KR20230106354A - Apparatus for recommending cosmetic contents based on artificial intelligence model and operating method thereof - Google Patents

Abstract

The present disclosure relates to a method for recommending cosmetic content based on an artificial intelligence model and an electronic device for performing the same. The method for an electronic device to recommend cosmetic content based on an artificial intelligence model according to an embodiment comprises the steps of: obtaining user information about at least one of the user's characteristics or the user's cosmetics consumption history based on user input to the electronic device; determining a user vector for the obtained user information; inputting the user vector to the artificial intelligence model which outputs cosmetic information including at least one recommended cosmetic product according to the user information and a score regarding the likelihood which the user will consume the at least one recommended cosmetic product when the user vector is input, and obtaining the cosmetics information from an artificial intelligence model; and outputting a predetermined number of cosmetics showing high scores within the cosmetics information as the cosmetics content. Accordingly, the present invention may improve the recommendation performance of cosmetics by using continuously accumulated data about cosmetics.

Description

Electronic device for recommending cosmetic contents based on artificial intelligence model and method for operating the same

The present disclosure relates to a method for recommending cosmetic contents based on an artificial intelligence model and an electronic device performing the same. More specifically, it relates to an electronic device for recommending customized cosmetics based on metadata analysis and an operating method thereof.

Despite the great demand to find cosmetics that are perfect for the skin or better than the cosmetics currently used, customers have limitations in comparing numerous cosmetics one by one, and the number of comparison groups per person is inevitably limited. There is a problem in that it takes a lot of time and money to find a cosmetic suitable for oneself among numerous cosmetics.

Conventionally, there have been attempts to develop technologies to recommend cosmetics suitable for customers, but conventional recommendation systems do not reflect information on the consumption history of cosmetics used by actual users over time, or it is difficult to apply these situations to the service. There was a limit to recommending necessary cosmetics.

On the other hand, as the development of artificial intelligence technology that can derive a predetermined output value for big data by learning a large amount of data on its own is activated, various technologies to improve user convenience by utilizing artificial intelligence technology in platform services are being developed

Therefore, there is a demand for technology development to provide user-customized cosmetic contents based on continuously accumulated data according to the user's consumption history by utilizing artificial intelligence technology.

Korean Registered Patent No. 2253750

According to an embodiment, a method for recommending cosmetics contents based on an artificial intelligence model by an electronic device and an electronic device performing the same may be provided.

According to an embodiment, a method for recommending cosmetics contents using an artificial intelligence model that can be continuously learned based on metadata analysis results and feedback on user consumption history, and an electronic device performing the same may be provided.

As a technical means for achieving the above-described technical problem, according to an embodiment, in a method for recommending cosmetics contents based on an artificial intelligence model by an electronic device, based on a user input to the electronic device, the user's characteristics or acquiring user information about at least one of the user's cosmetics consumption history; determining a user vector for the obtained user information; When the user vector is input, the user vector is sent to an artificial intelligence model that outputs cosmetic information including at least one recommended cosmetic product according to the user information and a score related to the possibility that the user will consume the at least one recommended cosmetic product. acquiring the cosmetics information from the artificial intelligence model by inputting the information; and outputting a predetermined number of cosmetics representing a high score in the cosmetics information as the cosmetics content. Including, a method may be provided.

According to another embodiment for achieving the above-described technical problem, an electronic device for recommending cosmetics contents based on an artificial intelligence model, comprising: a memory for storing one or more instructions; and at least one processor executing the one or more instructions; wherein, by executing the one or more instructions, the at least one processor acquires user information about at least one of a user's characteristic or a user's cosmetics consumption history based on a user input to the electronic device; Determines a user vector for the acquired user information, and if the user vector is input, cosmetics information including at least one recommended cosmetic product according to the user information and a score related to the possibility that the user will consume the at least one recommended cosmetic product An electronic device that obtains the cosmetics information from the artificial intelligence model by inputting the user vector into an artificial intelligence model that outputs, and outputs a predetermined number of cosmetics representing high scores in the cosmetics information as the cosmetics content. may be provided.

According to another embodiment, in a method for recommending cosmetics contents based on an artificial intelligence model by an electronic device, based on a user input to the electronic device, at least one of a user's characteristic or a user's cosmetics consumption history obtaining user information; determining a user vector for the obtained user information; When the user vector is input, the user vector is sent to an artificial intelligence model that outputs cosmetic information including at least one recommended cosmetic product according to the user information and a score related to the possibility that the user will consume the at least one recommended cosmetic product. acquiring the cosmetics information from the artificial intelligence model by inputting the information; and outputting a predetermined number of cosmetics representing a high score in the cosmetics information as the cosmetics contents. A computer-readable recording medium recording a program for executing the method on a computer, including, may be provided.

