Method and apparatus for personalizing content recommendation model

Abstract

A method and apparatus for personalizing a content recommendation model are provided. A method for personalizing a content recommendation model by an electronic device may include: acquiring a first content recommendation model used to recommend content to a user of the electronic device; personalizing the first content recommendation model based on the user's content usage history; receiving a second content recommendation model from a server; receiving a personalization model from the server for personalizing the second content recommendation model; personalizing the second content recommendation model using the input / output data of the personalized first content recommendation model and the personalization model; and providing a content recommendation service to the user using the personalized second content recommendation model.

Description

Technology Field

[0001] The present disclosure relates to a method and apparatus for personalizing a content recommendation model. Background Technology

[0002] Recently, services providing content to mobile devices via wireless internet have become commonplace, and the types of content provided to mobile devices in this manner are very diverse, including image content, music content, video content, game content, and real-time information content.

[0003] Content recommendation methods include personalized recommendations that utilize user-inputted preferred genre or category information, or purchase history. In particular, machine learning-based content recommendation methods are being used recently, as they improve recognition accuracy with usage and can more accurately understand user preferences through learning based on user usage history. Examples of existing content recommendation methods include Client, server, and client-server systems adapted for generating personalized recommendations ( International Publication WO2019 / 120578 (June 27, 2019), etc. The problem to be solved

[0004] Some embodiments may provide a method and apparatus for personalizing a content recommendation model received from a server based on a user's usage history. means of solving the problem

[0005] As a technical means for achieving the technical problem described above, the first aspect of the present disclosure may provide a method for an electronic device to personalize a content recommendation model, comprising: acquiring a first content recommendation model used to recommend content to a user of an electronic device; personalizing the first content recommendation model based on the user's content usage history; receiving a second content recommendation model from a server; receiving a personalization model from the server for personalizing the second content recommendation model; personalizing the second content recommendation model using the input / output data of the personalized first content recommendation model and the personalization model; and providing a content recommendation service to the user using the personalized second content recommendation model.

[0006] Additionally, a second aspect of the present disclosure may provide an electronic device for personalizing a content recommendation model, comprising a memory for storing one or more instructions and a processor for executing said one or more instructions, wherein the processor obtains a first content recommendation model used to recommend content to a user of an electronic device by executing said one or more instructions, personalizes said first content recommendation model based on the user's content usage history, receives a second content recommendation model from a server, receives a personalization model from said server for personalizing said second content recommendation model, personalizes said second content recommendation model using the input / output data of said personalized first content recommendation model and said personalization model, and provides a content recommendation service to said user using said personalized second content recommendation model.

[0007] Additionally, a third aspect of the present disclosure may provide a computer program device comprising a computer-readable recording medium having a program for executing the method of the second aspect on a computer. Brief explanation of the drawing

[0008] FIG. 1 shows an example of a content recommendation service provision system according to some embodiments. FIG. 2 is a flowchart illustrating how an electronic device according to some embodiments personalizes a content recommendation model. FIG. 3 is a flowchart illustrating how an electronic device according to some embodiment personalizes a content recommendation model using a personalization model. FIG. 4 is a diagram illustrating a method for an electronic device according to some embodiments to personalize a content recommendation model based on a user's content usage history. FIG. 5 is a diagram illustrating a method for an electronic device according to some embodiments to personalize a content recommendation model based on weight data obtained from a personalization model. FIG. 6 is a diagram illustrating a method for a server to generate a personalized model according to some embodiments. FIG. 7 is a diagram illustrating a method for a server to generate a personalized model according to some embodiments. FIG. 8 is a block diagram of an electronic device according to some embodiments. FIG. 9 is a block diagram of a server according to some embodiments. Specific details for implementing the invention

[0009] Embodiments of the present disclosure are described below in detail with reference to the attached drawings so that those skilled in the art can easily implement them. However, the present disclosure may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present disclosure in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0010] Throughout the specification, when a part is described as being "connected" to another part, this includes not only cases where they are "directly connected," but also cases where they are "electrically connected" with other components interposed between them. Furthermore, when a part is described as "including" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0011] Functions related to artificial intelligence according to the present disclosure are operated through a processor and memory. The processor may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. The one or more processors control the processing of input data according to predefined operation rules or artificial intelligence models stored in memory. Alternatively, if the one or more processors are artificial intelligence-dedicated processors, the artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

[0012] The predefined rules of operation or artificial intelligence models are characterized by being created through learning. Here, being created through learning means that a predefined rules of operation or artificial intelligence models configured to perform desired characteristics (or objectives) are created by a basic artificial intelligence model being trained using multiple learning data by a learning algorithm. Such learning may be performed on the device itself where the artificial intelligence according to the present disclosure is executed, or it may be performed through a separate server and / or system. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples described above.

[0013] An artificial intelligence model may be composed of multiple neural network layers. Each of the multiple neural network layers has multiple weight values ​​and performs neural network operations through operations between the results of previous layers and the multiple weights. The multiple weights possessed by the multiple neural network layers can be optimized based on the learning results of the artificial intelligence model. For example, the multiple weights may be updated so that the loss value or cost value obtained from the artificial intelligence model during the learning process is reduced or minimized. Artificial neural networks may include deep neural networks (DNNs), such as Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs), Restricted Boltzmann Machines (RBMs), Deep Belief Networks (DBNs), Bidirectional Recurrent Deep Neural Networks (BRDNNs), or Deep Q-Networks, but are not limited to the examples mentioned above.

