KR20210103735A - Method and device for recommending a poster - Google Patents

Abstract

Disclosed are a poster recommending method and a device thereof. According to an embodiment, a poster recommending method comprises: a step of receiving content purchase information and content inquiry information of a user; a step of applying the content purchase information and the content inquiry information to a neural network model; a step of predicting preference of one or more posters included in target content based on an output value of the neural network model; and a step of determining a poster predicted to have the highest user preference among the one or more posters as a main poster based on the preference of the posters.

Description

POSTER RECOMMENDATION METHOD AND DEVICE FOR RECOMMENDING A POSTER

The following embodiments relate to a poster recommendation method and apparatus, and more specifically, to a deep learning model architecture for determining a personalized poster.

In the past, there are several posters per VOD content, but the same poster is provided to all customers at once, so it is not possible to use factors that can increase the viewing/purchase rate compared to the exposure of the main screen by reflecting personal preferences.

As a related prior art, there is Korean Patent Application Laid-Open No. 10-2007-0079211 (Title of Invention: Device and Method for Recommending Multimedia Content, Applicant: Samsung Electronics Co., Ltd.). The publication discloses a technology for recommending content based on a user's preference for content calculated by learning physical characteristic information of content and user feedback information through a multi-layer neural network.

Embodiments intend to provide a personalized poster that can lead to viewing/purchase by analyzing IPTV service usage history.

Embodiments attempt to increase sales and customer service satisfaction by exposing personalized posters to induce VOD viewing.

Embodiments seek to design a deep learning model architecture for determining a personalized poster.

A poster recommendation method according to an embodiment includes: receiving content purchase information and content inquiry information of a user; applying the content purchase information and the content inquiry information to a neural network model; predicting a preference of one or more posters included in target content based on an output value of the neural network model; and determining, as a main poster, a poster predicted to have the highest user preference among the one or more posters of the target content, based on the preference of the poster.

The applying to the neural network model may include generating a poster preference vector for each of the plurality of contents of the user based on the contents purchase information and the contents inquiry information.

The predicting of the poster preference may include predicting the preference of one or more posters included in the target content based on the poster preference vector.

The applying to the neural network model may include: calculating a first weight based on the content purchase information and one or more poster information included in each of the plurality of contents; and calculating a second weight based on the first weight and the content purchase information.

Calculating the first weight may include calculating the first weight by cross-producting a content purchase information vector corresponding to the content purchase information and a poster information vector corresponding to the poster information.

Calculating the second weight may include calculating the second weight by cross-producting the first weight with a content inquiry information vector corresponding to the content inquiry information.

The predicting of the poster preference may include predicting the preference of one or more posters included in the target content based on the second weight and the content purchase information.

The poster recommendation method according to an embodiment may further include embedding the user's content purchase information and the content inquiry information.

A method for learning a neural network model according to an embodiment includes receiving one or more pieces of poster information included in each of a plurality of contents; receiving content purchase information and content inquiry information of a plurality of users; calculating a poster preference vector for each of the plurality of contents of the plurality of users based on the poster information, the contents purchase information, and the contents inquiry information; and learning a neural network model based on the poster preference vector.

Calculating the poster preference vector may include: calculating a first weight based on the content purchase information of the plurality of users and the poster information; and calculating a second weight based on the first weight and the content purchase information of the plurality of users.

Calculating the first weight may include calculating the first weight by cross-producting a content purchase information vector corresponding to the content purchase information of the plurality of users and a poster information vector corresponding to the poster information. have.

Calculating the second weight may include calculating the second weight by cross-producting the first weight with a content inquiry information vector corresponding to the content inquiry information of the plurality of users.

The poster recommendation service server according to an embodiment receives the user's content purchase information and content inquiry information, applies the content purchase information and the content inquiry information to a neural network model, and based on an output value of the neural network model, a target a processor for predicting a preference of one or more posters included in the content, and determining a poster predicted to have the highest preference of the user among the one or more posters of the target content as a main poster based on the preference of the poster can do.

