The Method And System For Social Language Conversion Based On Conversation Context using A Large Language Model

Abstract

The present invention relates to a conversation context-based social language conversion method and system using a large language model, and more specifically, to a conversation context-based social language conversion method and system using a large language model that, when a user inputs text, extracts the current situation, characteristics of the conversation, and the content of the previous conversation from the user's conversation history, and derives text converted through a large language model based on the extracted information.

Description

Technology Field

[0001] The present invention relates to a conversation context-based social language conversion method and system using a large language model, and more specifically, to a conversation context-based social language conversion method and system using a large language model that, when a user inputs text, extracts the current situation, characteristics of the conversation, and the content of the previous conversation from the user's conversation history, and derives text converted through a large language model based on the extracted information. Background Technology

[0003] With the development of digital communication today, text-based conversations via messengers, email, and social media are rapidly increasing, and non-face-to-face communication among users of various ages, genders, and social statuses is becoming commonplace. In this environment, while it is important to use appropriate tone or expressions depending on the sender's intent or the social context, in reality, such social language adjustments are frequently difficult to make or are overlooked.

[0004] In particular, in situations where the tone or formality of expression must be adjusted based on factors such as age differences, level of familiarity, gender, and the nature of the relationship, users must not only make direct judgments and modify the text themselves but also face the risk of making social errors. For instance, the use of honorifics considering hierarchical relationships, the appropriateness of emotional expression, and the expression of humor or emotional intimacy must be flexibly adjusted according to the context and the conversation partner.

[0005] Recently, advancements in natural language processing technologies and Large Language Models (LLMs) have enabled context-aware natural language generation, and research is underway to utilize these capabilities to automatically transform text style and tone. However, conventional technologies are mostly limited to simply altering sentence structure or tone, failing to adequately reflect the social relationship between the sender and receiver, the flow of the conversation, or emotional states.

[0006] In this context, there is a need for technology capable of transforming text into a natural and socially appropriate form based not only on personal information such as the user's gender and age but also on various social contextual elements derivable from the user's conversation history; such technology can improve the efficiency and quality of non-face-to-face communication. The problem to be solved

[0008] The present invention aims to provide a method and system for conversational context-based social language transformation using a large language model, and more specifically, to provide a method and system for conversational context-based social language transformation using a large language model that, when a user inputs text, extracts the current situation, characteristics of the conversation, and the content of the previous conversation from the user's conversation history, and derives text transformed through the large language model based on the extracted information. means of solving the problem

[0010] To solve the above-mentioned problem, in one embodiment of the present invention, a conversation context-based social language conversion method using a large language model through a service server comprising one or more processors and one or more memories comprises: a first text receiving step in which user information including the age and gender of a user received from a user terminal is stored in the service server, and a first text input from a conversion interface executed on the user terminal is received; a parameter derivation step in which conversation history information from a messenger app executed on the user terminal is collected, and the conversation history information is input into a parameter derivation model to derive a plurality of parameters that quantify the characteristics of the conversation by item; and a final context prompt generation step in which a final context prompt is generated by applying the user information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template. A social language conversion method is provided, comprising: a second text derivation step of deriving a second text converted from the first text based on response information output by inputting a final prompt including the first text, the final context prompt, and a preset conversion request prompt into a large language model outside or inside a service server; wherein the second text is displayed on the conversion interface and the second text is transmitted to the messenger app according to a selection input from a user terminal.

