JP7529344B2 - Method, system, and computer program for personalized recommendations based on topics of interest - Google Patents

Description

Application of Article 30, Paragraph 2 of the Patent Act - Published on November 23, 2021 at the following website address: "https://blog.naver.com/naver_search/222575844697" - Published on November 24, 2021 at the following website address: "https://tv.naver.com/v/23649833" - Published on November 24, 2021 at the following website address: "https://www.navercorp.com/promotion/pressReleaseView/30722" - Published on December 8, 2021 at the following website address. “https://www.ddaily.co.kr/news/article/?no=226991”

The following description relates to personalized recommendation techniques.

Product recommendations on the Internet typically involve selecting and recommending products that are most suitable for an individual based on that individual's purchasing history.

The main methods used for product recommendations include collaborative filtering, which selects users with similar purchasing histories as the target user to whom a product is to be recommended, and recommends products that the target user has not purchased among the products purchased by these users, and association rule mining, which uses the purchasing records of multiple users to recommend products that are likely to be purchased or clicked on simultaneously.

For example, Patent Document 1 (registered on February 2, 2009) discloses a technology that automatically recommends product information for a product using the product code, product name, or model of the product recognized by an information recommendation agent installed on a user terminal device.

Korean Patent No. 10-0882716

Language models can be used to generate new conceptual interest data that shows semantic relationships between keywords and topics.

The relationship between keywords and topics can be used to summarize a user's Internet activity into topics of interest that can be expressed in explicit words.

Users can be provided with personalized recommendations based explicitly on their topics of interest.

A method is provided that is executed on a computer system, the computer system including at least one processor configured to execute computer-readable instructions contained in a memory, the method including: generating topics for keywords used on the Internet using a language model and building relationships between the keywords and the topics, the at least one processor selecting at least one topic of interest corresponding to the user's activities on the Internet using the relationships between the keywords and the topics, and providing personalized recommendation information to the user based on the topics of interest, by the at least one processor.

According to one aspect, the step of establishing a relationship between the keyword and the topic may include constructing a prompt for the target keyword using sample data consisting of keyword samples and topics of the keyword, and using the prompt as an input to the language model and generating topics for the target keyword according to patterns in the sample data.

In another aspect, the step of establishing relationships between the keywords and the topics may include generating topics for keywords used in searching or classifying information on the Internet to represent interests of users who use the keywords.

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

According to another aspect, the step of generating the topic group may include a step of grouping similar topics by replacing each of a plurality of search terms included in the search record with a topic, linking the replaced topics to express them as a sentence, and expressing each of the search terms as a word and performing embedding learning.

According to another aspect, the step of generating the topic groups may include a step of expressing candidate topics generated by the language model for the keyword as sentences by linking them, and expressing each of the candidate topics as a word and performing embedding learning to group topics that are similar to each other.

According to another aspect, generating the topic groups may include grouping similar topics by performing matrix factorization (MF) on the keywords and the candidate topics generated by the language model.

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

According to another aspect, the step of selecting the topic of interest may select a topic that is related to at least one of the activity keywords among the search term entered by the user and the tag of the product clicked by the user as the topic of interest.

According to another aspect, the step of providing the user with personalized recommendation information may include the step of recommending at least one keyword related to the user's activity as a related keyword from among the keywords related to the topic of interest.

According to another aspect, the step of providing the user with personalized recommendation information may include selecting at least one keyword related to the user's activity from among the keywords related to the topic of interest as a related keyword, and recommending items on the Internet that are related to the related keyword.

According to yet another aspect, the step of recommending items related to the related keywords may include selecting recommended items using at least one of the user's profile information and the topics of interest.

A computer program for executing the method on the computer system is provided.

A computer system is provided, comprising at least one processor configured to execute computer-readable instructions contained in a memory, the at least one processor utilizing a language model to generate topics for keywords used on the Internet and build relationships between the keywords and the topics, utilizing the relationships between the keywords and the topics to select at least one topic of interest corresponding to the user's activities on the Internet, and providing personalized recommendation information to the user based on the topic of interest.