The processor may generate a poster preference vector for each of the plurality of contents of the user based on the contents purchase information and the contents inquiry information.

The processor may calculate a first weight based on the content purchase information and one or more poster information included in each of the plurality of contents, and calculate a second weight based on the first weight and the content purchase information. .

The processor may calculate the first weight by cross-producting a content purchase information vector corresponding to the content purchase information and a poster information vector corresponding to the poster information.

The processor may calculate the second weight by cross-producting the content inquiry information vector corresponding to the content inquiry information and the first weight.

The apparatus for training a neural network model for poster recommendation according to an embodiment receives at least one poster information included in each of a plurality of contents, receives contents purchase information and contents inquiry information of a plurality of users, and the poster information , a processor for calculating a poster preference vector for each of the plurality of contents of the plurality of users based on the contents purchase information and the contents inquiry information, and learning a neural network model based on the poster preference vector.

The processor may calculate a first weight based on the content purchase information and the poster information of the plurality of users, and calculate a second weight based on the first weight and the content purchase information of the plurality of users. have.

The processor may calculate the first weight by cross-producting a content purchase information vector corresponding to the content purchase information of the plurality of users and a poster information vector corresponding to the poster information.

The processor may calculate the second weight by cross-producting the content inquiry information vector corresponding to the content inquiry information of the plurality of users and the first weight.

Embodiments may provide a personalized poster that can lead to viewing/purchasing by analyzing IPTV service usage history.

Embodiments may increase sales and customer service satisfaction by exposing a personalized poster to induce VOD viewing.

Embodiments may design a deep learning model architecture for determining a personalized poster.

1 is a diagram for explaining a content service system according to an embodiment.
2 is a flowchart for explaining the operation of the service server 110 according to an embodiment.
3A to 3B are diagrams for explaining an operation and a learning method of a neural network model according to an embodiment.
4 is a diagram illustrating a block diagram of a service server according to an embodiment.

Specific structural or functional descriptions disclosed in this specification are merely illustrative for the purpose of describing embodiments according to technical concepts, and the embodiments may be embodied in various other forms and are limited to the embodiments described herein. doesn't happen

Terms such as first or second may be used to describe various elements, but these terms should be understood only for the purpose of distinguishing one element from another element. For example, a first component may be termed a second component, and similarly, a second component may also be termed a first component.

When a component is referred to as being “connected” or “connected” to another component, it is understood that the other component may be directly connected or connected to the other component, but other components may exist in between. it should be On the other hand, when it is mentioned that a certain element is "directly connected" or "directly connected" to another element, it should be understood that the other element does not exist in the middle. Expressions describing the relationship between elements, for example, “between” and “between” or “neighboring to” and “directly adjacent to”, etc. should be interpreted similarly.

The singular expression includes the plural expression unless the context clearly dictates otherwise. In this specification, terms such as "comprise" or "have" are intended to designate that the described feature, number, step, operation, component, part, or combination thereof exists, and includes one or more other features or numbers, It should be understood that the possibility of the presence or addition of steps, operations, components, parts or combinations thereof is not precluded in advance.

Unless defined otherwise, all terms used herein, including technical and scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present specification. does not

The embodiments may be implemented in various types of products, such as personal computers, laptop computers, tablet computers, smart phones, televisions, smart home appliances, intelligent cars, kiosks, wearable devices, and the like. Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. Like reference numerals in each figure indicate like elements.

1 is a diagram for explaining a content service system according to an embodiment.

Referring to FIG. 1 , a content service system 100 according to an embodiment includes a service server 110 and content reproducing devices 120-1 to 120-n.

The content service system 100 may support an Internet Protocol TV (IPTV) system. The IPTV system may be an interactive broadcasting and communication service that provides video contents such as movies and broadcast programs using a high-speed Internet network, or provides various multimedia services such as Internet searches and games to a television receiver. One or more of the content playback devices 120 - 1 to 120 - n may include a set-top box and a display. The service server 110 may transmit content (eg, real-time content and/or on-demand content) to a set-top box, and the set-top box may output content to a display. Accordingly, the user or viewer can watch the content.