[0014] The present disclosure will be described in detail below with reference to the attached drawings.

[0015] FIG. 1 shows an example of a content recommendation service provision system according to some embodiments.

[0016] Referring to FIG. 1, a content recommendation model update system according to some embodiments may include an electronic device (1000) and a server (2000).

[0017] In the present disclosure, the electronic device (1000) may be a device that provides content to a user. The electronic device (1000) according to the embodiment refers to a device for recommending appropriate content to a user using a content recommendation model, and may include, for example, a smartphone, a personal computer (PC), a laptop PC, a tablet PC, a smart TV, a smart speaker, and a smart audio, but is not limited thereto, and may include all devices capable of providing content to a user.

[0018] In the present disclosure, content may include information that can be consumed by a user by being played back by an electronic device. The content according to the embodiments is data including at least one of visual information or auditory information, such as, for example, digital newspapers, books, sound recordings, movies, and dramas, but is not limited thereto, and may include all information that a user can understand and consume.

[0019] The electronic device (1000) can recommend content to the user using a content recommendation model.

[0020] In the present disclosure, the recommendation of content may be to recommend at least one other content based on at least one of the characteristics of the selected content or the user's content usage history, as the user selects at least one content among the content provided by the electronic device (1000).

[0021] Here, the content recommendation model used by the electronic device (1000) to recommend content may be an artificial intelligence model for recommending content to a user based on information related to the user's content usage. The content recommendation model may, for example, recommend other content suitable for the user based on the user's content selection.

[0022] The electronic device (1000) may, for example, use a model including a Deep Neural Network (DNN) as a content recommendation model for recommending content. For example, the content recommendation model may be trained by taking information regarding selected content as input and outputting recommendation content information including at least one recommended content.

[0023] The deep neural networks that may be included in the content recommendation model in the present disclosure may include, for example, at least one of a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), and a Generative Adversarial Network (GAN), but are not limited thereto, and any type of deep neural network that can be used for content recommendation may be used.

[0024] A server (2000) according to some embodiments may generate a content recommendation model, and an electronic device (1000) may recommend content to a user based on the content recommendation model received from the server (2000).

[0025] The creation of a content recommendation model by the server (2000) can be achieved, for example, through learning based on big data that takes information about content as input and information about at least one recommended content as output. Here, the big data used by the server (2000) to create the content recommendation model may, for example, be a predetermined anonymized data obtained in advance to create the content recommendation model. Since a big data-based content recommendation system corresponds to a technology known in the field of content recommendation technology, a detailed explanation of such a method for creating a content recommendation model will be omitted.

[0026] The content recommendation model generated by the server (2000) can be distributed to each electronic device (1000), for example, and used to recommend content to the user of each electronic device (1000).

[0027] Meanwhile, the electronic device (1000) can recommend content to the user based on the content recommendation model received from the server (2000) and personalize the content recommendation model.

[0028] In the present disclosure, the personalization of the content recommendation model may involve the content recommendation model being additionally trained so that it can recommend content optimized for the user as the electronic device (1000) repeatedly recommends content to the user based on the content recommendation model. The electronic device (1000) can personalize the content recommendation model to correspond to the preferences and tastes of the user of the electronic device (1000) by repeatedly performing content recommendations using the content recommendation model received from the server (2000).

[0029] Referring again to FIG. 1, an electronic device (1000) according to some embodiment may receive a first content recommendation model from a server (2000). Based on the first content recommendation model received from the server (2000), the electronic device (1000) may recommend other content in response to the user's content selection.

[0030] As other content is recommended by the electronic device (1000), the user can select the desired content among the other content recommended by the electronic device (1000). At this time, as the content recommendation based on the first content recommendation model of the electronic device (1000) and the user's content selection are repeated, a personalized first content recommendation model (101) can be obtained.

[0031] Meanwhile, a server (2000) according to some embodiments may generate a second content recommendation model different from the first content recommendation model. In this disclosure, the generation of a second content recommendation model different from the previously generated first content recommendation model is referred to as an update of the content recommendation model.

[0032] The server (2000) may, for example, generate a second content recommendation model by updating the content recommendation model after transmitting the first content recommendation model to the electronic device (1000). Once the second content recommendation model is generated, the server (2000) may transmit the generated second content recommendation model to the electronic device (1000). In this case, for example, the first content recommendation model and the second content recommendation model may have the same or different network structures. If the first content recommendation model and the second content recommendation model have the same or different network structures, the input / output data of the first content recommendation model may be converted into a format suitable for the second content recommendation model.

[0033] An electronic device (1000) that performs content recommendation using a personalized first content recommendation model (101) can directly personalize the second content recommendation model using the input / output data of the personalized first content recommendation model (101) as it receives a second content recommendation model from a server (2000).

[0034] The electronic device (1000) of the present disclosure has the advantage of protecting the user's personal information because it can directly personalize the received second content recommendation model without transmitting usage history information regarding the user's content selection to the server (2000).