According to an embodiment of the present invention, by utilizing a large-scale language model, it is possible to generate new conceptual interest data that shows semantic relationships between keywords and topics.

According to an embodiment of the present invention, by utilizing the relationship between keywords and topics, a user's activity on the Internet can be summarized into topics of interest that can be expressed in explicit words.

Embodiments of the present invention allow for personalized recommendations to be provided to a user based explicitly on the user's topics of interest.

FIG. 1 illustrates an example of a network environment in accordance with an embodiment of the present invention. FIG. 1 is a block diagram illustrating an example computer system according to an embodiment of the present invention. FIG. 2 illustrates example components that may be included in a processor of a computer system in accordance with one embodiment of the present invention. 1 is a flow chart illustrating an example of a method that may be performed by a computer system in accordance with an embodiment of the present invention. 1 is a flow chart illustrating an example of a data generation process using prompts in accordance with an embodiment of the present invention. FIG. 13 is a diagram illustrating an example of data generation in one embodiment of the present invention. FIG. 13 is a diagram illustrating an example of data generation in one embodiment of the present invention. FIG. 2 illustrates an example of a topic selection process in accordance with an embodiment of the present invention. FIG. 2 illustrates an example of a topic selection process in accordance with an embodiment of the present invention. FIG. 2 illustrates an example of a topic selection process in accordance with an embodiment of the present invention. FIG. 2 illustrates an example of a topic selection process in accordance with an embodiment of the present invention. FIG. 2 illustrates an example of a personalized recommendation process in accordance with an embodiment of the present invention. FIG. 2 illustrates an example of a personalized recommendation process in accordance with an embodiment of the present invention. FIG. 2 illustrates an example of a personalized recommendation process in accordance with an embodiment of the present invention.

The following describes in detail an embodiment of the present invention with reference to the attached 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 into explicit words (i.e., topics of interest) and can utilize the user's topics of interest to personalize and recommend items on the Internet.

The personalized recommendation system according to the embodiment of the present invention may be realized by at least one computer system, and the personalized recommendation method according to the embodiment of the present invention may be executed by at least one computer system included in the personalized recommendation system. In this case, a computer program according to an embodiment of the present invention may be installed and executed in the computer system, and the computer system may execute the personalized recommendation method according to the embodiment of the present invention according to the control of the executed computer program. The above-mentioned computer program may be recorded in a computer-readable recording medium in combination with the computer system to cause the computer to execute the personalized recommendation method.

Figure 1 is a diagram showing an example of a network environment in one embodiment of the present invention. The network environment in Figure 1 shows an example including multiple electronic devices 110, 120, 130, 140, multiple servers 150, 160, and a network 170. Figure 1 is merely an example for explaining the invention, and the number of electronic devices and the number of servers are not limited to those shown in Figure 1. Furthermore, the network environment in Figure 1 is merely an example of an environment applicable to this embodiment, and the environment applicable to this embodiment is not limited to the network environment in Figure 1.

The electronic devices 110, 120, 130, and 140 may be fixed terminals or mobile terminals realized by computer devices. Examples of the electronic devices 110, 120, 130, and 140 include smartphones, mobile phones, navigation systems, personal computers (PCs), notebook PCs, digital broadcasting terminals, personal digital assistants (PDAs), portable multimedia players (PMPs), tablets, game consoles, wearable devices, Internet of Things (IoT) devices, virtual reality (VR) devices, and augmented reality (AR) devices. As an example, FIG. 1 shows a smartphone as an example of the electronic device 110, but in an embodiment of the present invention, the electronic device 110 may refer to one of a variety of physical computer systems that can communicate with other electronic devices 120, 130, 140 and/or servers 150, 160 via a network 170 using a substantially wireless or wired communication method.

