KR20230114440A - Method, system, and computer program for personalized recommendation based on topic of interest - Google Patents

Abstract

A method, system, and computer program for topic-of-interest-based personalized recommendations are disclosed. For keywords used on the Internet, a subject of the keyword is created using a language model to establish a relationship between the keyword and the subject, and a relationship between the keyword and the subject is used to determine the activity of the user on the Internet. At least one corresponding topic of interest may be selected, and personalized recommendation information for the user may be provided based on the topic of interest.

Description

METHOD, SYSTEM, AND COMPUTER PROGRAM FOR PERSONALIZED RECOMMENDATION BASED ON TOPIC OF INTEREST}

The description below relates to personalized recommendation technology.

In general, Internet product recommendation may select and recommend products optimized for the individual based on the individual's purchase history.

As a method mainly used for product recommendation, Collaborative Filtering, which selects users with similar purchase histories to the target user for product recommendation, and recommends products that the target user did not purchase among products purchased by similar users. , Algorithms such as Association Rule Mining, which recommends products with a high possibility of simultaneous purchase or simultaneous clicks for each product using purchase records of multiple users, may be used.

For example, Korean Patent Registration No. 10-0882716 (registered on February 2, 2009) automatically provides product information about a product by using the product code, product name, or model name of the product recognized by the information recommendation agent installed on the user terminal. A recommended technique is disclosed.

Interest data of a new concept representing a semantic relation between a keyword and a subject may be generated using a language model.

Using the relationship between keywords and topics, users' behavior on the Internet can be summarized into topics of interest that can be expressed in explicit words.

Personalized recommendations can be provided to the user based on the explicit reason that the subject is of interest to the user.

A method executed on a computer system, the computer system comprising at least one processor configured to execute computer readable instructions contained in a memory, the method comprising: a keyword used by the at least one processor on the Internet; constructing a relationship between the keyword and the subject by generating a subject of the keyword using a language model for the subject; and selecting, by the at least one processor, at least one topic of interest corresponding to the activity of the user on the Internet using the relationship between the keyword and the topic. and providing, by the at least one processor, personalized recommendation information for the user based on the subject of interest.

According to one aspect, the step of establishing a relationship between the keyword and the subject may include constructing a prompt for a target keyword using example data consisting of a keyword example and a subject of the keyword; and generating a subject of the target keyword according to a pattern of the example data by using the prompt as an input of the language model.

According to another aspect, the step of establishing a relationship between the keyword and the subject includes generating a subject to be expressed as an interest of a user using the keyword for a keyword used for information search or information classification on the Internet. can do.

According to another aspect, the constructing of the relationship between the keyword and the subject may include generating a subject group based on a similarity between the subjects generated through the language model; and selecting at least one representative topic for the keyword through embedding learning based on the relationship between the topics for the topic group.

According to another aspect, the generating of the subject group may include substituting each search word with a subject for a plurality of search words included in a search record, connecting the replaced subjects to express a sentence, and expressing each of the search words as a sentence. A step of grouping subjects having a similar relationship by performing embedding learning by expressing them in words may be included.

According to another aspect, the generating of the subject groups may include linking candidate subjects generated by the language model to the keywords and expressing them as sentences, and performing embedding learning by expressing each of the candidate subjects as words, thereby performing similarity relationship. It may include grouping the topics in .

According to another aspect, the generating of the topic groups includes grouping topics having a similar relationship by performing matrix factorization (MF) on the keyword and candidate topics generated from the language model. can do.

According to another aspect, the selecting of at least one representative subject for the keyword may include performing graph embedding by identifying at least one relationship between an inclusion relationship and an equivalence relationship between the topics; and selecting the representative topic based on a probability between the keyword and the topic using the graph embedding.

According to another aspect, in the step of selecting a topic of interest, a topic having a relationship with at least one activity keyword among a search word input by the user and a tag of a product clicked by the user may be selected as the topic of interest. there is.

According to another aspect, the providing of personalized recommendation information to the user may include recommending at least one keyword related to the activity of the user among keywords having a relationship with the topic of interest as a related keyword. there is.