[0035] In addition, the electronic device (1000) of the present disclosure has the advantage of being able to personalize a second content recommendation model using only the input / output data of the personalized first content recommendation model (101) without needing to store vast amounts of data regarding usage history used to personalize the first content recommendation model, thereby protecting the user's personal information while increasing the efficiency of personalization of the content recommendation model.

[0036] Meanwhile, the electronic device (1000) of the present disclosure may use a separate personalization model to directly personalize a second content recommendation model based on the input / output data of a personalized first content recommendation model. The personalization model may be an artificial intelligence model that outputs weight values ​​used to adjust weights between layers within the second content recommendation model in order to personalize the second content recommendation model. In this case, the personalization model may be generated on a server (2000) together with the second content recommendation model.

[0037] Hereinafter, based on the embodiments of FIGS. 2 to 7, a specific method for an electronic device of the present disclosure to personalize a content recommendation model and a method for a server to generate a personalization model will be described in detail.

[0038] FIG. 2 is a flowchart illustrating how an electronic device according to some embodiments personalizes a content recommendation model.

[0039] Referring to FIG. 2, a server (2000) according to some embodiment may generate a first content recommendation model in step S201. In step S202, the server (2000) may transmit the generated first content recommendation model to an electronic device (1000).

[0040] The electronic device (1000) that receives the first content recommendation model from the server (2000) can generate a personalized first content recommendation model by personalizing the first content recommendation model based on the user's content usage history in step S203.

[0041] Meanwhile, the server (2000) can generate a second content recommendation model in step S204. The server (2000) can generate a personalization model for personalizing the second content recommendation model in step S205.

[0042] The server (2000) can transmit the second content recommendation model and personalization model generated in step S206 to the electronic device (1000).

[0043] The electronic device (1000) that receives the second content recommendation model and the personalization model from the server (2000) can personalize the second content recommendation model using the input / output data of the personalized first content recommendation model and the personalization model in step S207.

[0044] The electronic device (1000) can provide a content recommendation service to the user using a personalized second content recommendation model in step S208.

[0045] FIG. 3 is a flowchart illustrating how an electronic device according to some embodiment personalizes a content recommendation model using a personalization model.

[0046] Referring to FIG. 3, an electronic device (1000) according to some embodiments can personalize a second content recommendation model using the input / output data of a personalized first content recommendation model and the personalized model in step S205.

[0047] At this time, the electronic device (1000) can input the input / output data of the personalized first content recommendation model into the personalized model in step S301.

[0048] The electronic device (1000) can obtain weight data, which is data representing weights applied between layers included in the second content recommendation model, in step S302.

[0049] The electronic device (1000) can personalize the second content recommendation model by changing the weights applied between the layers included in the second content recommendation model based on the acquired weight data in step S303.

[0050] FIG. 4 is a diagram illustrating a method for an electronic device according to some embodiments to personalize a content recommendation model based on a user's content usage history.

[0051] Referring to FIG. 4, an electronic device (1000) according to some embodiment may recommend content to a user based on a first content recommendation model (401). The electronic device (1000) may, for example, output output data Output_A_1, Output_A_2, and Output_A_3 corresponding to content recommendation in response to input data Input_A_1, Input_A_2, and Input_A_3 corresponding to the user's content selection by using the first content recommendation model (401). The input data corresponding to the user's content selection may include, for example, an identification value of the content selected by the user, a genre of the content, a time and place when the content is selected, an identification value of an application running on the electronic device (1000) when the content is selected, but is not limited thereto.

[0052] As the electronic device (1000) outputs output data Output_A_1, Output_A_2, and Output_A_3 to recommend content to the user, the user can select at least one of the recommended contents recommended by the electronic device (1000). For example, the output data output from the first content recommendation model (401) may include an identification value of the content, but is not limited thereto. As such user content selection and content recommendation by the electronic device (1000) are repeated, the first content recommendation model (401) may be personalized, and a personalized first content recommendation model (411) may be obtained. Additionally, the input data and output data used for content recommendation may be accumulated and stored within the electronic device (1000).

[0053] Meanwhile, a server (2000) according to some embodiments may update a first content recommendation model (401) to generate a second content recommendation model (402). Along with generating the second content recommendation model, the server (2000) may generate a personalization model (43) corresponding to the second content recommendation model. The generated second content recommendation model (402) and personalization model (43) may be transmitted to an electronic device (1000). Although the above description simply states that the server (2000) updates the first content recommendation model (401) to generate the second content recommendation model (402), it is not limited thereto. The content recommendation model of the server (2000) may be continuously updated using big data for content recommendation, and the content recommendation model before the update may be referred to as the first content recommendation model (401), and the content recommendation model after the update may be referred to as the second content recommendation model (402).

[0054] An electronic device (1000) that has received a second content recommendation model (402) and a personalization model (43) can obtain input / output data of a personalized first content recommendation model (411) in order to personalize the second content recommendation model (402). For example, the input / output data of the personalized first content recommendation model (411) may be accumulated and stored in the memory of the electronic device (411), and the electronic device (411) can extract the input / output data stored in the memory from the memory. Additionally, for example, the electronic device (411) can obtain output data output from the personalized first content recommendation model (411) by inputting a predetermined input data into the personalized first content recommendation model (411), and can obtain a set of input data and output data.