The communication method is not limited, and may include not only a communication method using a communication network that the network 170 can include (for example, a mobile communication network, a wired Internet, a wireless Internet, or a broadcast network), but also short-range wireless communication between devices. For example, the network 170 may include any one or more of a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a broadband network (BBN), the Internet, and the like. Furthermore, the network 170 may include any one or more of a network topology including, but not limited to, a bus network, a star network, a ring network, a mesh network, a star-bus network, a tree or a hierarchical network.

Each of the servers 150 and 160 may be realized by one or more computer devices that communicate with the multiple electronic devices 110, 120, 130, and 140 via the network 170 to provide instructions, codes, files, content, services, and the like. For example, the server 150 may be a system that provides a first service to the multiple electronic devices 110, 120, 130, and 140 connected via the network 170, and the server 160 may also be a system that provides a second service to the multiple electronic devices 110, 120, 130, and 140 connected via the network 170. As a more specific example, the server 150 may provide a service (such as a personalized recommendation service, for example) targeted by an application, which is a computer program installed and executed in the multiple electronic devices 110, 120, 130, and 140, as a first service to the multiple electronic devices 110, 120, 130, and 140. As another example, the server 160 may provide, as a second service, a service of distributing files for installing and executing the above-mentioned application to multiple electronic devices 110, 120, 130, and 140.

Figure 2 is a block diagram showing an example of a computer system in one embodiment of the present invention. Each of the multiple electronic devices 110, 120, 130, and 140 and each of the servers 150 and 160 described above may be realized by the computer system 200 shown in Figure 2.

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

The memory 210 is a computer-readable recording medium and may include a RAM (random access memory), a ROM (read only memory), and a permanent mass storage device such as a disk drive. Here, a permanent mass storage device such as a ROM or a disk drive may be included in the computer system 200 as a separate permanent storage device separate from the memory 210. In addition, the memory 210 may store an operating system and at least one program code. Such software components may be loaded into the memory 210 from a computer-readable recording medium separate from the memory 210. Such 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, a memory card, etc. In another embodiment, the software components may be loaded into the memory 210 through a communication interface 230 that is not a computer-readable recording medium. For example, software components may be loaded into the memory 210 of the computer system 200 based on a computer program that is installed by a file received over the network 170.

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

The communication interface 230 may provide a function for the computer system 200 to communicate with other devices (for example, the above-mentioned recording device) via the network 170. As an example, requests, commands, data, files, etc. generated by the processor 220 of the computer system 200 according to program code recorded in a recording device such as the memory 210 may be transmitted to other devices via the network 170 under the control of the communication interface 230. Conversely, signals, commands, data, files, etc. from other devices may be received by the computer system 200 through the communication interface 230 of the computer system 200 via the network 170. The signals, commands, data, etc. received through the communication interface 230 may be transmitted to the processor 220 or the memory 210, and the files, etc. may be recorded on a recording medium (the above-mentioned permanent recording device) that the computer system 200 may further include.

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

Also, in other embodiments, computer system 200 may include fewer or more components than those of FIG. 2. However, most prior art components need not be explicitly shown in the figures. For example, computer system 200 may be implemented to include at least some of the input/output devices 250 described above, and may further include other components such as a transceiver, a database, etc.

Specific embodiments of methods and systems for personalized recommendations based on topics of interest are described below.

Figure 3 is a block diagram showing an example of components that may be included in a processor of a computer system in one embodiment of the present invention, and Figure 4 is a flowchart showing an example of a personalized recommendation method that may be executed by a computer system in one embodiment of the present invention.

The computer system 200 of this embodiment may provide a personalized recommendation service to a client through a dedicated application installed on the client or through a connection to a web/mobile site associated with the computer system 200.

The processor 220 of the computer system 200 may include a data generation unit 310, an interest topic selection unit 320, and a personalized recommendation unit 330, as shown in FIG. 3, as components for performing the personalized recommendation method described below. Depending on the embodiment, the components of the processor 220 may be selectively included or excluded from the processor 220. Also, depending on the embodiment, the components of the processor 220 may be separated or combined to express the functions of the processor 220.