Also, the content service system 100 may support a mobile content system. One or more of the content playback devices 120-1 to 120-n may represent a mobile terminal (eg, a smart phone, a tablet, etc.), and the service server 110 may transmit content to the mobile terminal.

The content playback devices 120-1 to 120-n may communicate with other content playback devices and/or the service server 110 through a network using a wireless or wired communication method. The communication method is not limited, and not only a communication method using a communication network (eg, a mobile communication network, a wired Internet, a wireless Internet, a broadcasting network) that the network may include, but also short-range wireless communication between devices may be included. For example, the network is a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a broadband network (BBN), the Internet, etc. may include any one or more of the networks of Further, the network may include, but is not limited to, any one or more of a network topology including, but not limited to, a bus network, a star network, a ring network, a mesh network, a star-bus network, a tree or a hierarchical network, and the like. .

The service server 110 may be implemented as a computer device or a plurality of computer devices that communicate with the content playback devices 120-1 to 120-n through a network to provide commands, codes, files, content, services, etc. have.

In the past, there are several posters per content, but the same poster is provided to all customers at once, so it is not possible to use factors that can increase the viewing/purchasing rate compared to the exposure of the main screen by reflecting personal preferences.

According to an embodiment, the service server 110 may provide a personalized poster that can lead to viewing/purchase by analyzing the IPTV service usage history. As will be described in detail below, the service server 110 uses the poster predicted to have the highest user preference from among a plurality of posters of specific content based on the user's content purchase information and content inquiry information, as the main poster of the corresponding content. can decide Hereinafter, the operation of the service server 110 will be described in detail.

2 is a flowchart for explaining the operation of the service server 110 according to an embodiment.

Referring to FIG. 2 , steps 210 to 240 may be performed through the service server 110 described above with reference to FIG. 1 . The service server 110 according to an embodiment may be implemented by one or more hardware modules, one or more software modules, or various combinations thereof.

In step 210, the service server 110 receives the user's content purchase information and content inquiry information. The user's content purchase information may include, for example, information about the VOD content purchased by the customer through the VOD service provided by the service server 110 . Furthermore, the content purchase information may further include price information, discount information, and the like of the content at the time of purchase, as well as whether the content is purchased.

When a user selects a specific content from a list of a plurality of content consisting of a main poster representing the content, the user can check detailed information about the content and decide whether to purchase it. The user's content inquiry information may be information obtained by selecting specific content from a plurality of content lists and inquiring the corresponding content. Since the user determines whether or not to select content based on the main poster of the content list, the user's content inquiry information may include preference information for the main poster.

In step 220, the service server 110 may apply the content purchase information and the content inquiry information to the neural network model. The neural network model may be a pre-trained model to determine an optimal poster for each content of the user. The operation and learning method of the neural network model will be described in detail with reference to FIGS. 3A to 3B below. The service server 110 may generate a poster preference vector for each of the plurality of contents of the user based on the contents purchase information and the contents inquiry information.

In step 230, the service server 110 predicts the preference of one or more posters included in the target content based on the output value of the neural network model. The service server 110 may predict the preference of one or more posters included in the target content based on the poster preference vector.

In step 240 , the service server 110 determines, as the main poster, a poster that is predicted to have the highest user preference among one or more posters of the target content based on the poster preference. Through this, the service server 110 may provide a personalized poster that can lead to viewing/purchasing for each user.

3A to 3B are diagrams for explaining an operation and a learning method of a neural network model according to an embodiment.

Referring to FIG. 3A , the learning apparatus according to an embodiment may receive one or more pieces of poster information included in each of a plurality of contents (VOD 1, VOD 2, ??, VOD N). The poster information may include information on all posters of the corresponding content (eg, a plurality of posters such as a character-oriented poster, a background-oriented poster, and a letter-oriented poster).

