KR102443914B1 - Method and apparatus for recommending hehavior of user - Google Patents

Abstract

In order to recommend a target action to the user for the user's current state, obtain reference information on the user's action, generate training information by texturing or image the reference information, and at least among the time, place and situation where the action occurred The training information is vectorized based on one, the behavior recommendation algorithm for the user is updated based on the vectorized training information, and the target behavior for the current state of the user is recommended to the user based on the updated behavior recommendation algorithm.

Description

METHOD AND APPARATUS FOR RECOMMENDING HEHAVIOR OF USER

The following embodiments relate to a technique for recommending a user's action.

In general, a user may be provided with various contents or services through a user terminal, and the user using the user terminal may perform consumption activities such as searching for or purchasing contents or a specific service through a service server. In addition, recently, through the development of wireless communication, information exchange between various user terminals has become easy, and as the performance of the user terminals is improved, users can use contents or specific services regardless of place and time. On the other hand, while the number of contents provided by the service server is quite large and diverse, the contents required by the user are limited according to, for example, the taste of the user. , time and effort may be consumed unnecessarily to purchase.

One embodiment may provide a method of personalizing an algorithm for recommending actions to a user.

An embodiment may provide a method of recommending an action to a user through an algorithm.

According to one aspect, a method of personalizing a behavior recommendation algorithm, performed by a server, includes: obtaining reference information for a user's behavior; generating training information by texturing or imaging the reference information; the behavior Vectorizing the training information based on at least one of the generated time, place, and situation, and updating a behavior recommendation algorithm for the user based on the vectorized training information. Including, the behavior recommendation algorithm recommends a target behavior for the user's current state to the user.

The step of vectorizing the training information based on at least one of a time, a place, and a situation at which the behavior occurred may include: obtaining at least one of the time, the place, and the situation; and the training information and the time, the place and associating at least one of the situations.

The obtaining of the reference information on the user's behavior may include obtaining the reference information by receiving a voice uttered by the user through a microphone of the user terminal.

The generating of the training information by texturing or imaging the reference information may include generating the training information by texturing the voice through natural language processing (NLP).

The obtaining of the reference information on the user's behavior may include obtaining the reference information on the histories of one or more services used by the user.

The step of generating the training information by converting the reference information into text or image may include generating a spectrogram image of the target sound source as the training information when the service used by the user is a service that reproduces a target sound source. can

The vectorizing the training information based on at least one of a time, a place, and a situation in which the behavior occurred includes generating a life-style graph for the user based on the vectorized training information, wherein the vectoring includes: The updating of the behavior recommendation algorithm for the user based on the trained training information may include updating the behavior recommendation algorithm based on the life-style graph.

A method of personalizing a behavior recommendation algorithm includes: determining the current state of the user; determining the target behavior for the current state through the behavior recommendation algorithm; and recommending the target behavior to the user. may further include.

The method of personalizing the behavior recommendation algorithm may further include receiving feedback on the target behavior from the user, and updating the behavior recommendation algorithm based on the feedback.

According to another aspect, a server for performing a method for personalizing a behavior recommendation algorithm includes a memory in which a program for personalizing a behavior recommendation algorithm is recorded, and a processor for executing the program, wherein the program includes the reference information generating training information by texturing or imaging updating a behavior recommendation algorithm for the user based on the information, and the behavior recommendation algorithm recommends a target behavior for the user's current state to the user.

According to another aspect, a behavior recommendation method, performed by a server, includes: determining a current state of a user; determining a target behavior for the current state through a behavior recommendation algorithm personalized for the user; and recommending the target behavior to the user, wherein the behavior recommendation algorithm is vectorized training information based on at least one of a time, a place, and an occurrence of the user's behavior. is updated based on , and the training information is generated by texturing or imaging the reference information obtained for the behavior of the user.

The determining of the current status of the user may include determining the current status based on a history of one or more services used by the user.

Determining the current state of the user includes obtaining target information on the user's behavior, and generating search information by converting the target information into text or image, the time and place where the behavior occurred and the search information associated with at least one of a situation may be input to the action recommendation algorithm.

The determining of the target behavior for the current state through the behavior recommendation algorithm personalized for the user may include determining the target behavior for the current state based on a life-style graph generated for the user. may include

The determining of the target behavior for the current state based on the life-style graph generated for the user may include: Using at least one of a time, a place, and a situation for the behavior of the current state, determining the target behavior.