According to another aspect, the providing of the personalized recommendation information for the user may include selecting at least one keyword related to the activity of the user among keywords having a relationship with the topic of interest as a related keyword; and recommending an item related to the relevant keyword among items on the Internet.

According to another aspect, recommending an item related to the relevant keyword may include selecting a recommended item using at least one of profile information of the user and the subject of interest.

A computer program stored in a computer readable recording medium is provided to execute the method in the computer system.

A computer system comprising at least one processor configured to execute computer readable instructions contained in a memory, wherein the at least one processor generates a subject of a keyword using a language model for a keyword used on the Internet. to establish a relationship between the keyword and the subject, select at least one topic of interest corresponding to a user's activity on the Internet using the relationship between the keyword and the subject, and personalize the user based on the topic of interest It provides a computer system characterized by providing recommended information.

According to embodiments of the present invention, interest data of a new concept representing a semantic relationship between a keyword and a subject may be generated using a large-scale language model.

According to embodiments of the present invention, a user's behavior on the Internet can be summarized into a topic of interest that can be expressed in explicit words by using a relationship between a keyword and a topic.

According to embodiments of the present invention, a personalized recommendation may be provided to a corresponding user based on an explicit reason that the subject is of interest to the user.

1 is a diagram illustrating an example of a network environment according to an embodiment of the present invention.
2 is a block diagram illustrating an example of a computer system according to one embodiment of the present invention.
3 is a diagram showing an example of components that may be included in a processor of a computer system according to an embodiment of the present invention.
4 is a flowchart illustrating an example of a method that a computer system may perform according to an embodiment of the present invention.
5 is a flowchart illustrating an example of a data generation process using a prompt in an embodiment of the present invention.
6 to 7 show examples of data generation in one embodiment of the present invention.
8 to 11 show an example of a subject selection process in one embodiment of the present invention.
12 to 14 show examples of a personalized recommendation process in one embodiment of the present invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

Embodiments of the present invention relate to personalized recommendation technology.

Embodiments, including those specifically disclosed herein, can summarize a user's behavior on the Internet in explicit words (i.e., topics of interest) and use the user's topics of interest to personalize and recommend items on the Internet. .

The personalized recommendation system according to embodiments of the present invention may be implemented by at least one computer system, and the personalized recommendation method according to embodiments of the present invention is performed through at least one computer system included in the personalized recommendation system. It can be. At this time, a computer program according to an embodiment of the present invention may be installed and driven in the computer system, and the computer system may perform the personalized recommendation method according to the embodiments of the present invention under the control of the driven computer program. . The above-described computer program may be combined with a computer system and stored in a computer readable recording medium to execute a personalized recommendation method on a computer.

1 is a diagram illustrating an example of a network environment according to an embodiment of the present invention. The network environment of FIG. 1 shows an example including a plurality of electronic devices 110 , 120 , 130 , and 140 , a plurality of servers 150 and 160 , and a network 170 . 1 is an example for explanation of the invention, and the number of electronic devices or servers is not limited as shown in FIG. 1 . In addition, the network environment of FIG. 1 only describes one example of environments applicable to the present embodiments, and the environment applicable to the present embodiments is not limited to the network environment of FIG. 1 .

The plurality of electronic devices 110, 120, 130, and 140 may be fixed terminals implemented as computer devices or mobile terminals. Examples of the plurality of electronic devices 110, 120, 130, and 140 include a smart phone, a mobile phone, a navigation device, a computer, a laptop computer, a digital broadcast terminal, a personal digital assistant (PDA), and a portable multimedia player (PMP). ), tablet PC, game console, wearable device, internet of things (IoT) device, virtual reality (VR) device, augmented reality (AR) device, and the like. As an example, FIG. 1 shows the shape of a smartphone as an example of the electronic device 110, but in the embodiments of the present invention, the electronic device 110 substantially uses a wireless or wired communication method to transmit other information via the network 170. It may refer to one of various physical computer systems capable of communicating with the electronic devices 120 , 130 , and 140 and/or the servers 150 and 160 .