Such processor 220 and components of processor 220 may control computer system 200 to execute steps included in the personalized recommendation method described below. For example, processor 220 and components of processor 220 may be implemented to execute instructions from operating system code and at least one program code contained in memory 210.

Here, the components of the processor 220 may be representations of different functions executed by the processor 220 according to instructions provided by the program code recorded in the computer system 200. For example, the data generator 310 may be used as a functional representation of the processor 220 that controls the computer system 200 according to the above-mentioned instructions so that the computer system 200 generates interest data representing relationships between keywords and topics.

The processor 220 may read the necessary instructions from the memory 210 loaded with instructions related to the control of the computer system 200. In this case, the read instructions may include instructions for controlling the processor 220 to perform the personalized recommendation method described below.

The steps in the personalized recommendation method described below may be performed in a different order than shown in the figures, some steps may be omitted, or additional processes may be included.

Referring to FIG. 4, in step 410, the data generating unit 310 may generate interest data of a new concept that represents a semantic relationship between a keyword and a topic. The data generating unit 310 may generate a topic suitable for each keyword used on the Internet. In this case, the keyword may mean a comprehensive term that includes all keywords used for information search and information classification on the Internet, such as a search term that is a keyword used during a search, a keyword assigned as metadata for certain information, and a tag that indicates a classification. For example, the data generating unit 310 may generate interest data for a shopping search term used for the purpose of searching for a product to be purchased, a shopping tag used for shopping recommendations, and the like. Assuming that a relationship is established between the keyword "itokonnyaku" and the topic "diet", when user A searches for a product using the search term "itokonnyaku" or clicks on a product tagged with "itokonnyaku", the interest that can be read from the behavior of user A can be summarized as the explicit word "diet". As an example, the data generating unit 310 may generate interest data representing a relationship between a keyword and a topic by generating a topic that can be expressed as an interest of a user who uses a keyword used in a search term or tag in a language model. The data generating unit 310 may construct a relationship between a keyword and a topic by generating a topic appropriate for a keyword given in a prompt that is an input sentence of the language model. The relationship data between a keyword and a topic may be updated at regular intervals.

In step 420, the interest topic selection unit 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 interest topic selection unit 320 may select a topic related to a user activity keyword (such as a search term entered in real time, or a recent search term or clicked product tag within a certain period of time) as the user's interest topic.

In step 430, the personalized recommendation unit 330 may provide personalized recommendations based on the user's topics of interest. The personalized recommendation unit 330 may provide personalized recommendation information using keywords related to the user's topics of interest. As an example, the personalized recommendation unit 330 may recommend at least one keyword (search term, tag, etc.) belonging to the user's topics of interest as a related keyword. In this case, the personalized recommendation unit 330 may measure the similarity between keywords based on an embedding methodology (e.g., word2vec, FastText, etc.) and may select keywords similar to the user activity keywords from among the keywords belonging to the user's topics of interest as related keywords. In addition, the personalized recommendation unit 330 may select keywords preferred by the user from among the keywords belonging to the user's topics of interest as related keywords using a matrix factorization (MF) model. When a user is U and a user activity keyword is I, a user preference for each keyword may be obtained by performing matrix decomposition on interaction data of U×I size. After obtaining the preference for the keyword belonging to the user's topic of interest, a keyword having a preference level of a certain level or higher may be recommended as a related keyword of the user activity keyword. For a user whose user activity keyword is at a certain level or higher, a related keyword may be recommended using a matrix decomposition model, and for a user whose user activity keyword is below a certain level, a related keyword may be recommended using an embedding methodology. As another example, the personalized recommendation unit 330 may recommend an item on the Internet based on the related keyword according to the user's topic of interest. For example, the personalized recommendation unit 330 may provide some of the products that a large number of users have actually clicked by inputting the related keyword as a search term as an item related to the related keyword as a recommended product. In this case, a weight may be applied to the recently clicked products among the clicked products so that they are preferentially recommended. Additionally, from among the clicked products, products clicked by other users who match the user's profile information (e.g., gender, age, location, etc.) or who have the same topics of interest as the user may be selected as recommended products for that user.