The behavior recommendation method may further include receiving feedback on the target behavior from the user, and updating the behavior recommendation algorithm based on the feedback.

According to yet another aspect, a server for performing a method for recommending an action to a user includes a memory in which a program for recommending an action is recorded, and a processor for executing the program, wherein the program includes: determining a target behavior for the current state through a behavior recommendation algorithm personalized for the user, and recommending the target behavior to the user, wherein the behavior recommendation algorithm performs the steps of: is updated based on the vectorized training information based on at least one of a time, a place, and an occasion at which the behavior of It is created by texturing or imaging.

A method may be provided for personalizing an algorithm that recommends actions to a user.

A method of recommending actions to the user through an algorithm may provide.

1 is a configuration diagram of a system for recommending an action to a user according to an example.
2 is a configuration diagram of a server according to an embodiment.
3 is a flowchart of a method of operating a behavior recommendation algorithm according to an example.
4 is a flowchart of a method for initially updating a behavior recommendation algorithm according to an embodiment.
5 is a flowchart of a method of generating training information based on a user's voice according to an example.
6 is a flowchart of a method of generating training information based on a spectrogram image of a sound source according to an example.
7 is a flowchart of a method of updating a behavior recommendation algorithm for a user based on vectorized training information according to an example.
8 illustrates vectorized training information according to an example.
9 is a flowchart of a method of updating a behavior recommendation algorithm based on a life-style graph for a user according to an example.
10 illustrates a life-style graph for a user according to an example.
11 is a flowchart of a method of recommending a target behavior to a user through a behavior recommendation algorithm according to an embodiment.
12 is a flowchart of a method of determining a current state of a user according to an example.
13 illustrates a method of determining a target behavior for a current state based on a life-style graph according to an example.
14 is a flowchart of a method of additionally updating a behavior recommendation algorithm based on a user's feedback according to an example.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, since various changes may be made to the embodiments, the scope of the patent application is not limited or limited by these embodiments. It should be understood that all modifications, equivalents and substitutes for the embodiments are included in the scope of the rights.

Terms used in the examples are used for the purpose of description only, and should not be construed as limiting. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present specification, terms such as “comprise” or “have” are intended to designate that a feature, number, step, operation, component, part, or combination thereof described in the specification exists, but one or more other features It is to be understood that this does not preclude the possibility of the presence or addition of numbers, steps, operations, components, parts, or combinations thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the embodiment belongs. 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 application. does not

In addition, in the description with reference to the accompanying drawings, the same components are assigned the same reference numerals regardless of the reference numerals, and the overlapping description thereof will be omitted. In the description of the embodiment, if it is determined that a detailed description of a related known technology may unnecessarily obscure the gist of the embodiment, the detailed description thereof will be omitted.

In addition, in describing the components of the embodiment, terms such as first, second, A, B, (a), (b), etc. may be used. These terms are only for distinguishing the elements from other elements, and the essence, order, or order of the elements are not limited by the terms. When it is described that a component is “connected”, “coupled” or “connected” to another component, the component may be directly connected or connected to the other component, but between each component another component It will be understood that may also be "connected", "coupled" or "connected".

Components included in one embodiment and components having a common function will be described using the same names in other embodiments. Unless otherwise stated, descriptions described in one embodiment may be applied to other embodiments as well, and detailed descriptions within the overlapping range will be omitted.

1 is a configuration diagram of a system for recommending an action to a user according to an example.

According to an aspect, content may be recommended to a user based on collaborative filtering. However, the method of recommending content to a user based on collaborative filtering may have a problem of bias in popularity according to the preference of the group rather than the individual tendency of the user. In addition, the above method may have a scarcity problem because individuals belonging to the cluster do not consume all contents to be recommended. Additionally, the above method has a problem of a cold start of a recommendation range as a recommendation for newly registered content is possible after a predetermined time.

According to another aspect, in order to recommend content to the user, information related to the user's behavior is collected to reflect the user's personal inclination/preference, and based on the collected information, a behavior recommendation algorithm is trained, and the trained A method of recommending an action to a user through an algorithm may be considered.

Specifically, the above method generates a personalized life-style graph of the user based on personal inclination/preference, etc. through algorithm update technology (eg, deep learning) in the field of artificial intelligence (AI), and the user's time (Time), place (place), situation (occasion) (TPO) can recommend appropriate actions.