The communication method is not limited, and short-distance wireless communication between devices as well as a communication method utilizing a communication network (eg, a mobile communication network, a wired Internet, a wireless Internet, and a broadcasting network) that the network 170 may include may also be included. For example, the network 170 may include a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), and a broadband network (BBN). , one or more arbitrary networks such as the Internet. In addition, the network 170 may include any one or more of network topologies including a bus network, a star network, a ring network, a mesh network, a star-bus network, a tree or a hierarchical network, and the like. Not limited.

Each of the servers 150 and 160 communicates with the plurality of electronic devices 110, 120, 130, and 140 through the network 170 to provide commands, codes, files, contents, services, and the like, or a computer device or a plurality of computers. It can be implemented in devices. For example, the server 150 may be a system that provides a first service to a plurality of electronic devices 110, 120, 130, and 140 accessed through the network 170, and the server 160 may also include a network ( It may be a system that provides a second service to a plurality of electronic devices 110, 120, 130, and 140 accessed through 170). As a more specific example, the server 150 provides a service (for example, a personalized recommendation service, etc.) for which the application is intended through an application as a computer program that is installed and driven in the plurality of electronic devices 110, 120, 130, and 140. may be provided to the plurality of electronic devices 110, 120, 130, and 140 as the first service. As another example, the server 160 may provide a service for distributing files for installing and running the above-described application to the plurality of electronic devices 110, 120, 130, and 140 as a second service.

2 is a block diagram illustrating an example of a computer system according to one embodiment of the present invention. Each of the plurality of electronic devices 110 , 120 , 130 , and 140 or each of the servers 150 and 160 described above may be implemented by the computer system 200 shown in FIG. 2 .

As shown in FIG. 2 , the computer system 200 may include a memory 210, a processor 220, a communication interface 230, and an input/output interface 240.

The memory 210 is a computer-readable recording medium and may include a random access memory (RAM), a read only memory (ROM), and a permanent mass storage device such as a disk drive. Here, a non-perishable mass storage device such as a ROM and a disk drive may be included in the computer system 200 as a separate permanent storage device distinct from the memory 210 . Also, an operating system and at least one program code may be stored in the memory 210 . These software components may be loaded into the memory 210 from a computer-readable recording medium separate from the memory 210 . The separate computer-readable recording medium may include a computer-readable recording medium such as a floppy drive, a disk, a tape, a DVD/CD-ROM drive, and a memory card. In another embodiment, software components may be loaded into the memory 210 through the communication interface 230 rather than a computer-readable recording medium. For example, software components may be loaded into memory 210 of computer system 200 based on a computer program installed by files received over network 170 .

The processor 220 may be configured to process commands of a computer program by performing basic arithmetic, logic, and input/output operations. Instructions may be provided to processor 220 by memory 210 or communication interface 230 . For example, processor 220 may be configured to execute received instructions according to program codes stored in a recording device such as memory 210 .

The communication interface 230 may provide a function for the computer system 200 to communicate with other devices (eg, storage devices described above) through the network 170 . For example, a request, command, data, file, etc. generated according to a program code stored in a recording device such as the memory 210 by the processor 220 of the computer system 200 is controlled by the communication interface 230 to the network ( 170) to other devices. Conversely, signals, commands, data, files, etc. from other devices may be received into computer system 200 via communication interface 230 of computer system 200 via network 170 . Signals, commands, or data received through the communication interface 230 may be transferred to the processor 220 or the memory 210, and files may be stored as storage media that the computer system 200 may further include (described above). permanent storage).

The input/output interface 240 may be a means for interface with the input/output device 250 . For example, the input device may include a device such as a microphone, keyboard, or mouse, and the output device may include a device such as a display or speaker. As another example, the input/output interface 240 may be a means for interface with a device in which functions for input and output are integrated into one, such as a touch screen. The input/output device 250 may be configured as one device with the computer system 200 .

Also, in other embodiments, computer system 200 may include fewer or more elements than those of FIG. 2 . However, there is no need to clearly show most of the prior art components. For example, the computer system 200 may be implemented to include at least some of the aforementioned input/output devices 250 or may further include other elements such as a transceiver and a database.

