KR20230048413A - Community service management method, device and terminal - Google Patents

Abstract

A community service management method, device, and terminal are provided. A community service management method, device, and terminal according to an embodiment determines the activity of a community service based on a message input to a community service accessed by a plurality of terminals, and when the activity of the community service is less than a threshold value, the main topic Based on the above, a recommendation topic different from the main topic of the community service may be recommended by selecting a recommendation topic to be recommended and presented to the community service.

Description

Community service management method, device and terminal

The present invention relates to a community service management method and apparatus capable of managing and activating community services.

With the development of wireless communication technologies such as LTE and 5G and the spread of various smart terminals (e.g. smartphones and tablets), the use of community services using them is increasing. Such a community service provision service provides a community service in which a plurality of users can participate using terminals, and enables users who participate in the community service to exchange messages with each other using terminals.

On the other hand, the recent community service providing service not only provides a community service that allows a plurality of users to exchange messages with each other, but also automatically analyzes messages generated by users participating in the community service and generates messages based on it. It aims to provide additional information to users.

In particular, in the community service, in the process of real-time message exchange between multiple users online, topics discussed by users may change over time, and if there is no topic that users are constantly interested in, it is likely to become inactive over time. In this case, the corresponding community service is highly likely to be a community service that is forgotten by users, that is, a community service whose lifespan is over.

As such, various ideas for managing inactive community services due to low participation of users are being discussed. For example, Korean Patent Publication No. 10-2013-0115156 (title of invention: Cinderella server, Cinderella system, and Cinderella service method) and Korean Patent Publication No. 10-2016-0035197 (title of invention: messenger with automatic room deletion function) Service System) describes how to manage deactivated community services. However, these inventions mainly describe methods of deleting or discarding deactivated community services.

In addition, conventional community services only provide a platform for communication between users, but no means for encouraging users' activities within the community service is provided or implemented. Therefore, in the conventional community service, when users do not communicate in the community service, the number of users participating in the community service gradually decreases, and the community service is eventually closed.

Therefore, a method for inducing users to actively engage in community services is required.

Registered Patent Publication No. 10-1943137 (registered on January 22, 2019)

A technical problem to be solved in the present specification is to manage community services so that community services can be activated.

Another technical problem to be solved in the present specification is to induce active participation of users by activating inactive community services and to prevent community services from being stagnant due to low activity.

Another technical problem to be solved in the present specification is to induce the active participation of the above-mentioned users in consideration of the level of interest of the users.

Another technical problem to be solved in the present specification is to induce users to actively participate in community services, as described above, by taking into account even users' interest in a specific topic that has not been secured in advance.

The technical problems to be solved in this specification are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

A community service management method according to an embodiment includes determining an activity level of a community service from a message input to the community service by a plurality of users, and when the determined activity level is equal to or less than a threshold value, the activity level of the community service from the input message. The method may include extracting a main topic, selecting a recommended topic to be recommended and presented to the community service based on the main topic, and presenting the recommended topic to the community service.

In one embodiment, the step of determining a plurality of candidate topics related to the main topic, the step of collecting topic interests of each of the plurality of users with respect to the plurality of candidate topics, the topic of a first user for a first candidate topic If the interest level is not collected, the first user's topic interest level and the second user's topic interest level for the second candidate topic collected from the community service and/or other community services other than the community service, and the community service and/or or inferring the first user's topic interest in the first candidate topic by inputting the second user's topic interest in the first candidate topic collected from other community services other than the community service into an artificial neural network; and selecting a recommendation topic to be recommended and presented to the community service from among the plurality of candidate topics based on the collected topic interest level and the inferred topic interest level.

According to another embodiment, it includes a memory for storing instructions, and a processor operatively connected to the memory, wherein the processor executes the instructions so that the community service is generated from messages input to the community service by a plurality of users. An activity level is determined, and when the determined activity level is less than or equal to a threshold value, a main topic of the community service is extracted from the input message, a recommendation topic to be recommended and presented to the community service is selected based on the main topic, and the A recommended topic may be presented to the community service.

In another embodiment, the community service management device determines a plurality of candidate topics related to the main topic, collects topic interests of each of the plurality of users with respect to the plurality of candidate topics, and determines the first candidate topic for the first candidate topic. When the level of interest of a topic of one user is not collected, the level of topic interest of the first user and the level of topic interest of the second user for a second candidate topic, and the degree of topic interest of the second user for the first candidate topic are calculated using an artificial neural network. to infer the topic interest of the first user for the first candidate topic, and based on the collected topic interest and the inferred topic interest, recommendation to be presented to the community service among the plurality of candidate topics. You can choose a topic.

According to another embodiment, a communication module for transmitting and receiving data with a community service management server that manages community services.

A display module displaying the community service, a terminal memory storing a program for inputting a message to the community service and displaying the message input to the community service, and a processor executing the program, the processor comprising: the program As it is executed, the activity level of the community service accessed by the terminal is received through the communication module, and when the activity level of the received community service is less than or equal to a threshold value, a recommended topic to be recommended and presented to the community service is transmitted through the communication module. Received and displayed on the display module, the recommended topic may be selected based on a main topic extracted from a message input to the community service.

In another embodiment, the recommendation topic determines a plurality of candidate topics related to the main topic, collects topic interests of each of the plurality of users with respect to the plurality of candidate topics, and provides a first recommendation for a first candidate topic. When the user's topic interest level is not collected, the first user's topic interest level and the second user's topic interest level for the second candidate topic, and the second user's topic interest level for the first candidate topic are transmitted to the artificial neural network. The first user's topic interest level for the first candidate topic may be inferred by inputting the first candidate topic, and the candidate topic may be selected from among the plurality of candidate topics based on the collected topic interest level and the inferred topic interest level.

According to an embodiment, it is possible to manage community services so that inactive community services become more active by inducing activity of community services by users.

According to an embodiment, by suggesting a recommended topic to an inactive community service, it is possible to induce users' active participation in a conversation using the community service, thereby reducing the cost of maintaining the inactive community service.

In addition, according to an embodiment, since topic interests are taken into consideration, active community services can be induced by substantially presenting topics preferred by users to users.

In addition, according to an embodiment, there is an effect of inferring the topic interest of the missing user with only minimal information.

1 is a block diagram showing a schematic structure of a community service management server according to an embodiment.
2 is a flowchart illustrating an operation of recommending a recommended topic to users by a community service management server according to an exemplary embodiment.
3 is a diagram illustrating an example of detecting whether a community service is deactivated by a community service management server according to an exemplary embodiment.
4 to 5 are diagrams illustrating an example in which a community service management server determines a main topic according to embodiments.
6 is a diagram illustrating an example of a related word related to a main topic according to an exemplary embodiment.
7 is a diagram illustrating an example in which a community service management server generates related words related to a main topic according to an exemplary embodiment.
8 to 10 are diagrams illustrating an example in which a community service management server determines a recommended topic according to embodiments.
11 is a diagram illustrating an operation of re-recommending a new topic by a community service management server according to an embodiment.
12 is a block diagram showing a schematic structure of a terminal according to an embodiment.
13 is a diagram illustrating an example of an operation in which a terminal displays a state in which the activity level of a community service is less than or equal to a threshold value according to an embodiment.
14 is a diagram illustrating an example of an operation of displaying a recommended topic of a community service by a terminal according to an embodiment.
15 is a diagram illustrating an example of an operation of displaying a re-recommended topic by a terminal according to an embodiment.
16 is a diagram illustrating an example of an operation of displaying a community service corresponding to a newly recommended topic by a terminal according to an embodiment.
17 is a schematic diagram schematically illustrating an operation of managing a community service of a community service management method according to another embodiment.
18 is a flowchart of a community service management method according to another embodiment.
19 is a diagram for explaining an example of collecting related words related to a main topic using a topic dictionary.
20 is a diagram for explaining an example in which words having a degree of relevance to a main topic of a community service are determined as candidate topics.
21 and 22 are diagrams for explaining an example of collecting topic interests by using responses of users generated with respect to candidate topics in a plurality of community services.
FIG. 23 is a diagram expressing topic interest levels of each of a plurality of users for a plurality of candidate topics of community service using a matrix.
24 and 25 are diagrams for explaining an example of inferring uncollected topic interest using collected topic interest.
FIG. 26 is a diagram illustrating topic interest levels of each of a plurality of users for a plurality of candidate topics of community service using a graph.
27 and 28 are diagrams for explaining a method of inferring uncollected topic interests using a graph.
29 is a diagram schematically illustrating the structure of an artificial neural network for inferring topic interest.
30 is a diagram showing a configuration for learning a method of inferring uncollected topic interests by converting community services into a data structure of a graph and inputting them to an artificial neural network composed of GNNs.
31 is a hardware configuration diagram of a community service management device according to another embodiment.

Advantages and features of the present invention, and methods of achieving them, will become clear with reference to the detailed description of the following embodiments taken in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only these embodiments make the disclosure of the present invention complete, and common knowledge in the art to which the present invention belongs. It is provided to completely inform the person who has the scope of the invention, and the present invention is only defined by the scope of the claims.

In describing the embodiments of the present invention, if it is determined that a detailed description of a known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description will be omitted. In addition, terms to be described later are terms defined in consideration of functions in the embodiment of the present invention, which may vary according to the intention or custom of a user or operator. Therefore, the definition should be made based on the contents throughout this specification. Hereinafter, some embodiments of the present invention will be described with reference to the drawings.

In this specification, community means a space where users participate and share information and have conversations through text, video, voice, image, etc. . In addition, the community may be a platform for sharing information by inputting comments while providing a site supporting multinational and multi-institutional clinical research information and a web-based statistical program.

A community service is a service that provides conversation, communication, message transmission, replies or comments, or post writing among users to users accessing a community such as a chatting platform, discussion platform, meeting platform, portal site, or SNS. For example, community services may include chat rooms, one-to-one messages, group chat rooms, blogs, bulletin board posts, replies, or comments. The community service is not limited to the above types of services, and refers to various services in which users can share opinions or have conversations by inputting text or multi-media contents.

Users can access a community service using a user terminal, and can share information or have a conversation with other users through the community service.

Also, "community service" may refer to an application service in which messages are inputted and exchanged between users accessing the Internet or an online service, such as a conversation between a plurality of users who meet in person. Community services can be divided into general community services, secret community services, and open community services according to their characteristics. General community service refers to an application service in which users send and receive messages with mutually registered users, while open community service refers to an application service in which anonymous users send and receive messages. It means an application service that is encrypted and protected.

The term "community service" used herein is not limited to its type, but may preferably be a general community service or an open community service. In particular, community service herein refers to medical community service. The medical community service refers to a community service in which a user is a medical expert (doctor, nurse, etc.). In the case of such a medical open community service, the community service management server 100 according to the present embodiment may further include a user identification unit for verifying whether a user is a medical expert. The user identification unit may perform a process of verifying a user's medical professional's license before participating in the community service, and granting permission to participate in the community service if the license is confirmed. Since the user identification unit is not a major component of the community service management server according to the present embodiment, a detailed description thereof will be omitted.

In this specification, a topic refers to a topic of conversation that users are discussing within a community service. One topic or a plurality of topics can be extracted within one community service. The main topic means a topic determined by a predetermined criterion among topics. Here, the main topic and predetermined criteria will be described later.

The topic may be a text summarizing messages or conversations entered by users participating in the service. For example, when a plurality of conversations exist in a community service, a topic means a text composed of words, phrases, or sentences that can represent a plurality of conversations. 1 is a block diagram showing a schematic structure of a community service management server 100 according to an embodiment.