Although shopping items have been described as an example of items recommended based on a user's topics of interest, this is not intended to be limiting. For example, content from various platforms and channels such as cafes, blogs, news, videos, etc. may be used as recommended items.

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

Referring to FIG. 5, in step 501, the processor 220 may use a sample dataset to configure an input prompt for the language model. The processor 220 may use a predefined keyword and a topic of the keyword as sample data for which a correct answer is defined. In addition to the predefined sample data pair, a separate dataset may exist, and in some embodiments, a sample data pair consisting of a keyword and a topic of the keyword may be selected and used within the dataset. The processor 220 may create a dedicated prompt template that appropriately reflects the characteristics of a given NLP problem, and in this case, the prompt template may include a task definition and meta information. In other words, the processor 220 may configure a natural language prompt using sample data selected from the dataset. In this case, the prompt is created in a format that the language model can understand and is provided as an input for the language model. If the sample data is labeled data, the processor 220 may design the prompt so that an input sentence is generated together with label information. The format of the prompt itself may be configured in various ways, and as an example, the prompt may be configured with a keyword sample and a topic of the keyword as a sample. In this embodiment, the prompt may be composed of a sample data pair consisting of at least one keyword and a topic of the keyword, and a target keyword. A detailed method of constructing the prompt will be described with reference to FIG. 6.

In step 502, the processor 220 may input the prompt constructed in step 501 into a language model and understand the topic of the given target keyword from the language model. After inputting the prompt input sentence into the language model, the processor 220 may obtain the topic of the target keyword as a language generation result by the generation or completion function of the language model. The processor 220 may input the prompt into the language model and obtain a new result having the pattern by analyzing the sample natural language pattern included in the prompt in the language model.

Figures 6 and 7 show an example of data generation in one embodiment of the present invention.

Referring to FIG. 6, the processor 220 may use sample data pairs consisting of keyword samples selected as sample data and topics that can be expressed as interests for the keywords to construct a prompt 610 that serves as an input sentence for the language model.

The processor 220 may input a prompt 610 consisting of at least one sample data pair and a target keyword to a language model, and the language model may generate a topic for the target keyword. That is, a topic corresponding to the content of the target keyword may be generated by a completion function of the language model.

The processor 220 may use a prompt 610 consisting of a sample data pair in the form of [keyword + topic] and a [target keyword] to generate a topic for the target keyword according to the pattern of the sample data included in the prompt.

In other words, the processor 220 can provide keywords and topics of the keywords as samples and generate topics corresponding to the target keywords according to the patterns of the samples.

Referring to FIG. 7, for a given keyword 701, the processor 220 may use a language model 700 to generate candidate topics 702 appropriate for the keyword 701. For example, when the keyword "itokonnyaku" is given, the language model 700 may generate topics for the keyword such as diet, diet food, diet drink, salad chicken brand, body fat management, inbody, detox diet, etc.

Due to the characteristics of the language model 700 (diversity, randomness, etc.), the candidate topics 702 generated by the language model 700 have the problem of generating inconsistent words, such as words with different wordings but the same meaning, or words in different categories.

In order to build a relationship with the keyword 701, the processor 220 needs to group the candidate topics 702 generated by the language model 700 and select a representative topic.

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

First, the processor 220 may select topics that are similar to the candidate topics generated for a keyword based on the similarity between the topics.

As an example, referring to FIG. 8, the processor 220 may group similarly related topics using a method of modeling candidate topics generated by keywords and language models based on matrix factorization (MF).

As another example, referring to FIG. 9, the processor 220 may concatenate all the candidate topics generated for a keyword to represent them as a sentence, and then represent each candidate topic as an individual word to perform embedding learning, thereby grouping similar topics.

As another example, the processor 220 may replace the record searched by the user with a topic expressing the user's interest and then perform embedding learning. Referring to FIG. 10, when the user search record is, for example, <itokonnyaku, protein, cissus, chickpea snack, konnyaku jelly>, each search term may be replaced with a word corresponding to the topic, and all the replaced topics may be linked together to express the search terms as individual words and embedding learning may be performed to group similar topics.

