Private chat service providing method and device

Abstract

The embodiment of the invention relates to a private chat service providing method and a private chat service providing device, and relates to a private chat service providing method and a private chat service providing device which can execute message filtering by taking a large-scale user as an object and improve the private message receiving and sending processing speed by comprehensively using a database distributed to each user and a database for storing artist messages.

Description

Technical Field

[0001] This invention relates to a method and apparatus for providing a private chat service, specifically, to a method and apparatus for providing a private chat service to improve the speed of message sending and receiving between an artist and their numerous fans. Background Technology

[0002] With the booming development of the entertainment industry and the advancement of IT technology, social networking services and fan platforms designed to facilitate communication between celebrities, artists, and fans are increasingly prevalent. These social networking services enable the function of fan communities. For example, they can comprehensively provide interactive features with celebrities, live streaming services, and fan community functions.

[0003] Furthermore, the functions of such fan platforms have gone beyond those of traditional communities, gradually expanding into a wide range of areas such as merchandise sales and purchases, concert and event pre-sales, original content provision, and even the application of non-fungible tokens (NFTs).

[0004] As fan platforms expand their functionality, the notifications and messages offered to users are becoming increasingly diverse. For example, if a feature allows artists to directly write articles, posts, messages, videos, etc., and publish them to the community, users can immediately receive push notifications when artists write posts or comments, or receive various membership-specific offers. Furthermore, different content can be provided for notifications and messages depending on the specific features.

[0005] This demonstrates that not only are the functions of fan platforms constantly expanding, but also, with the globalization of the entertainment market, the number of artist communities and users on these platforms has increased dramatically. At the same time, as the influence of the Korean Wave has expanded the user base globally, the challenge of promptly handling messages between artists and multiple fans has emerged. Summary of the Invention

[0006] The problem the invention aims to solve

[0007] The purpose of the private chat service provision method and apparatus according to embodiments of the present invention is to improve the message filtering and processing speed in chats between artists and multiple fans.

[0008] Furthermore, the purpose of the private chat service provision method and apparatus according to embodiments of the present invention is to improve the message sending and receiving speed between artists and multiple fans.

[0009] Furthermore, the purpose of the private chat service provision method and apparatus according to embodiments of the present invention is to provide private messages to each user.

[0010] However, the technical issues to be addressed in this embodiment are not limited to those described above, and other technical issues may also exist.

[0011] Problem Solving Methods

[0012] As a technical means to solve the aforementioned technical problem, a method for providing a private chat service according to an embodiment of the present invention is a method for relaying messages between an artist terminal and multiple fan terminals in a private chat service relay server to provide a private chat service, comprising: a step of receiving an artist message from an artist terminal; a step of filtering whether the artist message includes preset prohibited words for the artist; a step of storing the artist message and artist information in the database of each DM when the artist message does not include the prohibited words for the artist; a step of receiving a private chat message update request from a first fan terminal that requests to query the artist message from the multiple fan terminals; a step of generating a private message corresponding to the first fan terminal based on the fan account information corresponding to the first fan terminal and the artist message; and a step of transmitting the private message to the first fan terminal, wherein the database of each DM is configured to store messages by artist unit, and the private message refers to a message after specific words or keywords of the message have been converted according to the receiving object.

[0013] Furthermore, according to an embodiment of the present invention, a method for providing a private chat service is provided by relaying messages between an artist terminal and multiple fan terminals in a private chat service relay server. The method includes: matching multiple independent databases corresponding to each of the multiple fan terminals; receiving fan messages from a first fan terminal among the multiple fan terminals; filtering whether the fan messages contain preset prohibited fan words; storing the fan messages in a first independent database corresponding to the first fan terminal among the multiple independent databases; classifying the context of the fan messages using a learning-based artificial intelligence model; classifying whether the fan messages are malicious messages based on the classified context; receiving artist messages from the artist terminal; storing the artist messages and artist information in the databases of each DM; receiving a private chat message update request from the first fan terminal; generating a private message corresponding to the first fan terminal based on fan account information corresponding to the first fan terminal and the artist messages; and transmitting the fan messages and the private messages stored in the first independent database to the first fan terminal. The databases of each DM are configured to store messages by artist unit, and the private message refers to a message whose specific words or keywords are transformed according to the receiving object.

[0014] Furthermore, a private chat service providing apparatus, for relaying messages between artist terminals and multiple fan terminals in a private chat service relay server to provide a private chat service, includes: a communication module that performs information sending and receiving between the terminals and the private chat service relay server; a memory that stores a private chat program; a database for each DM (DM), which stores messages by artist unit; multiple independent databases that are allocated to each user and store messages for each user; and a processor that executes the private chat program stored in the memory, wherein the processor receives fan messages from a first fan terminal among the multiple fan terminals and stores the fan messages in the first independent database corresponding to the first fan terminal among the multiple independent databases. The system receives artist messages from the artist terminal, filters whether the fan messages and artist messages include preset prohibited fan words and prohibited artist words, and stores the artist messages and artist information in the database of each DM when the fan messages do not include the prohibited fan words. When a private chat message update request is received from the first fan terminal, a private message corresponding to the first fan terminal is generated based on the fan account information corresponding to the first fan terminal and the artist messages, and the fan messages and the private messages stored in the first independent database are transmitted to the first fan terminal. The private message refers to a message after specific words or keywords of the message have been converted according to the recipient.

[0015] Invention Effects

[0016] The private chat service provision method and apparatus according to embodiments of the present invention can improve message filtering processing speed in chats between artists and multiple fans.

[0017] Furthermore, the private chat service provision method and apparatus according to embodiments of the present invention can improve the message sending and receiving speed between artists and multiple fans.

[0018] Furthermore, the private chat service provision method and apparatus according to embodiments of the present invention provide private messages to each user. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of the communication connection of a private chat service providing device according to an embodiment of the present invention.

[0020] Figure 2 This is a structural diagram of a private chat service relay server according to an embodiment of the present invention.

[0021] Figure 3 This is a structural diagram of a terminal according to an embodiment of the present invention.

[0022] Figure 4 This is a conceptual diagram illustrating the functionality of a processor according to an embodiment of the present invention.

[0023] Figure 5This is a conceptual diagram of data processing for storing fan messages according to an embodiment of the present invention.

[0024] Figure 6 This is a data flow diagram for storing artist messages according to an embodiment of the present invention.

[0025] Figure 7 This is a data flow diagram for private chat query on a fan terminal according to an embodiment of the present invention.

[0026] Figure 8 This is a data flow diagram for private chat query on a fan terminal according to an embodiment of the present invention.

[0027] Figure 9 This is a data flow diagram for private chat querying on an artist's terminal according to an embodiment of the present invention.

[0028] Figure 10 This is a sequence diagram of a fan message storage method according to an embodiment of the present invention.

[0029] Figure 11 This is a sequence diagram of the artist message storage method according to an embodiment of the present invention.

[0030] Figure 12 This is a sequence diagram of the private chat update method according to an embodiment of the present invention.

[0031] Figure 13 This is a sequence diagram of the private chat update method according to an embodiment of the present invention.

[0032] Figure 14 This is a sequence diagram of the private chat update method according to an embodiment of the present invention.

[0033] Figure 15 This is a conceptual diagram illustrating the functionality of a processor according to an embodiment of the present invention.

[0034] Figure 16 This is a data flow diagram for filtering fan messages according to an embodiment of the present invention.

[0035] Figure 17 This is a data flow diagram for artist message filtering according to an embodiment of the present invention.

[0036] Figure 18 This is a sequence diagram of a fan message filtering method according to an embodiment of the present invention.

[0037] Figure 19 This is a sequence diagram of the artist message filtering method according to an embodiment of the present invention. Detailed Implementation

[0038] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings to help those skilled in the art to readily implement the invention. The present invention can be implemented in various different forms and is not limited to the embodiments described herein. For the sake of clarity, content unrelated to the description is omitted, and the same or similar structures are given the same reference numerals throughout the specification.

[0039] Throughout the specification, when a part is described as "connected" to other parts, this includes not only direct connections but also indirect connections via other components. When a part is described as "including" a component, unless otherwise stated, it does not exclude other components but may also include them.

[0040] Furthermore, the accompanying drawings are only intended to help understand the embodiments disclosed in this specification, and are not intended to limit the technical ideas disclosed in this invention. They also include all modifications, equivalents, and even substitutions that fall within the scope of the ideas and techniques of this invention.

[0041] Ordinal numbers such as "first," "second," etc., indicating sequence, can be used to describe various structures, but structures are not limited by the terms mentioned above. The purpose of these terms is to distinguish one structure from another.

[0042] When one structure is "connected" or "accessed" to another structure, it means that the two structures are directly connected or accessed, or are connected or accessed through other structures. Conversely, when one structure is "directly connected" or "directly accessed" to another structure, it means that there are no other structures in between.

[0043] Unless the context clearly indicates otherwise, singular expressions include the meaning of plural expressions.

[0044] In this application, terms such as "comprising" or "possessing" indicate the presence of features, numbers, steps, actions, structures, components, or combinations thereof as described in the specification, rather than precluding the presence or additional possibility of one or more other features, numbers, steps, actions, structures, components, or combinations thereof.

[0045] See below Figure 1 This describes a private chat service providing apparatus according to an embodiment of the present invention.

[0046] Figure 1 This is a schematic diagram of the communication connection between the private chat service relay server 100 and the terminal 200 according to an embodiment of the present invention.

[0047] See Figure 1The private chat service relay server 100 establishes a communication connection with the terminal 200 via a communication network. Here, the private chat service relay server 100 refers to a device used for sending and receiving messages between users using fan platforms, social networking services, etc., and can correspond to fan platform servers, social networking service servers, etc. The terminal 200 includes artist terminals 200-1 used by artists and celebrities, and fan terminals 200-2 used by fans who follow artists and celebrities. The terminal 200 can be a device that displays messages and notifications received from the private chat service relay server 100 to the user.

[0048] The private chat service relay server 100 can generate and deliver messages and notification lists to users on fan platforms, social networking services, etc. In this case, it can not only provide users with messages and notifications, but also provide personalized notification messages and notification lists based on each user's filtering information. Furthermore, the private chat service relay server 100 can build independent databases for each user, allowing for independent message storage. The specific configuration of the private chat service relay server 100 will be described later. Figure 2 and Figure 4 Detailed explanation.