Learning apparatus An apparatus for learning a neural network model, it may be implemented by one or more hardware modules, one or more software modules, or various combinations thereof. The learning device may be provided in the service server 110 , or the service server 110 may serve as a learning device by itself.

The learning apparatus may receive content purchase information and content inquiry information of a plurality of users. The learning apparatus may learn the neural network based on one or more poster information included in each of the plurality of contents (VOD 1, VOD 2, ??, and VOD N), content purchase information of a plurality of users, and content inquiry information. .

When the learning device learns a neural network model using information of a plurality of users to predict a custom poster for a specific user, the learning device can predict a custom poster for a specific user using the learning information of users who have similar tastes to the specific user. have.

The neural network model may include an input layer, an interest layer, and a Deep Neural Network (DNN) layer. Each layer includes a plurality of nodes, and nodes between adjacent layers may be connected to each other with a connection weight. Each node may operate based on an activation model. An output value corresponding to an input value may be determined according to the activation model. An output value of an arbitrary node may be input to a node of the next layer connected to the node. A node of the next layer may receive values output from a plurality of nodes. In a process where an output value of an arbitrary node is input to a node of a next layer, a connection weight may be applied. The node of the next layer may output an output value corresponding to the input value to the node of the next layer connected to the node based on the activation model.

The learning apparatus may embed (eg, digitize) information to be applicable to the learning model by using the input layer. For example, the input layer may embed the user's content purchase information and content inquiry information in a vector form. Furthermore, the input layer may embed poster information in a vector form.

The learning apparatus may calculate a poster preference vector for each content of a plurality of users by using the preference layer. The learning apparatus may calculate a poster preference vector for each content of a plurality of users based on the poster information, the content purchase information, and the content inquiry information. _{For example, the learning apparatus may calculate a poster preference vector (R nk} _A) for the k-th poster of customer A's VOD N by using the preference layer.

Referring to FIG. 3B , the learning apparatus may calculate a first weight based on content purchase information and poster information of a plurality of users. For example, the learning apparatus may input content purchase information and poster information of a plurality of users into an interest unit provided for each content. The learning apparatus may calculate the first weight by cross-producting the content purchase information vector corresponding to the content purchase information of a plurality of users and the poster information vector corresponding to the poster information. For example, in the case of content actually purchased by the user, the weight of posters included in the content may be calculated to be high.

Furthermore, the learning apparatus may calculate the second weight based on the first weight and the content inquiry information of the plurality of users. For example, the learning apparatus may calculate the second weight by cross-producting the first weight and the content inquiry information vector corresponding to the content inquiry information of a plurality of users. For example, since there is a high probability that the main poster of the content searched by the user is a poster preferred by the user, the weight of the corresponding poster may be calculated to be high.

The poster preference vector according to an embodiment may be a vector composed of a second weight for a poster for each of the plurality of contents by a plurality of users.

The learning apparatus may learn the DNN layer based on the poster preference vector obtained from the preference layer and content purchase information of a plurality of users. For example, the learning apparatus may learn the weights of the DNN layer so that the difference between the optimal poster predicted based on the poster preference vector and the correct answer data is minimized.

Referring again to FIG. 3A , the service server 110 may further receive a user ID, a content to be determined whether the user purchases or not, and a poster ID. The service server 110 receiving the corresponding information may predict the user's preference (eg, purchase probability) for the corresponding content and poster by using the learned neural network model.

The service server 110 may embed (eg, digitize) information to be applicable to the learning model using the input layer. For example, the input layer may embed the user's content purchase information and content inquiry information in a vector form. Furthermore, the input layer may embed poster information in a vector form.

The service server 110 may calculate a poster preference vector for each content of a plurality of users by using the preference layer. The service server 110 may apply the user's content purchase information and content inquiry information to the neural network model learned as described above. The service server 110 according to an embodiment may calculate a first weight based on content purchase information and one or more poster information included in each of the plurality of contents, and may calculate a first weight based on the first weight and content purchase information. 2 weights can be calculated. Specifically, the service server 110 may calculate the first weight by cross-producting the content purchase information vector corresponding to the content purchase information and the poster information vector corresponding to the poster information, and the content inquiry information vector corresponding to the content inquiry information. and the first weight may be cross product to calculate the second weight.