[0055] Input / output data of the personalized first content recommendation model (411) can be input into the personalized model (43). The personalized model (43), having received the input / output data of the personalized first content recommendation model (411), can obtain information necessary to create a personalized second content recommendation model (422) by directly personalizing the second content recommendation model (402) in response to the received input / output data. In this case, for example, data regarding the second content recommendation model (402) and the input / output data of the personalized first content recommendation model (411) can be input into the personalized model (43). The data regarding the second content recommendation model (402) may include, for example, information regarding layers within the second content recommendation model (402) and weights between layers, but is not limited thereto.

[0056] The information used to directly personalize the second content recommendation model (402) may be, for example, weight data representing the weights applied between the layers included in the second content recommendation model.

[0057] FIG. 5 is a diagram illustrating a method for an electronic device according to some embodiments to personalize a content recommendation model based on weight data obtained from a personalization model.

[0058] Referring to FIG. 5, an electronic device (1000) according to some embodiments can obtain input data input to a personalized first content recommendation model (411) and output data output from the personalized first content recommendation model (411).

[0059] The input / output data of the personalized first content recommendation model (411) used by the electronic device (1000) may include, for example, input / output data corresponding to the user's content usage history. Here, the input / output data corresponding to the user's content usage history refers to the input / output data that was used to personalize the first content recommendation model (401) into the personalized first content recommendation model (411). Here, the input / output data corresponding to the user's content usage history may be data stored in the memory of the electronic device (1000) during the content recommendation process based on the first content recommendation model (401).

[0060] Meanwhile, the input / output data of the personalized first content recommendation model (411) used by the electronic device (1000) to directly personalize the second content recommendation model (402) may, for another example, be separate input / output data obtained for the purpose of personalizing the second content recommendation model (402) after receiving the second content recommendation model (402) from the server (2000). For example, the electronic device (1000) may obtain the input / output data of the personalized first content recommendation model (411) by inputting data into the personalized first content recommendation model (411) and obtaining the output data after receiving the second content recommendation model (402) from the server (2000), regardless of the user's existing usage history. At this time, the electronic device (1000) may not use the input / output data corresponding to the user's content usage history.

[0061] Meanwhile, the input / output data of the personalized first content recommendation model (411) used by the electronic device (1000) may, as another example, include both input / output data corresponding to the user's content usage history and separate input / output data obtained for the purpose of personalizing the second content recommendation model (402). The input data corresponding to the user's content usage history may include, for example, an identification value of content selected by the user, a genre of content, the time and place when the content was selected, and an identification value of an application running on the electronic device (1000) when the content was selected, but is not limited thereto. Additionally, the output data corresponding to the user's content usage history may be data output from the personalized first content recommendation model (411) based on the input data corresponding to the user's content usage history.

[0062] The electronic device (1000) can obtain an input / output data set (501) of a personalized first content recommendation model by combining input / output data extracted from a personalized first content recommendation model (411).

[0063] An electronic device (1000) that has obtained an input / output data set (501) of a personalized first content recommendation model can input the input / output data set (501) of the personalized first content recommendation model into a personalization model (43) to obtain weight data used to directly personalize a second content recommendation model (402) as an output. At this time, the weight data may be data representing weights applied between layers included in the second content recommendation model (402).

[0064] The electronic device (1000) can generate a personalized second content recommendation model (422) by obtaining weight data from the input / output data set (501) of the personalized first content recommendation model through the personalized model (43) and directly adjusting the weights applied between the included layers of the second content recommendation model (402) based on the obtained weight data.

[0065] FIG. 6 is a diagram illustrating a method for a server to generate a personalized model according to some embodiments.

[0066] A server (2000) according to some embodiments may generate a personalization model (63) based on a plurality of input data classified by a predetermined content category and a plurality of second content recommendation models (602) specialized by a predetermined content category to correspond to the classified input data. The content category may be classified based on at least one of, for example, the type of content, the attributes of the content, or the type of service provided to the user based on the content. The type of content may include, for example, music, movies, photos, etc., but is not limited thereto. Additionally, the attributes of the content may include, for example, the genre of the content, the playback time of the content, artists, creators, etc., but are not limited thereto. Additionally, the type of service may include, for example, broadcasting services, music streaming services, and video streaming services, but is not limited thereto.

[0067] Content categories can be determined based on the similarity between labels derived from the labels of the input data. For example, in the case of a music recommendation service, the input data may include a user profile and music content information. The user profile may consist of age, gender, region, etc., while the music content information may consist of genre, composer, singer, etc. For instance, age, gender, region, and genre may be classified as content categories based on the user profile, or content categories may be classified by grouping at least some of the content categories.

[0068] Referring to FIG. 6, the server (2000) may input a plurality of input data classified by a predetermined content category to, for example, a pre-generated second content recommendation model (602). The server (2000) may, for example, input input data classified as ballad, which is a content category related to music genres, into the second content recommendation model (602) to obtain recommendation results corresponding to the input data. In the process of repeatedly obtaining recommendation results corresponding to the input data classified as ballad, the second content recommendation model (602) may be trained to be specialized for ballads.

[0069] Although not shown in FIG. 6, the server (2000) can train the second content recommendation model (602) to be specialized for rock, Korean pop, and foreign music, respectively, for each input data classified into a predetermined content category related to music genres, such as rock, Korean pop, and foreign music, in the same way as for input data classified as ballad.