[0049] Terminal 200 can refer to any type of handheld wireless communication device, such as a laptop computer, desktop computer, or laptop computer equipped with a web browser, a portable and mobile wireless communication device, or a smartphone or tablet computer.

[0050] In addition, such as Figure 1 The communication network shown can be implemented as a wired network, such as a Local Area Network (LAN), a Wide Area Network (WAN), or a Value Added Network (VAN), or any type of wireless network, such as a mobile radio communication network or a satellite communication network.

[0051] See below Figure 2 The structure of the private chat service relay server according to an embodiment of the present invention is described.

[0052] Figure 2 This is a structural diagram illustrating the structure of a private chat service relay server 100 according to an embodiment of the present invention.

[0053] See Figure 2The private chat service relay server 100 includes a communication module 110, a memory 120, and a processor 140, and may also include a database 130. The communication module 110 performs the sending and receiving of information with the terminal 200. The communication module 110 may include hardware and software devices required to send control signals or data signals to other network devices via wired / wireless connections.

[0054] The memory 120 stores the private chat program. The name of the private chat program is only for ease of description and does not limit the function of the program. The memory 120 can store at least one of the following: information and data input to the communication module 110, information and data required for the functions executed by the processor 140, and data generated by the operation of the processor 140.

[0055] The memory 120 should be understood to encompass the concepts of non-volatile memory that can continuously retain stored information without a power supply, and volatile memory that requires power to retain stored information. The memory 120 may serve to temporarily or permanently store data processed by the processor 140. In addition to volatile memory that requires power to retain stored information, the memory 120 may include magnetic storage media or flash storage media, but the scope of the invention is not limited thereto.

[0056] Database 130 can store data related to artist information, user information, subscription information, artist messages, fan messages, and private messages provided by individual fan users. Database 130 can be part of storage 120, but does not necessarily have to be located inside the private chat service relay server 100, or it can be connected to the outside of the private chat service relay server 100 to perform data sending and receiving using a communication connection.

[0057] In addition, database 130 can store message data by artist unit and user unit. Detailed configuration of database 130 will be described later. Figure 4 Please provide a detailed explanation.

[0058] Processor 140 is configured to execute a private chat program in memory 120. Processor 140 may include various means for controlling and processing data. Processor 140 may refer to a data processing device built into the device, which has a circuitry of a physical structure for performing functions represented by code or instructions included in the program. In one example, processor 140 may be implemented as a microprocessor, central processing unit (CPU), processor core, multiprocessor, ASIC (application-specific integrated circuit), FPGA (field programmable gate array), etc., but the scope of the invention is not limited thereto.

[0059] Processor 140 is configured to perform the following functions and processes by running a private chat program.

[0060] See also Figure 1 and Figure 2 The processor 140 receives artist messages generated by the artist terminal 200-1 and transmits the artist messages to the fan terminals 200-2 that follow the artist. Additionally, the processor 140 can receive fan messages generated by the fan terminals 200-2 and provide the fan messages to the artist terminal 200-1.

[0061] Furthermore, the processor 140 can generate private messages for each fan user based on the content of the artist's message and provide them to the fan terminal 200-2. In this case, to improve the message conversion speed for multiple fan users, the processor 140 can generate private messages for each fan user and store them in each user's database. Therefore, even if multiple users use the private chat service, private messages can be provided promptly. Moreover, compared to generating private messages based on user message query requests, the speed of providing the private chat service can be significantly improved.

[0062] Furthermore, when the processor 140 stores artist messages in the database 130, it can read artist messages from the database 130 to generate and provide private messages during the process of generating and storing private messages in each user database. After the private messages are stored in each user database, the processor 140 can delete duplicate artist messages and receive message information stored in each user database to provide private chat services to users.

[0063] Figure 3 This is a block diagram illustrating the configuration of terminal 200.

[0064] See Figure 3The terminal 200 includes a memory 220, an input / output module 230, and a processor 240, and may also include a communication module 210.

[0065] The communication module 210 can send and receive information with an external database or external device. The external device can be the aforementioned private chat service relay server. Figure 1 (100 in the text). Memory 220 stores the private chat program. The name of the private chat program is only for ease of description and does not limit the program's functionality. Input / output module 230 can receive information, data, etc. transmitted from the outside to terminal 200, or output information, data, etc. held by terminal 200 to the outside. For example, input / output module 230 may include a display, touchpad, etc. Processor 240 executes the private chat program stored in memory 220. For supplementary descriptions of communication module 210, memory 220, and processor 240, please refer to the foregoing. Figure 2 Description of the communication module Figure 2 110, memory Figure 2 120 and processor ( Figure 2 The description in 140 is replaced.

[0066] Processor 240 is configured to perform the following functions and processes by running a private chat program.

[0067] The processor 240 can transmit artist messages or fan messages generated in the terminal 200 to the private chat service relay server 100. In addition, the processor 240 can receive artist messages, fan messages, private messages, push streams, alarms, etc. stored and generated in the private chat service relay server 100.

[0068] Below, please refer to Figure 4 The operation of the processor 140 and database 130 according to embodiments of the present invention is described in detail.

[0069] Figure 4 This is a conceptual diagram illustrating the function of a private chat service relay server according to an embodiment of the present invention.

[0070] See Figure 4 The processor 140 can be configured to perform the functions of the control unit 141 and the message queue unit, which will be described later. Furthermore, the database 130 can consist of databases 1310 for each DM (Direct Message), independent databases 1320, and sub-databases 1330.

[0071] The control unit 141 performs the following controller functions: storing artist messages and fan messages generated in artist terminal 200-1 and fan terminal 200-2 in database 130, or retrieving messages from database 130 for private chat updates. The message queue unit (not shown) performs the message publishing function, publishing artist messages or fan messages stored in database 130 to the receiving object.

[0072] Each DM's database 1310 can store messages by private chat unit or artist unit. Here, a private chat unit refers to the unit of the chat room where fans and artists send and receive messages. Independent databases 1320 are built for each user to store messages by user unit. Therefore, corresponding to the number of users, independent databases 1321 to Nth independent databases 1321 are formed.

[0073] In addition, the independent database 1320 may include an artist posting database 1322 that stores messages posted by artists and an artist receiving database 1323 that stores messages received by artists.

[0074] Sub-database 133 may include a subscription information storage unit 1331 that stores information related to fans who follow an artist, and a message body storage unit 1332 that stores the content of artist messages and fan messages.

[0075] The following will detail the interaction process between the processor 140 and the database 130 when providing private chat services between artists and fans.

[0076] The processor 140 can receive fan messages generated by a first fan terminal among multiple fan terminals 200-2. The processor 140 controls the storage of fan messages in a first independent database corresponding to the first fan terminal among multiple independent databases 1320. In addition, while fan messages are stored in the first independent database, the processor 140 can generate message push stream information for the fan messages and provide it to the artist terminal 200-1.

[0077] In addition, the processor 140 can receive artist messages from the artist terminal 200-1. When the processor 140 receives artist messages, it can control the storage of artist messages and artist information in the database 1310 of each DM. While the artist messages are stored in the database 1310 of each DM, the processor 140 generates a message push stream to be provided to multiple fan terminals and transmits the message push stream to the first fan terminal.

[0078] When a first fan terminal requests a private chat update, processor 140 receives the private chat message update request from the first terminal. Then, processor 140 generates a private message corresponding to the first fan terminal based on the fan account information and artist information for that first fan terminal. Processor 140 transmits the fan message stored in a first independent database and the generated private message to the first fan terminal. Here, the private message can refer to a message whose specific words or keywords have been transformed according to the recipient. Furthermore, the message push stream may include notifications, articles, etc., that notify the message recipient of message receipt.

[0079] In addition, when receiving artist messages, the processor 140 converts the artist messages into private messages and stores them in the database 1310 of each DM while storing them in an independent database 1320.

[0080] Specifically, processor 140 generates multiple private messages corresponding to each of the multiple fan terminals based on fan account information and artist messages. Then, processor 140 stores each private message in a corresponding independent database based on the fan account information. Once the private messages have been stored in the independent databases, processor 140 controls the deletion of artist messages stored in the databases 1310 of each DM (Distributor Manager). Furthermore, after the private messages are stored in the independent databases 1320, the processor receives the private messages from the independent databases (not the databases 1310 of each DM) and transmits them to the fan terminals.

[0081] In other words, when the private chat service relay server 100 composes an artist's message, it generates a corresponding private message for each subscribed fan and stores the private messages in an independent database allocated to each user. Therefore, when a user requests a message query or update, the message is retrieved from the independent database, thus enabling the rapid provision of private chat services to multiple users.

[0082] In addition, the processor 140 stores artist messages and artist information in the database 1310 of each DM. After a preset time, it deletes the artist messages stored in the database 1310 of each DM and receives private messages from the independent database 1320 and transmits them to the fan terminal.

[0083] At this point, the database 1310 of each DM can be built based on NoSQL (Not Only SQL), while the independent database 1320 can be built based on SQL (Structured Query Language).

[0084] Therefore, each DM's database 1310 has a non-relational database structure and can store unstructured data. In addition, the independent database 1320 has a relational database structure, which can create DB (DataBase) tables for each user and pre-apply filters for each user's DB table to store each user's message data.

[0085] Because the standalone database 1320 is built using SQL, it has slower write speeds but faster read speeds compared to the database 1310 of each DM. Conversely, compared to the standalone database 1320, the database 1310 of each DM has faster write speeds but slower read speeds.

[0086] The private chat service relay server 100 utilizes an independent database 1320 to provide stored private messages, thereby improving the speed of providing private chat services to multiple users. Furthermore, while private messages are stored in the independent database 1320, private messages can be provided using the databases 1310 of each DM, thus compensating for the operational issues when messages are stored in the independent database.

[0087] See below for further details. Figures 5 to 9 This section details the specific data processing flow between the processor 140 and the database 130.

[0088] Figure 5 This is a data processing flowchart for storing fan messages according to an embodiment of the present invention.

[0089] See Figure 5 When a fan composes and sends a fan message to an artist, the fan terminal 200-2 transmits fan message data to the control unit 141 of the private chat service relay server 100. The fan message data may include at least one of the following: message content, artist information of the message recipient, and artist subscription information of the fan.

[0090] The control unit 141 transmits the received fan message data to the message body storage unit 1332 and requests that the fan message data be stored in the message body storage unit 1332. The message body storage unit 1332 stores the fan message data, and after storing the fan message data, it generates a fan message ID corresponding to the fan message and responds to the control unit 141 with the fan message ID.