[0011] In one embodiment of the present invention, the parameter derivation step quantifies the characteristics of each item, including the ratio of expressions of hierarchical relationships, joke frequency, frequency of expressions of personal emotions, and emotional intimacy, based on the conversation text included in the conversation history information through the parameter derivation model and derives them as parameters, and detects sentence-ending morphemes, titles, and demonstrative vocabulary in the conversation text. It may include: a first parameter derivation step for deriving a first parameter for the ratio of the above-mentioned hierarchical relationship expressions; a second parameter derivation step for deriving a second parameter for the above-mentioned joke frequency by detecting phonological features, exaggerated expressions, and laughter expressions in the above-mentioned conversation text; a third parameter derivation step for deriving a third parameter for the above-mentioned personal emotion expression frequency by detecting emotional vocabulary and emotional intensity in the above-mentioned conversation text; and a fourth parameter derivation step for deriving a fourth parameter for the above-mentioned emotional intimacy by detecting language style, emotional synchronization, conversation frequency, and degree of information exchange in the above-mentioned conversation text.

[0012] In one embodiment of the present invention, the final context prompt generation step comprises: a current situation information derivation step that derives current situation information including the age difference, relationship, title, and personality of the other party and the user based on the conversation text included in the conversation history information; and a context application step that generates a final context prompt by applying the user information, the current situation information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template; wherein the current situation information derivation step derives the age difference, relationship, and title of the other party and the user by detecting honorifics and titles in the conversation text, and derives the personality of the other party based on conversation pattern indicators including emotional vocabulary, frequency of use of positive and negative words, sentence length, and conversation participation in the conversation text using a text-based personality analysis model.

[0013] In one embodiment of the present invention, the second text derivation step may use an embedding model to derive embedding vectors for each of the nouns and emotion vocabulary included in the first text and the conversation history information, derive an emotion type based on the similarity between a predefined emotion template and the embedding vectors, and generate a transformation request prompt including text requesting the derivation of response information corresponding to the emotion type, and input it into the large language model.

[0014] In one embodiment of the present invention, the second text derivation step derives an emotion type for the first text using an emotion analysis model and generates a plurality of candidate second texts with different styles and vocabulary intensities according to a plurality of emotion intensities for the emotion type, the plurality of candidate second texts are displayed on the conversion interface, and one of the plurality of candidate second texts can be transmitted to the messenger app according to a selection input from a user terminal.

[0015] To solve the above-mentioned problem, in one embodiment of the present invention, a conversation context-based social language conversion method using a large language model performed on a user terminal comprising one or more processors and one or more memories comprises: a first text receiving step in which user information including the age and gender of the user is stored on the user terminal and a first text input from a conversion interface displayed on the user terminal is received; a parameter derivation step in which conversation history information is collected from a messenger app of the user terminal and the conversation history information is input into a parameter derivation model to derive a plurality of parameters that quantify the characteristics of the conversation by item; and a final context prompt generation step in which a final context prompt is generated by applying the user information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template. A social language conversion method is provided, comprising: a second text derivation step of deriving a second text converted from the first text based on response information output by inputting a final prompt including the first text, the final context prompt, and a preset conversion request prompt into a large language model outside or inside a user terminal; and displaying the second text on the conversion interface and transmitting the second text to the messenger app according to a selection input from the user terminal. Effects of the invention

[0017] According to one embodiment of the present invention, when a user inputs a first text into a conversion interface, the user can receive a second text that has been automatically converted, thereby achieving the effect of being able to receive a second text that has been automatically converted.

[0018] According to one embodiment of the present invention, the effect of deriving a second text by transforming a first text by reflecting current situation information derived from conversation history information, a plurality of parameters, and the previous conversation text can be achieved.

[0019] According to one embodiment of the present invention, a parameter derivation model can be used to derive a plurality of parameters that quantify the characteristics of a conversation in the conversation history information by item.

[0020] According to one embodiment of the present invention, the effect of automatically deriving a second text by converting a first text through a large language model can be achieved.

[0021] According to one embodiment of the present invention, the derived second text is transmitted to a messenger app of a user terminal, thereby enabling the user to conveniently use social language conversion in the messenger app.

[0022] According to one embodiment of the present invention, it is possible to achieve the effect of performing social language conversion that comprehensively considers social context and conversational context, going beyond simple text conversion.