Hereinafter, specific embodiments of a method and system for subject-of-interest-based personalized recommendation will be described.

3 is a block diagram illustrating an example of components that may be included in a processor of a computer system according to an embodiment of the present invention, and FIG. It is a flowchart showing an example of a recommendation method.

The computer system 200 according to this embodiment may provide a personalized recommendation service to a client through a dedicated application installed on the client or access to a web/mobile site related to the computer system 200 .

The processor 220 of the computer system 200 is a component for performing a personalized recommendation method to be described later, and as shown in FIG. 3, the data generator 310, the topic of interest selection unit 320, and the personalized recommendation A portion 330 may be included. Depending on embodiments, components of the processor 220 may be selectively included in or excluded from the processor 220 . Also, components of the processor 220 may be separated or merged to express functions of the processor 220 according to embodiments.

The processor 220 and components of the processor 220 may control the computer system 200 to perform steps included in a personalized recommendation method to be described later. For example, the processor 220 and components of the processor 220 may be implemented to execute instructions according to an operating system code and at least one program code included in the memory 210 .

Here, elements of the processor 220 may be representations of different functions performed by the processor 220 according to instructions provided by program codes stored in the computer system 200 . For example, the data generator 310 is a functional expression of the processor 220 that controls the computer system 200 according to the above-described instructions so that the computer system 200 generates interest data representing a relationship between a keyword and a subject. can be used

The processor 220 may read necessary commands from the memory 210 loaded with commands related to the control of the computer system 200 . In this case, the read command may include a command for controlling the processor 220 to execute a personalized recommendation method to be described later.

Steps included in the personalized recommendation method to be described later may be performed in an order different from the shown order, and some of the steps may be omitted or additional processes may be further included.

Referring to FIG. 4 , in step S410, the data generating unit 310 may generate interest data of a new concept representing a semantic relationship between a keyword and a subject. The data generator 310 may generate a topic suitable for each keyword used on the Internet. In this case, the keyword may include all keywords used in information search or information classification on the Internet, such as a search word used in a search, a keyword given as metadata of certain information, or a tag indicating a classification. For example, the data generating unit 310 may generate interest data for a shopping property search word used for the purpose of finding a product to purchase, a shopping property tag used for shopping recommendation, and the like. Assuming that the relationship between the keyword 'konjac noodles' and the topic 'diet' is established, when user A searches for a product using the search term 'konjac noodles' or clicks a product with the tag 'konjac noodles', The concern expressed in A's behavior can be summed up in an explicit word: 'diet'. For example, the data generator 310 creates a topic that can be expressed as a user's interest using the keyword for a keyword used as a search term or tag using a language model, thereby interest data representing a relationship between a keyword and a topic. can create The data generator 310 may build a relationship between a keyword and a topic by creating a topic suitable for a given keyword through a prompt that is an input sentence of the language model. Relationship data between keywords and subjects may be updated at regular intervals.

In step S420, the interest topic selector 320 may select at least one topic corresponding to the user's activity on the Internet as the user's interest topic by using the relationship between the keyword and the topic. The topic of interest selection unit 320 selects a topic related to the keyword as a topic of interest for the user based on a user activity keyword (a search word entered in real time, a search word for a recent period of time, or a tag of a clicked product, etc.) can do.