The server 100 may collect information related to the behavior of the user 111 through the user terminals 112 and 113 used by the user 111 . With the user 111 as the center, the user terminals 112 and 113 used by the user 111 may be set as the group 110 . For example, the user terminal may be a smart phone, a connected car, an IPTV, an AI speaker, etc., and is not limited to the described embodiment. The server 100 may recommend actions to the user 111 through the action recommendation algorithm 105 personalized for the user 111 .

Through the connected car, sound source recommendation, content recommendation, destination identification and directions according to personal schedule, and nearby PoI recommendation can be performed to the user through the connected car.

Media content may be recommended to a user through IPTV.

Sound source recommendation, personalized alarm output, news content playback, etc. can be performed through the AI speaker.

A method of operating a behavior recommendation algorithm will be described in detail below with reference to FIGS. 2 to 14 .

2 is a configuration diagram of a server according to an embodiment.

The server 200 includes a communication unit 210 , a processor 220 , and a memory 230 . For example, the server 200 may be the server 100 described above with reference to FIG. 1 .

The communication unit 210 is connected to the processor 220 and the memory 230 to transmit and receive data. The communication unit 210 may be connected to another external device to transmit/receive data. Hereinafter, the expression “transmitting and receiving “A” may indicate transmission and reception of “information or data representing A”.

The communication unit 210 may be implemented as circuitry in the server 200 . For example, the communication unit 210 may include an internal bus and an external bus. As another example, the communication unit 210 may be an element that connects the server 200 and an external device. The communication unit 210 may be an interface. The communication unit 210 may receive data from an external device and transmit the data to the processor 220 and the memory 230 .

The processor 220 processes data received by the communication unit 210 and data stored in the memory 230 . A “processor” may be a data processing device implemented in hardware having circuitry having a physical structure for performing desired operations. For example, desired operations may include code or instructions included in a program. For example, a data processing device implemented as hardware includes a microprocessor, a central processing unit, a processor core, a multi-core processor, and a multiprocessor. , an Application-Specific Integrated Circuit (ASIC), and a Field Programmable Gate Array (FPGA).

Processor 220 executes computer readable code (eg, software) stored in memory (eg, memory 230 ) and instructions issued by processor 220 .

The memory 230 stores data received by the communication unit 210 and data processed by the processor 220 . For example, the memory 230 may store a program (or an application, software). The stored program may be a set of syntaxes coded to operate the behavior recommendation algorithm and executable by the processor 220 . The operation of the behavior recommendation algorithm may include a method of personalizing the behavior recommendation algorithm and a method of recommending a behavior to a user.

According to one aspect, memory 230 may include one or more volatile memory, non-volatile memory and random access memory (RAM), flash memory, hard disk drive, and optical disk drive.

The memory 230 stores an instruction set (eg, software) for operating the server 200 . The instruction set for operating the server 200 is executed by the processor 220 .

The communication unit 210 , the processor 220 , and the memory 230 will be described in detail below with reference to FIGS. 3 to 14 .

3 is a flowchart of a method of operating a behavior recommendation algorithm according to an example.

The operation method of the behavior recommendation algorithm may include the step 310 of initially updating the behavior recommendation algorithm, the step 320 of recommending a behavior through the behavior recommendation algorithm, and the step 330 of further updating the behavior recommendation algorithm. have. According to one aspect, the server 200 may perform all or part of steps 310 , 320 , and 300 .

The step 310 of initially updating the behavior recommendation algorithm will be described in detail with reference to FIGS. 4 to 10 below, and FIGS. 11 to 13 below for the step 320 of recommending a behavior through the behavior recommendation algorithm. The step 330 of further updating the action recommendation algorithm is described in detail with reference to FIG. 14 below.

4 is a flowchart of a method for initially updating a behavior recommendation algorithm according to an embodiment.

According to one aspect, the step 310 described above with reference to FIG. 3 includes the following steps 410 to 440 .

In step 410, the server 200 obtains reference information about the user's behavior. According to one aspect, reference information about the history of one or more services used by the user may be obtained. For example, the services used by the user are a search service through online through a user terminal such as a smartphone, AI speaker, IPTV, and a connected car (hereinafter, 'online search service' is abbreviated as a search service), It may include a voice command service, a video on demand (VoD) service, a sound reproduction service, a navigation service, and the like, but is not limited to the disclosed embodiment. A log of a provided service may be obtained as reference information. Examples of reference information are described in detail below with reference to FIGS. 5 and 6 .