[0023] According to one embodiment of the present invention, by reflecting complex social elements such as hierarchical relationships, emotional expressions, and intimacy extracted from the conversation history of an actual messenger app into a first text entered by a user, it is possible to derive a second text that is suitable for the conversation context and natural.

[0024] According to one embodiment of the present invention, by deriving a second text appropriate to the situation that reflects the relationship and emotional state between the other party and the user, it is possible to reduce the possibility of social errors or misunderstandings and to achieve the effect of supporting smooth communication.

[0025] According to one embodiment of the present invention, by providing a plurality of candidate second texts based on emotional intensity or expression style, the effect of providing a user-customized second text can be achieved.

[0026] According to one embodiment of the present invention, since it can provide a social language conversion function in real time by linking with various platforms including messenger apps, it can achieve the effect of providing various scalability and user accessibility. Brief explanation of the drawing

[0028] FIG. 1 schematically illustrates the internal configuration of a service server according to one embodiment of the present invention. FIG. 2 schematically illustrates the steps of performing a social language conversion method according to one embodiment of the present invention. FIG. 3 schematically illustrates the process of performing a social language conversion method according to one embodiment of the present invention. FIG. 4 schematically illustrates a conversion interface according to one embodiment of the present invention. FIG. 5 schematically illustrates the process of performing a parameter derivation step according to one embodiment of the present invention. FIG. 6 schematically illustrates the process of performing the final context prompt generation step according to one embodiment of the present invention. FIG. 7 schematically illustrates the process of performing the final context prompt generation step according to one embodiment of the present invention. FIG. 8 schematically illustrates the process of performing the second text derivation step according to one embodiment of the present invention. FIG. 9 schematically illustrates the process of performing a social language conversion method according to one embodiment of the present invention. FIG. 10 illustrates the internal configuration of a computing device according to one embodiment of the present invention. Specific details for implementing the invention

[0029] Hereinafter, various embodiments and / or aspects are disclosed with reference to the drawings. For illustrative purposes, numerous specific details are disclosed in the following description to aid in a general understanding of one or more aspects. However, it will also be recognized by those skilled in the art that these aspects may be practiced without such specific details. The following description and the accompanying drawings describe specific exemplary aspects of one or more aspects in detail. However, these aspects are exemplary, and some of the various methods in the principles of the various aspects may be used, and the description is intended to include all such aspects and their equivalents.

[0030] Additionally, terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but said components are not limited by said terms. Such terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. The term "and / or" includes a combination of a plurality of related described items or any of a plurality of related described items.

[0031] Furthermore, in the embodiments of the present invention, all terms used herein, including technical or scientific terms, unless otherwise defined, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in the embodiments of the present invention.

[0032] The "user terminal" mentioned below may be implemented as a computer or portable terminal capable of connecting to a server or other terminal via a network. Here, the computer includes, for example, a notebook, desktop, or laptop equipped with a web browser, and the portable terminal may include, for example, all types of handheld-based wireless communication devices that ensure portability and mobility, such as smartphones, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet), and BLE Beacon (Bluetooth Low Energy Beacon) terminals. In addition, the “network” can be implemented as a wired network such as a Local Area Network (LAN), Wide Area Network (WAN), or Value Added Network (VAN), or as any type of wireless network such as a mobile radio communication network or a satellite communication network.

[0034] FIG. 1 schematically illustrates the internal configuration of a service server (1000) according to one embodiment of the present invention.

[0036] As illustrated in FIG. 1, the service server (1000) comprises: a first text receiving unit (100) that performs a first text receiving step of receiving a first text input from a conversion interface executed on a user terminal (2000); a parameter derivation unit (200) that performs a parameter derivation step of collecting conversation history information from a messenger app executed on the user terminal (2000) and inputting the conversation history information into a parameter derivation model to derive a plurality of parameters that quantify the characteristics of the conversation by item; and a final context prompt generation unit (300) that performs a final context prompt generation step of generating a final context prompt by applying the user information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template. The system includes a second text derivation unit (400) that performs a second text derivation step of deriving a second text converted from the first text based on response information output by inputting a final prompt, which includes the first text, the final context prompt, and a preset conversion request prompt, into a large language model (3000) outside or inside the service server (1000).