[0091] The control unit 141, which receives fan message IDs, requests storage of fan message data from the independent database 1320. Here, the independent database 1320 refers to a database allocated per user. Therefore, when the first fan terminal receives a fan message, it requests storage of fan message data from the first independent database 1321 allocated to the first terminal. Furthermore, the fan message data stored in the independent database 1320 may include fan message IDs and subscription information.

[0092] When the fan message data is stored in the independent database 1320, the independent database 1320 sends a fan message data storage completion response to the control unit 141.

[0093] After receiving the response from the independent database 1320 indicating that the fan message data storage is complete, the control unit 141 requests the subscription information storage unit 1331 to query the fan's subscription information. Specifically, the message queue unit (not shown) included in the control unit 141 requests the subscription information from the subscription information storage unit 1331 based on the fan's subscriber ID information.

[0094] Upon receiving a subscription information query request, the subscription information storage unit 1331 will respond to the control unit 141 with the DM (Direct Message) ID corresponding to the subscriber ID information. At this time, the DM ID can correspond to the ID of the private chat room or artist information.

[0095] The control unit 141, which receives the DM ID from the subscription information storage unit 1331, requests the storage of fan message data from the artist receiving database 1323, which is the artist to whom the fan messages are received, based on the DM ID. At this time, the fan message data stored in the artist receiving database 1323 may include fan message ID and DM ID information.

[0096] After the fan message data is stored in the artist receiving database 1323, the artist receiving database 1323 responds to the fan message publishing request to the control department 141.

[0097] The control unit 141 receives fan message publishing requests from the artist receiving database 1323, generates an artist-specific fan push stream to be sent to the receiving artist, and transmits the one-person fan push stream to the terminal 200 of the receiving artist.

[0098] Therefore, when fans send messages to artists, the messages are stored in the artist's received database. Furthermore, when an artist needs to confirm the use of a fan push stream and query fan messages, the private chat relay server 100 can read the fan messages based on the fan message data stored in the artist's received database and provide them to the artist's terminal.

[0099] Below, please refer to Figure 6 Explain the data processing flow used to store artist messages.

[0100] Figure 6 This is a data flow diagram for storing artist messages according to an embodiment of the present invention.

[0101] See Figure 6When an artist composes and sends a message to a fan, the artist terminal 200-1 transmits the artist message data to the control unit 141 of the private chat service relay server 100. The artist message data may include at least one of the following: message content, message recipient information, and DM ID.

[0102] The control unit 141 transmits the received artist message data to the message body storage unit 1332 and requests that the fan message data be stored in the message body storage unit 1332. The message body storage unit 1332 stores the artist message data, and after storing the artist message data, it generates the corresponding artist message ID and responds to the control unit 141 with the artist message ID.

[0103] The control unit 141, which receives artist message IDs, requests the artist sending database 1322 to store artist message data. At this time, the artist sending database 1322 is an independent database assigned to the artist, meaning it stores messages written by the artist for sending. Furthermore, the artist sending database 1322 can be constructed together with the artist receiving database 1323, or it can be constructed independently. Additionally, the artist message data stored in the artist receiving database 1322 may include at least one or more related data from the artist message ID and DM ID.

[0104] When the artist's message data is stored in the artist receiving database 1322, the artist receiving database 1322 sends an artist message data storage completion response to the control unit 141.

[0105] After receiving the artist message data storage completion response from the artist receiving database 1322, the control unit 141 requests a query for artist message data from the database 1310 of each DM. Furthermore, the artist message data stored in the database 1320 of each DM may include at least one or more related data such as the artist message ID and the DM ID.

[0106] When the artist message data is stored in the database 1310 of each DM, the database 1310 of each DM sends an artist message data storage completion response to the control unit 141.

[0107] After receiving the artist message data storage completion response from the database 1310 of each DM, the control unit 141 requests the subscription information storage unit 1331 to query the subscriber list information of the corresponding DM ID. Specifically, the message queue unit (not shown) included in the control unit 141 requests the subscription information storage unit 1331 to query the subscriber ID list of the subscribed artist based on the DM ID.

[0108] When the control unit 141 receives the subscriber ID list from the subscription information storage unit 1331, it generates private messages for each subscriber based on the artist message data and the subscriber ID list. Furthermore, the control unit 141 requests storage of private messages in an independent database 1321 corresponding to each subscriber.

[0109] At this time, requesting the storage of private messages to the independent database 1321 may mean requesting the control unit 141 to store the private message content in the message body storage unit 1332, and to store at least one of the following data obtained from the message body storage unit 1332: private message ID, DM ID, and subscriber ID.

[0110] Therefore, when an artist sends a message to a fan, the private messages for each fan are stored in the fan's independent database 1321. Furthermore, when a fan wants to retrieve a message, the private message relay server 100 can read the private message data stored in the individual database 1321 and provide it to the fan's terminal.

[0111] Below, please refer to Figure 7 This section details the data processing flow when a fan requests to query a private chat before the private messages are stored in a separate database.

[0112] Figure 7 This is a data flow diagram of a fan querying messages according to an embodiment of the present invention.

[0113] See Figure 7 When a fan requests to query private chat messages or artist messages, the control unit 141 receives the private chat query request from the fan terminal 200-2. At this time, the private chat query request includes the fan's subscriber ID information.

[0114] When the control unit 141 receives a private chat query request from the fan terminal 200-2, the control unit 141 requests a message query for the corresponding subscriber ID from the independent database 1320 of the corresponding fan terminal 200-2 or fan account.

[0115] The independent database 1320 can respond to message query requests by providing a list of stored message IDs. At this time, the message responded to by the independent database 1320 to the control unit 141 may include one or more of fan message IDs and private message IDs. However, before private messages are stored in the independent database 1320, the message IDs responded to by the independent database 1320 to the control unit 141 only include fan message IDs.

[0116] In addition, the control unit 141 requests the subscription information storage unit 1331 to query the DMID of the corresponding fan's subscriber ID. Upon receiving the DM ID query request, the subscription information storage unit 1331 responds to the control unit 141 with the DMID stored for the corresponding subscriber ID.

[0117] The control unit 141, which receives the DMID, requests a message from the database 1310 of each DM to query the corresponding DMID. The database 1310 of each DM that receives the message query request responds to the control unit 141 with the artist message ID stored for the corresponding DMID.

[0118] Therefore, the control unit 141 receives fan message IDs from the independent database 1320 and artist message IDs from the databases 1310 of each DM. Then, based on the received fan message IDs and artist message IDs, the control unit 141 requests message body information corresponding to the message IDs and artist message IDs from the message body storage unit 1332. Therefore, the message body storage unit 1332 responds with message information corresponding to the received fan message IDs and artist message IDs.

[0119] The control unit 141 can transmit received fan messages and artist messages to the fan terminal 200-2. Furthermore, the control unit 141 can convert artist messages into private messages and transmit them to the fan terminal 200-2 based on the fan's subscription information.

[0120] Below, please refer to Figure 8 This section details the data processing flow when a fan requests to query a private chat after the private messages are stored in a separate database.

[0121] Figure 8 This is a data flow diagram of a fan querying messages according to an embodiment of the present invention.

[0122] See Figure 8 When a private message is stored in the independent database 1320, the control unit 141 requests the deletion of the completed artist message from the database 1310 of each DM. A completed artist message refers to an artist message that has been converted into a private message or its message ID.

[0123] Deleting the artist message ID stored in the database 1310 of each DM can prevent artist messages and private messages from being sent repeatedly to fan terminals.

[0124] After deleting the artist message ID from the database 1310 of each DM, when a request to query private chat or artist messages is made in the fan terminal 200, the control unit 141 receives the private chat query request from the fan terminal 200-2. At this time, the private chat query request includes the fan's subscriber ID information.

[0125] When the control unit 141 receives a private chat query request from the fan terminal 200-2, the control unit 141 requests a message query for the corresponding subscriber ID from the independent database 1320 of the corresponding fan terminal 200-2 or fan account.

[0126] The independent database 1320 can respond to message query requests by providing a list of stored message IDs. At this time, the message that the independent database 1320 responds to the control unit 141 may include fan message IDs and private message IDs.

[0127] Therefore, the control unit 141 receives fan message IDs and private message IDs from the independent database 1320. Then, based on the received fan message IDs and private message IDs, the control unit 141 requests message body information corresponding to the message IDs and private message IDs from the message body storage unit 1332. Therefore, the message body storage unit 1332 responds with the message information corresponding to the received fan message IDs and private message IDs.

[0128] The control unit 141 can transmit received fan messages and private messages to fan terminals 200-2. Therefore, even if multiple fan terminals 200-2 request to query private messages, the private message service relay server 100 reads the messages from the independent database 1320 of the corresponding fan terminal 200-2. Thus, even if multiple users request to query messages, a fast processing speed can be maintained.

[0129] Below, please refer to Figure 9 This document details the data processing procedure when an artist requests access to their private chat history.

[0130] Figure 9 This is a data flow diagram of an artist querying messages according to an embodiment of the present invention.

[0131] See Figure 9 When an artist requests to query private chats, the control unit 141 receives the private chat query request from the artist's terminal 200-1. At this time, the private chat query request includes DM ID information.

[0132] When the control unit 141 receives a private chat query request from the artist terminal 200-2, the control unit 141 sends a message query request to the artist via database 1322 for the corresponding DM ID. Therefore, the artist sends a list of artist message IDs stored in the corresponding artist's DM ID to database 1322 in response.

[0133] In addition, the control unit 141 requests a message from the artist receiving database 1323 to query the corresponding DM ID. Therefore, the artist receiving database 1323 responds with a list of fan message IDs stored in the corresponding artist's DM ID.

[0134] Therefore, the control unit 141 receives artist message IDs from the artist sending database 1322 and a list of fan message IDs from the artist receiving database 1323. Based on the received list of artist message IDs and the list of fan message IDs, the control unit 141 requests a query for message information from the message body storage unit 1332.

[0135] Therefore, the message body storage unit 1332 responds to the control unit 141 with message information corresponding to the artist message ID list and the fan message ID list. The control unit 141, which receives the artist message and fan message information, updates the message content to the artist terminal 200.

[0136] Below, in conjunction with Figures 10 to 14 This section details how the private chat service is provided.