Referring to FIG. 1 , a community service management server 100 may include a communication module 110 , a memory 120 and a processor 130 .

The communication module 110 may transmit and receive data with a plurality of terminals. Specifically, the communication module 110 can transmit/receive data based on various terminals (e.g. smart phone, tablet, PC, smart TV, wearable computer) and various communication methods (e.g. LTE, 5G, WLAN, Bluetooth). . A user who wants to participate in the community service can participate in the community service managed by the community service management server 100 through the aforementioned terminal.

The memory 120 may store a program (e.g. application service) managing a community service accessed by a plurality of terminals and information about the corresponding community service. Memory 120 may be non-volatile memory (e.g. HDD, SSD), volatile memory (e.g. SRAM, DRAM, SDRAM), or a combination of non-volatile and volatile memory.

At this time, the program that manages the community service may be a program that manages interactive messengers available on mobile/PC, such as KakaoTalk, or Facebook, Twitter, LinkedIn, and Instagram. It may be a program that manages social media/social network services that can provide message services based on posts/comments, such as Instagram and Snapchat. The program managing the community service may support a plurality of terminals accessing the community service to communicate while exchanging messages with each other.

In addition, the program managing the community service may support a service through which a terminal accessing the community service can perform real-time audio and video communication through the community service.

The processor 130 may execute a program stored in the memory 120 . Specifically, as the program is executed, the processor 130 may determine the activity level of the community service based on messages input by a plurality of terminals. The message input to the community service may be a message input by users participating in the community service to communicate with each other within services such as the aforementioned interactive messenger and social media.

According to this embodiment, active message input of users participating in the community service can be induced. When the processor 130 recommends a recommended topic to the community service, it means outputting the recommended topic to the community service running on the user's terminal or providing it to the user's terminal through 'text message' or 'e-mail' as needed. do.

2 is a flowchart illustrating an operation in which the community service management server 100 recommends a recommendation topic to users according to an exemplary embodiment.

Referring to FIG. 2 , the processor 130 of the community service management server 100 may determine the activity level of the community service from a message input to the community service (S210).

The processor 130 may determine whether the activity of the community service is less than or equal to a threshold value based on the message input to the community service (S220). If it is determined that the activity of the community service is less than the threshold value (S220-Y), the processor 130 extracts the main topic of the community service as described above in FIG. 1 (S230), and based on the main topic, the community service A recommended topic to be suggested is selected as a recommended topic (S240), and the recommended topic can be presented to the community service (S250). On the other hand, if it is determined that the activity of the community service is not less than the threshold (S220-N), the processor 130 may determine whether the activity of the community service is less than or equal to the threshold based on the input message (S220).

Hereinafter, an example in which the community service management server 100 determines whether the activity level of a community service is equal to or less than a threshold value in step S220 will be described.

3 is a diagram illustrating an example in which the community service management server 100 determines the activity level of a community service according to an exemplary embodiment.

The activity level of the community service is a measure of how actively a message is input in the community service. The activity of the community service may be measured based on the amount of messages transmitted within the community service. For example, the activity of the community service may be calculated based on the number of messages input to the community service during unit time, the number of texts included in the message, or the number of user accounts inputting the message.

The processor 130 may determine that the community service is not activated when the activity level of the community service is equal to or less than the threshold value, and determine that the community service is activated when the activity level exceeds the threshold value.

The processor 130 may recommend a recommendation topic different from the main topic of the community service to the community service when the activity of the community service is equal to or less than the threshold value.

For example, the processor 130 determines that the activity of the community service is less than or equal to the threshold value when the number of messages input to the community service during unit time is less than or equal to the set threshold number of messages and the number of terminals that input messages during unit time is less than or equal to the threshold number of users. It can be judged that

In this case, the unit time may be a fixed value or a value that is changed according to a specific criterion. For example, the unit time may change according to the creation time of the community service, the number of users, and the total number of messages input to the community service. If the community service has been created for a long time, since the activity of the corresponding community service is likely to be below the threshold, the size of the unit time can be reduced in order to quickly recommend a topic. Similarly, if the number of users is below a certain threshold or the total number of messages input to the community service is below a certain threshold, the community service is likely to be in an inactive state. can decrease

Hereinafter, a specific operation of determining a recommended topic by the community service management server will be described. First, an operation in which the community service management server determines the main topic of the community service will be described.

For example, the community service management server 100 may determine a main topic when a community service is created. At this time, the user who created the community service may determine and input the main topic of the community service.

As another example, the community service management server 100 may dynamically determine the main topic of the community service based on a message input to the community service. For example, the processor 130 of the community service management server 100 determines at least one of the number of input keywords and the number of users who input keywords, among one or more keywords extracted from messages input to the community service through a plurality of terminals. Based on this, the main topic can be determined. At this time, messages input to the community service may be messages input during a set reference time (eg 10 minutes). Meanwhile, the reference time may be different from the aforementioned unit time.

At this time, among the words included in the message, words that do not have independent meanings such as endings, postpositions, and conjunctions may be excluded, and only body words may be selected as keywords. On the other hand, common words (e.g. greetings) or words referring to users may be excluded from keywords.

Keywords may be extracted from the message by a syntax analyzer, morpheme analyzer, text mining engine, or other keyword extraction algorithm. For example, a syntax analyzer (e.g. Backus-Naur Form) may generate a syntax analysis tree from a message based on a preset grammar and extract keywords therefrom. As another example, the morpheme analyzer can extract keywords from the body words excluding endings, postpositions, and connecting words after identifying the linguistic structure of the message. As another example, a text mining engine can extract keywords by inputting a message to a pretrained machine learning/artificial neural network model.

In the present embodiments, the community service management server 100 determines at least one of the input frequency of keywords and the number of users who input keywords, among one or more keywords extracted from messages input to community services through a plurality of terminals. Based on this, the main topic can be determined.

4 to 5 are diagrams illustrating an example in which the community service management server 100 determines a main topic according to embodiments.

Referring to FIG. 4 , the processor 130 of the community service management server 100 may determine a main topic based on keywords for which the number of inputs to the community service is greater than or equal to a set threshold number of keywords. That is, keywords that are frequently entered into the community service may affect the determination of the main topic.

4, for the keywords A, B, C, D, E, and F, the number of inputs of A is 10, the number of inputs of B is 8, the number of inputs of C is 15, the number of inputs of D is 20, and the number of inputs of E is 8 is 7, and the number of inputs of F is 4. In this case, when the threshold number is set to 8, the processor 130 may determine the main topic from keywords A, B, C, and D for which the number of inputs is 8 or more.

Meanwhile, the processor 130 may determine a main topic based on a keyword input by the most users among keywords for which the number of inputs to the community service is equal to or greater than a set threshold number. This is because keywords that are evenly input by a large number of users are more likely to be of interest to more users than keywords that are intensively input by a small number of users.

Referring to FIG. 5 , for keywords A, B, C, and D having an input frequency of 8 or more, the number of users who input each keyword is 3 for A, 5 for B, 4 for C, and 2 for D. At this time, the processor 130 may determine the keyword B input by the most users as the main topic.

In the above, the operation of the community service management server 100 to determine the main topic has been described.

Hereinafter, an operation in which the community service management server 100 determines a recommended topic for a main topic will be described.

For example, the community service management server 100 may determine a recommended topic based on related words related to the main topic. The processor 130 of the community service management server 100 may determine a recommended topic based on any one of the related words, a combination of two or more of the related words, or a sentence using any one or more of the related words.

Hereinafter, related words will be described in FIGS. 6 to 8 . 6 is a diagram illustrating an example of a related word related to a main topic according to an exemplary embodiment.

Referring to FIG. 6, related words related to the main topic T are related words (S1, S2, S3, ...) to the main topic T, analogous words (T1, T2, T3, ??) to the main topic T, or main It can contain subwords (A, B, B1, B2, C, C1, C2, ...) for topic T. A related word for the main topic may be a set of keywords determined to be related to the main topic. In this case, whether a keyword is related to the main topic may be determined through whether it is registered as a related word in external dictionary information set as an example. For example, when the main topic is 'animal', if the keywords 'farm' and 'insect' are registered as related words of 'animal' in an external dictionary, 'farm' and 'insect' may be included in the related words.

A synonym for the main topic may be a set of keywords determined to be similar to the content of the main topic. At this time, whether a keyword is similar to the main topic may be determined through whether it is registered as a similar word in external dictionary information set as an example. For example, when the main topic is 'animal', if the keywords 'beast' and 'livestock' are registered as synonyms of 'animal' in an external dictionary, 'beast' and 'livestock' may be included in the synonyms.

Subwords for the main topic may be a set of keywords determined to be sub-concepts for the main topic. At this time, whether a keyword is a sub-concept of the main topic is registered as a sub-word in external dictionary information set as an example, or whether the keyword is stored as a sub-language of the main topic in a database indicating an upper-level relationship between keywords. can be determined through For example, when the main topic is 'animal', if information that 'animal' is a superordinate concept of 'dog', 'horse', and 'cow' is stored in the database, 'dog', 'horse', and 'cow' are stored in the database. ' may be included in the subword.

Meanwhile, the keyword included in the sub-word may be a direct sub-concept of the main topic or an indirect sub-concept passing through a plurality of layers. For example, when the main topic is 'animal', since 'bulldog' is a sub-concept of 'dog' and 'dog' is a sub-concept of 'animal', 'bulldog' may be included in the sub-word for the main topic.

Hereinafter, an example of generating the aforementioned related words will be described.

7 is a diagram illustrating an example in which the community service management server 100 generates related words related to a main topic according to an exemplary embodiment.

The processor 130 of the community service management server 100 may generate one or more related words to the main topic by performing web crawling on the main topic.

Specifically, the processor 130 may search for a web page corresponding to the main topic on the web and analyze the content of the searched web page to generate words related to the main topic.

In addition, the processor 130 searches for a main topic in a search engine (e.g. Google, Yahoo, Naver, Daum), extracts a related search word and similar search word sub-search words found for the main topic, and generates any one or more of the related words. can do.

Also, the processor 130 may search a preset database instead of web crawling to generate related words for the main topic. The database may store a history of previous searches for the main topic, and based on this history, related words for the main topic may be created.

Hereinafter, a specific method for the community service management server 100 to determine a recommended topic will be described.

8 to 10 are diagrams illustrating an example in which the community service management server 100 determines a recommended topic according to embodiments.

Referring to FIG. 8 , the processor 130 of the community service management server 100 selects a keyword from among one or more keywords extracted from messages input to the community service, the number of inputs to the community service or the largest number of input users. Based on this, recommended topics can be selected.

In FIG. 8, the number of inputs of the keywords K1, K2, K3, K4, K5, and K6 is 10, 8, 15, 20, 7, and 4, respectively, and the number of users who input the corresponding topic is 8, 2, 10, and 9, respectively. , 1, 3. In this case, the processor 130 may select a recommended topic based on the topic K4 with the greatest number of inputs or the topic K3 input by the most users among the keywords K1, K2, K3, K4, K5, and K6.

Referring to FIG. 9 , the processor 130 of the community service management server 100 may select a recommended topic based on a related word having the highest correlation with the main topic among related words. For example, the recommended topic may include a phrase including a related word having the highest degree of association with the main topic or a related word having the highest degree of association with the main topic.

Meanwhile, the degree of association between a certain related word and the main topic may be determined in various ways.

For example, the degree of association between a certain related word and the main topic may be determined based on the number of messages corresponding to both the related word and the main topic among a plurality of messages included in the community service. The higher the ratio of messages corresponding to both the related word and the main topic among all messages, the higher the value of the relevance.