[0038] Specifically, each component included in the service server (1000) illustrated in FIG. 1 performs the role of controlling the operation of the service server (1000) that performs the social language conversion method of the present invention.

[0039] More specifically, the first text receiving unit (100) of the service server (1000) can receive the first text input from a conversion interface executed on a user terminal (2000).

[0040] In one embodiment of the present invention, the conversion interface includes an interface that provides a converted second text when a user inputs a first text to be converted into a social language, and can be displayed on the user terminal (2000) in the form of an app that is linked with the service server (1000) and executed on the user terminal (2000).

[0042] The parameter derivation unit (200) of the above service server (1000) can collect conversation history information from a messenger app running on a user terminal (2000) and input the conversation history information into a parameter derivation model to derive multiple parameters that quantify the characteristics of the conversation by item.

[0043] In one embodiment of the present invention, the parameter derivation step can derive parameters by quantifying the characteristics of each item, including the ratio of expressions of hierarchical relationships, the frequency of jokes, the frequency of expressions of personal emotions, and emotional intimacy, based on the conversation text included in the conversation history information through the parameter derivation model.

[0044] At this time, the parameter derivation step may include: a first parameter derivation step for deriving a first parameter for the ratio of hierarchical relationship expressions by detecting sentence-ending morphemes, titles, and demonstrative vocabulary in the conversation text; a second parameter derivation step for deriving a second parameter for joke frequency by detecting phonological features, exaggerated expressions, and laughter expressions in the conversation text; a third parameter derivation step for deriving a third parameter for the frequency of personal emotion expressions by detecting emotion vocabulary and emotion intensity in the conversation text; and a fourth parameter derivation step for deriving a fourth parameter for emotional intimacy by detecting language style, emotional synchronization, conversation frequency, and degree of information exchange in the conversation text.

[0046] The final context prompt generation unit (300) of the service server (1000) can generate a final context prompt by applying the user information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a pre-configured context prompt template.

[0047] In one embodiment of the present invention, the final context prompt generation step may include: a current situation information derivation step that derives current situation information including the age difference, relationship, title, and personality of the counterpart and the user based on the conversation text included in the conversation history information; and a context application step that generates a final context prompt by applying the user information, the current situation information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template.

[0048] At this time, the current situation information derivation step can derive the age difference, relationship, and title between the other party and the user by detecting honorifics and titles in the conversation text, and derive the personality of the other party based on conversation pattern indicators including emotional vocabulary, frequency of use of positive and negative words, sentence length, and conversation participation in the conversation text using a text-based personality analysis model.

[0050] The second text extraction unit (400) of the service server (1000) can derive a second text converted from the first text based on response information output by inputting a final prompt including the first text, the final context prompt, and a preset conversion request prompt into a large language model (3000) outside or inside the service server (1000).

[0051] In one embodiment of the present invention, the second text derivation step may use an embedding model to derive embedding vectors for each of the nouns and emotion vocabulary included in the first text and the conversation history information, derive an emotion type based on the similarity between a predefined emotion template and the embedding vectors, and generate a conversion request prompt including text requesting to derive response information corresponding to the emotion type, and input it into the large language model (3000).

[0052] Additionally, the second text derivation step derives an emotion type for the first text using an emotion analysis model and generates multiple candidate second texts with different styles and vocabulary intensities according to multiple emotion intensities for the emotion type, the multiple candidate second texts are displayed on the conversion interface, and one of the multiple candidate second texts can be transmitted to the messenger app according to a selection input from the user terminal (2000).

[0054] Preferably, the second text is subsequently displayed on the conversion interface, and the second text can be transmitted to the messenger app according to a selection input at the user terminal (2000).