Unlike conventional cosmetics recommendation techniques, cosmetics recommendation performance can be improved by using continuously accumulated data on cosmetics.

According to an embodiment of the present disclosure, an increase in cosmetics consumption may be induced by providing user-customized cosmetics content.

1 is a diagram schematically illustrating a process in which an electronic device provides cosmetics contents using an artificial intelligence model according to an embodiment.
2 is a diagram schematically illustrating a process in which an electronic device collects and utilizes meta data according to an embodiment.
3 is a flowchart of a method of recommending cosmetics contents based on an artificial intelligence model by an electronic device according to an embodiment.
4 is a diagram illustrating a process of learning an artificial intelligence model by an electronic device according to an embodiment.
5 is a diagram illustrating a process of re-learning an artificial intelligence model based on cosmetics consumption history information by an electronic device according to an embodiment.
6 is a diagram for explaining user information used by an electronic device and user vectors corresponding thereto, according to an exemplary embodiment.
7 is a diagram for explaining a learning and evaluation process of an artificial intelligence model used by an electronic device according to an embodiment.
8 is a diagram for explaining an operation of a context-based language model used by an electronic device according to an exemplary embodiment.
9 is a block diagram of an electronic device according to an embodiment.

Terms used in this specification will be briefly described, and the present disclosure will be described in detail.

The terms used in the present disclosure have been selected from general terms that are currently widely used as much as possible while considering the functions in the present disclosure, but they may vary according to the intention or precedent of a person skilled in the art, the emergence of new technologies, and the like. In addition, in a specific case, there is also a term arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the description of the invention. Therefore, terms used in the present disclosure should be defined based on the meaning of the term and the general content of the present disclosure, not simply the name of the term.

When it is said that a certain part "includes" a certain component throughout the specification, it means that it may further include other components without excluding other components unless otherwise stated. In addition, terms such as "...unit" and "module" described in the specification mean a unit that processes at least one function or operation, which may be implemented as hardware or software or a combination of hardware and software. .

Hereinafter, with reference to the accompanying drawings, embodiments of the present disclosure will be described in detail so that those skilled in the art can easily carry out the present disclosure. However, the present disclosure may be implemented in many different forms and is not limited to the embodiments described herein. And in order to clearly describe the present disclosure in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

1 is a diagram schematically illustrating a process in which an electronic device provides cosmetics contents using an artificial intelligence model according to an embodiment.

According to an embodiment, the electronic device 1000 may provide cosmetics content 140 optimized for user information 120 by analyzing meta data of user information using the artificial intelligence model 130 . According to an embodiment, the electronic device 1000 recommends cosmetics suitable for the user's skin according to the user information 120 using an artificial intelligence model that can be repeatedly learned based on user information and feedback data. ) can be provided.

The electronic device 1000 according to the present disclosure overcomes the limitation that it is difficult for an individual to compare information on a large number of cosmetics, and recommends a predetermined number of cosmetics according to the individual user's skin type, taste, or product characteristics, thereby providing a cosmetic user It is possible to improve the satisfaction of cosmetics and improve the sales volume of cosmetics. In addition, according to an embodiment, the electronic device 1000 may further provide an online sampling service to the user while providing cosmetics and cosmetic review contents suitable for user information using an artificial intelligence model.

After providing cosmetic contents, the electronic device 1000 outputs a sample information input interface including at least one of shipping information for providing cosmetic samples, sample quantity information, service user name, or cosmetic sample type information on the screen. Then, based on the user's input on the screen of the electronic device, an online sampling service for delivering predetermined cosmetics is provided. After the cosmetic sample is delivered, the electronic device 1000 may induce purchase of the corresponding cosmetic by further providing contents for a satisfaction survey on the cosmetic sample and purchasing contents for the corresponding cosmetic sample on the screen.

According to an embodiment, the electronic device 1000 may obtain user information 120 based on a user input to the electronic device. For example, the electronic device 1000 may include at least one user input interface, and based on a user input to the user input interface, information about at least one of a user's characteristic or a user's cosmetics consumption history may be included. user information can be obtained. According to an embodiment, user information may include at least one of user characteristic information and cosmetics consumption history information.

According to an embodiment, the user characteristic information may include information about at least one of the user's skin type information, skin trouble information, and taste information. According to an embodiment, the taste information may include at least one of the user's preferred scent information, the user's used number of cosmetics information, and the user's preferred cosmetics price range information. In addition, according to an embodiment, the consumption history information includes information on cosmetics consumed by the user over time and feedback information on the cosmetics consumed, and the feedback information is a review message transmitted by the user to a server connected to the electronic device. Information and horoscope information may be included.

According to an embodiment, the cosmetics contents 140 provided by the electronic device 1000, when the user information 120 is input, the cosmetics type information output from the artificial intelligence model, and the user's consumption according to the user information on the corresponding cosmetics It may contain score information about the likelihood. According to another embodiment, the cosmetics content 140 may further include user characteristic information used to recommend the corresponding cosmetics.