Next, the processor 220 may select a representative topic from the topic group that has a similar relationship. The processor 220 may select a representative topic by understanding the relationship between topics in the grouped topic group based on the similarity between the topics.

As an example, the processor 220 may use categories (e.g., search term categories) that classify keywords on the Internet to build relationships with topics. For example, as shown in FIG. 11, it is assumed that the search terms <protein shake bottle, QUEST protein bar, Nike Crew, Fila performance pants, Zexy mix joggers, Adidas T-shirt, dumbbell, burpee test mat, barbell 5 kg, kettlebell, burn machine> all fall under the category "home training", of which <Nike Crew, Fila performance pants, Zexy mix joggers, Adidas T-shirt> fall under the category "sportswear", and <dumbbell, burpee test mat, barbell 5 kg, kettlebell, burn machine> fall under the category "exercise equipment". If the topics "home workout," "sportswear," and "exercise equipment" are generated by a language model, a relationship can be built between the topics "home workout" and "sportswear" because search terms belonging to "sportswear" are included in "home workout," and a relationship can be built between the topics "home workout" and "exercise equipment" because search terms belonging to "exercise equipment" are included in "home workout."

As another example, the processor 220 may use the inclusion and equivalence relationships of words to build relationships with topics. First, the processor 220 searches for topics in an inclusion relationship from a topic group that is in a similar relationship. For example, topic A "diet" is a word that is completely contained in topic B "diet menu", so it is assumed that topic B belongs to topic A (inclusion relationship). Also, if topic A contains words a1, a2, and a3, and topic B contains words a1, a2, a3, b1, and b2, topic B completely contains topic A, so it is assumed that topic A belongs to topic B (inclusion relationship). Next, the processor 220 searches for topics in an equivalence relationship from the remaining topics excluding the topics in the inclusion relationship. For example, if topics C1, C2, and C3 are generated for keyword C, it is assumed that topics C1, C2, and C3 are in an equivalence relationship. The processor 220 may perform graph embedding (e.g., Knowledge Graph Embedding, etc.) using the inclusion and equivalence relationships between topics. In other words, the inclusion and equivalence relationships may be applied as the relationships used in graph embedding. After embedding the graph expressing the relationships between topics, it may be determined probabilistically which topic each keyword belongs to, and the representative topic may be determined based on a probability criterion. In addition to probability, a topic of a certain depth in a hierarchical structure based on the relationships between topics may be selected as the representative topic.

Here, the processor 220 may tune the criteria for selecting representative topics. For example, the processor 220 may perform quantitative and qualitative quality tuning, such as determining the degree of overlap of words between topics and comparing the similarity between topics.

The processor 220 may also execute post-processing logic to remove a wide range of topics based on the relationships between topics. As an example, the processor 220 may exclude from the selection of representative topics those topics that have a threshold number of topics in a lower hierarchy in a hierarchical structure based on the relationships between topics.

When multiple topics are generated for each keyword, the processor 220 may ultimately select at least one topic that can be representative by embedding learning based on the relationship between the topics, and build a relationship with the keyword.

Figures 12-14 show an example of a personalized recommendation process in one embodiment of the present invention.

Referring to FIG. 12, when a user inputs a search term 1201 on a search service screen 1200, the processor 220 may provide search results corresponding to the search term 1201 including personalized recommendation information 1210.

As an example, the processor 220 may select topics related to the keywords corresponding to the search terms 1201 as topics of interest 1211 of the user, and then provide related keywords 1212, which are keywords similar to the search terms 1201 or preferred by the user among the keywords belonging to the topics of interest 1211, as personalized recommendation information 1210.

As another example, the processor 220 may provide items 1213 related to related keywords 1212 for the search term 1201 as personalized recommendation information 1210. For example, the items 1213 may be selected products for which the related keywords 1212 are registered as tags, products that other users have searched for and clicked on using the related keywords 1212, or products that other users with a similar profile (gender, age, etc.) to the user have clicked on.