[0070] When multiple input data classified by content categories related to music genres and multiple second content recommendation models (602) specialized to correspond to the classified multiple input data are obtained, the server (2000) can perform learning of the personalization model (63) by using the multiple input data classified by content categories related to music genres as input values ​​and using the weights of each specialized multiple second content recommendation model (602) as output values.

[0071] Here, the training of the personalization model (63) can be performed, for example, by adjusting the weights between layers included in the personalization model during the process of outputting the probability of the label of the weight of the second content recommendation model (602) corresponding to the category, in response to input data classified into a specific content category. Additionally, for example, the personalization model (63) can be trained to determine the weight to be output from the personalization model (63) among the weights between labels within the specialized second content recommendation model (602) using input / output data corresponding to a specific content category. In this case, the weight output from the personalization model (63) among the weights within the specialized second content recommendation model (602) may be a weight that has a high influence on specializing the second content recommendation model (602). For example, among the weights within the specialized second content recommendation model (602), a value that differs significantly from the weight of the second content recommendation model (602) before specialization may be determined as the weight to be output from the personalization model (63). Alternatively, for example, among the weights within the specialized second content recommendation model (602), weights that have a high probability of causing the specialized second content recommendation model (602) to output an output value corresponding to a specific content category from input data corresponding to a specific content category may be selected as weights to be output from the personalization model (63). In this case, the personalization model (63) may be trained, for example, using a set of input / output data used to specialize the second content recommendation model (602), and weights of the specialized second content recommendation model based on the input / output data set.

[0072] In this way, the server (2000) according to some embodiments can generate a second content recommendation model (602) and generate a personalization model (63) corresponding to the generated second content recommendation model (602), thereby enabling the electronic device (1000) to receive the second content recommendation model (602) and the personalization model (63) from the server (2000) and to directly personalize the second content recommendation model (602).

[0073] Meanwhile, the electronic device (1000) can obtain an input / output data set of the personalized first content recommendation model (611) by combining input data input to the personalized first content recommendation model (611) and output data output from the personalized first content recommendation model (611).

[0074] The electronic device (1000) that has acquired the input / output data set of the personalized first content recommendation model (611) can input the input / output data set of the personalized first content recommendation model (611) into the personalized model (63) received from the server (2000). As the input / output data set of the personalized first content recommendation model (611) is input into the personalized model (63), the electronic device (1000) can acquire as an output weight data that is used to directly personalize the second content recommendation model (602), that is, weight data representing the weights applied between the layers included in the second content recommendation model (602). At this time, the weight data may be data representing the weights applied between the layers included in the second content recommendation model (602).

[0075] When weight data is obtained through the personalization model (63), the electronic device (1000) can directly personalize the second content recommendation model (602) by changing the weights applied between the layers included in the second content recommendation model (602) based on the obtained weight data.

[0076] FIG. 7 is a diagram illustrating a method for a server to generate a personalized model according to some embodiments.

[0077] Referring to FIGS. 6 and FIGS. 7, the second content recommendation model (702) generated by the server according to some embodiment of FIGS. 7 may include a recommendation layer and a personalization layer, unlike the second content recommendation model (602) generated by the server according to some embodiment of FIGS. 6.

[0078] The recommendation layer included in the second content recommendation model (702) is a layer for content recommendation, and the personalization layer included in the second content recommendation model (702) may be a layer for personalizing output data from the recommendation layer. The recommendation layer included in the second content recommendation model (702) may, for example, be a layer used for performing content recommendation, related learning, and recommendation. Meanwhile, the personalization layer included in the second content recommendation model (702) may, for example, be a layer separate from the recommendation layer and may be a layer used for personalizing the second content recommendation model (702) in addition to content recommendation, related learning, and recommendation. The personalization layer may be used to personalize the second content recommendation model (702) by changing at least some of the weights output from the recommendation layer. Additionally, among the recommendation layer and the personalization layer within the second content recommendation model (702), the personalization layer may be specialized for a specific user, thereby creating a second content recommendation model (722) personalized for a specific user.

[0079] A server (2000) according to some embodiments can generate a personalization model (73) based on a plurality of input data classified by a predetermined user category and a plurality of personalization layers specialized by a predetermined user category to correspond to the classified plurality of input data.

[0080] Referring to FIG. 7, the server (2000) may input a plurality of input data classified by a predetermined user category to, for example, a pre-generated second content recommendation model (702). The server (2000) may, for example, input input data classified as a first user, which is a content category related to the type of user, into the second content recommendation model (702) to obtain recommendation results corresponding to the input data. In the process of repeatedly obtaining recommendation results corresponding to the input data classified as a first user, the personalization layer of the second content recommendation model (702) may be trained to be specialized for the first user.

[0081] Although not illustrated in FIG. 7, the server (2000) can train the personalization layer of the second content recommendation model (702) to be specialized for each of the second user, third user, and fourth user, for each of the input data classified as a second user, third user, and fourth user, for example, in the same way as the input data classified as a first user, for each of the input data classified as a second user, third user, and fourth user.