For example, when the number of messages corresponding to a certain related word R is 30 and the number of messages corresponding to both the related word R and the main topic is 15, the degree of association between the related word R and the main topic can be calculated as 15/30 = 0.5. .

As another example, the degree of association between a certain related word and the main topic may be determined based on information about the degree of association stored in an internal or external database of the community service management server 100 .

At this time, the relevance information may be stored in the form of various data structures (e.g. tables, trees, graphs).

In FIG. 9 , the degree of association of the related words R1, R2, R3, R4, R5, and R6 with the main topic is 0.8, 0.6, 0.4, 0.5, 0.9, and 0.1, respectively. In this case, the processor 130 may select a recommended topic based on related word R5 having the highest degree of association with the main topic among related words R1, R2, R3, R4, R5, and R6.

As another example, the degree of association between a certain related word and the main topic may be determined based on a weight given according to the relationship between the related word and the main topic (e.g. technology field, occurrence of domestic and international issues, relationship with user propensity).

Meanwhile, when selecting a recommended topic based on the aforementioned related word or keyword, the related word or keyword itself may be presented as a recommended topic, but the recommended topic may be presented as a combination of related words or keywords or a sentence using them.

For example, if the main topic is 'Coronavirus' and the related word is 'Corona vaccine', the recommended topic could be created as 'Corona vaccine', but also 'How to develop a vaccine for Corona virus' or 'Corona virus vaccine development companies' '.

Referring to FIG. 10 , the processor 130 of the community service management server 100 may select a recommended topic based on propensity information of a plurality of users participating in the community service.

In FIG. 10, it is assumed that N users (user 1, user 2, user 3, ??, user N) participate in the community service. At this time, the processor 130 may select a recommended topic based on the propensity information of each of the N users (User 1: CCC / User 2: DDD / User 3: EEE / ?? / User N: ZZZ).

In this case, the propensity information may selectively include personal information (e.g. age, gender, occupation, region).

Meanwhile, the propensity information may include information such as portal search history, previously participated community service information, and areas of interest in addition to personal information. At this time, the personal information is information directly input by the user, and the field of interest, portal search history, and community service information previously participated may be results generated by the community service management server 100 tracking the user's previous activities. The processor 130 of the community service management server 100 may select a recommended topic based on at least one of personal information, portal search history, previously participated community service information, and areas of interest.

When the community service management server 100 recommends a recommended topic to users in the above-described method, the users may accept the recommended topic, but the users may reject the recommended topic and request re-recommendation of a new topic. there is.

11 is a diagram illustrating an operation of re-recommending a new topic by the community service management server 100 according to an embodiment.

In FIG. 11 , a user participating in the community service may transmit a message requesting re-recommendation of a topic different from the corresponding recommended topic to the recommended topic recommended by the community service management server 100 through a terminal.

When the processor 130 of the community service management server 100 receives a message requesting re-recommendation of a different topic from the recommended topic through the communication module 110, a topic different from the previously recommended recommended topic is sent to the community service. can be re-recommended.

Meanwhile, the processor 130 of the community service management server 100 may create a new community service corresponding to the recommended topic and allow one or more terminals among a plurality of terminals accessing the community service to participate in the newly created new community service. there is. In this case, the user participating in the new community service may be, for example, a user who has input a message corresponding to the recommended topic one or more times.

After a new community service corresponding to the recommended topic is created, the community service management server 100 may maintain or delete the existing community service.

The community service management server 100 described above summarizes the contents of the messages input during the conversation process when a conversation is made on the main topic and recommended topic (or re-recommended topic) in the existing community service or new community service. can do. In addition, the community service management server 100 may allow users to search for a history of an input message or content summary.

12 is a block diagram showing a schematic structure of a terminal 200 according to an embodiment.

First, the terminal 200 may include a display module 210 that displays the accessed community service. The display module 210 may display various types of information related to community services (e.g. names of community services, participating users, messages). At this time, the information of the community service that can be displayed on the display module 210 may vary according to the resolution of the display module 210 or the setting of a program displaying a message input to the community service.

Also, the terminal 200 may include a terminal communication module 220 that transmits/receives data to/from the community service management server 100 that manages community services. That is, the terminal 200 may transmit data to the community service management server 100 through the terminal communication module 220 and receive data from the community service management server 100 . At this time, the terminal communication module 220 may transmit and receive data to and from the community service management server 100 based on various communication methods such as LTE, 5G, WLAN, and Bluetooth.

Also, the terminal 200 may include a terminal memory 230 storing a program for inputting a message to the community service and displaying the message input to the community service. The terminal memory 230 may be, for example, a hard disk, a non-volatile memory (e.g. ROM, NAND flash memory), or a volatile memory (e.g. SRAM, DRAM).

The terminal 200 may include a terminal processor 240 that executes a program for inputting a message to the community service and displaying the message input to the community service. The terminal processor 240 may be, for example, a CPU installed in a PC or a System On Chip (SoC) installed in a mobile device.

As a program for inputting a message to the community service and displaying the message input to the community service is executed, the terminal processor 240 first detects whether the activity of the community service accessed by the terminal 200 is equal to or less than a threshold value. It can be received through the terminal communication module 220.

Further, the terminal processor 240 may display on the display module 210 after receiving information on a recommendation topic different from the main topic of the community service through the terminal communication module 220 when the activity of the community service is equal to or less than the threshold value. .

As described above with reference to FIGS. 4 and 5 , the main topic may be determined based on, for example, the number of input times of keywords and the number of users who input keywords among one or more keywords extracted from messages input to community services.

As described above with reference to FIGS. 8 to 10 , a recommended topic is 1) selected as a keyword with the largest input frequency or the largest number of input users among one or more keywords extracted from messages input to the community service; 2) Among related words (e.g. synonyms, subwords, related words) for the main topic, a related word with the highest degree of relevance to the main topic is selected, or 3) based on the propensity information of a plurality of users participating in the community service can be selected.

13 is a diagram illustrating an example of an operation of displaying a state in which the activity level of a community service is less than or equal to a threshold value by the terminal 200 according to an embodiment.

Referring to FIG. 13 , the terminal 200 may display community services on the display module 210 . At this time, information indicating a user participating in the community service (e.g. user's name or nickname) may be displayed on the display module 210 . In addition, messages input to the community service by users participating in the community service may also be displayed on the display module 210 .

Meanwhile, the terminal 200 may display on the display module 210 information indicating that the activity level of the community service is less than or equal to a threshold value. Through this, the user using the terminal 200 can know that the activity level of the current community service is below the threshold value.

14 is a diagram illustrating an example of an operation of displaying a recommended topic of a community service by the terminal 200 according to an embodiment.

Referring to FIG. 14, the terminal 200 displays information TTT' of a recommended topic different from the main topic of the community service on the display module 210, and displays a window asking whether to participate in a new community service corresponding to the new recommended topic. can be displayed

If a user using the terminal 200 rejects a recommended topic and requests re-recommendation of a new topic, the terminal 200 may display the re-recommended topic as shown in FIG. 15 .

On the other hand, when a user using the terminal 200 requests to participate in a community service corresponding to a recommended topic, the terminal 200 may display a community service corresponding to the newly recommended topic as shown in FIG. 16 .

15 is a diagram illustrating an example of an operation of displaying a re-recommended topic by the terminal 200 according to an embodiment.

Referring to FIG. 15, the terminal 200 displays a new re-recommended topic TTT" different from the recommended topic TTT' recommended in FIG. 14 on the display module 210, and participates in a new community service corresponding to the re-recommended topic. A window may be displayed asking whether

16 is a diagram illustrating an example of an operation of displaying a community service corresponding to a newly recommended topic by the terminal 200 according to one embodiment.

Referring to FIG. 16 , the terminal 200 may display on the display module 210 whether to participate in a new community service corresponding to the recommended topic TTT' recommended in FIG. 14 .

So far, a community service management method and apparatus according to an embodiment have been described. Hereinafter, a community service management method and device according to another embodiment will be described.

17 is a schematic diagram schematically illustrating an operation of managing a community service of a community service management method according to another embodiment.

Referring to FIG. 17 , the community service management method may provide community services to users using a community service management device 1700 . Also, the community service management device 1700 may manage community services so that community services can be activated.

The community service management device 1700 may extract a plurality of topics by clustering the messages based on the meanings of the messages input from the plurality of users in the community service. Specifically, when text is input from the community service by users, the community service management device 1700 processes the input text in natural language to perform semantic analysis, and words and phrases having similar meanings based on the meaning. Alternatively, sentences may be clustered and grouped into a plurality of groups. At this time, the topic becomes a word, phrase, or sentence with a similar meaning indicating each of a plurality of clustered groups.

As described above, the community service management device 1700 may extract topics by clustering conversations based on meanings from conversations input by users. These topics can be textual information summarized in sentences, phrases or words.

A topic could be a word like 'severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)', a phrase like 'how to treat severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)', or 'severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)'. It could also be a sentence like 'When will the respiratory syndrome coronavirus 2 (SARS-CoV-2) end?' On the other hand, a topic has nothing to do with the type of language. For example, a topic may be composed of Korean, English, Japanese, or Chinese, or may be composed of a combination of the above languages.

The community service management device 1700 may extract a main topic from a plurality of classified topics according to a predetermined criterion as follows. The predetermined criterion may be the number of times a community service is mentioned during an arbitrarily determined time period. The community service management device 1700 may determine the main topic based on the number of times it is mentioned during a predetermined time period among topics from immediately after the community service starts to the present analysis process. Here, the predetermined time may be arbitrarily determined. For example, the predetermined time may be a predetermined time such as 1 hour after the community service is created or 30 minutes after a conversation starts in the community service, but is not limited thereto.

In addition, determining the main topic based on the number of mentions means that the topic with the largest number of times is determined as the main topic, or the top several topics or the top one topic according to the order when the number of times is arranged in descending or ascending order This may mean that it is determined as the main topic. However, it is not limited thereto.

As described above, the community service management apparatus 1700 may extract a topic using the conversations input by the users u1 to u4 in the community service, and comprehensively consider the extracted topics and select among the topics. A main topic that best meets a predetermined criterion can be extracted. The predetermined criterion is not limited to the above-described embodiment, and may mean various criteria capable of extracting a main topic from among topics.

The community service management device 1700 may extract a plurality of related words related to the main topic using words included in the main topic, and determine a plurality of candidate topics related to the main topic by combining the plurality of related words. A related word means a word related to a word included in the main topic. A detailed method of extracting related words will be described later. The candidate topic may be text information composed of a word, phrase, or sentence generated by a combination of at least one related word.

Words included in the topic may be extracted from the message by a syntax analyzer, a morpheme analyzer, a text mining engine, or other word extraction algorithm. For example, a syntax analyzer (e.g., Backus-Naur Form) may generate a syntax analysis tree from a message based on a preset grammar and extract words from it. As another example, the morpheme analyzer identifies the linguistic structure of the message and then extracts words from the body words excluding endings, postpositions, and conjunctions. Words, as another example, can be extracted by a text mining engine by inputting a message into a pretrained machine learning/artificial neural network model.

The community service management device 1700 may determine at least one recommended topic from among a plurality of candidate topics and present it to the community service. The community service management device 1700 may present a recommended topic in the form of text or image. At this time, the community service management device 1700 searches for a recommended topic through a search engine, etc., and when a paper, news article, or patent for the recommended topic is found through a search engine, etc., a link to or related to the paper, news article, or patent is found. You can submit the file to community services.