In step 420, the server 200 generates training information by texturing or imaging the reference information. According to an aspect, when the reference information is not text or an image, the reference information may be displayed by converting the reference information into text or an image. For example, when the reference information is a voice, the voice may be represented as text. As another example, when the reference information is sound or music, the sound or music may be represented as a spectrogram image. A method of converting reference information into text or an image and displaying it will be described in detail below with reference to FIGS. 5 and 6 .

In step 430 , the server 200 vectorizes the training information based on at least one of a time, a place, and a situation in which the user's behavior occurs. The server 200 may vectorize the relationship between words in the training information. For example, training information may be vectorized using Word2Vec based on frequency of use of words and associations between words. Specifically, the training information may be vectorized for Skip-Gram model with Negative Sampling (SGNS) and Continuous Bag Of Words (CBOW). A method of vectorizing training information is described in detail below with reference to FIGS. 7 to 9 .

Additionally, the vectorized training information may be stored in the database of the server 200 as search information to recommend a target action to a user later.

In step 440, the server 200 updates the behavior recommendation algorithm for the user based on the vectorized training information. For example, a behavior recommendation algorithm may be updated using a neural network.

According to an aspect, a behavior recommendation algorithm may be updated using a convolution neural network (CNN) model. For example, when the training information is a spectrogram image, a CNN model may be used for algorithm update.

According to another aspect, the behavior recommendation algorithm may be updated using a recurrent neural network (RNN) model. For example, when training information can be expressed in a time series, an RNN model can be used for algorithm update. More specifically, a long short term memory (LSTM) model may be used.

When the behavior recommendation algorithm is updated using the RNN model, a life-style graph for the user may be generated. When the behavior recommendation algorithm is iteratively updated, the generated life-style graph may also be iteratively changed.

Various learning methods may be applied to update the behavior recommendation algorithm, and it is not limited to a specific embodiment or method.

5 is a flowchart of a method of generating training information based on a user's voice according to an example.

According to one aspect, the step 410 described above with reference to FIG. 4 may include the following step 510 , and the step 420 may include the following step 520 .

In step 510 , the server 200 obtains reference information by receiving the voice uttered by the user through the microphone of the user terminal. The user terminal may be a smartphone, an AI speaker, an IPTV, or a connected car, but is not limited to the described embodiment. The user terminal may transmit the acquired user's voice to the server 200 . For example, the user may obtain a weather inquiry, a schedule inquiry, a navigation setting command, a voice search, etc. as reference information, and is not limited to the described embodiment.

In step 520, the server 200 generates training information by converting the voice acquired from the user terminal into text through natural language processing (NLP).

According to one aspect, the server 200 secures a script by converting the acquired voice into text through text to speech (TTS), and builds a corpus by removing adjectives, adverbs, and propositions unnecessary for learning in the script. can create The training information may include a corpus.

A word with a high frequency of appearance can be identified through the corpus, and a ranking based on the number of times used for each word can be secured, which can be used as learning data. Additionally, based on the secured ranking, raw data (eg, sound source, VoD information, article information, point of interest (POI)) centered on a noun may be additionally secured, which is additional learning data can be used as That is, various pieces of information related to words in the corpus (the corresponding information may also be converted into the corpus) may be secured as raw data. For example, movie information, cast information, script, user reviews, and related music information related to the word "zoo next to an art museum" may be secured as raw data. The training information may further include raw data.

6 is a flowchart of a method of generating training information based on a spectrogram image of a sound source according to an example.

According to one aspect, step 410 described above with reference to FIG. 4 may further include step 610 below, and step 420 may further include step 620 below.

In step 610, the server 200 obtains reference information about the history of one or more services used by the user. The server 200 may check the service used by the user through the user terminal. For example, the user can check the sound source reproduction service and VoD reproduction service used through the user terminal as a history, and reference information about the sound source reproduced through the sound source reproduction service and VoD reproduced through the VoD reproduction service is obtained. can

In step 620, when the service used by the user is a service for reproducing a sound source, the server 200 may generate a spectrogram image of the reproduced target sound source as training information. As another example, when the service used by the user is a service for reproducing VoD, a spectrogram image of an OST (original sound track) of the reproduced target VoD may be generated as training information.