According to an embodiment, the artificial intelligence model used by the electronic device 1000 may be a model that outputs cosmetics contents based on user information when user information is input. According to an embodiment, the artificial intelligence model may include a deep neural network (DNN). Also, according to an embodiment, the artificial intelligence model may be a multi-layer perceptron (MLP). In addition, artificial intelligence models can be Convolutional Neural Network (CNN), Deep Neural Network (DNN), Recurrent Neural Network (RNN), Restricted Boltzmann Machine (RBM), Deep Belief Network (DBN), Bidirectional Recurrent Deep Neural Network (BRDNN) or Deep Q-Networks, etc., but are not limited thereto.

According to an embodiment, the artificial intelligence model 130 is based on a learning vector generated based on a combination of a user characteristic vector and a cosmetics vector, and scores (or probability value) may be pre-learned to output.

2 is a diagram schematically illustrating a process in which an electronic device collects and utilizes meta data according to an embodiment.

Referring to figures 210 and 220 of FIG. 2 , a process of collecting meta data for user information by the electronic device 1000 and a method of using the meta data are illustrated. According to an embodiment of the present disclosure, the electronic device 1000 may provide predetermined recommendation data by obtaining collected data based on a customer's skin type, customer taste, and customer review, and database of the acquired collected data. . For example, the electronic device 1000 may create a database based on cosmetic property data and a database based on review data by processing and classifying the acquired collected data.

The electronic device 1000 builds a database for a system for providing a cosmetic recommendation service based on an artificial intelligence model based on the collected data, and when user information similar to the learned information is input based on the built database, the user According to the information, cosmetic products suitable for the user's taste are recommended.

3 is a flowchart of a method of recommending cosmetics contents based on an artificial intelligence model by an electronic device according to an embodiment.

In operation S310, the electronic device 1000 may obtain user information about at least one of the user's characteristics and the user's cosmetics consumption history based on the user's input to the electronic device. According to an embodiment, the electronic device 1000 may output a user information input interface for obtaining user information on a screen and obtain user information based on a user input to the output user information input interface.

The electronic device 1000 according to an embodiment of the present disclosure may store previously acquired user information. According to an embodiment, the electronic device 1000 identifies whether there is pre-stored user information in the electronic device 1000, and if there is pre-stored user information, the electronic device 1000 determines whether or not there is consumption history information in the pre-stored user information. It is possible to identify whether the user has experience in using the service. For example, when the electronic device 1000 reads pre-stored user information and identifies that consumption history information including feedback information is included in the user information, the electronic device 1000 identifies a user according to the user information as a user with a consumption history. , and if it is identified that the consumption history information is not included in the user information, the user according to the user information may be identified as a user having no consumption history.

Although not shown in FIG. 3 , the electronic device 1000 according to an embodiment displays a quiz-type user input interface on the screen for a predetermined time, and user inputs to the quiz-type user input interface displayed on the screen. User information may be obtained based on.

In S320, the electronic device 1000 may determine a user vector for user information. For example, the electronic device 1000 may determine a user vector based on at least one of a user characteristic vector according to user characteristic information in user information and a consumption history vector according to user consumption history information. According to an embodiment, the user vector for a user who uses the service for the first time may include only the user characteristic vector according to user characteristic information in the form of metadata about skin type and taste, excluding user consumption history information. However, according to another embodiment, a user vector for a user who has used a service in the past may include a combination of a user characteristic vector based on user characteristic information and a consumption history vector based on user consumption history information.

In addition, according to an embodiment, the consumption history vector may include a combination of a cosmetics vector according to user's consumption cosmetics information over time and a feedback vector according to feedback information on consumed cosmetics. A method of determining a user vector according to user information will be described in detail with reference to FIG. 6 to be described later.

In S330, when the user vector is input, the electronic device 1000 outputs cosmetic information including at least one recommended cosmetic product according to user information and a score related to a possibility that the user will consume the at least one recommended cosmetic product to an artificial intelligence model that outputs cosmetic information. , cosmetics information can be obtained from an artificial intelligence model by inputting a user vector. For example, the electronic device 1000 generates a user vector that can be used as an input of an artificial intelligence model by preprocessing user information, and obtains cosmetic information from the artificial intelligence model by inputting the generated user vector to the artificial intelligence model. can do. According to an embodiment, the cosmetics information output from the artificial intelligence model may include the type of cosmetics matched with user information and a probability value for the possibility of consuming the cosmetics (or a score for the possibility of consumption).