[0137] Figure 10 This is a sequence diagram of a fan message storage method according to an embodiment of the present invention.

[0138] See Figure 10 In the private chat service of this invention, the method for storing fan messages sent by fan users to artists includes: a fan message receiving step S110, a fan message storage step S120, and a message push stream providing step S130.

[0139] Specifically, in the fan message receiving step S110, when a fan composes and sends a fan message to an artist, the fan terminal 200-2 transmits fan message data to the control unit 141 of the private chat service relay server 100. The fan message data may include at least one of the following: message content, artist information of the message recipient, and artist subscription information of the fan.

[0140] In the fan message storage step S120, the control unit 141 stores the received fan message data in the independent database 1320. Specifically, the control unit 141 transmits the received fan message data to the message body storage unit 1332 and requests that the fan message data be stored in the message body storage unit 1332. Then, the message body storage unit 1332 stores the fan message data. Upon completion of storing the fan message data, it generates a fan message ID corresponding to the fan message and responds to the control unit 141 with the fan message ID.

[0141] The control unit 141, which receives fan message IDs, requests storage of fan message data from the independent database 1320. Here, the independent database 1320 refers to a database allocated per user. Therefore, when the first fan terminal receives a fan message, it requests storage of fan message data from the first independent database 1321 allocated to the first terminal. Furthermore, the fan message data stored in the independent database 1320 may include fan message IDs and subscription information.

[0142] In the message push stream provision step S130, the control unit 141 generates a message push stream to be transmitted to the artist who is the recipient of the fan messages and transmits it to the artist terminal 200. Specifically, when the fan message data is stored in the independent database 1320, the independent database 1320 sends a fan message data storage completion response to the control unit 141.

[0143] After receiving the response from the independent database 1320 indicating that the fan message data storage is complete, the control unit 141 requests the subscription information storage unit 1331 to query the fan's subscription information. Specifically, the message queue unit (not shown) included in the control unit 141 requests the subscription information from the subscription information storage unit 1331 based on the fan's subscriber ID information.

[0144] Upon receiving a subscription information query request, the subscription information storage unit 1331 will respond to the control unit 141 with the DM (Direct Message) ID corresponding to the subscriber ID information. At this time, the DM ID can correspond to the ID of the private chat room or artist information.

[0145] The control unit 141, which receives the DM ID from the subscription information storage unit 1331, requests the storage of fan message data from the artist receiving database 1323, which is the artist to whom the fan messages are received, based on the DM ID. At this time, the fan message data stored in the artist receiving database 1323 may include fan message ID and DM ID information.

[0146] After the fan message data is stored in the artist receiving database 1323, the artist receiving database 1323 responds to the fan message publishing request to the control department 141.

[0147] The control unit 141 receives fan message publishing requests from the artist receiving database 1323, generates an artist-specific fan push stream to be sent to the receiving artist, and transmits the one-person fan push stream to the terminal 200 of the receiving artist.

[0148] Therefore, when fans send messages to artists, the messages are stored in the artist's received database. Furthermore, when an artist needs to confirm the use of a fan push stream and query fan messages, the private chat relay server 100 can read the fan messages based on the fan message data stored in the artist's received database and provide them to the artist's terminal.

[0149] Below, please refer to Figure 11 Explain the method for storing artist messages.

[0150] Figure 11 This is a sequence diagram of the artist message storage method according to an embodiment of the present invention.

[0151] See Figure 11 The artist message storage method includes: receiving artist messages in step S210, storing artist messages in an independent database in step S220, storing artist messages in the databases of each DM in step S230, generating private messages in step S240, and storing private messages in step S250.

[0152] In step S210 of receiving artist messages, when an artist user composes and sends an artist message to a fan, the artist terminal 200-1 transmits artist message data to the control unit 141 of the private chat service relay server 100. The artist message data may include at least one of the following: message content, message recipient information, and DM ID.

[0153] In step S220, where artist messages are stored in an independent database, the control unit 141 stores the received artist message data in the artist sending database 1322 of the artist's independent database 1320.

[0154] Specifically, the control unit 141 transmits the received artist message data to the message body storage unit 1332 and requests that the fan message data be stored in the message body storage unit 1332. The message body storage unit 1332 stores the artist message data, and after storing the artist message data, it generates the corresponding artist message ID and responds to the control unit 141 with the artist message ID.

[0155] The control unit 141, which receives artist message IDs, requests the artist sending database 1322 to store artist message data. At this time, the artist sending database 1322 is an independent database assigned to the artist, meaning it stores messages written by the artist for sending. Furthermore, the artist sending database 1322 can be constructed together with the artist receiving database 1323, or it can be constructed independently. Additionally, the artist message data stored in the artist receiving database 1322 may include at least one or more related data from the artist message ID and DM ID.

[0156] When the artist's message data is stored in the artist receiving database 1322, the artist receiving database 1322 sends an artist message data storage completion response to the control unit 141.

[0157] In step S230, where artist messages are stored in the database of each DM, the control unit 141 stores the received artist messages in the database 1310 of each DM.

[0158] Specifically, the control unit 141 requests the storage of artist message data from the database 1310 of each DM. The database 1310 of each DM stores artist messages or message IDs respectively, categorized by DM ID. Furthermore, the artist message data stored in the database 1320 of each DM may include at least one or more related data, such as the artist message ID and the DM ID.

[0159] In step S240 of generating a private message, the control unit 141 generates a private message for transmission to fan users who have subscribed to the artist, based on the received artist message.

[0160] Specifically, after receiving the artist message data storage completion response from the database 1310 of each DM, the control unit 141 requests the subscription information storage unit 1331 to query the subscriber list information of the corresponding DM ID. Specifically, the message queue unit (not shown) included in the control unit 141 requests the subscription information storage unit 1331 to query the subscriber ID list of the subscribed artist based on the DM ID.

[0161] When the control unit 141 receives the subscriber ID list from the subscription information storage unit 1331, the control unit generates private messages for each subscriber based on the artist message data and the subscriber ID list.

[0162] In step S250 of storing private messages, the control unit 141 stores the generated private messages in the corresponding independent database 1320.

[0163] Specifically, private messages are generated individually for each fan user who subscribes to the artist. Furthermore, each fan user is assigned an independent database 1321. Therefore, the control unit 141 stores the generated private messages in each fan user's independent database 1321.

[0164] At this time, requesting the storage of private messages to the independent database 1321 may mean requesting the control unit 141 to store the private message content in the message body storage unit 1332, and to store at least one of the following data obtained from the message body storage unit 1332: private message ID, DM ID, and subscriber ID.

[0165] Therefore, when an artist sends a message to a fan, the private messages for each fan are stored in the fan's independent database 1321. Furthermore, when a fan wants to retrieve a message, the private message relay server 100 can read the private message data stored in the individual database 1321 and provide it to the fan's terminal.

[0166] Below, please refer to Figure 12 and Figure 13 This explains how to update privately to fans.

[0167] Figure 12 This is a sequence diagram of the fan terminal private chat update method according to an embodiment of the present invention.

[0168] See Figure 12 The private chat update method based on fans' private chat updates and query requests includes: step S310 of receiving private chat update requests, step S320 of confirming the storage of private messages, step S340 of querying the independent database, step S332 of querying the database of each DM, and step S350 of updating private chats.

[0169] In step S310 of receiving a private chat update request, when a fan requests to query a private chat or artist message, the control unit 141 receives a private chat query request from the fan terminal 200-2. At this time, the private chat query request includes the fan's subscriber ID information.

[0170] In step S320 of confirming the storage of private messages, it is confirmed whether the private messages are being stored or have been stored in the independent database 1321.

[0171] While private messages are being stored in independent database 1321, artist messages will be read from the databases of each DM instead of the independent database and the private chat content will be updated.

[0172] Specifically, when the private message is not stored in the independent database 1321, the steps of querying the independent database S331, querying the databases of each DM S332, and generating the private message S333 are performed.

[0173] In step S331 of querying the independent database, the control unit 141 requests a message query for the corresponding subscriber ID from the independent database 1320 of the corresponding fan terminal 200-2 or fan account.

[0174] The independent database 1320 can respond to message query requests by providing a list of stored message IDs. At this time, the message responded to by the independent database 1320 to the control unit 141 may include one or more of fan message IDs and private message IDs. However, before private messages are stored in the independent database 1320, the message IDs responded to by the independent database 1320 to the control unit 141 only include fan message IDs. Therefore, in step S331 of querying the independent database, fan message data written by fans can be received.

[0175] In step S332, which queries the database of each DM, the control unit 141 requests the subscription information storage unit 1331 to query the DM ID of the subscriber ID of the corresponding fan. Upon receiving the DM ID query request, the subscription information storage unit 1331 responds to the control unit 141 with the DM ID stored for the corresponding subscriber ID.

[0176] The control unit 141, which receives the DMID, requests a message from the database 1310 of each DM to query the corresponding DMID. The database 1310 of each DM that receives the message query request responds to the control unit 141 with the artist message ID stored for the corresponding DMID.

[0177] Therefore, the control unit 141 receives fan message IDs from the independent database 1320 and artist message IDs from the databases 1310 of each DM. Then, based on the received fan message IDs and artist message IDs, the control unit 141 requests message body information corresponding to the message IDs and artist message IDs from the message body storage unit 1332. Therefore, the message body storage unit 1332 responds with message information corresponding to the received fan message IDs and artist message IDs.

[0178] The control unit 141 can transmit the received fan messages and artist messages to the fan terminal 200-2.

[0179] Furthermore, in step S333 of generating a private message, the control unit 141 may convert the artist's message into a private message and transmit it to the fan terminal 200-2 based on the fan's subscription information.

[0180] In step S320, when the private message has been stored in the independent database 1320, step S340, which queries the independent database, is performed only for the private chat update for the fan.

[0181] In step S340, which queries the independent database, the control unit 141 requests the deletion of completed artist messages from the database 1310 of each DM. Completed artist messages refer to artist messages that have been converted to private messages or their message IDs.

[0182] Deleting the artist message ID stored in the database 1310 of each DM can prevent artist messages and private messages from being sent repeatedly to fan terminals.

[0183] After deleting the artist message ID in the database 1310 of each DM, when requesting to query private chat or artist messages in the fan terminal 200, in step S340 of querying the independent database, the control unit 141 requests the message query corresponding to the subscriber ID from the independent database 1320 of the corresponding fan terminal 200-2 or fan account.