According to one aspect, the server 200 may generate a corpus related to music and VoD similar to the above-described step 520 .

7 is a flowchart of a method of updating a behavior recommendation algorithm for a user based on vectorized training information according to an example.

According to one aspect, the step 430 described above with reference to FIG. 4 may include the following steps 710 and 720 , wherein the following steps 710 and 720 are to be performed before the training information is vectorized. can

In step 710 , the server 200 acquires at least one of a time, a place, and a situation where the user's action with respect to the reference information occurred. For example, at least one of a time, a place, and a situation in which the user uttered a voice may be acquired. The situation may be a type of an application being executed on the user terminal or actions occurring before and after a target action.

In step 720, the server 200 associates the training information with at least one of the acquired time, place and situation. After step 720 is performed, the server 200 may vectorize training information associated with at least one of a time, a place, and a situation. The vectorized training information is described in detail below with reference to FIG. 8 .

8 illustrates vectorized training information according to an example.

According to one aspect, each word in the training information may be vectorized. Words that are frequently mentioned with the target word and have a TPO similar to the TPO of the target word may be vectorized to be closer to the target word. For example, a vector table may be created indicating the frequency with which each of the words is mentioned with the target word. In order to visually represent the vector table, the size of the corresponding word may be adjusted to reflect the frequency.

9 is a flowchart of a method of updating a behavior recommendation algorithm based on a life-style graph for a user according to an example.

According to one aspect, the step 430 described above with reference to FIG. 4 may include the following step 910 , and the step 440 may further include the following step 920 .

In step 910, the server 200 generates a life-style graph for the user based on the vectorized training information. For example, the life-style graph may represent the behaviors classified according to the user's behavior type in time series. The behavior type may correspond to the type of service. A plurality of life-style graphs may be generated for each day. The life-style graph may be generated for each day of the week to reflect the user's periodic life-style, but is not limited thereto. The generated life-style graph may be an initial life-style graph.

A life-style graph is described in detail below with reference to FIG. 10 .

In step 920, the server 200 updates the behavior recommendation algorithm based on the life-style graph.

For example, the behavior recommendation algorithm may be updated through a long short term memory (LSTM) model that analyzes training information in a time series based on a life-style graph. The life-style graph may be updated through the updated behavior recommendation algorithm. The updated life-style graph may be a predictive life-style graph.

As an important factor for updating the behavior recommendation algorithm, the TPO of the behavior, the type of the user terminal in which the behavior occurred, the number of utterances uttering voice, the number of times of service use, and the like may be considered.

The server 200 may iteratively update the behavior recommendation algorithm based on the updated life-style graph.

10 illustrates a life-style graph for a user according to an example.

According to one aspect, the user's actions may be classified according to action types. For example, actions on AI speaker, actions on calendar, actions on news & weather, actions on VoD, actions on connected car, actions on search, actions on sound source , and may be classified by types of actions for the service log.

A life-style graph may be generated so that the user's actions classified by action type appear in time series. A plurality of life-style graphs may be generated for each day based on the user's daily behavior history. A plurality of life-style graphs generated on a daily basis may be used to update the behavior recommendation algorithm.

11 is a flowchart of a method of recommending a target behavior to a user through a behavior recommendation algorithm according to an embodiment.

According to one aspect, with reference to FIG. 3 , step 320 includes the following steps 110 to 1130 .

In step 1110, the server 200 determines the current status of the user.

According to an aspect, the current state may be determined based on a history of one or more services used by the user. For example, the current state may be determined based on the most recent behavior of the user searching for news through the user terminal. As another example, the current state may be determined based on which user terminal the user uses the service.

A method of determining the user's current status is described in detail below with reference to FIG. 12 .

In step 1120 , the server 200 determines a target action for the current state through a personalized action recommendation algorithm for the user.

According to an aspect, the server 200 may determine a target behavior for the current state through a behavior recommendation algorithm based on the life-style graph generated for the user. For example, the target action for the current situation may be determined using at least one of a time, a place, and a situation for the action of the current state. A method of determining a target behavior based on the life-style graph is described in detail below with reference to FIG. 13 .

In step 1130, the server 200 recommends the target action to the user. The server 200 recommends a target action to the user through the user terminal. For example, the server 200 may output a voice for recommending a target action to the user through the AI speaker. As another example, the server 200 may output an image for recommending a target action to a user through a user terminal to which a display is connected.