In S340, the electronic device 1000 may output a predetermined number of cosmetics representing high scores in cosmetics information as cosmetics content. According to an embodiment, the electronic device 1000 may output, as cosmetic content, a preset number of cosmetics in the order of scores most likely to be consumed by the user among the cosmetics included in the cosmetics information. According to an embodiment, the electronic device 1000 outputs cosmetics content and learns an artificial intelligence model according to an ATTENTION technique to inform the basis of recommended cosmetics within the cosmetics content, thereby allowing the artificial intelligence model to , among user characteristic information, related user characteristic information used to select recommended cosmetics may be provided together. That is, cosmetic information output from the artificial intelligence model according to the present disclosure may include user characteristic information used to recommend a predetermined cosmetic product among user characteristic information.

According to another embodiment, the electronic device 1000 outputs a predetermined number of cosmetics as cosmetics contents and related user characteristic information used to select recommended cosmetics in order to inform the basis of the cosmetics recommended in the cosmetics contents. And by further providing user review information uploaded on the Internet or SNS server for the corresponding cosmetic product, it is possible to easily induce the user to purchase the recommended cosmetic product or to request a sample.

According to an embodiment, the electronic device 1000 may express the importance of five types of information representing the user as a probability value using an attention technique. According to an embodiment, the importance of each piece of information may be assigned a value between 0 and 1, the sum of the importance of all five pieces of information may be expressed as a probability value, and the probability value of adding all the importance of the five pieces of information is 1 can be provided with When five types of user information are input to the artificial intelligence model, the electronic device 1000 may set the output of the corresponding artificial intelligence model to be the importance of each information, and set the importance value of each information, which is the output of the artificial intelligence model, to each information. After reflecting the importance by multiplying the indicated vector, the end user vector can be obtained by combining the calculated five vectors. The electronic device 1000 may generate a cosmetic vector and a feedback vector as well as a user vector using the same attention technique.

In addition, although not shown in FIG. 3 , the electronic device 1000 may further provide a sample information input interface for providing cosmetic samples for predetermined cosmetics in the cosmetic contents on the screen after outputting the cosmetic contents. For example, when the electronic device 1000 outputs cosmetics content including a predetermined number of cosmetics lists and then identifies a user input for selecting a predetermined cosmetic product from the cosmetic list, the electronic device 1000 displays a 'sample' for the selected cosmetic product. Would you like to receive?' may be displayed on the screen.

When the user input for the notification message is identified, the electronic device 1000 outputs a sample information input interface for corresponding sample delivery on the screen, and then the sample type and sample quantity through the sample information input interface output on the screen. , sample delivery address, and online sampling service information including at least one of a user name may be obtained.

In addition, when the electronic device 1000 identifies that user feedback information on the cosmetic sample has been uploaded from the user receiving the cosmetic sample after delivery of the cosmetic sample, the electronic device 1000 may share the feedback information to the electronic device or a server connected to the electronic device. can

In addition, the electronic device 1000 may share upload of feedback information in real time between a user who has received the cosmetic sample and users linked to SNS among users who use the cosmetics service according to the present disclosure. For example, the electronic device 1000 transmits the feedback information when a user who has received the cosmetic sample and one of the users linked through SNS uploads feedback information on the same product as the delivered cosmetic sample. Obtained in real time from a server connected to the electronic device 1000, and based on the obtained feedback information, a notification phrase such as 'one user among users linked to SNS has uploaded feedback information' may be displayed on the screen. there is. The electronic device 1000 according to the present disclosure may induce rapid upload of feedback information of service subscribers according to the present disclosure by sharing user feedback information linked to SNS and providing notification information through the sharing.

In addition, although not shown in FIG. 3 , the electronic device 1000 determines whether or not to purchase the same type of cosmetics as the delivered cosmetic sample when feedback information is obtained from the user of the electronic device with respect to the delivered cosmetic sample. The input interface may be further output on the screen. When a user input related to 'purchase' is obtained through a user input interface for confirming whether or not to purchase cosmetics, the electronic device 1000 provides online sampling service information obtained in advance for a cosmetics sample delivery service (e.g., cosmetics sample delivery). By sharing the acquired sample type, sample quantity, sample delivery address, and user name) to the online sales server connected to the electronic device 1000, the user completes the cosmetics purchase process without entering separate information for purchasing cosmetics can do.

4 is a diagram illustrating a process of learning an artificial intelligence model by an electronic device according to an embodiment.

In S410, the electronic device 1000 may obtain learning user information and learning cosmetics information for learning the artificial intelligence model. For example, the electronic device 1000 obtains learning user information and learning cosmetics information from a server connected to the electronic device prior to utilizing the artificial intelligence model, and uses the obtained learning user information and learning cosmetics information as learning data. to train artificial intelligence models.

In S420, the electronic device 1000 may learn the artificial intelligence model based on the learning user vector according to the learning user information and the learning cosmetics vector according to the learning cosmetics information. For example, the electronic device 1000 generates at least one of learning user information or learning cosmetics information as a ONE-HOT vector, generates a k-dimensional learningable learning vector from the generated one-hot vector, and generates the An artificial intelligence model can be trained based on the learned learning vector.