In step S430, the personalized recommendation unit 330 may provide personalized recommendations based on the user's subject of interest. The personalized recommendation unit 330 may provide personalized recommendation information using a keyword related to a subject of interest of the user. For example, the personalized recommendation unit 330 may recommend at least one keyword among keywords (search words, tags, etc.) belonging to a topic of interest of the user as a related keyword. At this time, the personalized recommendation unit 330 may measure the similarity between keywords through an embedding methodology (eg, word2vec, FastText, etc.), and among keywords belonging to the user's interest topic, keywords similar to user activity keywords may be used as related keywords. can be selected In addition, the personalized recommendation unit 330 may select a keyword preferred by the user among keywords belonging to a subject of interest of the user as a related keyword through a matrix factorization (MF) model. Assuming that the user is U and the user activity keywords are I, it is possible to obtain the user's preference for each keyword by performing matrix decomposition on the interaction data of size U×I. After obtaining preference for keywords belonging to the user's interest topic, keywords having a preference level of a certain level or higher may be recommended as related keywords of user activity keywords. For users whose user activity keywords have accumulated at or above a certain level, relevant keywords may be recommended using a matrix decomposition model, and for users whose user activity keywords are below a certain level, related keywords may be recommended using an embedding methodology. As another example, the personalized recommendation unit 330 may recommend items on the Internet based on related keywords according to the subject of interest of the user. For example, the personalized recommendation unit 330 is an item related to a related keyword, and may provide some of products clicked after various users input a related keyword as a search term as recommended products. At this time, among the click products, the most recent click product may be weighted so that it is recommended first. In addition, among clicked products, products clicked by other users who match the user's profile information (eg, gender, age, location, etc.) or by other users who have the same interest topic as the user can be selected as the user's recommended product. .

Items recommended based on the subject of interest of the user are described as shopping products, but are not limited thereto. For example, contents such as various platforms or channels such as cafes, blogs, news, and videos may be used as recommended items.

5 is a flowchart illustrating an example of a data generation process using a prompt in an embodiment of the present invention.

Referring to FIG. 5 , in step S501, the processor 220 may construct an input prompt of a language model using an example dataset. The processor 220 may use a predetermined keyword and a subject of the keyword as example data for which a correct answer is determined. A separate data set may exist in addition to a predetermined example data pair, and depending on an embodiment, an example data pair consisting of a keyword and a subject of the corresponding keyword may be selected and used in the data set. The processor 220 may create a dedicated prompt template in which characteristics of a given NLP problem are well reflected. In this case, the prompt template may include a task definition or meta information. In other words, the processor 220 may construct a prompt in the form of a natural language using example data selected from the dataset, and in this case, the prompt is produced in a format that the language model can understand and given as an input to the language model. When the example data is labeled data, the processor 220 may design a prompt so that an input sentence can be created together with label information. The form itself of the prompt may be configured in various ways, and for example, the prompt may be composed of an example keyword and an example subject of the keyword. In this embodiment, the prompt may be composed of at least one keyword, an example data pair consisting of a subject of the keyword, and a target keyword. A detailed method of configuring the prompt will be described with reference to FIG. 6 below.

In step S502, the processor 220 inputs the prompt configured in step S501 to the language model to understand the subject of the given target keyword from the language model. The processor 220 may obtain the subject of the target keyword as a language generation result through a language model generation or completion function after inputting the prompt input sentence into the language model. The processor 220 may obtain a new result having a corresponding pattern by inputting a prompt into a language model and analyzing an example natural language pattern included in the prompt through the language model.

6 to 7 show examples of data generation in one embodiment of the present invention.

Referring to FIG. 6 , the processor 220 configures a prompt 610 serving as an input sentence of a language model using a pair of example data consisting of a keyword example selected as example data and a subject that can be expressed as an interest in the keyword. can do.

The processor 220 may use the prompt 610 including at least one example data pair and the target keyword as an input of the language model, and cause the language model to generate a topic for the target keyword. That is, a topic corresponding to the content of the target keyword may be generated through the completion function of the language model.

The processor 220 may generate a topic for the target keyword according to a pattern of example data included in the prompt using the prompt 610 composed of example data pairs in the form of [keyword+topic] and [target keyword].

In other words, the processor 220 may provide a keyword and a subject of the corresponding keyword as an example, and generate a subject corresponding to the target keyword according to the pattern of the example.

Referring to FIG. 7 , the processor 220 may create a candidate topic 702 suitable for a given keyword 701 by using a language model 700 . For example, when the keyword 'konjac noodles' is given, the language model 700 may create diet, diet food, diet drink, chicken breast brand, body fat management, in-body, and detox diet as the subject of the keyword.

Candidate topics 702 created from the language model 700 generate inconsistent words, such as words with the same meaning or words with different categories, even though the wording is different due to the characteristics (diversity, contingency, etc.) of the language model 700 there is

The processor 220 needs to select a representative topic by grouping the candidate topics 702 created in the language model 700 in order to establish a relationship with the keyword 701 .