[0056] In another embodiment of the present invention, user information including the age and gender of the user is stored in the user terminal (2000), and the user terminal (2000) performs a first text receiving unit (100) that performs a first text receiving step of receiving a first text input from a conversion interface displayed on the user terminal (2000); a parameter derivation unit (200) that performs a parameter derivation step of collecting conversation history information from a messenger app of the user terminal (2000) and inputting the conversation history information into a parameter derivation model to derive a plurality of parameters that quantify the characteristics of the conversation by item; and a final context prompt generation unit (300) that performs a final context prompt generation step of generating a final context prompt by applying the user information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template. The second text derivation unit (400) performs a second text derivation step of deriving a second text converted from the first text based on response information output by inputting a final prompt including the first text, the final context prompt, and a preset conversion request prompt into a large language model (3000) outside or inside the user terminal (2000).

[0057] In addition, the second text can be displayed on the conversion interface, and the second text can be transmitted to the messenger app according to the selection input on the user terminal (2000).

[0059] At this time, each component included in the user terminal (2000) performs the role of controlling the operation of the user terminal (2000) that performs the social language conversion method of the present invention, and the social language conversion method is performed as a language conversion app in the form of an application executed on the user terminal (2000), and the conversion interface can be displayed on the user terminal (2000).

[0061] FIG. 2 schematically illustrates the steps of performing a social language conversion method according to one embodiment of the present invention.

[0063] As illustrated in FIG. 2, the social language conversion method has user information including the age and gender of a user received from a user terminal (2000) stored in a service server (1000), and the service server (1000) receives (S10) a first text input from a conversion interface executed on the user terminal (2000), collects conversation history information from a messenger app executed on the user terminal (2000), and inputs the conversation history information into a parameter derivation model to derive (S11) a plurality of parameters that quantify the characteristics of the conversation by item.

[0064] Preferably, the service server (1000) can periodically collect conversation history information from the messenger app and store it in the service server (1000).

[0065] At this time, a first parameter regarding the ratio of hierarchical relationship expressions can be derived by detecting sentence-ending morphemes, titles, and demonstrative vocabulary in the above dialogue text; a second parameter regarding joke frequency can be derived by detecting phonological features, exaggerated expressions, and laughter expressions in the above dialogue text; a third parameter regarding the frequency of personal emotion expressions can be derived by detecting emotion vocabulary and emotion intensity in the above dialogue text; and a fourth parameter regarding emotional intimacy can be derived by detecting language style, emotional synchronization, conversation frequency, and degree of information exchange in the above dialogue text.

[0067] In addition, a final context prompt (S12) can be generated by applying the user information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template.

[0068] In one embodiment of the present invention, based on the conversation text included in the conversation history information, current situation information including the age difference, relationship, title, and personality of the other party and the user is derived, and then a final context prompt can be generated by applying the user information, the current situation information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template.

[0070] Afterwards, a second text converted from the first text can be derived (S13) based on the response information output by inputting a final prompt including the first text, the final context prompt, and a preset conversion request prompt into a large language model (3000) outside or inside the service server (1000).

[0071] At this time, a conversion request prompt containing text requesting the derivation of embedding vectors for each of the nouns and emotion vocabulary included in the first text and the conversation history information using an embedding model, the derivation of an emotion type based on the similarity between a predefined emotion template and the embedding vector, and the derivation of response information corresponding to the emotion type can be generated and input into the large language model (3000).

[0073] Preferably, the second text is displayed on the conversion interface, and the second text can be transmitted to the messenger app according to a selection input at the user terminal (2000).

[0074] At this time, an emotion type for the first text is derived using an emotion analysis model, and multiple candidate second texts are generated by adjusting the intensity of different writing styles and vocabulary according to multiple emotional intensities for the emotion type, and the multiple candidate second texts are displayed on the conversion interface, and one of the multiple candidate second texts can be transmitted to the messenger app according to a selection input from the user terminal (2000).

[0076] FIG. 3 schematically illustrates the process of performing a social language conversion method according to one embodiment of the present invention.