To determine the recommendation topic, the topic interest level of each of the users u1 to u4 for each candidate topic may be used. To this end, the community service management device 1700 may collect the above-mentioned topic interest. Here, the topic interest level means a preference indicating how interested the users (u1 to u4) are in the corresponding topic, and may be a vector value expressed as a number. The degree of interest in a topic is not limited to the described terms and may be substituted with various words indicating a value of preferring a topic. For example, topic interest may be replaced with topic preference, user preference, or user concentration.

The community service management device 1700 may collect topic interests by using responses generated by users u1 to u4 in community services in which candidate topics are discussed or other community services. In an embodiment, the community service management device 1700 uses a plurality of user responses generated in advance with respect to the plurality of candidate topics in a community service or another community service other than the community service to search for the plurality of candidate topics. It is possible to calculate the degree of topic interest of each of the plurality of users.

Response is data for calculating the degree of interest in a topic, and refers to all input actions input by the user. For example, the reaction refers to text, emoticon, data file transfer, multimedia file transfer, touch, drag, swipe, or moving operation of the user terminal that the user inputs by accessing the community service, but is not limited thereto, and the user It may refer to various input actions input through a user terminal.

Here, collection is an act of securing a level of interest in a topic from a user's response. The level of interest in a topic may be collected in a manner calculated using a user's reaction. At this time, the topic interest level may be calculated according to various embodiments using the user's response and is not limited to any one implementation method. In various embodiments, the degree of interest in a topic may be calculated as 'traffic generated by the user's reaction * weight according to the degree of association of the meaning of the reaction'.

The degree of interest in a topic may be calculated using traffic based on a user's response, speed at which a user's response to a topic is input, analysis of the meaning of a user's response, and the like. A detailed example of collecting topic interests using user responses will be described later.

In this way, the community service management device 1700 may collect topic interest for each candidate topic by using the user's response. In this case, while the topic interest level of some users for a specific candidate topic is collected, the topic interest level of other users for the same candidate topic may not be collected.

This is because the above-mentioned 'other users' among all users may not have mentioned the specific topic even once due to various factors. At this time, the community service management device 1700 may infer the level of interest in the corresponding candidate topic of 'other users' who have never mentioned the specific topic described above. The type of information used for reasoning and the reasoning method will be described in detail later.

The community service management device 1700 may comprehensively consider the inferred topic interest level and the collected topic interest level, determine a recommended topic within the community service, and recommend and present the determined recommended topic to the community service. A detailed method of determining the recommended topic will be described later.

Even if an additional topic is presented in the community service, it is not guaranteed that the added topic is a user-preferred topic. Accordingly, community services may not be activated in some cases because the added topic cannot guarantee to actively induce participation of users.

The community service management device 1700 according to the present embodiment presents topics preferred by the users u1 to u4 to the community service. Therefore, active participation of users can be induced. Also, the community service management device 1700 presents a recommended topic related to the main topic to the community service. Therefore, community service can be continued in a direction related to the main topic being discussed.

Recommendation systems such as collaborative filtering were difficult to use for activating community services. The service management apparatus 1700 according to the present embodiment may recommend a topic in which users are most interested in in order to activate a community service.

In addition, according to the present embodiment, the recommendation system can be applied and used to community services, but the recommendation system can be operated more smoothly because the range of reasoning is minimized while inferring uncollected data.

Hereinafter, a community service management method according to another embodiment will be described with reference to FIGS. 18 to 31 . This embodiment may be performed by a computing device. For example, the computing device may be the community service management device described with reference to FIG. 1 . In describing the present embodiment, description of a performer of some operations may be omitted. At this time, the performing entity is the computing device. This computing device will be described in more detail with reference to FIG. 31 .

18 is a flowchart of a community service management method according to another embodiment.

As shown in FIG. 18, the main topic of the community service may be extracted in step S1810. As described above, the main topic may be extracted based on messages input to the community service by a plurality of users, but is not limited thereto.

In step S1820, a plurality of candidate topics related to the main topic may be determined. Hereinafter, an example in which a plurality of candidate topics are determined will be described with reference to FIGS. 19 and 20 .

19 is a diagram for explaining an example of collecting related words related to a main topic using a topic dictionary.

A related word is a word related to a word included in the main topic. For example, a related word means a word having a meaning similar to that of the main topic among words included in the main topic. That is, related words are words related to words included in the main topic. When there are a plurality of related words related to the main topic, there may be a plurality of related words.

For example, if the main topic is 'similarity and correlation between Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) and Severe Acute Respiratory Syndrome (SARS)', the related word is 'Severe Acute Respiratory Syndrome Coronavirus 2 ( SARS-CoV-2)', 'Severe Acute Respiratory Syndrome (SARS)', 'similarity', 'correlation', 'relationship', 'virus', 'SARS', etc.

Also, related words do not distinguish between types of language. For example, a word composed of Korean and a word composed of English may be related to each other. A more detailed example of related words is described below, for example. Related words may be determined according to various embodiments. The various embodiments below are illustrative and not limited thereto.

In one embodiment, related words may be determined using a pre-constructed topic corpus as shown in FIG. 19 . The topic dictionary may be a database in which natural language is stored. At this time, the degree of association and relationship between words may be stored in the form of various data structures (e.g., tables, trees, graphs).

Words stored in the topic dictionary may be mapped and stored with a degree of association between the respective words. In this embodiment, the degree of association may be a pre-calculated value based on meanings between words stored in the topic dictionary. For example, as shown in FIG. 19 , word 1 and word 2 are mapped with an association degree of 0.7 and stored in the topic dictionary, word 1 and word 3 are mapped with an association degree of 0.8 and stored in the topic dictionary, and word 1 and word 3 are mapped with an association degree of 0.8 and stored in the topic dictionary. 4 may be stored with an association degree mapped to 0.4, and word 1 and word 5 may be mapped with an association degree 0.2 and stored.

When a new word is added to the topic dictionary, the topic dictionary may add the new word by mapping the newly added word with a degree of association calculated for each of the words previously stored in the topic dictionary. An entity operating the topic dictionary may be a community service providing device or an external device, but the operating entity of the database is not limited thereto. After words included in the main topic are searched for in the topic dictionary, words included in a pre-specified relevance range with the searched words may be determined as related words.

In one embodiment, related words may be determined using a clustering algorithm. For example, each word may be represented as a vector in a Euclidean space based on the meaning of the word, and clustered based on a distance in the Euclidean space. As a result of clustering through the clustering, words located in the same group may be related to each other because they are located adjacent to each other based on meaning. In this case, the clustering algorithm may be any one of a K-Means algorithm, a DBSCAN algorithm, or an H-Clustering algorithm, and may be an algorithm in which the clustering algorithm is combined, but is not limited thereto.

Also, in an embodiment, a related word may be a word having a similar meaning to a word included in the main topic, or a word included in the same or similar category, although not similar.

For example, if the main topic is 'Infection with Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)', the related word is a word similar to 'Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)'. It could be 'Severe Acute Respiratory Syndrome Coronavirus 1 (SARS-CoV-1)'.

In addition, the related word is not a word similar to 'Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)', but is a word of the same or similar category, and is a word of 'Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)'. It may be a word included in the same category as 'disease', such as 'medicine' or 'vaccine'.

Also, in one embodiment, a related word may be a synonym or a subword. For example, the topic dictionary may be a database in which structures of synonyms and subwords of stored words are pre-constructed. In this way, the related word may be a synonym or a subword, but is not limited to such a grammatical name, and may mean various words related to the main topic or derived from the main topic.

For example, when the main topic is 'Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)', the topic dictionary is 'Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)' as a synonym for 'CoV'. If the words of 'Bid 19', 'Corona 19', and 'New Corona' are constructed, 'Covid 19', 'Corona 19', and 'New Corona' can be synonyms, and 'Covid 19', 'Corona 19' ' and 'new corona' are synonyms that can eventually become related.

In addition, when the main topic is 'Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)', 'Structure of SARS-CoV-2', 'Treatment of SARS-CoV-2', 'SARS-CoV-2' If the information 'infectivity of SARS-CoV-2' is stored as a subword in the topic dictionary, 'Structure of SARS-CoV-2', 'treatment of SARS-CoV-2', and 'infectiousness of SARS-CoV-2' can be subwords. , the subword can be a related word.

Meanwhile, words included in the subwords may be words of a direct subconcept of the main topic, or may be words of an indirect subconcept passing through a plurality of layers. For example, when the main topic is 'Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)', 'Structure of SARS-CoV-2' is a direct sub-concept and 'Structure of SARS-CoV-2's spike protein'. ' is an indirect sub-concept, so the word 'structure of the spike protein of SARS-CoV-2' can also be a sub-word.

In addition, related words may be extracted from a database such as a topic dictionary or may be extracted through web crawling.

20 is a diagram for explaining an example in which related words having a degree of relevance to a main topic of a community service are determined as candidate topics. Related words, synonyms, and subwords of the main topic are words to which a degree of relevance is assigned to words included in the main topic, and may become candidate topics. That is, similar words and subwords described as examples of related words and related words may become candidate topics. Candidate topics are texts consisting of words, phrases or sentences. The candidate topic may be a related word itself, or may be a phrase or sentence in which a postposition or suffix is added to the related word.

For example, if the related term is 'Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)', the candidate topic is 'Can accurately and quickly diagnose Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)'. It can be a sentence like 'How is it?' That is, candidate topics may be generated using words included in the main topic, words similar to words included in the main topic, or words related to words included in the main topic, and candidate topics may be words, phrases, or words. It can be a sentence. Also, when there are a plurality of related words, the candidate topic may be a new phrase or sentence created by combining a plurality of related words. The above-described examples of candidate topics are examples, and are not limited thereto.

As shown in FIG. 20 , related words 1 to 3 may be determined from the main topic. In this embodiment, it is assumed that related words are determined as words included in the main topic and a degree of association within 1. Related word 1 is a word with a degree of relevance of 0.8 to a word included in the main topic, related word 2 is a word with a degree of relevance of 0.4 to a word included in the main topic, and related word 3 is a word with a degree of relevance of 0.7 to a word included in the main topic. is a word that is Related word 1 may be a first candidate topic, related word 2 may be a second candidate topic, and related word 3 may be a third candidate topic.

For example, if the main topic is 'Infection with Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2)', words with a degree of relevance within a predetermined criterion are 'Severe Acute Respiratory Syndrome Coronavirus 1 (SARS-CoV-1)'. )', 'therapeutic agent', and 'vaccine'. Related word 1 may be 'severe acute respiratory syndrome coronavirus 1 (SARS-CoV-1)', related word 2 may be 'therapeutic agent', and related word 3 may be 'vaccine'. In one embodiment, the first candidate topic is 'Severe Acute Respiratory Syndrome Coronavirus 1 (SARS-CoV-1)' and the second candidate topic is 'Treatment for Severe Acute Respiratory Syndrome Coronavirus 1 (SARS-CoV-1)'. , The third candidate topic may be determined as 'vaccination rate of severe acute respiratory syndrome coronavirus 1 (SARS-CoV-1)'.

The process of determining a plurality of candidate topics related to the main topic in step S1820 has been described with reference to FIGS. 19 and 20 so far. Hereinafter, with reference to FIGS. 21 and 22, a specific example of collecting topic interest using users' responses to candidate topics will be described.

In step S1830, topic interests of each of a plurality of users for a plurality of candidate topics may be collected. As shown in FIG. 21, the degree of interest in a topic may be collected by using a response generated from each user with respect to a candidate topic in a community service. For example, the user's topic interest in the first candidate topic may be collected by comprehensively considering quantitative evaluation and qualitative evaluation of a response input by the user in community service 1.