[0184] The independent database 1320 can respond to message query requests by providing a list of stored message IDs. At this time, the message that the independent database 1320 responds to the control unit 141 may include fan message IDs and private message IDs.

[0185] Therefore, the control unit 141 receives fan message IDs and private message IDs from the independent database 1320. Then, based on the received fan message IDs and private message IDs, the control unit 141 requests message body information corresponding to the message IDs and private message IDs from the message body storage unit 1332. Therefore, the message body storage unit 1332 responds with the message information corresponding to the received fan message IDs and private message IDs.

[0186] In step S350 of the private chat update, the control unit 141 transmits the received fan messages and private messages to the fan terminal 200-2 to update the private chat of the fan terminal 200-2.

[0187] Therefore, even if multiple fan terminals 200-2 request to query private messages, the private message service relay server 100 reads the message from the independent database 1320 of the corresponding fan terminal 200-2. Thus, even with multiple users requesting to query messages, a fast processing speed can be maintained.

[0188] Furthermore, in the fan terminal private chat update method according to an embodiment of the present invention, the step S320 of confirming the storage of private messages can be changed to the step S321 of confirming whether the reference time has elapsed.

[0189] In step S321, which confirms whether the baseline time has elapsed, the control unit 141 determines that the private message is being stored in the independent database 1321 before the preset baseline time expires. Therefore, when the control unit 141 receives a private chat update request from the fan terminal 200 after receiving an artist's message, it will execute step S331 of querying the independent database, step S332 of querying the databases of each DM, and step S333 of generating the private message.

[0190] That is, after receiving the artist's message, the control unit 141 retrieves the artist's message information stored in the database 1310 of each DM before the end of the preset base time, and generates a private message based on the artist's message and provides it to the fan terminal 200.

[0191] Furthermore, in step S321, which confirms whether the reference time has elapsed, the control unit 141 determines that the private message has been stored in the independent database 1321 when the preset reference time has elapsed. Therefore, in step S340, after receiving an artist's message and the preset time has elapsed, the control unit 141 receives the message in the independent database when the fan terminal 200 requests a private chat update.

[0192] That is, after receiving the artist's message and setting a base time, the control unit 141 provides the artist's message and private message stored in the independent database 1320 to the fan terminal 200.

[0193] Below, please refer to Figure 14 This section details how to update private chat messages when an artist requests to view them.

[0194] Figure 14 This is a sequence diagram of the artist terminal private chat update method according to an embodiment of the present invention.

[0195] See Figure 14 The private chat update method of artist terminal 200 includes: step S410 of receiving private chat update request, step S420 of querying artist sending database, step S430 of receiving artist database and step S440 of updating private chat.

[0196] In step S410 of receiving a private chat update request, when an artist requests to query private chat information, the control unit 141 receives the private chat query request from the artist terminal 200-1. At this time, the private chat query request includes DM ID information.

[0197] In step S420 of querying the artist message database, the control unit 141 requests the artist message database 1322 to query the message corresponding to the DM ID. Therefore, the artist message database 1322 responds with a list of artist message IDs stored in the DM ID of the corresponding artist.

[0198] In step S430 of querying the artist receiving database, the control unit 141 requests the artist receiving database 1323 to query the message corresponding to the DM ID. Therefore, the artist receiving database 1323 responds with a list of fan message IDs stored for the corresponding artist's DM ID.

[0199] Therefore, the control unit 141 receives artist message IDs from the artist sending database 1322 and a list of fan message IDs from the artist receiving database 1323. Based on the received list of artist message IDs and the list of fan message IDs, the control unit 141 requests a query for message information from the message body storage unit 1332.

[0200] Therefore, the message body storage unit 1332 responds to the control unit 141 with message information corresponding to the artist message ID list and fan message ID list. Then, in the private chat update step S440, the control unit 141, which receives the artist message and fan message information, updates the message content to the artist terminal 200.

[0201] Below, in conjunction with Figures 15 to 19 This section details the methods for filtering artist and fan messages.

[0202] Figure 15 This is a conceptual diagram illustrating the functionality of a processor according to an embodiment of the present invention.

[0203] See Figure 15, in addition to having the functions of the aforementioned control unit 141, the processor 140 may further include a prohibited word filtering module 142 for filtering artist messages and fan messages, and a malicious message filtering module 143 for filtering malicious messages.

[0204] The prohibited word filtering module 142 determines whether the artist messages and fan messages include preset prohibited words. At this time, the prohibited words can be separately set for artists and fans respectively. In addition, for fans, different prohibited words can be set according to the artist who is the recipient of the message.

[0205] The prohibited word filtering module 142 identifies the words included in the artist messages and fan messages, and determines whether the identified words correspond to the preset prohibited words.

[0206] In addition, the prohibited word filtering module 142 can convert the fan message and artist message data into image data. After that, the prohibited word filtering module 142 extracts the text corresponding to the converted image data.

[0207] The prohibited word filtering module 142 can identify the words included in the text extracted from the image, and determine whether the identified words correspond to the prohibited words, so as to filter the prohibited words in the artist messages and fan messages.

[0208] The prohibited word filtering module 142 uses a CNN (Convolution Neural Network) learning model to convert the artist message data and fan message data into image data.

[0209] Specifically, when the prohibited word filtering module 142 extracts the text corresponding to the image data, it can be respectively converted into Korean consonants and vowels. For example, when the artist message or fan message includes contents such as "んЙ刀│○ㅑ", "人ㅣ凹卜", etc., the prohibited word filtering module 142 converts the message into an image, and extracts the consonants "ㅅ, ㄲ, ㅇ" and "ㅅ, ㅂ", and the vowels "ㅐ, ㅣ, ㅑ" and "ㅣ, ㅏ". After that, it extracts "새끼야" and "시바" as the combinations of vowels and consonants.

[0210] In addition, the CNN may include a part for extracting image features by effectively identifying and enhancing adjacent image features while maintaining the spatial information of the image, and a part for classifying the image. The feature extraction part may include a convolution layer that uses filters to find image features while minimizing the number of shared parameters, and a pooling layer that enhances and aggregates features. Based on the existing network, the CNN is re-learned for a new recognition task and used, so the CNN learning model can filter new types of malicious messages that are creatively deformed to avoid prohibited words.

[0211] Furthermore, the prohibited word filtering module 142 can utilize the BERT (Bidirectional Encoder Representation of Transformer) learning model to filter whether the extracted text contains prohibited words. The BERT learning model can be fine-tuned based on the accuracy of the judgment results.

[0212] BERT is a pre-trained learning model developed by Google for labeling and training large amounts of articles on the internet using semi-supervised learning. BERT is used in the embedding process when performing tasks such as object name recognition and text classification.

[0213] The prohibited word filtering module 142 determines whether prohibited words are included by means of: an algorithm that uses user-labeled messages containing preset words, and a method that makes a judgment based on preset criteria.

[0214] The pre-defined standards can be categorized as follows: expressions containing general insults, vulgar and low-level expressions that cause discomfort to others, sexually provocative expressions, expressions involving physical threats, discriminatory expressions based on region / race / country / religion, and degrading expressions that cause the other party to feel humiliated and ashamed.

[0215] The disabled word filtering module 142 outputs a value of "0" when a message contains abusive or defamatory expressions, and outputs a value of "1" when a message does not contain abusive or defamatory expressions, in order to filter messages.

[0216] The malicious message filtering module 143 analyzes the context of the text extracted by the banned word filtering module 142 and filters whether the fan message is malicious based on the analyzed context.

[0217] The malicious message filtering module 143 performs encoding using bidirectional contextual analysis of fan messages to classify the context of fan messages. Specifically, the malicious message filtering module 143 can input fan messages into a BERT (Bidirectional Encoder Representation of Transformer) learning model to filter whether fan messages belong to malicious messages. Specifically, the BERT learning model used in the malicious message filtering module 143 obtains input with a single embedding that integrates word embeddings, segment embeddings, and position embeddings for fan messages, and analyzes the context of fan messages.

[0218] In addition, the malicious message filtering module 143 uses Word Piece embedding, which segments words into sub-words, to prevent out-of-vocabulary words from constituting the fan message.

[0219] Specifically, the malicious message filtering module 143 can utilize Word Piece embedding to process word embeddings. In Word Piece embedding, embedding can be done in units smaller than a word, with the longest sub-word set as a unit. That is, common sub-words can be used as embedding units, while rare words can be segmented into sub-words. Existing word embedding methods suffer from out-of-vocabulary (OV) word problems, leading to difficulties in learning and translating rare words, proper nouns, numbers, or words not found in the vocabulary. Word embedding, however, is applicable to all languages, and by segmenting words into sub-word units, it effectively solves the OV word problem and improves accuracy.

[0220] Furthermore, the malicious message filtering module 143 can receive the sentence separator [SEP] along with the two sentences. At this point, due to input length limitations, the total length of the two sentences must be controlled within 512 words. That is, because the learning time increases quadratically with increasing input length, the input length needs to be set reasonably. Korean on average consists of 20 words, and 99% of sentences do not exceed 60 words; therefore, the combined length of the two sentences is limited to 128. However, considering the possibility of long sentences, the input length can be initially limited to 128 for learning, while longer inputs (greater than 128) can be supplemented in the final step.

[0221] Furthermore, the malicious message filtering module 143 can use position encoding. For this, a Transformer model can be applied, which does not require a CNN or RNN model but can use a Self-Attention model. However, since the Self-Attention model cannot reflect the position of the input, additional word embeddings with relevant positional information are needed. Therefore, in the Transformer model, positional encoding based on the Sinusoid function is used, which the malicious message filtering module 143 can transform before using positional encoding. In positional encoding, it can simply be encoded according to the word (token) order, in the order of 0, 1, 2, ...

[0222] Furthermore, the malicious message filtering module 143 can integrate word embeddings, fragment embeddings, and positional embeddings to generate a single embedding value (i.e., a single embedding). This single embedding can be generated in the form of a feature map and then converted to an appropriate form according to the subsequent workflow. For example, the malicious message filtering module 143 can use the single embedding result after layer normalization and dropout processing as input.

[0223] Furthermore, the malicious message filtering module 143 can provide the output malicious message encoding value to the fine-tuning network to improve the accuracy of malicious message judgment. That is, the malicious message filtering module 143 can receive the sentence separator [SEP] together with two sentences related to the user's message, and generate a malicious message encoding value about the probability of the malicious message as output through pre-learning and internal reasoning steps.