According to an embodiment, the learning user information may include at least one of learning skin type information, learning skin trouble information, or learning skin taste information, and the learning cosmetic information is learning cosmetic ingredient information, learning cosmetic scent information, or learning cosmetic type. It may contain at least one piece of information. According to an embodiment, the learning skin taste information may include at least one of learning preference fragrance information, learning number of used cosmetics information, and learning preference cosmetics price range information.

In addition, according to an embodiment, the learned skin type information may include information on at least one of dry, normal, oily, or combination skin types, and the learned skin concern information may include at least one of sensitive, acne, atopic, or not applicable. It may include information on one, and the learning preferred cosmetics price range information may include low, medium, low, medium, medium, or high range information, and the learning cosmetic ingredient information may include natural moisturizers, plant-extracted oils, and the like. Ingredient information may be included, the learning cosmetic scent information may include information for identifying the type of fragrance such as a flower scent or a fruit scent, and the learning cosmetic type information may include cosmetics such as moisture cream, eye cream, lotion, or toner. It may contain information about the type.

More specifically, the electronic device 1000 determines the value of each information (e.g., the learning skin type information, the learning skin concern information, the learning preference scent information, the learning used cosmetics number information, or the learning preference cosmetics price range information), respectively. In the case of type information, in order to obtain a total of 5 vectors by assigning embedding vectors that can be learned to voice), and to obtain one end-user vector representing a user based on the obtained 5 vectors, A user vector may be obtained from the artificial intelligence model by inputting the obtained five vectors to the artificial intelligence model. The electronic device 1000 inputs the obtained five vectors to the artificial intelligence model, thereby providing a high-dimensional (hi) between learning skin type information, learning skin trouble information, learning preference fragrance information, learning number of used cosmetics information, or learning preferred cosmetics price range information. -order) A user vector representing a user may be obtained by considering the correlation.

Also, according to an embodiment, the electronic device 1000 may generate a learning cosmetics vector in a manner similar to the process of generating the learning user vector described above. For example, the electronic device 1000 assigns an embedding vector similarly to a process of obtaining a user vector from each of three pieces of information, namely learning cosmetic ingredient information, learning cosmetic scent information, or learning cosmetic type information, in the learning cosmetics information. , vectors obtained from each of the three pieces of information, and the learning cosmetics vector can be obtained by inputting the obtained three vectors into an artificial intelligence model.

The electronic device 1000 learns an artificial intelligence model based on the above-described learning user vector and learning cosmetics vector, and specifically, layers including a plurality of nodes in the artificial intelligence model and connection strengths of the layers and nodes. An artificial intelligence model can be trained by modifying and updating the weight of . The process of generating the learning user vector and the learning cosmetics vector used by the electronic device 1000 to learn the artificial intelligence model corresponds to the process of generating a user vector based on user information and a cosmetics vector based on cosmetics information, which will be described later. It can be.

5 is a diagram illustrating a process of re-learning an artificial intelligence model based on cosmetics consumption history information by an electronic device according to an embodiment.

In operation S510, the electronic device 1000 may acquire new cosmetics consumption history information based on a user input to the electronic device after outputting cosmetics content. For example, after outputting cosmetics content, the electronic device 1000 may provide an online cosmetics sample delivery service and obtain cosmetics consumption history information from the user based on a user input to the electronic device 1000. .

According to an embodiment, the cosmetic consumption history information may include cosmetic consumption information for which a user has requested a sample and feedback information on the consumed cosmetic product. The feedback information may include review writing information input by the user for the electronic device and star rating information regarding cosmetics preference.

In S520, the electronic device 1000 may determine a new consumption history vector according to the new cosmetics consumption history information. In S530, the electronic device 1000 may relearn the artificial intelligence model based on the new consumption history vector. According to an embodiment, the consumption history vector is a combination of a plurality of cosmetics vectors for cosmetics consumed by the user over time and a feedback vector provided by the user over time, and the user's interest over time can indicate change.

In addition, according to an embodiment, the feedback vector may be a combination of a review article vector according to review article information and a star score vector according to star rating information, and the review article vector inputs review text in the review article information to a context-based language model, It may be a vector value in which a context obtained from a context-based language model is reflected. The electronic device 1000 according to the present disclosure may re-learn an artificial intelligence model based on a consumption history vector according to cosmetics consumption history information newly acquired from a user using a sample, and the artificial intelligence model according to the present disclosure may be deep learning. As a model, it may be a sequential recommendation model that can accurately identify the user's preference for cosmetics over time.

6 is a diagram for explaining user information used by an electronic device and a user vector corresponding thereto, according to an exemplary embodiment.