The service server 110 may predict the preference of one or more posters included in the target content by using the DNN layer. The service server 110 may predict the preference of one or more posters included in the target content based on the second weight and the user's content purchase information. For example, the service server 110 may output the VOD-poster combination preferred by the user as a preference score by using the DNN layer.

4 is a diagram illustrating a block diagram of a service server according to an embodiment.

Referring to FIG. 4 , the service server 110 according to an embodiment includes a processor 410 . The service server 110 may further include a communication interface and a memory 420 . The processor 410 , the memory 420 and the communication interface may communicate with each other via a communication bus.

The processor 410 receives the user's content purchase information and content inquiry information, applies the content purchase information and content inquiry information to the neural network model, and selects one or more posters included in the target content based on the output value of the neural network model. A preference is predicted, and a poster predicted to have the highest user preference among one or more posters of the target content is determined as the main poster based on the poster preference.

The processor 410 may generate a poster preference vector for each of the plurality of contents of the user based on the contents purchase information and the contents inquiry information.

The processor 410 may calculate a first weight based on the content purchase information and one or more poster information included in each of the plurality of contents, and may calculate a second weight based on the first weight and the content purchase information.

The processor 410 may calculate the first weight by cross-producting the content purchase information vector corresponding to the content purchase information and the poster information vector corresponding to the poster information.

The processor 410 may calculate the second weight by cross-producting the content purchase information vector corresponding to the content purchase information and the first weight.

The memory 420 may store various pieces of information generated in the process of the above-described processor 410 . In addition, the memory 420 may store various data and programs. The memory 420 may include a volatile memory or a non-volatile memory. The memory 420 may include a mass storage medium such as a hard disk to store various data.

In addition, the processor 410 may perform at least one method described above with reference to FIGS. 1 to 3B or an algorithm corresponding to the at least one method. The processor 410 may execute a program and control the service server 110 . The program code executed by the processor 410 may be stored in the memory 420 . The service server 110 may be connected to an external device (eg, a personal computer or a network) through an input/output device (not shown) and exchange data.

The embodiments described above may be implemented by a hardware component, a software component, and/or a combination of a hardware component and a software component. For example, the apparatus, methods and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate (FPGA). array), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For convenience of understanding, although one processing device is sometimes described as being used, one of ordinary skill in the art will recognize that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that can include For example, the processing device may include a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as parallel processors.

The software may comprise a computer program, code, instructions, or a combination of one or more thereof, which configures a processing device to operate as desired or is independently or collectively processed You can command the device. The software and/or data may be any kind of machine, component, physical device, virtual equipment, computer storage medium or device, to be interpreted by or to provide instructions or data to the processing device. , or may be permanently or temporarily embody in a transmitted signal wave. The software may be distributed over networked computer systems, and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and carry out program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like.

As described above, although the embodiments have been described with reference to the limited drawings, those skilled in the art may apply various technical modifications and variations based on the above. For example, the described techniques are performed in a different order than the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (22)