[0224] The malicious message filtering module 143 can use the BERT learning model directly by default to perform malicious comment filtering, and the accuracy of malicious message judgment can be improved by appending learning. Therefore, the malicious message filtering module 143 can be defined to combine the BERT learning model with a fine-tuning network for the malicious message judgment process.

[0225] The malicious message filtering module 143 can provide the malicious message encoding value to a one-dimensional convolutional layer, or generate a feature map by calculating an ensemble based on a variable kernel size, and provide the feature map to a BiLSTM (Bidirectional Long Short-Term Memory) network to determine the likelihood of a malicious message targeting a follower.

[0226] Subsequently, the malicious message filtering module 143 allows the output of the BiLSTM network to pass through an affine layer and a softmax layer, thereby performing a binary classification of the probability of malicious messages. The BiLSTM (Bidirectional Long Short-Term Memory) network and the bidirectional LSTM can be viewed as structures sharing two independent LSTM architectures. First, the bidirectional LSTM sequentially receives the sentence input.

[0227] That is, just like humans, sentences are received sequentially from left to right. Furthermore, the bidirectional LSTM can be used in conjunction with a reverse LSTM that reads backward from right to left to further consider the context behind the sentence. During prediction at the output layer, the bidirectional LSTM connects the outputs of the forward and reverse LSTMs, thus utilizing both types of information simultaneously.

[0228] These bidirectional LSTMs can achieve end-to-end learning of all parameters simultaneously while minimizing output value loss, and internalize the similarity between words and phrases into the input vector to improve performance. Furthermore, in the case of bidirectional LSTMs, by incorporating the basic performance characteristics of LSTMs and the attention mechanism, they have the advantage of not degrading performance even with long datasets.

[0229] Furthermore, the malicious message filtering module 143 can combine the BERT learning model with the CNN (Convolutional Neural Network) learning model to achieve malicious message filtering. Specifically, the malicious message filtering module 143 inputs the encoded value of the malicious message into a one-dimensional convolutional layer, inputs the output of the one-dimensional convolutional layer into a GeLU layer, inputs the output of the GeLU layer into a max pooling layer, and generates a binary classification result based on the output of the max pooling layer. At this time, the binary classification can use a linear classification method.

[0230] For example, linear classification methods may include linear regression or linear classification models.

[0231] Furthermore, the malicious message filtering module 143 can combine the BERT learning model with an ensemble CNN model to achieve malicious message filtering. More specifically, the malicious message filtering module 143 can perform the following steps: First, perform ensemble computation based on the malicious message encoding value using a variable kernel size; Second, independently input the ensemble computation result into the GeLU layer; Third, independently input the output of the GeLU layer into the MaxPooling layer; Fourth, sequentially connect the outputs of the MaxPooling layer to generate intermediate values; Fifth, input the intermediate values ​​into the Affine layer; Sixth, input the output of the Affine layer into the softmax layer; Seventh, generate a binary classification result based on the softmax layer result.

[0232] The ensemble computation based on variable kernel size can be a series of convolutions performed by applying different kernel sizes to the same malicious message encoding value. Furthermore, the convolutions can include ID convolution operations. That is, the malicious comment judgment unit 230 can independently apply convolution kernels of different sizes to the feature map belonging to the malicious comment encoding value, performing multiple convolution operations respectively. Then, the malicious message filtering module 143 performs a classification operation based on the malicious probability of the fan message by integrating the results of each convolution operation into a single result (i.e., the fourth step).

[0233] Furthermore, the malicious message filtering module 143 can combine the BERT learning model with an ensemble CNN model and a BiLSTM to achieve malicious message filtering. More specifically, the malicious message filtering module 143 can perform the following steps: First, perform ensemble computation based on the malicious comment encoding value using a variable kernel size; Second, independently input the ensemble computation result into a GeLU layer; Third, independently input the output of the GeLU layer into a max pooling layer; Fourth, sequentially connect the outputs of the max pooling layer to generate intermediate values; Fifth, apply the intermediate values ​​to a BiLSTM (Bidirectional LSTM (Long Short-Term Memory)) model; Sixth, input the application result of the BiLSTM model into an affine layer; Seventh, input the output of the affine layer into a softmax layer; Eighth, generate a binary classification result based on the softmax layer result.

[0234] That is, the malicious message filtering module 143 can achieve malicious message filtering by adding a step of applying intermediate values ​​to BiLSTM (Bidirectional LSTM (Long Short-Term Memory)) between the fourth and fifth steps of the method of combining the BERT learning model and the ensemble CNN model.

[0235] Furthermore, to filter malicious messages, the control unit 141 derives a malicious message prediction score based on the frequency of banned words used by fans, the contextual information of fan messages, and the frequency of malicious message writing by fans. The control unit 141 may provide a fan message review request notification to the fan terminal 200 when the malicious message prediction score exceeds a preset first reference score. Additionally, the control unit 141 may provide a fan message cessation notification to the fan terminal when the malicious message prediction score exceeds a preset second reference score. In this case, the second reference score is set to a value higher than the first reference score.

[0236] Below, please refer to Figure 16 This document details the data processing workflow used for filtering fan messages.

[0237] Figure 16 This is a data flow diagram for filtering fan messages according to an embodiment of the present invention.

[0238] See Figure 16 In the case of fan messages, further steps such as filtering for banned words and filtering for malicious messages can be performed to achieve message filtering.

[0239] Specifically, when the control unit 141 receives fan messages from the fan terminal 200, the control unit 141 requests the banned word filtering module 142 to filter banned words in the fan messages.

[0240] Upon receiving a request to filter prohibited words, the prohibited word filtering module 142 will filter whether the fan messages contain prohibited words based on the prohibited words set by the receiving artist.

[0241] In addition, see also Figure 15 The prohibited word filtering module 142 can convert fan messages into images. The prohibited word filtering module 142 can extract text from the converted image data, identify words contained in the extracted text, and determine whether the identified words correspond to prohibited words, thereby filtering fan messages for prohibited words.

[0242] The banned word filtering module 142 responds to the control unit 141 with the fan message filtering results. When the fan message does not contain banned words, the control unit 141 initiates a fan message storage request to the message body storage unit 1332 and to the independent database 1320.

[0243] When fan messages are stored in the independent database 1320, the control unit 141 requests malicious message filtering from the malicious message filtering module 143. Upon receiving the malicious message filtering request, the malicious message filtering module 143 categorizes the context of the fan messages and determines whether the filtered fan messages are malicious based on the categorized filtering.

[0244] See also Figure 15The malicious message filtering module 143 performs encoding using bidirectional contextual analysis of fan messages to classify the context of fan messages. The malicious message filtering module 143 can input fan messages into a BERT (Bidirectional Encoder Representation of Transformer) learning model to filter whether fan messages are malicious. Specifically, the BERT learning model used in the malicious message filtering module 143 receives input with a single embedding that integrates token embedding, segment embedding, and position embedding for fan messages, and analyzes the context of fan messages to classify them as malicious.

[0245] The control unit 141, which receives the malicious message filtering results from the malicious message filtering module 143, derives a malicious message prediction score based on the frequency of banned words used by fans, the contextual information of fan messages, and the frequency of malicious message writing by fans.

[0246] Furthermore, the control unit 141 may provide a fan message review request notification to the fan terminal 200 when the malicious message prediction score is greater than a preset first reference score. Additionally, the control unit 141 may provide a fan message stop sending notification to the fan terminal when the malicious message prediction score is greater than a preset second reference score. In this case, the second reference score is set to a value higher than the first reference score.

[0247] When the malicious message filtering results indicate that the fan message is not malicious, the control unit 141 will subscribe to the data query through the subscription information storage unit 1332 and initiate a fan message storage request to the artist receiving database 1323.

[0248] Below, please refer to Figure 17 This describes the data processing flow used for filtering artist messages.

[0249] Figure 17 This is a data flow diagram for artist message filtering according to an embodiment of the present invention.

[0250] See Figure 17 In the case of artist messages, further steps can be taken to filter prohibited words in order to achieve message filtering.

[0251] Specifically, when the control unit 141 receives artist messages from the artist terminal 200, the control unit 141 requests the banned word filtering module 142 to filter banned words in the artist messages.

[0252] Upon receiving a request to filter prohibited words, the prohibited word filtering module 142 will check whether the artist's message contains prohibited words based on the prohibited words set by the artist who wrote the message.

[0253] In addition, see also Figure 15 The prohibited word filtering module 142 can convert artist messages into images. The prohibited word filtering module 142 can extract text from the converted image data, identify words contained in the extracted text, and determine whether the identified words correspond to prohibited words, thereby filtering the artist messages for prohibited words.

[0254] The banned word filtering module 142 responds to the artist message filtering results to the control unit 141. When the artist message does not contain banned words, the control unit 141 initiates an artist message storage request to the message body storage unit 1332, an artist message storage request to the artist sending database 1322, an artist message storage request to the databases 1310 of each DM, and an artist message storage request to the independent database 1320.

[0255] Below, please refer to Figure 18 Provide a detailed explanation of the methods for filtering fan messages.

[0256] Figure 18 This is a sequence diagram of a fan message filtering method according to an embodiment of the present invention.

[0257] See Figure 18 In the case of fan message filtering methods, it is possible to Figure 10 The fan message storage method shown further includes a banned word filtering step S510 and a malicious message filtering step S520.

[0258] In the banned word filtering step S510, when the control unit 141 receives fan messages from the fan terminal 200, the control unit 141 requests the banned word filtering module 142 to filter banned words in the fan messages.

[0259] In the next step S510, the banned word filtering module 142, which receives the banned word filtering request, will filter whether the fan message contains banned words based on the banned words set by the receiving artist.

[0260] In the prohibited word filtering step S510, the prohibited word filtering module 142 can convert fan messages into images. The prohibited word filtering module 142 can extract text from the converted image data, identify the words contained in the extracted text, and determine whether the identified words correspond to prohibited words, so as to perform prohibited word filtering on fan messages.

[0261] The stopword filtering module 142 responds the fan message filtering result to the control unit 141. When the fan message does not include stopwords, the control unit 141 executes the fan message storage step S120 in the independent database.

[0262] In the stopword filtering step S510, the stopword filtering module 142 uses a CNN (Convolution Neural Network) learning model to convert the artist message data and the fan message data into image data.