According to an embodiment, the user information 610 may include at least one of user characteristic information 612 and user consumption history information 626 . According to an embodiment, when a user has no history of using a cosmetics recommendation service, user information 610 for the corresponding user may include only user characteristic information 612 .

According to an embodiment, the user characteristic information 612 may include at least one of skin type information 614 , skin trouble information 616 , and taste information 618 . According to an embodiment, the taste information 618 may include at least one of preferred scent information 620 , used cosmetics number information 62 , and preferred cosmetics price range information 624 .

Also, according to an embodiment, the user consumption history information 626 may include at least one of consumption cosmetics information 628 and feedback information 630 . According to an embodiment, the feedback information 630 may include at least one of review writing information 632 and star rating information 634 . The user information 610 shown in FIG. 6 may correspond to the user learning information described above with reference to FIG. 4 .

According to an embodiment, the electronic device 1000 may determine a user vector 640 from user information 610 . According to an embodiment, the user vector 640 may be determined based on a combination of a user characteristic vector 642 according to user characteristic information and a user consumption history vector 644 according to user consumption history information 626 .

According to an embodiment, the user consumption history vector 644 may be determined based on a combination of a cosmetics vector 646 according to cosmetics information and a feedback vector 648 according to feedback information 630 . Also, according to an embodiment, the feedback vector 648 may be determined based on a combination of a review article vector 650 based on review article information and a star score vector 652 based on star rating information.

7 is a diagram for explaining a learning and evaluation process of an artificial intelligence model used by an electronic device according to an embodiment.

In S702, the electronic device 1000 acquires data owned by the company. According to an embodiment, in-house data may correspond to learning user data. In S704, the electronic device 1000 may pre-process the acquired in-house data. According to an embodiment, the electronic device 1000 may generate verification data in S706 and learning data in S708 by pre-processing in-house data.

In S710, the electronic device 1000 may train a recommendation system (eg, an artificial intelligence model) based on the learning data and the verification data. In S712, after learning of the artificial intelligence model is completed, the electronic device 1000 may obtain evaluation data by inputting verification data to the artificial intelligence model in S714. The electronic device 1000 may determine recommendation accuracy of an artificial intelligence model (eg, a recommendation system) based on the evaluation data obtained in S716 and S714. As a result of evaluating the recommendation accuracy of the artificial intelligence model, the electronic device 1000 may improve the accuracy of the recommendation system by iteratively learning the recommendation system when the accuracy is not higher than a preset threshold accuracy.

8 is a diagram for explaining an operation of a context-based language model used by an electronic device according to an exemplary embodiment.

Referring to FIG. 8 , an electronic device 1000 provides a context-based language model for providing appropriate resources and services to a user using user information in a cloud computing environment and a context-based recommendation algorithm (210) for operating the context-based language model. ) is shown.

According to an embodiment, the electronic device 1000 may store a context-based language model in a memory, perform a document semantic analysis on a postscript input by a user by using the context-based language model stored in the memory, and Based on this, it is possible to recommend predetermined contents (for example, cosmetics contents). In addition, based on the user consumption history information of the electronic device 1000, preference-based content may be recommended based on a cosmetic consumption history pattern used by the user over time.

For example, the electronic device 1000 may obtain feedback information including review writing information and star rating information based on a user input to the electronic device. The electronic device 1000 may obtain a late writing vector from the context language based model by inputting the late text according to the late writing information to the context language based model learned with many Korean writings. The late writing vector is a vector format representing the characteristics of late writing information in consideration of the context around a specific word of the context-based language model, and may be information indicating the context of the user's late writing information, not simply the meaning of a word.

Also, according to an embodiment, in order to obtain a star vector, the electronic device 1000 generates an embedding vector that can be learned for each score (eg, 1 point, 5 points, etc.), similar to a method of acquiring a user vector and a cosmetics vector. By assigning, a star vector can be obtained. The electronic device 1000 generates a feedback vector by inputting the postscript vector and the star vector to the multi-artificial neural network model.

9 is a block diagram of an electronic device according to an embodiment.

According to an embodiment, the electronic device 1000 may include a processor 1300, a network interface 1500, and a memory 1700 storing one or more instructions. However, it is not limited to the above example, and the electronic device 1000 may further include other components for providing a cosmetics recommendation service. According to another embodiment, the electronic device 1000 may further include a display and a user input interface, and of course may be provided with fewer components than those shown in FIG. 9 .

According to an embodiment, a user input interface (not shown) means a means through which a user using a cosmetics recommendation service controls an electronic device or inputs data to input user information. For example, a user input interface (not shown) includes a key pad, a dome switch, a touch pad (contact capacitive method, pressure resistive film method, infrared sensing method, surface ultrasonic conduction method, An integral tension measurement method, a piezo effect method, etc.), a jog wheel, a jog switch, etc. may be included, but are not limited thereto.