[0078] Schematically, FIG. 3(a) illustrates the process of performing a social language conversion method, and FIG. 3(b) illustrates the process of performing a step of converting a first text into a second text and displaying it.

[0080] As illustrated in FIG. 3(a), the social language conversion method can be performed in a service server (1000) comprising one or more processors and one or more memories, and as illustrated in FIG. 3(a), the service server (1000) can communicate with a user terminal (2000) and a large language model (3000) to perform the social language conversion method.

[0081] At this time, the service server (1000) can collect a first text and conversation history information from a user terminal (2000), input a final prompt generated based on the first text and conversation history information into a large language model (3000), receive response information or a second text from the large language model (3000), and transmit the second text to the user terminal (2000).

[0083] In one embodiment of the present invention, the connection configuration of the service server (1000) and the user terminal (2000) may include a form of accessing the service server (1000) through the user terminal (2000), which includes a computer or portable terminal used by the user, and a form of accessing the service server through a language conversion app installed on the user terminal (2000). The large language model (3000) is located outside or inside the service server (1000).

[0085] In another embodiment of the present invention, the social language conversion method may be performed on a user terminal (2000) comprising one or more processors and one or more memories, and the user terminal (2000) may communicate with a large language model (3000) and perform the social language conversion method.

[0086] Accordingly, the user terminal (2000) inputs a final prompt generated based on the first text entered by the user and conversation history information into the large language model (3000), receives response information or a second text from the large language model (3000), and can display the second text on the conversion interface of the user terminal (2000).

[0087] At this time, the social language conversion method can be performed through a language conversion app installed on the user terminal (2000), and the large language model (3000) can be located outside the user terminal (2000).

[0089] As illustrated in FIG. 3(b), when a user inputs a first text to be converted into a conversion interface running on a user terminal (2000), a second text converted from the first text can be displayed in the conversion interface.

[0090] If the user inputs a first text, "I'm in a bad mood today. Just leave me alone.", a second text, "Today is a bit of a tough day. I need some time alone," which is converted based on the first text and conversation history information in the messenger app of the user terminal (2000), may be displayed.

[0091] Preferably, when a second text is displayed on the conversion interface, the second text is transmitted to the messenger app according to the user's selection input, so that the user can conveniently proceed with a conversation through the messenger app.

[0093] Accordingly, the social language conversion method of the present invention can provide a service that, when a user inputs content expressed in their own natural language, converts it into a form of communication that is more socially appropriate, polite, and effective while maintaining the essence and intent of the content through artificial intelligence.

[0095] FIG. 4 schematically illustrates a conversion interface according to one embodiment of the present invention.

[0097] In summary, a first text input can be received from a conversion interface executed on a user terminal (2000) through the first text receiving step, and after the second text is derived, the second text is displayed on the conversion interface, and the second text can be transmitted to the messenger app according to a selection input on the user terminal (2000).

[0099] As illustrated in FIG. 4, the conversion interface may be provided to the user in the form of an app executed on the user terminal (2000), displayed on the user terminal (2000) in the form of a separate app, or displayed on the user terminal (2000) in the form of a widget so as to be used on a messenger app running on the user terminal (2000).

[0100] Preferably, an E1 element can be displayed on the messenger app shown in FIG. 4, and when a user selects and inputs the E1 element, an L3 layer can be displayed overlaid on the messenger app in the form of a widget.

[0101] Accordingly, the L1 and L2 layers correspond to a messenger app running on a user terminal (2000), and the E1 element and L3 layer may correspond to a conversion interface provided through the present invention. At this time, the conversation history information may include all conversation texts displayed on the L1 layer.

[0103] In one embodiment of the present invention, the messenger app may include an L1 layer for displaying conversation history and an L2 layer for inputting and transmitting conversations. When a user wishes to convert language using the social language conversion method of the present invention while conversing with another party through the messenger app, the user terminal (2000) may select and input an E1 element to display an L3 layer on the messenger app, and when a first text to be converted is input into the L3 layer, the converted second text may be displayed on the L3 layer.