12 is a flowchart of a method of determining a current state of a user according to an example.

According to one aspect, the step 1110 described above with reference to FIG. 11 may include the following steps 1210 to 1230 .

In step 1210 , the server 200 obtains target information about the user's behavior. The target information may correspond to reference information. That is, information used in the method of updating the behavior recommendation algorithm may be named reference information, and information used in the method of recommending the target behavior may be named target information. The description for step 410 may be similarly applied to the description for step 1210 .

In step 1220, the server 200 generates the search information by converting the target information into text or image. The search information may correspond to the training information. That is, the information used in the method of updating the behavior recommendation algorithm may be named training information, and the information used in the method of recommending the target behavior may be named search information. The description of step 420 may be similarly applied to the description of step 1220 .

Additionally, the server 200 may associate the search information with at least one of a time, a place, and a situation in which the user's action occurred.

The server 200 may determine the current state of the user by generating search information about the action.

The server 200 may input the related search information to the behavior recommendation algorithm and determine the target behavior as an output.

13 illustrates a method of determining a target behavior for a current state based on a life-style graph according to an example.

According to one aspect, the server 200 may determine a point corresponding to the current state on the life-style graph. For example, the determined current state of the user may be the action of the news search 1310 . Specifically, what news was searched for may be generated as search information.

An action following the action of the news search 1310 may be determined as the target action. For example, a continuous search for similar news ( 1320 ), confirmation of today's schedule ( 1330 ), and playing music ( 1340 ) may be determined as target actions. One or more actions may be determined as the target action. When the similar news continuous search 1320 is determined as the target action, news similar to the news searched or confirmed by the user may be determined based on the search information. For example, news similar to the news searched by the user may be determined based on the vectorized words stored in the database in the server 200 .

As another example, when the music reproduction 1340 is determined as the target action, the target sound source reflecting the user's taste may be determined. The user's taste may be determined based on spectrogram images of sound sources previously reproduced by the user. A sound source having a waveform similar to that of spectrogram images of previously reproduced sound sources may be determined as the target sound source.

Since the TPO for the current state of the user is considered to determine the target behavior, the determined target behavior may be changed when at least one of time, location, and situation changes.

14 is a flowchart of a method of additionally updating a behavior recommendation algorithm based on a user's feedback according to an example.

According to one aspect, the following steps 1410 and 1420 may be performed after the step 1130 described above with reference to FIG. 11 is performed.

In step 1410, the server 200 receives feedback on the target behavior from the user. The server 200 may receive the user's feedback through the user terminal. For example, feedback may be in the form of stars and scores, and is not limited to the examples described. As another example, the feedback may be a playback time of a recommended sound source or VoD.

In step 1420, the server 200 updates the action recommendation algorithm based on the feedback. For example, when the feedback is positive, a portion in which the target behavior is determined in the behavior recommendation algorithm may be strengthened, and when the feedback is negative, a portion in the behavior recommendation algorithm in which the target behavior is determined may be weakened.

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 execute 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. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

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.

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.

200: server
210: communication unit
220: processor
230: memory

Claims (19)