[0263] Specifically, when the stopword filtering module 142 extracts the text corresponding to the image data, it can be respectively converted into Korean consonants and vowels. For example, when the artist message or the fan message includes contents such as "んЙ刀│○ㅑ" and "人ㅣ凹卜", the stopword filtering module 142 converts the message into an image and extracts the consonants "ㅅ, ㄲ, ㅇ" and "ㅅ, ㅂ", and the vowels "ㅐ, ㅣ, ㅑ" and "ㅣ, ㅏ". Then, it extracts "새끼야" and "시바" which are combinations of vowels and consonants.

[0264] In addition, the CNN can include a part for extracting image features by effectively identifying and strengthening adjacent image features while maintaining the spatial information of the image, and a part for classifying the image. The feature extraction part can include a convolution layer that uses filters to find image features while minimizing the number of shared parameters, and a pooling layer that strengthens and aggregates features. Based on the existing network, the CNN is re - learned for a new recognition task, so the CNN learning model can filter new malicious messages that are creatively deformed to avoid stopwords.

[0265] In addition, in the stopword filtering step S510, the stopword filtering module 142 can use a BERT (Bidirectional Encoder Representation of Transformer) learning model to filter whether the extracted text includes stopwords. The BERT learning model can be fine - tuned using the accuracy of the judgment result.

[0266] BERT is a pre - trained learning model used by Google to label and train a large number of articles on the Internet using semi - supervised learning. BERT is used in the embedding process when performing tasks such as object name recognition and text classification.

[0267] The method by which the stopword filtering module 142 determines whether stopwords are included includes: a method using an algorithm that labels messages containing preset vocabulary by users, and a method of making a judgment according to preset criteria.

[0268] The pre-defined standards can be categorized as follows: expressions containing general insults, vulgar and low-level expressions that cause discomfort to others, sexually provocative expressions, expressions involving physical threats, discriminatory expressions based on region / race / country / religion, and degrading expressions that cause the other party to feel humiliated and ashamed.

[0269] In the disabled word filtering step S510, the disabled word filtering module 142 outputs a value of "0" when the message contains abusive or defamatory expressions, and outputs a value of "1" when the message does not contain abusive or defamatory expressions, in order to filter the message.

[0270] When fan messages are stored in the independent database 1320, in the malicious message filtering step S520, the control unit 141 requests malicious message filtering from the malicious message filtering module 143. In the malicious message filtering step S520, the malicious message filtering module 143, having received the malicious message filtering request, classifies the context of the fan messages and determines whether the filtered fan messages are malicious based on the classified filtering criteria.

[0271] In the malicious message filtering step S520, the malicious message filtering module 143 analyzes the context of the text extracted by the banned word filtering module 142 and filters whether the fan message belongs to malicious messages based on the analyzed context.

[0272] To this end, the malicious message filtering module 143 performs encoding using bidirectional contextual analysis of fan messages to classify the context of fan messages.

[0273] In the malicious message filtering step S520, the malicious message filtering module 143 can input the fan message into the BERT (Bidirectional Encoder Representation of Transformer) learning model to filter whether the fan message belongs to malicious messages. Specifically, the BERT learning model used in the malicious message filtering module 143 obtains a single embedding input that integrates the word embedding, segment embedding, and position embedding for the fan message, and analyzes the context of the fan message to classify it as a malicious message.

[0274] In the malicious message filtering step S520, the malicious message filtering module 143 performs encoding using bidirectional contextual analysis of fan messages to classify the context of fan messages. Specifically, the malicious message filtering module 143 can input fan messages into a BERT (Bidirectional Encoder Representation of Transformer) learning model to filter whether fan messages belong to malicious messages. Specifically, the BERT learning model used by the malicious message filtering module 143 obtains input with a single embedding that integrates word embeddings, segment embeddings, and position embeddings for fan messages, and analyzes the context of fan messages.

[0275] In addition, the malicious message filtering module 143 uses Word Piece embedding, which segments words into sub-words, to prevent out-of-vocabulary words from constituting the fan message.

[0276] Specifically, the malicious message filtering module 143 can utilize Word Piece embedding to process word embeddings. In Word Piece embedding, embedding can be done in units smaller than a word, with the longest sub-word set as a unit. That is, common sub-words can be used as embedding units, while rare words can be segmented into sub-words. Existing word embedding methods suffer from out-of-vocabulary (OV) word problems, leading to difficulties in learning and translating rare words, proper nouns, numbers, or words not found in the vocabulary. Word embedding, however, is applicable to all languages, and by segmenting words into sub-word units, it effectively solves the OV word problem and improves accuracy.

[0277] Furthermore, in the malicious message filtering step S520, the malicious message filtering module 143 can receive the sentence separator [SEP] along with the two sentences. At this time, due to the input length limit, the total length of the two sentences needs to be controlled within 512 words. That is, since the learning time increases quadratically with the input length, the input length needs to be set reasonably. Korean is composed of an average of 20 words, and 99% of sentences do not exceed 60 words. Therefore, the two sentences are combined and limited to 128 words. However, considering the possibility of long sentences, the input length can be limited to 128 words for learning first, and long inputs greater than 128 words can be supplemented for learning in the final step.

[0278] Furthermore, in the malicious message filtering step S520, the malicious message filtering module 143 can use position encoding. For this, a Transformer model can be applied, which does not require a CNN or RNN model but can use a Self-Attention model. Since the Self-Attention model cannot reflect the position of the input, additional word embeddings with relevant positional information are needed. Therefore, in the Transformer model, positional encoding based on the Sinusoid function is used, and the malicious message filtering module 143 can transform it and use positional encoding. In positional encoding, it can simply be encoded according to the word (token) order, in the order of 0, 1, 2, ...

[0279] Furthermore, in the malicious message filtering step S520, the malicious message filtering module 143 can integrate word embeddings, fragment embeddings, and position embeddings to generate a single embedding value (i.e., a single embedding). At this time, the single embedding can be generated in the form of a feature map and then converted to an appropriate form according to the subsequent workflow. For example, the malicious message filtering module 143 can use the single embedding result after layer normalization and dropout processing as input.

[0280] Furthermore, in the malicious message filtering step S520, the malicious message filtering module 143 can provide the output malicious message encoding value to the fine-tuning network to improve the accuracy of malicious message judgment. That is, the malicious message filtering module 143 can receive the sentence separator [SEP] together with two sentences related to the user's message, and generate a malicious message encoding value about the probability of the malicious message as output through a step of internal reasoning constructed through pre-learning.

[0281] In the malicious message filtering step S520, the malicious message filtering module 143 can directly use the BERT learning model to perform malicious comment filtering by default, and then improve the accuracy of malicious message judgment through append learning. To this end, the malicious message filtering module 143 can define a combination of the BERT learning model and a fine-tuning network for the malicious message judgment process.

[0282] The malicious message filtering module 143 can provide the malicious message encoding value to a one-dimensional convolutional layer, or generate a feature map by calculating an ensemble based on a variable kernel size, and provide the feature map to a BiLSTM (Bidirectional Long Short-Term Memory) network to determine the likelihood of a malicious message targeting a follower.

[0283] Subsequently, in the malicious message filtering step S520, the malicious message filtering module 143 allows the output of the BiLSTM network to pass through an affine layer and a softmax layer, thereby performing a binary classification of the probability of malicious messages. The BiLSTM (Bidirectional Long Short-Term Memory) network and the bidirectional LSTM can be considered as structures sharing two independent LSTM architectures. First, the bidirectional LSTM can sequentially obtain the sentence input.

[0284] That is, just like humans, sentences are received sequentially from left to right. Furthermore, the bidirectional LSTM can be used in conjunction with a reverse LSTM that reads backward from right to left to further consider the context behind the sentence. During prediction at the output layer, the bidirectional LSTM connects the outputs of the forward and reverse LSTMs, thus utilizing both types of information simultaneously.

[0285] These bidirectional LSTMs can achieve end-to-end learning of all parameters simultaneously while minimizing output value loss, and internalize the similarity between words and phrases into the input vector to improve performance. Furthermore, in the case of bidirectional LSTMs, by incorporating the basic performance characteristics of LSTMs and the attention mechanism, they have the advantage of not degrading performance even with long datasets.

[0286] Furthermore, in the malicious message filtering step S520, the malicious message filtering module 143 can combine the BERT learning model and the CNN (Convolutional Neural Network) learning model to achieve malicious message filtering. Specifically, the malicious message filtering module 143 inputs the malicious message encoded value into a one-dimensional convolutional layer, inputs the output of the one-dimensional convolutional layer into a GeLU layer, inputs the output of the GeLU layer into a max pooling layer, and generates a binary classification result based on the output of the max pooling layer. At this time, the binary classification can use a linear classification method.

[0287] For example, linear classification methods may include linear regression or linear classification models.

[0288] Furthermore, in the malicious message filtering step S520, the malicious message filtering module 143 can combine the BERT learning model with the ensemble CNN model to achieve malicious message filtering. More specifically, the malicious message filtering module 143 can perform the following steps: First, perform ensemble computation based on the malicious message encoding value using a variable kernel size; Second, independently input the ensemble computation result into the GeLU layer; Third, independently input the output of the GeLU layer into the MaxPooling layer; Fourth, sequentially connect the outputs of the MaxPooling layer to generate intermediate values; Fifth, input the intermediate values ​​into the Affine layer; Sixth, input the output of the Affine layer into the softmax layer; Seventh, generate a binary classification result based on the softmax layer result.

[0289] The ensemble computation based on variable kernel size can be a series of convolutions performed by applying different kernel sizes to the same malicious message encoding value. Furthermore, the convolutions can include 1D convolution operations. That is, the malicious comment judgment unit 230 can independently apply convolution kernels of different sizes to the feature map containing the malicious comment encoding value, performing multiple convolution operations respectively. Then, the malicious message filtering module 143 performs a classification operation based on the malicious probability of the fan message by integrating the results of each convolution operation into a single result (i.e., the fourth step).