A display (not shown) may output cosmetics content. According to another embodiment, the display (not shown) may also output recommended cosmetics and user information used to determine the recommended cosmetics. Also, according to an embodiment, the display may further output a user input interface for receiving user information and a user input interface for receiving feedback information. In addition, according to an embodiment, the display inputs at least one of address information, delivery address information, user name, type of sample cosmetics, or information on the quantity of sample cosmetics required for online sample application under the control of the processor 1300. An input interface may be further provided.

The processor 1300 typically controls overall operations of the electronic device 1000 . According to an embodiment, the processor 1300 may generally control a user input interface, a display, and a network interface by executing programs stored in the memory 1700. Also, the processor 1300 may perform the functions of the electronic device described in FIGS. 1 to 9 by executing programs stored in the memory 1700 .

According to an embodiment, the processor 1300 obtains user information about at least one of the user's characteristics and the user's cosmetics consumption history based on the user's input to the electronic device, and obtains user information on the obtained user information. An artificial intelligence model that determines a vector and, when the user vector is input, outputs cosmetic information including at least one recommended cosmetic product according to the user information and a score related to a possibility that the user will consume the at least one recommended cosmetic product. , By inputting the user vector, the cosmetics information is acquired from the artificial intelligence model, and a predetermined number of cosmetics representing high scores in the cosmetics information can be output as the cosmetics content.

According to an embodiment, the processor 1300 obtains learning user information and learning cosmetics information for learning the artificial intelligence model by executing one or more instructions stored in the memory 1700, and learning according to the learning user information. The cosmetics information may be acquired by learning the artificial intelligence model based on a user vector and a learning cosmetics vector according to the learning cosmetics information, and inputting the user vector to the learned artificial intelligence model.

According to an embodiment, the processor 1300 generates a one-hot vector from at least one of the learning user information and the learning cosmetics information, generates a learning vector capable of k-dimensional learning from the one-hot vector, and uses the generated learning vector Based on this, the artificial intelligence model can be learned.

According to an embodiment, the processor 1300 obtains new cosmetics consumption history information based on a user input to the electronic device after the step of outputting the cosmetics contents, and obtains new cosmetics consumption history information according to the new cosmetics consumption history information. A consumption history vector may be determined, and the artificial intelligence model may be re-learned based on the new consumption history vector.

The network interface (not shown) may include one or more components that allow the electronic device 1000 to communicate with a server or another electronic device connected to the electronic device. The other electronic device (not shown) may be a computing device such as the guardian terminal 1000, a sensing device, or a server device, but is not limited thereto. For example, the network interface (not shown) may include a short-distance communication unit, a mobile communication unit, and a broadcast reception unit.

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

The mobile communication unit transmits and receives radio signals with at least one of a base station, an external terminal, and a server on a mobile communication network. Here, the radio signal may include a voice call signal, a video call signal, or various types of data according to text/multimedia message transmission/reception. The broadcast reception unit receives a broadcast signal and/or broadcast-related information from the outside through a broadcast channel.

The memory 1700 may store programs for processing and control of the processor 1300 and may store information about a context-based language model and an artificial intelligence model used by the electronic device 1000 .

According to an embodiment, the memory 1700 may further store information about weight information, nodes, and layers in the artificial intelligence model. In addition, according to an embodiment, the memory may store information about a deep learning model and a deep neural network (DNN), a convolutional neural network (CNN), a deep neural network (DNN), and a recurrent neural network (RNN). ), Restricted Boltzmann Machine (RBM), Deep Belief Network (DBN), Bidirectional Recurrent Deep Neural Network (BRDNN), or deep Q-Networks.

The memory 1700 may include a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (eg SD or XD memory, etc.), RAM (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, magnetic disk , an optical disk, and at least one type of storage medium.

A method of recommending cosmetics contents based on an artificial intelligence model by an electronic device according to an 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. Program instructions recorded on the medium may be those specially designed and configured for the present invention or those known and usable to those skilled in computer software.

Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler.

Also, some embodiments may be implemented as a computer program or computer program product including instructions executable by a computer, such as a computer program executed by a computer. Although the embodiments of the present invention have been described in detail above, the scope of the present invention is not limited thereto, and various modifications and improvements made by those skilled in the art using the basic concept of the present invention defined in the following claims are also included in the scope of the present invention. fall within the scope of the right

Claims (15)