A method of personalizing an action recommendation algorithm, performed by the server, includes:
obtaining reference information about the user's behavior;
generating training information by texturing or imaging the reference information;
vectorizing the training information based on at least one of a time, a place, and an occurrence of the behavior; and
updating a behavior recommendation algorithm for the user based on the vectorized training information
including,
The behavior recommendation algorithm recommends a target behavior for the user's current state to the user,
Vectorizing the training information based on at least one of a time, a place, and a situation where the behavior occurred,
acquiring the situation, the situation including a type of an application being executed on the user terminal of the user or actions occurring before and after the action; and
associating the training information with the context;
containing,
How to personalize a behavioral recommendation algorithm.
delete According to claim 1,
The step of obtaining reference information about the user's behavior includes:
obtaining the reference information by receiving the voice uttered by the user through a microphone of the user terminal
containing,
How to personalize a behavioral recommendation algorithm.
4. The method of claim 3,
The step of generating training information by texturing or imaging the reference information includes:
Generating the training information by texturing the voice through natural language processing (NLP)
containing,
How to personalize a behavioral recommendation algorithm.
According to claim 1,
The step of obtaining reference information about the user's behavior includes:
obtaining the reference information for the history of one or more services used by the user
containing,
How to personalize a behavioral recommendation algorithm.
6. The method of claim 5,
The step of generating training information by texturing or imaging the reference information includes:
When the service used by the user is a service for reproducing a target sound source, generating a spectrogram image of the target sound source as the training information
containing,
How to personalize a behavioral recommendation algorithm.
According to claim 1,
Vectorizing the training information based on at least one of a time, a place, and a situation where the behavior occurred,
generating a life-style graph for the user based on the vectorized training information;
including,
Updating the behavior recommendation algorithm for the user based on the vectorized training information comprises:
updating the behavior recommendation algorithm based on the life-style graph
containing,
How to personalize a behavioral recommendation algorithm.
According to claim 1,
determining the current state of the user;
determining the target action for the current state through the action recommendation algorithm; and
recommending the target action to the user
further comprising,
How to personalize a behavioral recommendation algorithm.
9. The method of claim 8,
receiving feedback on the target behavior from the user; and
updating the action recommendation algorithm based on the feedback.
further comprising,
How to personalize a behavioral recommendation algorithm.
10. A computer-readable recording medium storing a program for performing the method of any one of claims 1 and 3 to 9.
A server that performs a method of personalizing a behavioral recommendation algorithm comprises:
a memory in which a program for personalizing a behavior recommendation algorithm is recorded; and
a processor that executes the program
including,
The program is
obtaining reference information about the user's behavior;
generating training information by texturing or imaging the reference information;
vectorizing the training information based on at least one of a time, a place, and an occurrence of the behavior; and
updating a behavior recommendation algorithm for the user based on the vectorized training information
do,
The behavior recommendation algorithm recommends a target behavior for the user's current state to the user,
The step of generating training information by texturing or imaging the reference information includes:
generating a script based on the textualized reference information, and generating a corpus based on the script;
determining a frequency-of-use ranking for words in the corpus;
determining raw data based on the frequency of use ranking; and
generating the corpus and the raw data as training information
containing,
server.
The action recommendation method, performed by the server, is:
determining the current status of the user;
determining a target action for the current state through a personalized action recommendation algorithm for the user; and
recommending the target action to the user
including,
The behavior recommendation algorithm is
obtaining reference information about the user's behavior;
generating training information by texturing or imaging the reference information;
vectorizing the training information based on at least one of a time, a place, and an occurrence of the behavior; and
updating a behavior recommendation algorithm for the user based on the vectorized training information
updated through
The step of generating training information by texturing or imaging the reference information includes:
generating a script based on the textualized reference information, and generating a corpus based on the script;
determining a frequency-of-use ranking for words in the corpus;
determining raw data based on the frequency of use ranking; and
generating the corpus and the raw data as training information
containing,
How to recommend action.
13. The method of claim 12,
The step of determining the current state of the user,
determining the current status based on a history of one or more services used by the user
containing,
How to recommend action.
13. The method of claim 12,
The step of determining the current state of the user,
obtaining target information about the user's behavior; and
generating search information by converting the target information into text or an image;
including,
The search information associated with at least one of a time, place, and situation in which the action occurred is input to the action recommendation algorithm,
How to recommend action.
13. The method of claim 12,
Determining the target behavior for the current state through a behavior recommendation algorithm personalized for the user,
determining the target behavior for the current state based on a life-style graph generated for the user;
containing,
How to recommend action.
16. The method of claim 15,
Determining the target behavior for the current state based on the life-style graph generated for the user comprises:
determining the target behavior for the current state by using at least one of a time, place, and situation for the behavior in the current state.
containing,
How to recommend action.
13. The method of claim 12,
receiving feedback on the target behavior from the user; and
updating the action recommendation algorithm based on the feedback.
further comprising,
How to recommend action.
A computer-readable recording medium containing a program for performing the method of any one of claims 12 to 17.
The server, which performs the method of recommending an action to the user,
a memory in which a program for recommending an action is recorded; and
a processor that executes the program
including,
The program is
determining the current status of the user;
determining a target action for the current state through a personalized action recommendation algorithm for the user; and
recommending the target action to the user
do,
The behavior recommendation algorithm is
obtaining reference information on the user's behavior;
generating training information by texturing or imaging the reference information;
acquiring an occurrence including the type of application being executed in the user terminal of the user or actions occurring before and after the action;
vectorizing the training information by associating the training information with the context;
updating the behavior recommendation algorithm for the user based on the vectorized training information;
updated through
server.

Images