8 to 11 show an example of a subject selection process in one embodiment of the present invention.

First, the processor 220 may select a topic having a similar relationship based on a degree of similarity between the topics among the candidate topics generated for the keyword.

For example, referring to FIG. 8 , the processor 220 may group topics having a similar relationship through a method of modeling candidate topics generated from a keyword and a language model based on matrix decomposition (MF).

As another example, referring to FIG. 9 , the processor 220 concatenates all candidate topics generated for a keyword, expresses them as a sentence, expresses each candidate topic as an individual word, and performs embedding learning, thereby performing embedding learning. You can group related topics.

As another example, the processor 220 may perform embedding learning after substituting a record searched by a user for a subject expressing interest. Referring to FIG. 10, when the user's search record is equal to <konjac noodles, protein, cissus flour, chickpea snacks, and konjac jjondugi>, each search word is replaced with a word corresponding to the subject, and then all replaced subjects are connected to form a sentence. , and each search word is expressed as an individual word to perform embedding learning, thereby grouping topics in a similar relationship.

Next, the processor 220 may select a representative subject from a subject group having a similar relationship. The processor 220 may select a representative topic by identifying a relationship between topics in a subject group grouped based on a degree of similarity between the topics.

For example, the processor 220 may build relationships between subjects by using categories (eg, search word categories, etc.) for classifying keywords on the Internet. For example, as shown in FIG. 11, the search term <protein shake barrel, quest protein bar, Nike Crew, Fila Performance Pants, Jeximix Jogger, Adidas T-shirt, dumbbell, vertitest mat, disc 5kg, kettlebell, burn machine> All of them belong to the category 'Homet', and among them, <Nike Crew, Fila Performance Pants, Jexi Mix Jogger, Adidas T-shirt> fall under the category 'Sportswear', and <Dumbbell, Vertitest Mat, Disc 5kg, Kettlebell, Burn Machine> It is assumed that corresponds to the category 'exercise equipment'. When the topics 'Holmt', 'Sportswear', and 'Sportswear' are created through the language model, it is possible to build a relationship between the topics 'Holmt' and 'Sportswear' because search words belonging to 'Sportswear' are included in 'Holmt'. And since the search word belonging to 'sports equipment' is included in 'Holmt', it is possible to build a relationship between the subject 'Holmt' and 'Sports equipment'.

As another example, the processor 220 may establish a relationship between subjects by using an inclusion relationship and an equivalence relationship of words. First, the processor 220 searches for a subject having an inclusion relationship in a subject group having a similar relationship. For example, we assume that topic B belongs to topic A (inclusion relationship) because topic A 'diet' is a word that is completely included in topic B 'diet diet'. Also, if topic A contains the words a1, a2, a3 and topic B contains the words a1, a2, a3, b1, b2, then topic A belongs to topic B because topic B completely contains topic A. (inclusion relationship) is assumed. Next, the processor 220 searches for a subject having an equivalent relationship among subjects other than subjects having an inclusive relationship. For example, if topics C1, C2, and C3 are generated for keyword C, it is assumed that topics C1, C2, and C3 are equivalent. The processor 220 may perform graph embedding (eg, knowledge graph embedding, etc.) using the inclusive relationship and the equivalence relationship between subjects. In other words, as a relationship used in graph embedding, an inclusive relationship and an equivalence relationship can be applied. After embedding a graph representing the relationship between topics, it is possible to determine a representative topic based on probability by probabilistically obtaining which topic belongs to each keyword. In addition to probability, a topic of a certain depth may be selected as a representative topic in a hierarchical structure based on relationships between topics.

Here, the processor 220 may tune a criterion for selecting a representative topic, and perform quantitative and qualitative quality tuning, such as determining how much words overlap between topics or comparing similarities between topics, for example.

Also, the processor 220 may perform post-processing logic that removes a wide range of subjects based on relationships between subjects. For example, the processor 220 may exclude topics in a hierarchical structure based on relationships between topics from a representative topic selection target, in which the number of topics in a lower layer is greater than or equal to a threshold value.

