can be collected. The music tag data may include a public playlist of a sound source service or a set of tags put by users for individual music in a social network service.

In operation 609, the processor 250 may store and manage the personalized music recommendation list in the memory 240. In this operation, the processor 250 may store the named K music clusters in the memory 240 as a personalized music recommendation list. The K music cluster may include groups listing music data having a plurality of different similarities.

As described above, this specification includes details of a number of specific implementations, but these should not be understood as limiting with respect to any invention or claimable scope, but rather may be specific to a particular embodiment of a particular invention. It should be understood as a description of the features.

Also, although the operations are depicted in the drawings in a particular order, it should not be understood that such operations should be performed in the particular order shown or in sequential order, or that all shown actions should be performed in order to obtain desirable results. In certain cases, multitasking and parallel processing may be advantageous. In addition, the separation of various system components of the above-described embodiments should not be understood as requiring such separation in all embodiments, and the described program components and systems will generally be integrated together into a single software product or packaged in multiple software products. You should understand that you can.

The described description presents the best mode of the present invention, and provides examples for explaining the present invention and to enable a person skilled in the art to make and use the present invention. This written specification is not intended to limit the invention to the specific terms presented. Therefore, the present invention has been described in detail with reference to the above-described examples, but a person skilled in the art can make modifications, alterations, and modifications to these examples without departing from the scope of the present invention.

Therefore, the scope of the present invention should not be determined by the described embodiments, but should be determined by the claims.

According to the present invention, the present invention enables the results to be consistent in providing music personalized recommendation results, thereby maximizing the quality of experience, and helping the user to easily select music according to the context, and the operator has a high quality listening history Support to secure data.

In addition, the present invention allows the user to easily recognize differences between music clusters by naming the clusters with tags, so that there is no gap between the description and the user's perception using tags that are the user's language to name the clusters. You can.

Accordingly, the present invention can contribute to the establishment of a virtuous cycle logistics system that can increase the desire to purchase music and thereby provide users with a higher quality sound source by improving the reliability of music recommendation in the music service market.

150: communication network
100: terminal device
200: service device
210: communication circuit
240: memory
250: processor
251: preference prediction unit
252: recommended music provider
253: recommended music cluster processing unit
255: cluster naming processing unit

Claims (13)

A communication circuit forming a communication channel with the terminal device;
A memory for storing at least one music item; And
And a processor functionally connected to the communication circuit and the memory.
The processor
Create a music recommendation list related to the terminal device based on a history of using music items,
A service device for providing a personalized music recommendation list, characterized in that the music items included in the music recommendation list are set to be classified into a plurality of clusters and stored as a music recommendation list. According to claim 1,
The processor
A service for providing a personalized music recommendation list characterized in that it is configured to generate an item vector corresponding to a numeric vector based on the history of using the music item, and to generate the music recommendation list based on the similarity with the item vector. Device. According to claim 1,
The processor
A service device providing a personalized music recommendation list, characterized in that the music items are set to be classified into a plurality of clusters based on a designated clustering algorithm. According to claim 1,
The processor
A service device providing a personalized music recommendation list, characterized in that it is set to collect music tag data for the music items from an external server device. According to claim 1,
The processor
A service device providing a personalized music recommendation list, characterized in that the classified clusters are configured to name cluster information using music tag data, and the clusters to which the cluster information is named are stored as a personalized music recommendation list. . The method of claim 5,
The processor
A service device providing a personalized music recommendation list characterized in that it is configured to score music tag data for the clusters based on a text mining technique and to write music tag data within a designated rank as the cluster information. According to claim 1,
The processor
If the history of using the music item does not satisfy a specified condition, it is configured to provide a music recommendation list prepared based on a preset rule or a music recommendation list prepared based on a history of using a plurality of users' music items to the terminal device. A service device that provides a personalized music recommendation list. According to claim 1,
The processor
Personalization characterized in that the personalized music recommendation list is set to provide the personalized music recommendation list to the terminal device when the personalized music recommendation list is stored in the memory when the music item usage history of the terminal device satisfies a specified condition. Device providing a list of recommended music. Service device,
Generating a music recommendation list related to the terminal device based on the history of using the music item;
And classifying the music items included in the music recommendation list into a plurality of clusters and providing the personalized music recommendation list to the terminal device as a personalized music recommendation list. The method of claim 9,
Providing the terminal device with a music recommendation list prepared based on a preset rule or a music recommendation list prepared based on a history of using a plurality of users' music items when the music item use history does not satisfy a specified condition; Personalized music recommendation method further comprising. The method of claim 9,
And naming cluster information using music tag data with respect to the classified clusters. The method of claim 11,
The step of naming the cluster information is
Collecting music tag data for the music items from an external server device; And
And scoring music tag data for the clusters based on a text mining technique and writing music tag data within a designated rank as the cluster information. A computer readable recording medium stored in memory and having at least one instruction executable by a processor, the computer readable recording medium comprising:
Generating a music recommendation list related to the terminal device based on the history of using the music item; And
Classifying the music items included in the music recommendation list into a plurality of clusters and providing the personalized music recommendation list to the terminal device; a computer readable recording medium storing a computer program including at least one instruction set to perform .

Images