A method for an electronic device to recommend cosmetic contents based on an artificial intelligence model,
obtaining user information about at least one of a user's characteristics and a cosmetic consumption history of the user, based on the user's input to the electronic device;
determining a user vector for the obtained user information;
When the user vector is input, the user vector is sent to an artificial intelligence model that outputs cosmetic information including at least one recommended cosmetic product according to the user information and a score related to the possibility that the user will consume the at least one recommended cosmetic product. acquiring the cosmetics information from the artificial intelligence model by inputting the information; and
outputting a predetermined number of cosmetics representing a high score in the cosmetics information as the cosmetics contents; Including, method. The method of claim 1, wherein the method
acquiring learning user information and learning cosmetics information for learning the artificial intelligence model; and
learning the artificial intelligence model based on a learning user vector according to the learning user information and a learning cosmetics vector according to the learning cosmetics information; including,
The step of obtaining the cosmetic information is
obtaining the cosmetics information by inputting the user vector to the learned artificial intelligence model; Including, method. The method of claim 2, wherein the step of training the artificial intelligence model
generating at least one of the learning user information and the learning cosmetics information as a one-hot vector;
generating a learning vector capable of learning in k-dimension from the one-hot vector; and
learning the artificial intelligence model based on the generated learning vector; Including, method. 3. The method of claim 2, wherein the method
obtaining new cosmetic consumption history information based on a user input to the electronic device after outputting the cosmetic contents; and
determining a new consumption history vector according to the new cosmetics consumption history information;
re-learning the artificial intelligence model based on the new consumption history vector; including,
Characterized in that the artificial intelligence model is a serial recommendation model that tracks a change in interest of a user over time by being learned based on the new consumption history vector. The method of claim 4, wherein obtaining the user information
obtaining user characteristic information about at least one of skin type information, skin trouble information, and taste information of the user; and
acquiring user consumption history information about the user's cosmetics consumption history; Including, method. The method of claim 5, wherein the taste information
characterized in that it includes at least one of the user's preferred scent information, the user's used cosmetics number information, and the user's preferred cosmetics price range information. According to claim 6,
The consumption history information includes information on cosmetics consumed by the user over time and feedback information on the cosmetics consumed,
The feedback information may include review writing information and star rating information transmitted by the user to a server connected to the electronic device. 8. The method of claim 7, wherein determining the user vector comprises:
determining the user vector based on at least one of a user characteristic vector according to the user characteristic information and a consumption history vector according to the user consumption history information; Including, method. The method of claim 8, characterized in that the consumption history vector is a combination of a cosmetics vector according to information about cosmetics consumed by the user over time and a feedback vector according to feedback information on the cosmetics consumed. 10. The method of claim 9, wherein the feedback vector is
A combination of a review article vector according to the review article information and a star vector according to the star point information,
The method of claim 1 , wherein the latter vector is obtained from the context-based language model by inputting the latter text in the latter information into the context-based language model. The method of claim 1, wherein acquiring the cosmetic information comprises:
obtaining cosmetic information including user characteristic information related to the recommended cosmetic among the user characteristic information, using the artificial intelligence model learned according to an attention technique; Including, method. In an electronic device that recommends cosmetic contents based on an artificial intelligence model,
a memory that stores one or more instructions; and
at least one processor to execute the one or more instructions; including,
By executing the one or more instructions, the at least one processor:
Obtaining user information on at least one of a user's characteristic or a user's cosmetic consumption history based on a user input to the electronic device;
determining a user vector for the obtained user information;
When the user vector is input, the user vector is sent to an artificial intelligence model that outputs cosmetic information including at least one recommended cosmetic product according to the user information and a score related to the possibility that the user will consume the at least one recommended cosmetic product. By inputting, obtaining the cosmetics information from the artificial intelligence model,
An electronic device that outputs a predetermined number of cosmetics representing a high score in the cosmetics information as the cosmetics content. 13. The method of claim 12, wherein the at least one processor
Acquiring learning user information and learning cosmetics information for learning the artificial intelligence model;
Learning the artificial intelligence model based on a learning user vector according to the learning user information and a learning cosmetics vector according to the learning cosmetics information;
The electronic device for obtaining the cosmetics information by inputting the user vector to the learned artificial intelligence model. 14. The method of claim 13, wherein the at least one processor
generating at least one of the learning user information or the learning cosmetics information as a one-hot vector;
generating a k-dimensional learnable learning vector from the one-hot vector;
The electronic device for learning the artificial intelligence model based on the generated learning vector. A method for an electronic device to recommend cosmetic contents based on an artificial intelligence model,
obtaining user information about at least one of a user's characteristics and a cosmetic consumption history of the user, based on the user's input to the electronic device;
determining a user vector for the obtained user information;
When the user vector is input, the user vector is sent to an artificial intelligence model that outputs cosmetic information including at least one recommended cosmetic product according to the user information and a score related to the possibility that the user will consume the at least one recommended cosmetic product. acquiring the cosmetics information from the artificial intelligence model by inputting the information; and
outputting a predetermined number of cosmetics representing a high score in the cosmetics information as the cosmetics content; A computer-readable recording medium recording a program for executing a method on a computer, including a.

Images