When various topics are generated for each keyword, the processor 220 may finally select at least one representative topic through embedding learning based on a relationship between topics to establish a relationship with the corresponding keyword.

12 to 14 show examples of a personalized recommendation process in an embodiment of the present invention.

Referring to FIG. 12 , when a user inputs a search word 1201 on a search service screen 1200, the processor 220 includes personalized recommendation information 1210 in search results corresponding to the search word 1201 and provides the same. can

For example, the processor 220 selects a topic related to a keyword corresponding to the search term 1201 as a topic of interest 1211 of the corresponding user, and then selects a keyword similar to the search term 1201 or a user belonging to the topic of interest 1211. The related keyword 1212, which is a preferred keyword, may be provided as the personalized recommendation information 1210.

As another example, the processor 220 may provide an item 1213 related to a keyword 1212 related to the search word 1201 as the personalized recommendation information 1210 . For example, the item 1213 is a product registered as a tag with the related keyword 1212, a product that other users searched for and clicked on with the related keyword 1212, or another user whose profile (gender, age, etc.) is similar to that of the user. may be selected as a clicked product.

Referring to FIG. 13 , when the user accesses the service screen 1300 including the personalized recommendation area 1310, the processor 220 selects a topic of interest 1311 of the user based on user activity during a recent period. can provide At this time, the processor 220 selects a subject related to a user's activity keyword (a search term for a recent period or a tag of a clicked product, etc.) as a subject of interest 1311 of the user and exposes it to the personalized recommendation area 1310. can

When a specific topic is selected from among the topics of interest 1311 exposed on the personalized recommendation area 1310, the processor 220, as shown in FIG. A related keyword 1412 that is a preferred keyword may be provided through the personalized recommendation area 1310 .

The processor 220 may expose the relevant keyword 1412 and an item 1413 related to the relevant keyword 1412 to the personalized recommendation area 1310 . For example, the item 1413 is a product registered as a tag with a related keyword 1412, a product clicked by other users after searching for the related keyword 1412, or another user whose profile or interest topic 1311 is similar to that of the user. may be selected as a clicked product.

As described above, according to the embodiments of the present invention, it is possible to generate new conceptual data representing a semantic relationship between a keyword and a subject using a large-scale language model, and based on the data on the relationship between a keyword and a subject, users' behavior on the Internet can be determined. It can be summarized in explicit words (topics of interest), based on which it can provide personalized recommendations to that user.

The devices described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, devices and components described in the embodiments include a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), and a programmable PLU (programmable logic unit). logic unit), microprocessor, or any other device capable of executing and responding to instructions. The processing device may run an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. You can command the device. The software and/or data may be embodied in any tangible machine, component, physical device, computer storage medium or device to be interpreted by or to provide instructions or data to a processing device. there is. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. In this case, the medium may continuously store a program executable by a computer or temporarily store the program for execution or download. In addition, the medium may be various recording means or storage means in the form of a single or combined hardware, but is not limited to a medium directly connected to a certain computer system, and may be distributed on a network. Examples of the medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical recording media such as CD-ROM and DVD, magneto-optical media such as floptical disks, and ROM, RAM, flash memory, etc. configured to store program instructions. In addition, examples of other media include recording media or storage media managed by an app store that distributes applications, a site that supplies or distributes various other software, and a server.

As described above, although the embodiments have been described with limited examples and drawings, those skilled in the art can make various modifications and variations from the above description. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

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

Claims (20)