For example, as a quantitative evaluation of the user's reaction, when the amount of information on the reaction generated by the user with respect to the first candidate topic is large or the reaction is quickly generated, the user's interest in the topic may be calculated as a high value.

For example, when traffic generated by a user during unit time is large, the user's interest in a topic may be determined as a relatively high value in proportion to the amount of traffic. When community service 1 is a community service, responses such as text, media files, emoticons, and likes input by users participating in the community service to the first candidate topic are displayed at the time when the community service discusses the first candidate topic. Based on this, topic interest may be determined.

As the amount of traffic due to the user's response to the first candidate topic increases, the degree of interest in the topic may be determined as a relatively high value in proportion to this. In an embodiment, as the time at which the user's response to the first candidate topic is input is faster, the user's interest in the topic may be determined as a relatively higher value in proportion to the time at which the response is input.

If community service 1 is a service provided in content such as book reviews, papers, SNS, postings, or blogs, the user's response may be a comment, retweet, or reply. In this case, when it is determined that the first candidate topic is discussed in the corresponding content, the degree of interest in the topic may be determined based on a user's reaction to the first candidate topic in the corresponding content, such as a comment, retweet, or reply.

As the amount of traffic of the user's response to the first candidate topic increases, the degree of interest in the topic may be determined as a relatively high value in proportion thereto. Also, when the time for inputting the user's response to the first candidate topic is short, the user's interest in the topic may be determined as a relatively high value in proportion to the quicker input time for the user's response.

Also, as a qualitative evaluation of the user's response, the higher the meaning of the user's response to the first candidate topic is, the higher the degree of interest in the topic may be determined in proportion thereto.

Specifically, by embedding the meaning of the information of the reaction generated by the user for the first candidate topic and embedding the meaning with the first candidate topic, the higher the degree of relevance to each other's meaning, the higher the user's interest in the topic. can For example, a first candidate topic may be arranged in a Euclidean space according to its meaning, and a response of a user input to the first candidate topic may be arranged in a Euclidean space in which the first candidate topic is arranged according to its meaning. At this time, the user's reaction may be a text, media file, like mark, dislike mark, comment, retweet or reply.

Relevance can then be measured using the distance at which the first candidate topic is arranged in the Euclidean space and the distance at which the user's response is arranged in the Euclidean space. However, the embodiment of performing qualitative evaluation through the meaning of the user's response is not limited thereto and may use various algorithms, but a more detailed description thereof will be omitted so as not to obscure the point of the embodiment.

In addition, the user's propensity information may be input in advance, and the user's topic interest may be collected by referring to the propensity information. For example, the propensity information may include information such as user's personal information, portal search history, previously participated community service information, and fields of interest. In this case, the personal information may be information directly input by the user, and the field of interest, portal search history, and community service information previously participated in may be information generated by tracking the user's previous activities. The topic interest level may be calculated for the first candidate topic based on the user's propensity information.

Otherwise, if there is no response from the user to the candidate topic 1, it can be regarded that the user's topic interest level to the first candidate topic has not been collected. For example, when the first candidate topic is being discussed in community service 1, no response is generated from the user, and if text, media files, likes, dislikes, comments, retweets, or replies from the user respond If it is determined that the meaning of the user's reaction is not at all related to the meaning of the first candidate topic even when is input, the user's topic interest in the first candidate topic is not collected. At this time, an algorithm that determines that the meaning of the user's response and the meaning of the first candidate topic are not related at all may use a previously known semantic analysis algorithm, but is not limited to such an algorithm.

As shown in FIG. 21, in a situation in which the first candidate topic is discussed within community service 1, responses may be input from 3 users out of 4 users. In this case, among the 4 users, the topic interest level for 3 users who input a response is collected, and the topic interest level for the other 1 user is not collected.

Also, in a situation in which the second candidate topic is being discussed within community service 1, responses may be input from two users among four users. In this case, among the four users, the topic interest level of two users who input a response is collected, and the topic interest level of the other two users is not collected.

In addition, responses may be input from all four users in a situation in which the third candidate topic is being discussed within community service 1. In this case, topic interests of all four users may be collected.

As shown in FIG. 21 , the level of interest in a topic may be collected using responses generated in advance with respect to the plurality of candidate topics within a community service in which users participate. That is, in FIG. 21, an embodiment of collecting topic interests in community service 1 while all users are currently having a conversation in community service 1 has been described.

Referring to FIG. 22 , reactions of users may be collected not only from community service 1 but also from community service 2 and community service 3 . Here, community service 2 and community service 3 refer to various types of community services different from community service 1, and the opening time or opening subject is not limited.

Community service 2 and community service 3 mean a plurality of community services, and mean various community services that are different from community service 1 and can collect user responses. In addition, community service 2 and community service 3 may be community services in which users who participated in community service 1 participated in the past and input responses, communities in which current users remain, or community services that have left the community service after participating previously. can be

As shown in FIG. 22, in a situation in which the first candidate topic is discussed within community service 1, 3 out of 4 users may input a response. In this case, the topic interest level of the three users for the first candidate topic may be collected, but the topic interest level of the other one user is not collected.

Also, in a situation in which the first candidate topic is being discussed within community service 2, two users out of four users may input a response. At this time, the user's topic interest level for the first candidate topic that has not been collected in community service 1 may be collected using responses of users input in community service 2.

Also, in a situation in which the first candidate topic is being discussed within community service 3, responses may be input from two users among four users. All users have already input responses to the first candidate topic in community service 1 and community service 2, but some users may input responses to the first candidate topic again in community service 3. In this case, the topic interest level for the first candidate topic of several users who repeatedly input responses may be collected in consideration of the responses generated in the community service 3 together.

In other words, users' topic interest in the candidate topic is collected using conversations that users have entered in the past within the community service in which the users are currently participating, or users who have participated in other community services in the past, even if they are not currently participating, are collected. It can also be collected using the conversation you entered.

An embodiment of collecting user interest in a candidate topic has been described with reference to FIGS. 21 and 22 so far. In step S1840, a step of inferring the user's topic interest may be performed. This operation will be described in detail with reference to FIGS. 23 to 25.

23 to 25 are diagrams for explaining examples of inferring uncollected topic interest using collected topic interest.

23 , in the matrix expressing the collected topic interests of the first and second users for the first candidate topic and the second candidate topic, the first user's topic interest for the first candidate topic was not collected; The second user's topic interest level for the first candidate topic was collected as 0.5. The topic interest levels of the first user and the second user for the second candidate topic were collected as 0.8 and 0.9, respectively.

Here, the first user's topic interest in the first candidate topic may be inferred using the topic interest of the first user and the second user in the second candidate topic and the second user's topic interest in the first candidate topic. . That is, the inclination of the topic interest level of the first user and the second user may be determined using the topic interest level of the first user and the second user for the second candidate topic.

The degree of interest in the uncollected topic may be inferred using the identified inclinations of the first user and the second user. Specifically, the topic interest level of the first user for the first candidate topic may be inferred through the previously collected topic interest level of the second user for the first candidate topic and the propensity of the topic interest level between the first user and the second user. there is.

For example, the topic interests of the first user and the second user for the second candidate topic are 0.8 and 0.9, which are not identical but can be regarded as almost similar. That is, in this step, it may be determined that the first user and the second user have similar inclinations to the topic interest level with respect to the second candidate topic.

As such, the uncollected topic interest may be inferred by using the propensity or trend of the topic interest of the first user and the second user for the candidate topic. It may be inferred that the inclinations of the first user and the second user are the same, similar, dissimilar, unrelated, or opposite through the topic interest levels of the first user and the second user for the second candidate topic. However, since there are two candidate topics in FIG. 23, it can be inferred that the user's propensity is the same, similar, or opposite, but it is difficult to infer that the user's propensity is dissimilar or unrelated through the matrix presented in FIG. 23. . That is, the smaller the number of candidate topics and the lower the collected topic interests, the more difficult it is to deduce user propensity.

If the first user's topic interest level and the second user's topic interest level for the second candidate topic are the same, the propensity of the first user and the second user may be regarded as the same.

In addition, when the first user's topic interest in the second candidate topic is the lowest score (eg, 0.1) and the second user's topic interest is the highest score (eg, 1.0), the relationship between the first user and the second user is Inclinations can be treated as opposites.

However, it is not known that users' inclinations are dissimilar or irrelevant using only the topic interest level collected for the second candidate topic. At this time, the user's topic interest in the additional candidate topic may be further referred to in order to analyze the user's propensity in more detail. Hereinafter, with reference to FIGS. 24 and 25, a method of analyzing a user's propensity by additionally referring to a third candidate topic and inferring topic interest according to the analyzed result will be described.

As shown in FIG. 24 , candidate topics include first to third candidate topics, and users may include first to fourth users. That is, the second candidate topic described in FIG. 23 is not limited to one candidate topic, and may mean a plurality of candidate topics as shown in FIG. 24 . In addition, the second user described in FIG. 23 is not limited to one user, and may refer to at least one of a plurality of users other than the first user, such as a second user and a third user, as shown in FIG. 24 .

As shown in FIG. 24 , the topic interests of the second to fourth users for the first candidate topic were collected as 0.5, 0.9, and 0.7, respectively, and the topic interest of the first user for the first candidate topic was not collected. The topic interests of the first to fourth users for the second candidate topic were collected as 0.8, 0.9, 0.4, and 0.3, respectively. The topic interests of the first user, the second user, and the fourth user for the third candidate topic are collected as 0.1, 0.2, and 0.7, respectively, and the topic interest of the third user is not collected.

Here, in order to determine the candidate topic having the highest topic interest among the candidate topics as a recommended topic, inference of the degree of interest of a topic not collected in the first candidate topic and the degree of interest in a topic not collected in the third candidate topic is required. Among several candidate topics, a candidate topic having the largest sum of topic interests, the largest product of topic interests, or the largest sum of topic interests multiplied by a weight corresponding to each user may be a recommended topic. However, the method of determining the recommended topic is not limited to this calculation method.

In this step, the propensity of topic interest between users 1 to 4 in the topic interest levels of users for the second candidate topic and the third candidate topic is analyzed, and through this, the first user for the first candidate topic as shown in FIG. 25 of topic interests can be inferred. For example, the topic interest levels of the first user and the second user for the second candidate topic are 0.8 and 0.9, respectively, indicating a difference of 0.1 between the respective topic interests.

In comparison, the topic interest level of the fourth user for the second candidate topic is 0.3, which is 0.5 and 0.6 difference from the first user and the second user, respectively. Through this, first user and second user's topic interest inclination can be treated as similar, and the fourth user's topic interest inclination can be treated as dissimilar to that of the first user and the second user. That is, it may be determined that the first user and the second user are interested in the second candidate topic, but the fourth user is not interested in the second candidate topic.

Also, when the third candidate topic is referred to, the topic interests of the first user and the second user for the third candidate topic are 0.1 and 0.2, respectively, which differ by 0.1 from each other. In contrast, the topic interest level of the fourth user for the third candidate topic is 0.7, which is 0.6 and 0.5 different from those of the first user and the second user, respectively.

Through this, once the first user and the second user have a relatively small difference in topic interest compared to other users, the first user and the second user may be treated as having similar topic interest inclinations. In addition, since the fourth user's topic interest level is relatively different from that of the first user and the second user, the fourth user's topic interest tendency is treated as similar to that of the first user and the second user. It can be. That is, it may be determined that the first user and the second user are interested in the second candidate topic, but the fourth user is interested in the second candidate topic.