[0082] When multiple input data classified by content categories related to the type of user and multiple personalized layers specialized to correspond to the classified multiple input data are obtained, the server (2000) can perform training of the personalized model (73) by taking the multiple input data classified by content categories related to the type of user as input and the weights of each specialized multiple personalized layer as output. Accordingly, the personalized model (73) can take the input data of the second content recommendation model (702) as input and output the weights of the personalized layer to change at least some of the weights output from the recommendation layer within the second content recommendation model (702).

[0083] Here, the learning of the personalization model (73) can be performed by adjusting the weights between the layers included in the personalization model (73) in the process of outputting the weights of the personalization layer of the second content recommendation model (702) corresponding to the category, for example, in response to input data classified into a specific content category.

[0084] In this way, the server (2000) according to some embodiments can generate a second content recommendation model (702) and generate a personalization model (73) corresponding to the personalization layer of the generated second content recommendation model (702), thereby enabling the electronic device (1000) to receive the second content recommendation model (702) and the personalization model (73) from the server (2000) and to directly personalize the second content recommendation model (702), while simultaneously performing the personalization of the second content recommendation model (702) using relatively small weight data.

[0085] Meanwhile, the electronic device (1000) can obtain an input / output data set of the personalized first content recommendation model (711) by combining input / output data extracted from the personalized first content recommendation model (711).

[0086] The electronic device (1000) that has acquired the input / output data set of the personalized first content recommendation model (711) can input the input / output data set of the personalized first content recommendation model (711) into the personalized model (73) received from the server (2000) to obtain the weight data of the personalized layer used to directly personalize the second content recommendation model (702) as an output. At this time, the weight data of the personalized layer may be data representing the weights applied between the personalized layers included in the second content recommendation model (702).

[0087] When weight data is obtained through the personalization model (73), the electronic device (1000) can directly personalize the second content recommendation model (702) by changing the weights applied between the personalization layers included in the second content recommendation model (702) based on the weight data of the obtained personalization layer.

[0088] FIG. 8 is a block diagram of an electronic device according to some embodiments.

[0089] Referring to FIG. 8, an electronic device (1000) according to some embodiments may include a communication unit (1001), an input / output unit (1002), a processor (1003), and a memory (1004).

[0090] The communication unit (1001) may include one or more communication modules for communication with the server (2000). For example, the communication unit (1001) may include at least one of a short-range communication unit or a mobile communication unit.

[0091] The short-range wireless communication unit may include, but is not limited to, a Bluetooth communication unit, a BLE (Bluetooth Low Energy) communication unit, a Near Field Communication unit, a WLAN (Wi-Fi) communication unit, a Zigbee communication unit, an infrared (IrDA, infrared Data Association) communication unit, a WFD (Wi-Fi Direct) communication unit, an UWB (ultra wideband) communication unit, an Ant+ communication unit, etc.

[0092] A mobile communication unit transmits and receives wireless signals with at least one of a base station, an external terminal, and a server on a mobile communication network. Here, the wireless signal may include various forms of data resulting from voice call signals, video call call signals, or the transmission and reception of text / multimedia messages.

[0093] The input / output unit (1002) receives input from a user to control the operation of the electronic device (1000) and can output data related to content that can be played on the electronic device (1000) as information that the user can perceive visually and / or audibly.

[0094] The input / output unit (1002) can receive user input by being connected to input devices such as, for example, a key pad, a microphone, a dome switch, a touch pad (contact capacitive type, pressure resistive type, infrared detection type, surface ultrasonic conduction type, integral tension measurement type, piezo effect type, etc.), a jog wheel, a jog switch, etc., but is not limited thereto.

[0095] Additionally, the input / output unit (1002) may be connected to an output device, such as a speaker capable of outputting sound related to a function performed by an electronic device (1000) (e.g., a call signal reception sound, a message reception sound, a notification sound) and content being played, or a display capable of displaying information processed by the electronic device (1000) and content being played, to output data related to the content as information that a user can perceive visually and / or audibly, but is not limited thereto.

[0096] The processor (1003) can control the overall operation of the electronic device (1000). For example, the processor (1003) can control the overall operation of the communication unit (1001), the input / output unit (1002), and the memory (1004) by executing programs stored in the memory (1004).

[0097] The processor (1003) can obtain a first content recommendation model (1102) used to recommend content to a user of an electronic device. The processor (1003) can personalize the first content recommendation model (1102) based on the user's content usage history.

[0098] The processor (1003) can receive a second content recommendation model (1104) and a personalization model (1101) for personalizing the second content recommendation model from the server.

[0099] The processor (1003) can personalize the second content recommendation model (1104) using the input / output data of the personalized first content recommendation model (1103) and the personalization model (1101). The processor (1003) can perform personalization of the second content recommendation model (1104) based on the input / output data of the personalized first content recommendation model (1103) and the personalization model (1101) by, for example, executing a personalization module (1106), which is a program module stored in memory (1004).

[0100] The processor (1003) can provide a content recommendation service to the user using a personalized second content recommendation model (1105).

[0101] A processor (1003) according to some embodiments may, for example, perform artificial intelligence operations. The processor (1003) may be, for example, any one of a CPU (Central Processing Unit), GPU (Graphics Processing Unit), NPU (Neural Processing Unit), FPGA (Field Programmable Gate Array), or ASIC (application specific integrated circuit), but is not limited thereto.