In a method executed on a computer system,
The computer system includes at least one processor configured to execute computer readable instructions contained in a memory;
The method,
generating, by the at least one processor, a subject of the keyword using a language model for a keyword used on the Internet, and establishing a relationship between the keyword and the subject; and
selecting, by the at least one processor, at least one topic of interest corresponding to an activity of a user on the Internet using a relationship between the keyword and the topic; and
providing, by the at least one processor, personalized recommendation information for the user based on the subject of interest;
How to include. According to claim 1,
Establishing a relationship between the keyword and the subject,
configuring a prompt for a target keyword using example data consisting of a keyword example and a topic of the keyword; and
generating a subject of the target keyword according to a pattern of the example data by using the prompt as an input of the language model;
How to include. According to claim 1,
Establishing a relationship between the keyword and the subject,
A step of creating a topic to be expressed as an interest of a user who uses the keyword for a keyword used for information search or information classification on the Internet.
How to include. According to claim 1,
Establishing a relationship between the keyword and the subject,
generating a topic group based on a similarity between the topics generated through the language model; and
Selecting at least one representative topic for the keyword through embedding learning based on the relationship between the topics for the topic group
How to include. According to claim 4,
The step of creating the subject group,
For a plurality of search words included in the search record, each search word is replaced with a subject, the replaced subject is connected and expressed as a sentence, and each of the search words is expressed as a word to perform embedding learning. Steps to group topics in
How to include. According to claim 4,
The step of creating the subject group,
Grouping topics having a similar relationship by connecting candidate topics generated by the language model to the keywords to express sentences and performing embedding learning by expressing each of the candidate topics as words.
How to include. According to claim 4,
The step of creating the subject group,
Grouping topics having a similar relationship by performing matrix factorization (MF) on candidate topics generated from the keyword and the language model.
How to include. According to claim 4,
Selecting at least one representative topic for the keyword,
performing graph embedding by identifying at least one relationship between an inclusion relationship and an equivalence relationship between the subjects; and
Selecting the representative topic based on the probability between the keyword and the topic using the graph embedding
How to include. According to claim 1,
The step of selecting the topic of interest is,
Selecting, as the topic of interest, a topic having a relationship with at least one activity keyword among a search word input by the user and a tag of a product clicked by the user.
A method characterized by. According to claim 1,
Providing personalized recommendation information for the user,
recommending at least one keyword related to the activity of the user among keywords having a relationship with the topic of interest as a related keyword;
How to include. According to claim 1,
Providing personalized recommendation information for the user,
selecting at least one keyword related to the activity of the user among keywords having a relationship with the subject of interest as a related keyword; and
Step of recommending an item related to the related keyword among items on the Internet
How to include. According to claim 11,
The step of recommending an item related to the related keyword,
Selecting a recommended item using at least one of the user's profile information and the subject of interest
How to include. A computer program stored in a computer readable recording medium to execute the method of any one of claims 1 to 12 in the computer system. In a computer system,
at least one processor configured to execute computer readable instructions contained in memory;
including,
The at least one processor,
Creating a subject of the keyword using a language model for a keyword used on the Internet to establish a relationship between the keyword and the subject;
selecting at least one topic of interest corresponding to a user's activity on the Internet using the relationship between the keyword and the topic;
Providing personalized recommendation information for the user based on the topic of interest.
Characterized by a computer system. According to claim 14,
The at least one processor,
Construct a prompt for a target keyword using example data consisting of keyword examples and topics of the keyword;
Generating a subject of the target keyword according to a pattern of the example data by using the prompt as an input of the language model.
Characterized by a computer system. According to claim 14,
The at least one processor,
For keywords used in information retrieval or information classification on the Internet, creating topics to express the interest of users using the keywords.
Characterized by a computer system. According to claim 14,
The at least one processor,
Create a topic group based on a similarity between topics generated through the language model;
Selecting at least one representative topic for the keyword through embedding learning based on the relationship between the topics targeting the topic group
Characterized by a computer system. According to claim 17,
The at least one processor,
For a plurality of search terms included in the search record, after substituting each search term with a topic, connecting the replaced topic to express it as a sentence, and expressing each of the search terms as a word to perform embedding learning, thereby grouping topics with similar relationships
Characterized by a computer system. According to claim 17,
The at least one processor,
Graph embedding is performed by identifying at least one relationship between an inclusion relationship and an equivalence relationship between the subjects,
Selecting the representative topic based on the probability between the keyword and the topic using the graph embedding
Characterized by a computer system. According to claim 14,
The at least one processor,
Recommending at least one keyword related to the activity of the user among keywords having a relationship with the topic of interest as a related keyword or recommending an item related to the related keyword among items on the Internet
Characterized by a computer system.

Images