Accordingly, when the second candidate topic and the third candidate topic are comprehensively referred to, the following user's propensity can be inferred. It may be inferred that the first user and the second user have similar topic interest inclinations through the topic interests of the second candidate topic and the third candidate topic. In addition, it may be inferred that the inclinations of the topic interests of the first user and the fourth user or between the second user and the fourth user are dissimilar to each other through the topic interests of the second candidate topic and the third candidate topic.

Although not shown, if it is not known whether the user's topic interests are the same, similar, dissimilar, or opposite even after referring to several candidate topics, it may be determined that the topic interests of the users are irrelevant.

The degree of interest in a topic that has not been collected for each candidate topic can be inferred using the propensity of users inferred as described above. The propensity of the first user may be inferred using the degree of topic interest for the second and third candidate topics, and the propensity of the second user and the fourth user may be the degree of topic interest for the first to third candidate topics. It can be inferred using , and the propensity of the third user can be inferred using the topic interest level for the first candidate topic and the second candidate topic.

The first user's topic interest level for the first candidate topic may be inferred using the first user's propensity and the topic interest level of the other users for the first candidate topic. The inclination of the first user may be determined to be similar to the inclination of the second user and simultaneously determined to be dissimilar to the inclination of the fourth user. Through this, the first user's topic interest level for the first candidate topic can be inferred similarly to the topic interest level of the second user and at the same time similarly inferred to the topic interest level of the fourth user.

Here, it can be inferred that the first user's topic interest in the first candidate topic is lower than the second user's candidate topic by 0.1, and 0.4, which is smaller than the second user's topic interest 0.5, can be inferred. In addition, the topic interest level of the first user for one candidate topic is 0.7, which is the degree of topic interest of the fourth user for the first candidate topic, so that it is dissimilar to the disposition of the fourth user by further referring to the degree of topic interest of the fourth user. It can be inferred that there are many differences. For example, the first user's topic interest in the first candidate topic may be inferred to be 0.35, which is slightly smaller than 0.4.

In one embodiment, the first user's topic interest level for the first candidate topic that has not been collected may be inferred by inputting the collected topic interest level to an artificial neural network.

The first user's topic interest in the first candidate topic is the first user's topic interest and the second user's topic interest collected in the second candidate topic and the second user's topic interest in the first candidate topic. It can be inferred by inputting topic interest into an artificial neural network.

However, this may be a minimum example for inference of the first user's topic interest level for the first candidate topic that has not been collected, and may be inferred by further using additional information.

For example, as defined above, the second candidate topic may mean a candidate topic other than the first candidate topic, and the second user may mean a user other than the first user. According to this definition, the first user's topic interest in the first candidate topic that has not been collected is the first user's topic interest collected for topics other than the first candidate topic and the topic interest of users other than the first user and the first user's topic interest in other topics than the first candidate topic. It can be inferred by inputting the topic interests of users other than the first user collected for one candidate topic to the artificial neural network.

The first user's level of interest in the first candidate topic may be inferred by further referring to the level of topic interest in the first candidate topic and the second candidate topic of the third user. For example, a first user's topic interest level that has not been collected for a first candidate topic is a topic interest level of the first user and second user for a second candidate topic, and a first user's and second user interest level for a third candidate topic. It may be inferred using the user's topic interest level and the second user's topic interest level for the first candidate topic.

In addition, the first user's topic interest level that has not been collected for the first candidate topic is the topic interest level of the first to third users for the second candidate topic and the topic interest level of the second and third users for the first candidate topic. It can be collected using the degree of interest.

In this way, when inference is made by further using the degree of topic interest, the accuracy of inference can be further increased, and this embodiment is not limited to the number of candidate topics and the number of users described above.

In addition, when uncollected topic interest is inferred, the first user's topic interest for the first candidate topic may be used for other inference. For example, when inferring a first user's topic interest for a first candidate topic, the interest level is determined by combining the first user's topic interest and the second user's topic interest collected for a second candidate topic with respect to the first candidate topic. The second user's topic interest level collected for is needed. Among them, the second user's topic interest in the second candidate topic may be omitted.

In this case, the second user's topic interest in the second candidate topic is the second user's topic interest in the third candidate topic (or the fourth candidate topic) and the third user's topic interest (or the fourth user's topic interest) can be inferred using Alternatively, the degree of topic interest of the second user for the second candidate topic is the second user to the nth user (n>3, n is an integer) for the third candidate topic to the mth candidate topic (m>3, m is an integer). It can be inferred using the candidate topic of

In this way, when the topic interest required for inference of topic interest is again missing, inference is performed using another collected topic interest, and eventually, the topic interest required for inference can be inferred in the same way. This reasoning method is an example, and the reasoning method is not limited thereto.

Compared to existing recommendation systems inferring the degree of interest in all content and recommending content, according to the present embodiment, the degree of interest in the content to be recommended is firstly collected using data that has already been reliably collected, and insufficient data that is not collected is collected. If there is, the method of reasoning as a ratio is used if necessary. Accordingly, by minimizing inferred data, the accuracy of the user's topic interest level is substantially increased.

Further reference is made to FIGS. 26 to 28 to describe a method of inferring uncollected topic interests according to various embodiments.

FIG. 26 is a diagram illustrating topic interest levels of each of a plurality of users for a plurality of candidate topics of community service using a graph.

The community service may be expressed as a graph including nodes representing a plurality of users, nodes representing a plurality of candidate topics, and links connecting the nodes representing a plurality of users and nodes representing a plurality of candidate topics. Here, the link may include topic interests of each of a plurality of users for each of a plurality of candidate topics.

The topic interest level assigned to the link pointing to the first candidate topic from the first user is NULL, and the topic interest level assigned to the link pointing to the first candidate topic from the second user is 0.5.

The degree of topic interest assigned to the link pointing to the second candidate topic from the first user is 0.8, and the degree of topic interest assigned to the link pointing to the second candidate topic from the second user is 0.9.

The topic interest level assigned to the link pointing to the third candidate topic from the first user is 0.1, and the topic interest level assigned to the link pointing to the third candidate topic from the second user is 0.2.

Here, the propensity of the first user and the second user may be analyzed based on the degree of topic interest of the first user and the second user for the second candidate topic and the degree of topic interest of the first user and the second user for the third candidate topic. The topic interest level of the first user for the first candidate topic may be inferred by learning the topic interest level of the second user for the first candidate topic through the artificial neural network. Refer to FIGS. 27 and 28 for a detailed method of inferring topic interest.

27 and 28 are views for explaining a method of inferring uncollected topic interests using a graph.

As shown in FIG. 27, the topic interests assigned to the links pointing to the first candidate topic from the second to fourth users are 0.5, 0.9, and 0.7, respectively, and the topic interests assigned to the links pointing to the first candidate topic from the first user are is NULL.

Topic interests assigned to links pointing to the second candidate topic from the first to fourth users are 0.8, 0.9, 0.4, and 0.3, respectively, and topics assigned to links pointing to the third candidate topic from the second user and the fourth user. The interest levels are 0.1, 0.2, and 0.7, respectively, and the topic interest level assigned to the link pointing to the third candidate topic from the third user is NULL.

In this case, in order to determine a candidate topic having the highest topic interest, inference of a topic interest not collected in the first candidate topic and a topic interest not collected in the third candidate topic is required.

In this step, the propensity between users 1 to 4 is analyzed using the topic interest levels of users for the second candidate topic and the third candidate topic, and through this analysis, as shown in FIG. Topic interest can be inferred.

Specifically, the topic first user's topic interest in the second candidate topic is 0.8, the first user's topic interest in the third candidate topic is 0.1, the second user's topic interest in the second candidate topic is 0.9, and the topic interest is 0.9. The second user's topic interest level for the three candidate topics is 0.2.

The first user and the second user's topic interest in the second candidate topic was relatively high, and the first user and the second user's topic interest in the third candidate topic were relatively low. Looking at this, it can be seen that the first user and the second user have similar topic interests in the candidate topic. When only the graph of FIG. 26 is viewed, it can be determined that there is a high possibility that the first user's and the second user's interest in most of the candidate topics are similar.

Conversely, unlike the first user (or the second user), the fourth user had a low level of interest in the second candidate topic and a high level of interest in the third candidate topic, unlike the first user (or the second user). It may be determined that the interest level of the first user (or the second user) is highly likely to be opposite to each other.

In addition, while it is important whether each user has similar or opposite tendencies through topic interest, it is also important which candidate topic the topic interest was collected for. The range of influence on which user has a relationship with which user may vary through each user's topic interest level collected for each candidate topic.

The first user has no direct topic interest in the first candidate topic, but indirect topic interest in the first candidate topic through the second user because the topic interest in the second candidate topic was collected together with the second user. can be inferred

Likewise, although the third user does not have direct topic interest in the third candidate topic, indirect topic interest in the third candidate topic may be inferred through the fourth user.

That is, although the first user has only topic interest in the second candidate topic and the third candidate topic, the topic interest in the first candidate topic can be inferred through the second to fourth users. In addition, although the third user has only topic interest in the first candidate topic and the second candidate topic, the topic interest in the third candidate topic can be inferred through the first user, the second user, and the fourth user.

In this way, as shown in FIG. 28 , the first user's topic interest level for the first candidate topic can be inferred as 0.7, and the third user's topic interest level for the third candidate topic can be inferred as 0.8.

For example, it is assumed that the first candidate topic is 'necessity for molecular diagnosis', the second candidate topic is 'necessity for vaccine', and the third candidate topic is 'necessity for therapeutic agent'. .

At this time, the first user's topic interest in the second candidate topic 'necessity for vaccine' is 0.8, the second user's topic interest is 0.9, and the first user's topic interest in the third candidate topic 'necessity for treatment' is 0.9. The interest level is 0.1 and the second user's topic interest level is 0.2. When only looking at this graph, since the tendency of the first user's and the second user's topic interests are similarly collected, it can be determined that there is a high possibility that the first user's and the second user's topic interests for most of the candidate topics are similar.

Therefore, the topic relevance for the first candidate topic, 'the need for molecular diagnosis', is inferred by determining based on the information that the second user's topic relevance is 0.5 and the information that the second user has a similar disposition to the first user. It can be. In addition, since the propensity of the first user may be opposite to that of the fourth user, it may be opposite to 0.7, which is the degree of topic relevance of the fourth user to the first candidate topic, 'the need for molecular diagnosis'. In addition, the first user's topic for the first candidate topic is based on the information of 0.9, which is the third user's topic relevance to the first candidate topic 'necessity for molecular diagnosis', and the propensity between the third user and the first user. Relevance can be judged. When the above description is collectively determined, the first user's topic relevance to the first candidate topic is inferred to be 0.7, and the third user's topic interest to the third candidate topic is inferred to be 0.8.

According to the present embodiment, it is possible to infer uncollected interest in a topic by recognizing disposition among users with minimal information. A method of inferring uncollected topic interests has been described with reference to FIGS. 23 to 28 so far. In FIGS. 23 to 28, uncollected topic interest was inferred using the inferred user propensity. However, in the case of such a user's propensity, it is difficult to specify any one formula, and if it is specified with any one formula, it is difficult to know the exact propensity, and it is difficult to infer users' propensity in various situations. In order to overcome such difficulties, a method of more easily inferring user propensity and inferring topic interest using the inferred user propensity will be described below with reference to FIGS. 29 and 30 .

29 is a diagram schematically illustrating the structure of an artificial neural network for inferring topic interest.

As shown in FIG. 29 , the first user and the second user's topic interest level collected for the second candidate topic and the second user's topic interest level collected for the first candidate topic may be input to the artificial neural network.

A user's level of interest in a candidate topic collected from community services may be input to an artificial neural network (ANN). In this case, the artificial neural network receives the first user's topic interest in the second candidate topic, the second user's topic interest in the second candidate topic, and the second user's topic interest in the first candidate topic, and The first user's topic interest level for one candidate topic may be output.