Referring to FIG. 13, when a user accesses a service screen 1300 including a personalized recommendation area 1310, the processor 220 may provide the user with topics of interest 1311 based on the user's activity within a recent certain period of time. In this case, the processor 220 may select topics related to user activity keywords (such as search terms and clicked product tags within a recent certain period of time) as the user's topics of interest 1311 and display them in the personalized recommendation area 1310.

When a specific topic is selected from among the topics of interest 1311 displayed in the personalized recommendation area 1310, the processor 220 may provide in the personalized recommendation area 1310, as shown in FIG. 14, related keywords 1412, which are keywords belonging to the selected topic 1411 that are similar to the user activity keywords or are preferred by the user.

The processor 220 may display items 1413 related to the related keywords 1412 in the personalized recommendation area 1310 together with the related keywords 1412. For example, the items 1413 may be selected products for which the related keywords 1412 are registered as tags, products that other users have searched for and clicked on using the related keywords 1412, or products that other users who have a similar profile or topic of interest 1311 to the user have clicked on.

Thus, according to an embodiment of the present invention, a large-scale language model can be used to generate new conceptual data that represents semantic relationships between keywords and topics, and based on the keyword-topic relationship data, user behavior on the Internet can be summarized into explicit words (topics of interest), and personalized recommendations can be provided to the user based on this.

The above-described devices may be realized by hardware components, software components, and/or a combination of hardware and software components. For example, the devices and components described in the embodiments may be realized using one or more general-purpose or special-purpose computers, such as a processor, controller, arithmetic logic unit (ALU), digital signal processor, microcomputer, field programmable gate array (FPGA), programmable logic unit (PLU), microprocessor, or various devices capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications that run on the OS. The processing device may also respond to the execution of the software and access, record, manipulate, process, and generate data. For ease of understanding, the description may be given as if one processing device is used, but one skilled in the art will understand that the processing device may include multiple processing elements and/or multiple types of processing elements. For example, a processing unit may include multiple processors or one processor and one controller. Other processing configurations, such as parallel processors, are also possible.

The software may include computer programs, codes, instructions, or a combination of one or more of these, and may configure or instruct the processing device to operate as desired, either independently or collectively. The software and/or data may be embodied in any type of machine, component, physical device, computer storage medium, or device to be interpreted based on the processing device or to provide instructions or data to the processing device. The software may be distributed and stored or executed in a distributed manner on computer systems connected by a network. The software and data may be stored on one or more computer-readable storage media.

The method according to the embodiment may be realized in the form of program instructions executable by various computer means and recorded on a computer-readable medium. Here, the medium may be one that continuously records the computer-executable program or one that temporarily records it for execution or download. The medium may be one of various recording means or storage means in the form of a single or multiple hardware combined, and is not limited to a medium directly connected to a certain computer system, but may be one that is distributed over a network. Examples of the medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and ROMs, RAMs, flash memories, etc., configured to record program instructions. Other examples of the medium include recording media or storage media managed by application stores that distribute applications, or sites, servers, etc. that supply or distribute various other software.

Although the embodiments have been described above based on limited embodiments and drawings, those skilled in the art will appreciate that various modifications and variations can be made from the above description. For example, the described techniques may be performed in an order different from that described, and/or the components of the described systems, structures, devices, circuits, etc. may be combined or combined in a manner different from that described, or may be counterbalanced or replaced by other components or equivalents, and still achieve suitable results.

Therefore, different embodiments that are equivalent to the scope of the claims are within the scope of the attached claims.