[0102] The memory (1004) can store a program for controlling the operation of the electronic device (1000). The memory (1004) may include at least one instruction for controlling the operation of the electronic device (1000). Programs stored in the memory (1004) may be classified into a plurality of modules according to their functions.

[0103] The memory (1004) can store, for example, a first content recommendation model (1103), a second content recommendation model (1104), and a personalization model (1101) received from a server (2000).

[0104] The memory (1004) can store, for example, a personalized first content recommendation model (1103) generated by the electronic device (1000) repeatedly performing content recommendations based on the first content recommendation model (1103).

[0105] The memory (1004) can store, for example, input / output data of a personalized first content recommendation model (1103) and a personalization module (1106) for performing personalization of a second content recommendation model (1104) using a personalization model (1101).

[0106] The memory (1004) can store, for example, a personalized second content recommendation model (1105) generated by the electronic device (1000) personalizing the second content recommendation model (1104) through the personalization module (1106).

[0107] The memory (2004) may include, for example, a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, a magnetic disk, an optical disk, at least one type of storage medium, but is not limited thereto.

[0108] FIG. 9 is a block diagram of a server according to some embodiments.

[0109] Referring to FIG. 9, a server (2000) according to some embodiments may include a communication unit (2001), a processor (2002), and a memory (2003).

[0110] The communication unit (2001) may include one or more communication modules for communication with the electronic device (1000). For example, the communication unit (2001) may include at least one of a short-range communication unit or a mobile communication unit.

[0111] The short-range wireless communication unit may include, but is not limited to, a Bluetooth communication unit, a BLE (Bluetooth Low Energy) communication unit, a Near Field Communication unit, a WLAN (Wi-Fi) communication unit, a Zigbee communication unit, an infrared (IrDA, infrared Data Association) communication unit, a WFD (Wi-Fi Direct) communication unit, an UWB (ultra wideband) communication unit, an Ant+ communication unit, etc.

[0112] A mobile communication unit transmits and receives wireless signals with at least one of a base station, an external terminal, and a server on a mobile communication network. Here, the wireless signal may include various forms of data resulting from voice call signals, video call call signals, or the transmission and reception of text / multimedia messages.

[0113] The processor (2002) can control the overall operation of the server (2000). For example, the processor (2002) can control the overall operation of the communication unit (2001) and the memory (2003) by executing programs stored in the memory (2003).

[0114] The processor (2002) can generate a content recommendation model to deliver to the electronic device (1000).

[0115] The processor (2002) can generate a content recommendation model to be delivered to the electronic device (1000). The processor (2002) can generate a content recommendation model to be delivered to the electronic device (1000) by executing, for example, a content recommendation model generation module (2101), which is a program module stored in memory (2003). The content recommendation model generated by the processor (2002) is delivered to the electronic device (1000) through the communication unit (2001) and can be used to provide a content recommendation service for the electronic device (1000).

[0116] The processor (2002) can generate a personalized model to be delivered to the electronic device (1000). The processor (2002) can generate a personalized model to be delivered to the electronic device (1000) by executing, for example, a personalized model generation module (2102), which is a program module stored in memory (2003). The personalized model generated by the processor (2002) is delivered to the electronic device (1000) through the communication unit (2001) and can be used for personalizing the content recommendation model of the electronic device (1000).

[0117] The memory (2003) can store a program for controlling the operation of the server (2000). The memory (2003) may include at least one instruction for controlling the operation of the server (2000). The programs stored in the memory (2003) may be classified into multiple modules according to their functions.

[0118] The memory (2003) can store, for example, a content recommendation model generation module (2101) used to generate a content recommendation model to be delivered to an electronic device (1000).

[0119] The memory (2003) can store, for example, a personalization model generation module (2102) used to generate a personalization model to be delivered to an electronic device (1000).

[0120] The memory (2003) may include, for example, a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), RAM (Random Access Memory), SRAM (Static Random Access Memory), ROM (Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), PROM (Programmable Read-Only Memory), magnetic memory, a magnetic disk, an optical disk, at least one type of storage medium, but is not limited thereto.

[0121] Some embodiments may also be implemented in the form of a recording medium containing computer-executable instructions, such as program modules executed by a computer. A computer-readable medium may be any available medium accessible by a computer and includes both volatile and non-volatile media, and both removable and non-removable media. Additionally, a computer-readable medium may include a computer storage medium. A computer storage medium includes both volatile and non-volatile, removable and non-removable media implemented by any method or technique for storing information, such as computer-readable instructions, data structures, program modules, or other data.

[0122] Additionally, computer-readable storage media may be provided in the form of non-transitory storage media. Here, 'non-transitory storage media' simply means that it is a tangible device and does not contain a signal (e.g., electromagnetic waves), and this term does not distinguish between cases where data is stored semi-permanently and cases where it is stored temporarily. For example, 'non-transitory storage media' may include a buffer in which data is stored temporarily.

[0123] According to one embodiment, the method according to the various embodiments disclosed herein may be provided by being included in a computer program product. The computer program product may be traded between a seller and a buyer as a product. The computer program product may be distributed in the form of a device-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or distributed online (e.g., download or upload) through an application store (e.g., Play Store™) or directly between two user devices (e.g., smartphones). In the case of online distribution, at least a portion of the computer program product (e.g., downloadable app) may be temporarily stored or temporarily created on a device-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or a relay server.