Here, the second candidate topic is not limited to one candidate topic and may mean a plurality of candidate topics. For example, the second candidate topic may include a second candidate topic and a third candidate topic. Also, here, the second user is not limited to one user, and may include at least one of a second user and a third user, which are at least one arbitrary user other than the first user.

The artificial neural network may be a multilayer perceptron or a deep neural network (DNN) including a plurality of hidden layers between an input layer and an output layer, and nodes representing each neuron are weighted It may be a structure connected by a link. The artificial neural network may be composed of various artificial neural network structures capable of inferring topic interest, and is not limited to any one structure. In order not to obscure the gist of the present embodiment, a detailed description of the previously disclosed artificial neural network will be omitted.

In this embodiment, the artificial neural network can be learned based on information about a plurality of candidate topics and topic interests of a plurality of users for each of the plurality of candidate topics. The artificial neural network may be learned based on a plurality of candidate topics collected from a community service and/or other community services other than the community service, and topic interests of a plurality of users for each of the plurality of candidate topics. Other community services other than community services may be secondary community services. Here, the second community service may refer to one service or two or more services.

When learning the artificial neural network, learning data may be collected within a community service or may be learning data collected from other community services other than the community service. Also, the learning data may be learning data collected from a community service and other community services other than the community service.

Other community services other than the community service may be community services different from community services in which the users participate. For example, a community service related to a community service may be a chat room different from a chat room in which users participate. When the community service is the first chat room, the community service related to the community service may be the second chat room.

Other community services other than the community service may be in a state in which users participating in the community service participate or in a state in which the users do not participate.

The input as learning data may include information about topic interest each user has for each candidate topic. Among the above information, information about the degree of topic interest that a specific user has for a specific candidate topic is missing, and a correct answer as learning data may include the missing information.

Also, the artificial neural network may be a graph neural network that receives information in the form of a graph. Refer to FIG. 30 for a more detailed description of this. 30 is a diagram showing a configuration for learning a method of inferring uncollected topic interests by converting community services into a data structure of a graph and inputting them to an artificial neural network composed of GNNs.

As shown in FIG. 30, the community service is converted into a graph-type structure, learned through an artificial neural network, and then uncollected topic interests can be inferred.

Specifically, the community service may be converted into a graph structure through a graph conversion module. The graph transformation module converts community services into graphs based on the users who participated in the community service, the main topic being discussed in the community service, the candidate topic related to the main topic, and the user's response generated in the community service and the topic interest collected through the response. It can be converted into a data structure of the form.

The graph transformation module generates a plurality of users as nodes of a structure of the first type and a plurality of candidate topics as nodes of a structure of the second type, respectively, user nodes and candidate topic nodes are connected by links, and the links are connected to the corresponding candidates. A user's level of interest in a topic may be assigned. The graph conversion module may input the converted graph to the artificial neural network.

In this embodiment, the input layer of the artificial neural network may receive a graph including links with missing topic interest, and the output layer may output a graph including links with uncollected topic interest. That is, the output layer may output links in which some missing topics of interest are inferred, or the entire graph including the links may be output.

Information input to the artificial neural network is a graph including a node representing the plurality of users, a node representing the plurality of candidate topics, and a link connecting the node representing the plurality of users and the node representing the plurality of candidate topics. may consist of In this case, the link may include a topic interest level of each of the plurality of users for each of the plurality of candidate topics.

In this embodiment, the artificial neural network may be composed of a neural network that learns the topic interest of the link through a connection structure between a node representing the plurality of users and a node representing the plurality of candidate topics when a component of the graph is input. It is not limited.

The artificial neural network may be composed of a neural network that learns characteristics of the structure and components of the graph. In one embodiment, the artificial neural network may be implemented as a graph neural network (GNN). A GNN may be a neural network that receives input data for a graph, learns the components and structure of the graph in a hidden layer, and outputs output data for the graph.

An artificial neural network may be composed of an auto encoder. Specifically, the artificial neural network compresses input data using an encoder having fewer nodes than the input layer, restores the compressed input data using a decoder in the output layer, and outputs the same data as the input data. learning can be done. In the process of performing self-supervised learning, a value assigned to a link indicating a level of interest to a missing topic may be inferred. For example, a graph in which a community service is converted is input through an input layer, and each user's topic interest is inferred using components of the graph in an artificial neural network composed of autoencoders, and a graph including the topic interest through an output layer. can be output.

In one embodiment, a graph containing noise, a graph partially masked, or a graph in which community services are converted without noise and masking may be input to the input layer. The autoencoder can learn the trend of the links assigned according to the structure of the graph through the structure of the input graph and the degree of topic interest assigned to the links. The output layer may output a graph including links (part or all of the missing topic interest) from which topic interest, which is the value of the missing link, is inferred through the learned trend of the graph.

In addition, the data of the matrix structure described above with reference to FIGS. 23 to 25 can also be self-supervised learning using an artificial neural network in a similar manner.

An artificial neural network composed of GNNs, when a graph including the degree of topic interest of the second user collected for the first candidate topic and the degree of topic interest of the first user and the second user collected for the second candidate topic is input, the artificial neural network It is possible to learn the tendency of the degree of topic interest between users, and to infer the degree of topic interest of the first user to the first topic using a graph including the degree of topic interest of the second user to the first topic based on this tendency. there is.

In some embodiments, the artificial neural network is composed of an input layer, a hidden layer, and an output layer, and is composed of a layer that learns the structure and components of the graph to infer the topic interest of a link whose topic interest is missing. The activation function may be composed of Relu. , and the activation function is not limited thereto.

Returning to FIG. 2 again, step S1850 will be described. Based on the level of interest in the topic collected in step S1850 and the level of interest in the inferred topic, a recommended topic to be recommended and presented to the community service may be selected from among a plurality of candidate topics.

At this time, in the community service in which users participate, one recommendation topic may be recommended and suggested, or a plurality of recommended topics may be recommended and suggested. If a plurality of recommended topics are suggested, the number of recommended topics to be recommended for the community service among a plurality of candidate topics may be determined. Then, according to the number of recommended topics to be recommended, among the plurality of candidate topics, the recommended topics may be determined in order of candidate topics having the highest topic interest for the plurality of users. The determined plurality of recommended topics may be recommended and suggested to the community service so that users participating in the community service may be notified.

When the recommended topic is recommended and suggested to the community service in the above-described method, users may accept the recommended topic, but users may reject the recommended topic and request re-recommendation of a new topic.

In addition, in this step, a new community service corresponding to the recommended topic is additionally created, and an account of one or more users among a plurality of users accessing the community service may participate in the new community service. In this case, the user participating in the new community service may be, for example, a user who has input a message corresponding to the recommended topic one or more times. In an embodiment, after a new community service corresponding to a recommended topic is created, an existing community service may be maintained, but may be deleted according to an embodiment.

In addition, when providing community services to users, the community service may provide users with information on a main topic on which topic the users are having a conversation in the community service while inputting a conversation. Also, information on candidate topics can be provided to users simultaneously with or separately from the main topic.

In order to vitalize the community service, it will be difficult to actually activate the community by simply adding the number of topics discussed within the community service. Therefore, there is a need for a method of proposing a topic that is of high interest to users and can actually activate community services.

According to this embodiment, a recommendation topic of which the user's topic interest is high among candidate topics related to the main topic may be presented to the community service. In addition, a recommended topic highly related to the main topic may be presented to the community service, and accordingly, the community service may continue to provide services in a direction related to the main topic being discussed.

Hereinafter, an exemplary computing device 3100 capable of implementing the community service management device described in various embodiments will be described with reference to FIG. 31 . In this embodiment, the computing device 3100 may be the community service management device 1700 of FIG. 17 .

31 is an exemplary hardware configuration diagram illustrating a computing device 3100 .

As shown in FIG. 31, the computing device 3100 loads one or more processors 3110, a bus 3150, a communication interface 3170, and a computer program 3191 executed by the processor 3110. It may include a memory 3130 for storing and a storage 3190 for storing the computer program 3191 . However, only components related to the present embodiment are shown in FIG. 31 . Accordingly, those skilled in the art may know that other general-purpose components may be further included in addition to the components shown in FIG. 31 .

The processor 3110 controls the overall operation of each component of the computing device 3100 . The processor 3110 includes at least one of a Central Processing Unit (CPU), a Micro Processor Unit (MPU), a Micro Controller Unit (MCU), a Graphic Processing Unit (GPU), or any type of processor well known in the art. can be configured. Also, the processor 3110 may perform an operation for at least one application or program for executing a method/operation according to various embodiments. Computing device 3100 may include one or more processors.

Memory 3130 stores various data, commands and/or information. The memory 3130 may load one or more programs 3191 from the storage 3190 to execute methods/operations according to various embodiments. For example, when the computer program 3191 is loaded into the memory 3130, the logic (or module) shown in FIG. 18 may be implemented on the memory 3130. An example of the memory 3130 may be RAM, but is not limited thereto.

The bus 3150 provides a communication function between components of the computing device 3100 . The bus 3150 may be implemented as various types of buses such as an address bus, a data bus, and a control bus.

The communication interface 3170 supports wired and wireless Internet communication of the computing device 3100 . The communication interface 3170 may support various communication methods other than Internet communication. To this end, the communication interface 3170 may include a communication module well known in the art.

Storage 3190 may non-temporarily store one or more computer programs 3191. The storage 3190 may be non-volatile memory such as read only memory (ROM), erasable programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, etc., a hard disk, a removable disk, or any other device well known in the art. It may be configured to include a computer-readable recording medium in the form of.

Computer program 3191 may include one or more instructions in which methods/operations according to various embodiments are implemented. When the computer program 3191 is loaded into the memory 3130, the processor 3110 may perform methods/operations according to various embodiments by executing the one or more instructions.

Meanwhile, the method according to various embodiments described above can be implemented in the form of a computer program stored in a computer readable recording medium programmed to perform each step of the method, and also a computer programmed to perform each step of the method. It may be implemented in the form of a computer readable recording medium for storing the program.

The above description is merely an example of the technical idea of the present invention, and various modifications and variations can be made to those skilled in the art without departing from the essential qualities of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention, but to explain, and the scope of the technical idea of the present invention is not limited by these embodiments. The protection scope of the present invention should be construed according to the following claims, and all technical ideas within the scope equivalent thereto should be construed as being included in the scope of the present invention.