receiving the user's content purchase information and content inquiry information;
applying the content purchase information and the content inquiry information to a neural network model;
predicting a preference of one or more posters included in target content based on an output value of the neural network model; and
determining, as a main poster, a poster predicted to have the highest user preference among the one or more posters of the target content, based on the preference of the poster
How to recommend a poster that includes.
According to claim 1,
The step of applying the neural network model is
generating a poster preference vector for each of the plurality of contents of the user based on the contents purchase information and the contents inquiry information;
Including, poster recommendation method.
3. The method of claim 2,
The step of predicting the preference of the poster
Predicting the preference of one or more posters included in the target content based on the poster preference vector
Including, poster recommendation method.
According to claim 1,
The step of applying the neural network model is
calculating a first weight based on the content purchase information and one or more poster information included in each of the plurality of contents; and
calculating a second weight based on the first weight and the content purchase information
Including, poster recommendation method.
5. The method of claim 4,
The step of calculating the first weight is
calculating the first weight by cross-producting a content purchase information vector corresponding to the content purchase information and a poster information vector corresponding to the poster information;
Including, poster recommendation method.
5. The method of claim 4,
Calculating the second weight includes:
calculating the second weight by cross-producting the content inquiry information vector corresponding to the content inquiry information and the first weight;
Including, poster recommendation method.
5. The method of claim 4,
The step of predicting the preference of the poster
Predicting the preference of one or more posters included in the target content based on the second weight and the content purchase information
Including, poster recommendation method.
According to claim 1,
Embedding the user's content purchase information and the content inquiry information
A poster recommendation method that further includes.
receiving one or more pieces of poster information included in each of a plurality of contents;
receiving content purchase information and content inquiry information of a plurality of users;
calculating a poster preference vector for each of the plurality of contents of the plurality of users based on the poster information, the contents purchase information, and the contents inquiry information; and
learning a neural network model based on the poster preference vector.
A training method of a neural network model for poster recommendation, including
10. The method of claim 9,
The step of calculating the poster preference vector is
calculating a first weight based on the content purchase information of the plurality of users and the poster information; and
calculating a second weight based on the first weight and the content purchase information of the plurality of users
A training method of a neural network model for poster recommendation, including
11. The method of claim 10,
The step of calculating the first weight is
calculating the first weight by cross-producting a content purchase information vector corresponding to the content purchase information of the plurality of users and a poster information vector corresponding to the poster information;
Including, poster recommendation method.
11. The method of claim 10,
Calculating the second weight includes:
calculating the second weight by cross-producting the content inquiry information vector corresponding to the content inquiry information of the plurality of users and the first weight;
Including, poster recommendation method.
A computer program stored on a medium in combination with hardware to execute the method of any one of claims 1 to 12.
Receive the user's content purchase information and content inquiry information, apply the content purchase information and the content inquiry information to a neural network model, and determine the preference of one or more posters included in the target content based on the output value of the neural network model A processor that predicts, and determines, as a main poster, a poster predicted to have the highest user preference among the one or more posters of the target content based on the poster preference.
A poster recommendation service server that includes.
15. The method of claim 14,
the processor
A poster recommendation service server for generating a poster preference vector for each of the plurality of contents of the user based on the contents purchase information and the contents inquiry information.
15. The method of claim 14,
the processor
A poster recommendation service server that calculates a first weight based on the content purchase information and one or more poster information included in each of the plurality of contents, and calculates a second weight based on the first weight and the content purchase information .
17. The method of claim 16,
the processor
and calculating the first weight by cross-producting a content purchase information vector corresponding to the content purchase information and a poster information vector corresponding to the poster information.
17. The method of claim 16,
the processor
A poster recommendation service server for calculating the second weight by cross-producting the content inquiry information vector corresponding to the content inquiry information and the first weight.
Receive at least one poster information included in each of a plurality of contents, receive contents purchase information and contents inquiry information of a plurality of users, and receive the plurality of pieces of information based on the poster information, the contents purchase information, and the contents inquiry information A processor that calculates a poster preference vector for each of the plurality of contents of users and learns a neural network model based on the poster preference vector
A training device for a neural network model for poster recommendation, including.
20. The method of claim 19,
the processor
poster recommendation, calculating a first weight based on the content purchase information of the plurality of users and the poster information, and calculating a second weight based on the first weight and the content purchase information of the plurality of users A training device for neural network models for
21. The method of claim 20,
the processor
and calculating the first weight by cross-producting a content purchase information vector corresponding to the content purchase information of the plurality of users and a poster information vector corresponding to the poster information.
21. The method of claim 20,
the processor
The apparatus for learning a neural network model for poster recommendation, wherein the second weight is calculated by cross-producting the first weight and a content inquiry information vector corresponding to the content inquiry information of the plurality of users.

Images