[0290] Furthermore, in the malicious message filtering step S520, the malicious message filtering module 143 can combine the BERT learning model with the ensemble CNN model and BiLSTM to achieve malicious message filtering. More specifically, the malicious message filtering module 143 can perform the following steps: First, perform ensemble computation based on the malicious comment encoding value using a variable kernel size; Second, independently input the ensemble computation result into the GeLU layer; Third, independently input the output of the GeLU layer into the max pooling layer; Fourth, sequentially connect the outputs of the max pooling layer to generate intermediate values; Fifth, apply the intermediate values ​​to the BiLSTM (Bidirectional LSTM (Long Short-Term Memory)) model; Sixth, input the application result of the BiLSTM model into the affine layer; Seventh, input the output of the affine layer into the softmax layer; Eighth, generate a binary classification result based on the softmax layer result.

[0291] That is, the malicious message filtering module 143 can achieve malicious message filtering by adding a step of applying intermediate values ​​to BiLSTM (Bidirectional LSTM (Long Short-Term Memory)) between the fourth and fifth steps of the method of combining the BERT learning model and the ensemble CNN model.

[0292] Furthermore, in the malicious message filtering step S520, in order to perform malicious message filtering, the control unit 141 derives a malicious message prediction score based on the frequency of banned words used by fans, the contextual information of fan messages, and the frequency of malicious message writing by fans.

[0293] In the malicious message filtering step S520, the control unit 141 may provide a fan message review request notification to the fan terminal 200 when the malicious message prediction score is greater than a preset first reference score. Furthermore, the control unit 141 may provide a fan message stop sending notification to the fan terminal when the malicious message prediction score is greater than a preset second reference score. In this case, the second reference score is set to a value higher than the first reference score.

[0294] When the malicious message filtering results indicate that the fan message is not a malicious message, the message push stream provision step S130 is executed.

[0295] Below, in conjunction with Figure 19 This section details the methods for filtering artist messages.

[0296] Figure 19 This is a sequence diagram of the artist message filtering method according to an embodiment of the present invention.

[0297] See Figure 19 In the case of artist message filtering methods, it is possible to Figure 11 The artist message storage method shown further includes a banned word filtering step S511.

[0298] In the banned word filtering step S511, when the control unit 141 receives artist messages from the artist terminal 200, the control unit 141 requests the banned word filtering module 142 to filter banned words in the artist messages.

[0299] In the banned word filtering step S511, the banned word filtering module 142, which receives the banned word filtering request, checks whether the artist's fan message contains banned words based on the artist's banned words set by the artist who wrote the message.

[0300] As described above, the banned word filtering module 142 can convert artist messages into images. The banned word filtering module 142 can extract text from the converted image data, identify the words contained in the extracted text, and determine whether the identified words correspond to banned words, so as to filter the artist messages for banned words.

[0301] The banned word filtering module 142 responds to the artist message filtering result to the control unit 141. When the artist message does not contain banned words, the control unit 141 performs the following steps: storing the artist message in the independent database 1320 S220; storing the artist message in the database 1310 of each DM S230; generating a private message S240; and storing the private message in the independent database 1321 S250.

[0302] In the foregoing description of the present invention, artist terminal 200-1 and fan terminal 200-2 can correspond to a terminal logged in using an artist account and a terminal logged in using a fan account, respectively.

[0303] The above embodiments are merely illustrative of the invention and not intended to limit it. Those skilled in the art should understand that modifications, variations, or equivalent substitutions can be made to the invention. Such modifications, variations, or equivalent substitutions should all be covered within the scope of the claims without departing from the spirit and scope of the invention. For example, the structural elements of a single type can be implemented separately; similarly, the dispersed structural elements can be implemented in combination.

[0304] The scope of this invention is defined more by the scope of the claims than by the description, and all modifications or variations derived from the meaning, scope and equivalent concepts of the claims should be interpreted as falling within the scope of this invention. Detailed Implementation

[0306] The present invention can be implemented in the same manner as the preferred embodiment described herein.

[0307] Industrial applicability

[0308] This invention relates to a method and apparatus for providing private chat services, which can be applied to the entertainment industry and communication-related industries, thus having industrial applicability.

Claims

1. A method for providing a private chat service, characterized in that, a method for relaying messages between an artist's terminal and multiple fan terminals in a private chat service relay server to provide a private chat service, wherein... include: Steps for receiving artist messages from an artist's terminal; The step of filtering whether the artist messages contain preset banned words for artists; When the artist message does not contain the artist's banned words, the step of storing the artist message and artist information in the database of each DM; The step of receiving a private chat message update request from the first fan terminal that requests to query the artist's message from the plurality of fan terminals; The step of generating a private message corresponding to the first fan terminal based on the fan account information corresponding to the first fan terminal and the artist message; as well as The step of transmitting the private message to the first fan terminal. The database configuration of each DM is to store messages by artist unit, and the private message refers to the message after specific words or keywords are converted according to the recipient.

2. The method for providing private chat service according to claim 1, characterized in that, The steps for filtering whether the artist's messages contain banned words for the artist include: The step of converting the artist's message into an image; The step of extracting text based on the image; and The step of determining whether the text filter includes the artist's banned words.

3. The method for providing private chat service according to claim 2, characterized in that, The steps to convert the artist's message into an image include: The steps for converting the artist's message into an image using a CNN learning model.

4. The method for providing private chat service according to claim 3, characterized in that, The steps for filtering whether the artist's banned words are included include: The steps of using the BERT learning model to filter whether the text contains the artist's banned words.

5. The method for providing private chat service according to claim 4, characterized in that, The steps of storing the artist messages and artist information in the databases of each DM include: The step of matching the multiple independent databases corresponding to each of the multiple fan terminals; The steps of generating multiple private messages corresponding to each of the multiple fan terminals based on fan account information and artist messages; and The step of storing the multiple private messages in the independent database corresponding to the fan's terminal based on the fan account information.

6. The method for providing private chat service according to claim 5, characterized in that, The steps for transmitting the private message to the first fan terminal include: The step of deleting artist messages stored in the databases of each DM when the multiple private messages have been stored in the independent databases; and The steps of receiving private messages from a first independent database corresponding to the first fan terminal and transmitting them to the first fan terminal.

7. The method for providing private chat service according to claim 6, characterized in that, The steps for transmitting the private message to the first fan terminal include: The steps include: storing artist messages and information in the databases of each DM and deleting the artist messages stored in the databases of each DM after a preset base time; and... The steps of receiving private messages from the first independent database and transmitting them to the first fan terminal.

8. A method for providing a private chat service, characterized in that, a method for relaying messages between an artist's terminal and multiple fan terminals in a private chat service relay server to provide a private chat service, wherein... include: The step of matching the multiple independent databases corresponding to each of the multiple fan terminals; The steps to receive fan messages from the first fan terminal out of multiple fan terminals; The step of filtering whether the fan messages contain preset banned words for fans; The step of storing the fan messages in the first independent database corresponding to the first fan terminal among the multiple independent databases; The step of classifying the context of the fan messages using a learning-based artificial intelligence model; The steps to determine whether a fan message is malicious based on the contextual classification after categorization; The steps of receiving artist messages from the artist's terminal; The step of storing the artist messages and artist information in the database of each DM; The steps of receiving private chat message update requests from the first fan terminal; The step of generating a private message corresponding to the first fan terminal based on the fan account information corresponding to the first fan terminal and the artist message; as well as The step of transmitting the fan messages and the private messages stored in the first independent database to the first fan terminal. The database configuration of each DM is to store messages by artist unit, and the private message refers to the message after specific words or keywords are converted according to the recipient.

9. The method for providing private chat service according to claim 8, characterized in that, The steps for filtering whether the fan messages contain banned words include: The step of converting the fan message into an image; The step of extracting text based on the image; and The step of determining whether the text filter includes the banned words for fans.

10. The method for providing private chat service according to claim 9, characterized in that, The steps to convert the fan message into an image include: The steps for converting the fan messages into images using a CNN learning model.

11. The method for providing private chat service according to claim 10, characterized in that, The steps for filtering whether the banned words are included include: The steps of using the BERT learning model to filter whether the text contains the banned words for fans.

12. The method for providing private chat service according to claim 9, characterized in that, The steps for classifying the context of the fan messages include: The step involves performing encoding that utilizes bidirectional contextual analysis of the fan messages to classify the context of the fan messages.

13. The method for providing private chat service according to claim 12, characterized in that, The step of classifying the context of the fan messages further includes: The process of encoding through the bidirectional context analysis involves integrating single embeddings of word embeddings, fragment embeddings, and positional embeddings into the BERT training model.

14. The method for providing private chat service according to claim 11 or 13, characterized in that, The steps for classifying fan messages as malicious include: The steps for deriving a malicious message prediction score based on the frequency of banned words used by the fans, the contextual information of the fans' messages, and the frequency of malicious messages written by the artist. The step of providing a review request notification for the fan message when the malicious message prediction score is greater than a preset first reference score; and When the malicious message prediction score is greater than a preset second reference score, a step is provided to stop sending the fan message.

15. The method for providing private chat service according to claim 14, characterized in that, The steps for transmitting the private message to the first fan terminal include: The step of deleting artist messages stored in the databases of each DM when the private messages have been stored in the independent database; and The steps of receiving the private message from the first independent database and transmitting it to the first fan terminal.

16. The method for providing private chat service according to claim 15, characterized in that, The steps for transmitting the private message to the first fan terminal include: The steps include: storing artist messages and information in the databases of each DM and deleting the artist messages stored in the databases of each DM after a preset base time; and... The steps of receiving private messages from the first independent database and transmitting them to the first fan terminal.

17. A private chat service providing apparatus, for relaying messages between an artist's terminal and multiple fan terminals in a private chat service relay server to provide a private chat service, characterized in that, include: The communication module performs the sending and receiving of information between the terminal and the private chat service relay server; The storage device contains the private chat program; Each DM's database stores messages according to artist unit; Multiple independent databases are assigned to each user and store each user's messages; as well as The processor executes the private chat program stored in the memory. The processor receives fan messages from a first fan terminal among multiple fan terminals, and stores the fan messages in a first independent database corresponding to the first fan terminal among the multiple independent databases. It also receives artist messages from an artist terminal, filtering the fan messages and artist messages to determine if they contain preset prohibited fan and artist words. If the fan messages do not contain the prohibited fan words, the processor stores the artist messages and artist information in the databases of each DM (Distributor Manager). When receiving a private chat message update request from the first fan terminal, the processor generates a private message corresponding to the first fan terminal based on the fan account information corresponding to the first fan terminal and the artist message. Finally, the processor transmits the fan messages and the private message stored in the first independent database to the first fan terminal. The private message refers to a message whose specific words or keywords have been transformed according to the recipient.

Images