220: Processor 310: Data Generation Unit 320: Interest Topic Selection Unit 330: Personalized Recommendation Unit

Claims (15)

1. A method implemented on a computer system, comprising:
The computer system includes at least one processor configured to execute computer-readable instructions contained in a memory;
The method comprises:
inputting a prompt, which is composed of a sample data pair of a keyword sample and a topic of the keyword, and a target keyword, into a language model by the at least one processor for a keyword used on the Internet , to generate a plurality of candidate topics corresponding to the target keyword, selecting candidate topics having a similar relationship based on a similarity between the candidate topics, performing graph embedding to express a relationship between the plurality of candidate topics by using an inclusion relationship between the similar candidate topics and an equivalence relationship between the candidate topics excluding the inclusion relationship, and constructing a relationship between the target keyword and the plurality of candidate topics by using the graph embedding; selecting at least one interest topic corresponding to a user's activity on the Internet by the at least one processor by using the relationship between the target keyword and the plurality of candidate topics; and providing personalized recommendation information to the user based on the interest topics by the at least one processor. The topics of interest that are related to the keywords include:
The method of claim 1 , further comprising: indicating , in association with the keywords used to search for or categorize information on the Internet, the interests of a user using the keywords. The constructing step includes:
The method of claim 1 , further comprising: selecting at least one representative topic based on relationships between the plurality of candidate topics grasped using inclusion relationships between the candidate topics and equivalence relationships between the candidate topics. The candidate topics having the similar relationship are:
The method of claim 1 , wherein the candidate topics are grouped by performing matrix factorization (MF) on the keywords and the candidate topics generated by the language model. The step of selecting a representative topic includes:
The method according to claim 3 , further comprising the step of selecting a candidate topic having a certain depth or more in a relationship between the plurality of candidate topics forming a hierarchical structure as a representative topic . The step of selecting a topic of interest includes:
The method according to claim 1 , further comprising: selecting, as the topic of interest, a topic that is related to at least one of the activity keywords among the search term entered by the user and the tag of the product clicked by the user. The step of providing the user with personalized recommendation information comprises:
The method of claim 1 , further comprising: recommending, as a related keyword, at least one keyword related to the user's activity among the keywords related to the topic of interest. The step of providing the user with personalized recommendation information comprises:
The method of claim 1 , comprising: selecting at least one keyword related to the user's activity among the keywords related to the topic of interest as a related keyword; and recommending items on the Internet that are related to the related keyword. The step of recommending an item related to the related keyword includes:
The method of claim 8 , comprising: selecting recommended items utilizing at least one of the user's profile information and the topics of interest. A computer program for causing a computer system to carry out the method according to any one of claims 1 to 9 . 1. A computer system comprising:
at least one processor configured to execute computer readable instructions contained in a memory;
The at least one processor
For a keyword used on the Internet, a prompt consisting of a sample data pair consisting of a keyword sample and a topic of the keyword, and a target keyword is input to a language model to generate a plurality of candidate topics corresponding to the target keyword, candidate topics having a similar relationship are selected from the candidate topics based on the similarity between the candidate topics, and a graph embedding is performed to express the relationship between the plurality of candidate topics using an inclusion relationship between the similar candidate topics and an equivalence relationship between the candidate topics excluding the inclusion relationship, and a relationship is constructed between the target keyword and the plurality of candidate topics using the graph embedding ,
selecting at least one topic of interest corresponding to the user's activity on the Internet using the relationship between the target keyword and the plurality of candidate topics;
providing personalized recommendations to the user based on the topics of interest. The topics of interest that are related to the keywords include:
12. The computer system of claim 11 , associated with the keywords used to search for or categorize information on the Internet, to indicate interests of a user using the keywords. The at least one processor
The computer system according to claim 11, further comprising: selecting at least one representative topic based on the relationships between the plurality of candidate topics grasped using the inclusion relationships between the candidate topics and the equivalence relationships between the candidate topics. The at least one processor
The computer system according to claim 13 , further comprising : selecting a candidate topic at a certain depth or higher as a representative topic in a relationship between the plurality of candidate topics forming a hierarchical structure. The at least one processor
The computer system according to claim 11, further comprising: recommending at least one keyword related to the user's activity among the keywords related to the topic of interest as a related keyword; or recommending an item related to the related keyword among items on the Internet.