[0124] Additionally, in this specification, “part” may be a hardware component, such as a processor or circuit, and / or a software component executed by a hardware component, such as a processor.

[0125] The foregoing description of the present disclosure is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present disclosure. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0126] The scope of the present disclosure is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and equivalent concepts thereof should be interpreted as being included within the scope of the present disclosure.

Claims

Claim 1 A method for personalizing a content recommendation model by means of at least one processor of an electronic device, comprising: acquiring a first content recommendation model used to recommend content to a user of the electronic device; personalizing the first content recommendation model based on the user's content usage history; receiving a second content recommendation model from a server; receiving a personalization model from the server for personalizing the second content recommendation model; combining input / output data extracted from the personalized first content recommendation model to acquire an input / output data set of the personalized first content recommendation model; inputting the input / output data set of the personalized first content recommendation model into the personalization model to acquire weight data to be applied between a plurality of neural network layers of the second content recommendation model as an output; personalizing the second content recommendation model by applying weights between a plurality of neural network layers of the second content recommendation model according to the weight data; and providing a content recommendation service to the user using the personalized second content recommendation model; wherein the content recommendation model, the first content recommendation model, the second content recommendation model, and the personalization model are artificial intelligence models. Claim 2 A method according to claim 1, wherein the input / output data of the personalized first content recommendation model includes input / output data corresponding to the user's content usage history. Claim 3 A method according to claim 1, wherein the personalization model is generated based on a plurality of input data classified by a predetermined content category and a plurality of second content recommendation models specialized by a predetermined content category to correspond to the classified plurality of input data. Claim 4 delete Claim 5 A method according to claim 1, wherein the second content recommendation model includes a content recommendation layer and a personalization layer, and the personalization model is generated based on a plurality of input data classified by a predetermined user category and a plurality of personalization layers specialized by a predetermined user category to correspond to the classified plurality of input data. Claim 6 In claim 5, the operation of personalizing the second content recommendation model comprises: the operation of inputting input / output data of the personalized first content recommendation model into the personalization model; the operation of obtaining weight data of a personalization layer, which is data representing weights applied between personalization layers included in the second content recommendation model; and the operation of personalizing the second content recommendation model by changing the weights applied between personalization layers included in the second content recommendation model based on the obtained weight data of the personalization layer. Claim 7 In an electronic device for personalizing a content recommendation model, a memory for storing one or more instructions; The electronic device comprises: a processor that executes one or more of the above instructions; wherein the processor, by executing the one or more of the above instructions, obtains a first content recommendation model used to recommend content to a user of the electronic device, personalizes the first content recommendation model based on the user's content usage history, receives a second content recommendation model from a server, receives a personalization model for personalizing the second content recommendation model from the server, combines input / output data extracted from the personalized first content recommendation model to obtain an input / output data set of the personalized first content recommendation model, inputs the input / output data set of the personalized first content recommendation model into the personalization model to obtain weight data to be applied between a plurality of neural network layers of the second content recommendation model, and personalizes the second content recommendation model by applying weights between a plurality of neural network layers of the second content recommendation model according to the weight data, and provides a content recommendation service to the user using the personalized second content recommendation model, wherein the content recommendation model, the first content recommendation model, the second content recommendation model, and the personalization model are artificial intelligence models. Claim 8 An electronic device according to claim 7, wherein the input / output data of the personalized first content recommendation model includes input / output data corresponding to the user's content usage history. Claim 9 An electronic device according to claim 7, wherein the personalization model is generated based on a plurality of input data classified by a predetermined content category and a plurality of second content recommendation models specialized by a predetermined content category to correspond to the classified plurality of input data. Claim 10 delete Claim 11 An electronic device according to claim 7, wherein the second content recommendation model includes a content recommendation layer and a personalization layer, and the personalization model is generated based on a plurality of input data classified by a predetermined user category and a plurality of personalization layers specialized by a predetermined user category to correspond to the classified plurality of input data. Claim 12 An electronic device according to claim 11, wherein the processor inputs input / output data of the personalized first content recommendation model into the personalized model, obtains weight data of a personalized layer which is data representing weights applied between personalized layers included in the second content recommendation model, and personalizes the second content recommendation model by changing the weights applied between personalized layers included in the second content recommendation model based on the obtained weight data of the personalized layer. Claim 13 A computer-readable recording medium comprising: an operation of obtaining a first content recommendation model used to recommend content to a user of an electronic device; an operation of personalizing the first content recommendation model based on the user's content usage history; an operation of receiving a second content recommendation model from a server; an operation of receiving a personalization model from the server for personalizing the second content recommendation model; an operation of obtaining an input / output data set of the personalized first content recommendation model by combining input / output data extracted from the personalized first content recommendation model; an operation of obtaining weight data to be applied between a plurality of neural network layers of the second content recommendation model by inputting the input / output data set of the personalized first content recommendation model into the personalization model; an operation of personalizing the second content recommendation model by applying weights between a plurality of neural network layers of the second content recommendation model according to the weight data; and an operation of providing a content recommendation service to the user using the personalized second content recommendation model; wherein the content recommendation model, the first content recommendation model, the second content recommendation model, and the personalization model are a computer-readable recording medium that records a program for executing on a computer a method for personalizing a content recommendation model which is an artificial intelligence model.

Images