Claims (52)

In a method performed by a computing device,
determining the activity level of the community service from messages input to the community service by a plurality of users;
extracting a main topic of the community service from the input message when the determined activity level is equal to or less than a threshold value;
selecting a recommendation topic to be recommended and presented to the community service based on the main topic; and
Including the step of presenting the recommended topic to the community service,
How to administer community services. According to claim 1,
The step of determining the activity of the community service,
Determining the activity of the community service based on at least one of (i) the number of messages input to the community service during unit time and (ii) the number of users inputting messages during unit time,
How to administer community services. According to claim 1,
The step of extracting the main topic of the community service,
clustering the messages based on meanings of the messages input from the plurality of users in the community service and determining a plurality of topics using a plurality of clustered groups; and
Extracting a main topic from among the plurality of topics according to a predetermined criterion,
How to administer community services. According to claim 1,
In the step of selecting the recommended topic,
Among one or more keywords extracted from messages related to the main topic, selecting the recommended topic based on a keyword with a frequency input to the community service or a keyword with the largest number of input users,
How to administer community services. According to claim 1,
In the step of selecting the recommended topic,
Selecting the recommended topic based on related words related to the main topic,
How to administer community services. According to claim 5,
In the step of selecting the recommended topic,
Selecting a recommendation topic including at least one of a word included in the main topic and the related word,
How to administer community services. According to claim 5,
In the step of selecting the recommended topic,
Selecting the recommended topic based on a related word having the highest degree of association with the main topic among the related words,
How to administer community services. According to claim 1,
In the step of selecting the recommended topic,
Selecting the recommended topic based on the propensity information of each of the plurality of users participating in the community service,
How to administer community services. According to claim 1,
Based on the main topic, the step of selecting a recommended topic to be recommended and presented to the community service includes the following steps:
determining a plurality of candidate topics related to the main topic;
collecting topic interests of each of the plurality of users with respect to the plurality of candidate topics;
When the first user's topic interest in the first candidate topic is not collected,
The first user's topic interest level and the second user's topic interest level for the second candidate topic collected from the community service and/or other community services other than the community service;
The second user's topic interest in the first candidate topic collected from the community service and/or other community services other than the community service
inferring a topic interest level of the first user for the first candidate topic by inputting the input to an artificial neural network; and
Selecting a recommendation topic to be recommended and presented to the community service from among the plurality of candidate topics based on the collected topic interest level and the inferred topic interest level. According to claim 9,
The step of determining a plurality of candidate topics related to the main topic,
Determining the plurality of candidate topics from words included in the main topic using related words of a topic dictionary in which words that are related to each other are clustered,
How to administer community services. According to claim 10,
The step of determining a plurality of candidate topics related to the main topic,
Determining a candidate topic including at least one of a word included in the main topic and the related word,
How to administer community services. According to claim 9,
The step of collecting topic interest of each of the plurality of users with respect to the plurality of candidate topics,
Calculate topic interest of each of the plurality of users with respect to the plurality of candidate topics by using responses of each of the plurality of users generated in advance with respect to the plurality of candidate topics in the community service or other community services other than the community service. Including the steps of
How to administer community services. According to claim 9,
In the step of selecting the recommended topic,
determining the number of recommendation topics to be recommended for the community service among the plurality of candidate topics; and
Determining the recommended topic in order of a candidate topic having the highest topic interest for the plurality of users among the plurality of candidate topics according to the number of recommended topics to be recommended.
How to administer community services. According to claim 9,
The artificial neural network,
Learned based on a plurality of candidate topics collected from the community service and / or other community services other than the community service and a plurality of users' topic interests for each of the plurality of candidate topics,
How to administer community services. According to claim 14,
When learning the artificial neural network,
Using learning data collected within the community service and/or learning data collected from other community services other than the community service,
The correct answer as the learning data includes a topic interest level of a predetermined user for a predetermined candidate topic among topic interest degrees for each of a plurality of candidate topics of a plurality of users;
The input as the learning data includes the degree of topic interest of each of the plurality of users for the remaining candidate topics excluding the predetermined candidate topic and the degree of topic interest of each user excluding the predetermined user for the predetermined candidate topic. doing,
How to administer community services. According to claim 15,
The information input to the artificial neural network,
It is composed of a graph including a node representing the plurality of users, a node representing the plurality of candidate topics, and a link connecting the node representing the plurality of users and the node representing the plurality of candidate topics;
The link is
Including topic interest of each of the plurality of users for each of the plurality of candidate topics,
How to administer community services. According to claim 16,
The artificial neural network,
Including a graph neural network that receives the information in the form of a graph,
How to administer community services. According to claim 1,
Further comprising, when receiving a message requesting re-recommendation of a topic different from the recommended topic, re-recommending a topic different from the recommended topic in response to the message requesting re-recommendation of the topic.
How to administer community services. memory for storing instructions; and
a processor operatively coupled with the memory;
The processor executes the command,
The activity level of the community service is determined from messages input to the community service by a plurality of users, and when the determined activity level is less than or equal to a threshold value, a main topic of the community service is extracted from the input message, and based on the main topic Thus, selecting a recommended topic to be recommended and presented to the community service, and presenting the recommended topic to the community service,
Community Services Management Unit. According to claim 19,
The activity of the community service is
Determined based on at least one of the number of messages input to the community service during unit time and the number of users inputting messages during unit time,
Community Services Management Unit. According to claim 19,
The main topic is,
In the community service, a plurality of topics are determined using a plurality of groups in which the messages are clustered based on the meaning of messages input from the plurality of users, and extracted according to a predetermined criterion among the plurality of topics,
Community Services Management Unit. According to claim 19,
The recommended topics are:
Among one or more keywords extracted from messages related to the main topic, selected based on the keyword with the largest input frequency or the largest number of input users to the community service,
Community Services Management Unit. According to claim 19,
The recommended topics are:
Selected based on related words related to the main topic,
Community Services Management Unit. According to claim 23,
The recommended topics are:
At least one of the word included in the main topic and the related word is included,
Community Services Management Unit. According to claim 23,
The recommended topics are:
Among the related words, selected based on the related word having the highest degree of relevance with the main topic,
Community Services Management Unit. According to claim 19,
The community service management device,
Selecting the recommended topic based on the propensity information of each of a plurality of users participating in the community,
Community Services Management Unit. According to claim 19,
The community service management device,
determining a plurality of candidate topics related to the main topic, collecting topic interests of each of the plurality of users for the plurality of candidate topics, and when a first user's topic interest for a first candidate topic is not collected; The first user's topic interest level and the second user's topic interest level for a second candidate topic, and the second user's topic interest level for the first candidate topic are input to an artificial neural network to obtain information about the first candidate topic. Inferring a first user's level of interest in a topic, and selecting a recommended topic to be recommended and presented to the community service from among the plurality of candidate topics based on the collected level of interest and the inferred level of topic interest.
Community Services Management Unit. According to claim 27,
The plurality of candidate topics,
Determined from words included in the main topic using related words of a topic dictionary in which words that are related to each other are clustered,
Community Services Management Unit. 29. The method of claim 28,
The candidate topic is
At least one of the words and related words included in the main topic is included,
Community Services Management Unit. According to claim 27,
The topic interest level of each of the plurality of users for the plurality of candidate topics,
Calculated using responses of each of the plurality of users generated in advance with respect to the plurality of candidate topics in the community service or other community services other than the community service,
Community Services Management Unit. According to claim 27,
The community service management device,
The number of recommended topics to be recommended for the community service is determined from among the plurality of candidate topics, and according to the number of recommended topics to be recommended, among the plurality of candidate topics, a candidate topic having the highest topic interest for the plurality of users is ranked in order. determining the recommended topic,
Community Services Management Unit. According to claim 27,
The artificial neural network,
Learned based on a plurality of candidate topics collected from the community service and / or other community services other than the community service and a plurality of users' topic interests for each of the plurality of candidate topics,
Community Services Management Unit. 33. The method of claim 32,
When learning the artificial neural network,
Using learning data collected within the community service and/or learning data collected from other community services other than the community service,
The correct answer as the learning data includes a topic interest level of a predetermined user for a predetermined candidate topic among topic interest degrees for each of a plurality of candidate topics of a plurality of users;
The input as the learning data includes the degree of topic interest of each of the plurality of users for the remaining candidate topics excluding the predetermined candidate topic and the degree of topic interest of each user excluding the predetermined user for the predetermined candidate topic. doing,
Community Services Management Unit. 34. The method of claim 33,
The information input to the artificial neural network,
It is composed of a graph including a node representing the plurality of users, a node representing the plurality of candidate topics, and a link connecting the node representing the plurality of users and the node representing the plurality of candidate topics;
The link is
Including topic interest of each of the plurality of users for each of the plurality of candidate topics,
Community Services Management Unit. 35. The method of claim 34,
The artificial neural network,
Including a graph neural network that receives the information in the form of a graph,
Community Services Management Unit. According to claim 19,
The community service management device,
re-recommending a topic different from the recommended topic in response to the message requesting re-recommendation of the topic when receiving a message requesting re-recommendation of a topic different from the recommended topic;
Community Services Management Unit. A communication module for transmitting and receiving data with a community service management server that manages community services;
a display module displaying the community service;
a terminal memory storing a program for inputting a message to the community service and displaying the message input to the community service; and
Including a processor that executes the program,
As the program is executed, the processor receives the activity of the community service accessed by the terminal through the communication module,
When the activity of the received community service is less than or equal to the threshold value, a recommendation topic to be recommended and presented to the community service is received through the communication module and displayed on the display module;
The recommended topic is selected based on a main topic extracted from a message input to the community service,
Terminal. 38. The method of claim 37,
The activity of the community service is
Determined based on at least one of the number of messages input to the community service during unit time and the number of users inputting messages during unit time,
Terminal. 38. The method of claim 37,
The main topic is,
In the community service, a plurality of topics are determined using a plurality of groups in which the messages are clustered based on the meaning of messages input from the plurality of users, and extracted according to a predetermined criterion among the plurality of topics,
Terminal. 38. The method of claim 37,
The recommended topics are:
Among one or more keywords extracted from messages related to the main topic, selected based on the keyword with the largest input frequency or the largest number of input users to the community service,
Terminal. 38. The method of claim 37,
The recommended topics are:
Selected based on related words related to the main topic,
Terminal. The method of claim 41 ,
The recommended topics are:
At least one of the word included in the main topic and the related word is included,
Terminal. 41. The method of claim 41,
The recommended topics are:
Among the related words, selected based on the related word having the highest degree of relevance with the main topic,
Terminal. 38. The method of claim 37,
The recommended topics are:
Selected based on the propensity information of each of the plurality of users participating in the community service,
Terminal. 38. The method of claim 37,
The recommended topics are:
determining a plurality of candidate topics related to the main topic, collecting topic interests of each of the plurality of users for the plurality of candidate topics, and when a first user's topic interest for a first candidate topic is not collected; The first user's topic interest level and the second user's topic interest level for a second candidate topic, and the second user's topic interest level for the first candidate topic are input to an artificial neural network to obtain information about the first candidate topic. The first user's topic interest is inferred, and the plurality of candidate topics are selected based on the collected topic interest and the inferred topic interest.
Terminal. 46. The method of claim 45,
The artificial neural network,
Learned based on a plurality of candidate topics collected from the community service and / or other community services other than the community service and a plurality of users' topic interests for each of the plurality of candidate topics,
Terminal. 47. The method of claim 46,
When learning the artificial neural network,
Using learning data collected within the community service and/or learning data collected from other community services other than the community service,
The correct answer as the learning data includes a topic interest level of a predetermined user for a predetermined candidate topic among topic interest degrees for each of a plurality of candidate topics of a plurality of users;
The input as the learning data includes the degree of topic interest of each of the plurality of users for the remaining candidate topics excluding the predetermined candidate topic and the degree of topic interest of each user excluding the predetermined user for the predetermined candidate topic. doing,
Terminal. 39. The method of claim 38,
The information input to the artificial neural network,
It is composed of a graph including a node representing the plurality of users, a node representing the plurality of candidate topics, and a link connecting the node representing the plurality of users and the node representing the plurality of candidate topics;
The link is
Including topic interest of each of the plurality of users for each of the plurality of candidate topics,
Terminal. The method of claim 39,
The artificial neural network,
Including a graph neural network that receives the information in the form of a graph,
Terminal. 38. The method of claim 37,
The display module,
Displaying new community services corresponding to the recommended topics and information of users participating in the new community services;
Terminal. A computer readable recording medium storing a computer program programmed to perform each step included in the method according to claims 1 to 18. A computer program stored in a computer readable recording medium programmed to perform each step included in the method according to claims 1 to 18.

Images