[0104] At this time, the user can transmit the second text to the messenger app based on the selection input at the L3 layer, and in this case, the second text may be automatically displayed on the L2 layer of the messenger app. Alternatively, the user can transmit the second text to the other party by copying the second text displayed on the L3 layer and pasting it into the L2 layer of the messenger app.

[0105] Accordingly, when a user has a conversation with another person in an existing messenger app running on a user terminal (2000), the user can quickly receive the second text converted from the first text through the social language conversion method provided in the form of a widget, and conveniently send the second text to the existing messenger app and proceed with a conversation using the converted second text simply by selecting input.

[0107] Preferably, the messenger app may include KakaoTalk, and the service server (1000) may collect conversation history information from one or more messenger apps running on a user terminal (2000), including KakaoTalk.

[0109] In another embodiment of the present invention, the user terminal (2000) may include a smartphone and a PC, and in the case of a smartphone, a conversion interface may be displayed by overlaying it on a messenger app in the form of a widget, and in the case of a PC, the messenger app and the conversion interface may be displayed in a left-right screen split form.

[0111] FIG. 5 schematically illustrates the process of performing a parameter derivation step according to one embodiment of the present invention.

[0113] In summary, the parameter derivation step may collect conversation history information from a messenger app running on a user terminal (2000) and input the conversation history information into a parameter derivation model to derive multiple parameters that quantify the characteristics of the conversation by item.

[0115] Specifically, the parameter derivation step can derive parameters by quantifying the characteristics of each item, including the ratio of expressions of hierarchical relationships, the frequency of jokes, the frequency of expressions of personal emotions, and emotional intimacy, based on the conversation text included in the conversation history information through the parameter derivation model.

[0116] More specifically, the parameter derivation step may include: a first parameter derivation step for deriving a first parameter for the ratio of hierarchical relationship expressions by detecting sentence-ending morphemes, titles, and demonstrative vocabulary in the conversation text; a second parameter derivation step for deriving a second parameter for joke frequency by detecting phonological features, exaggerated expressions, and laughter expressions in the conversation text; a third parameter derivation step for deriving a third parameter for the frequency of personal emotion expressions by detecting emotion vocabulary and emotion intensity in the conversation text; and a fourth parameter derivation step for deriving a fourth parameter for emotional intimacy by detecting language style, emotional synchronization, conversation frequency, and degree of information exchange in the conversation text.

[0118] As illustrated in FIG. 5, the service server (1000) can periodically collect conversation history information from a messenger app running on a user terminal (2000), and input the conversation text included in the conversation history information into a parameter derivation model to derive a plurality of parameters that quantify the characteristics of the conversation by item.

[0119] At this time, the plurality of parameters may include a first parameter for the ratio of expression of hierarchical relationships, a second parameter for the frequency of jokes, a third parameter for the frequency of expression of personal emotions, and a fourth parameter for emotional intimacy.

[0120] In addition, by deriving multiple parameters that quantify social and emotional characteristics and reflecting them in language conversion, the said multiple parameters can play a key role in finely adjusting the nuances of the conversation.

[0122] Preferably, the language conversion app displaying the conversion interface may periodically collect conversation history information from the messenger app of the user terminal (2000) and transmit it to the service server (1000), or the service server (1000) may access the messenger app and collect conversation history information through an API.

[0124] In addition, the above parameter derivation model may correspond to a model that extracts characteristics of a conversation based on input text and derives parameters by quantifying each characteristic, including the ratio of expressions of hierarchical relationships, joke frequency, frequency of expressions of personal emotions, and emotional intimacy, into a value between 0 and 100.

[0125] For example, based on the input conversation text, the parameter derivation model can derive a ratio of expressions of hierarchical relationships of 70%, a joke frequency of 15%, a personal emotion expression frequency of 20%, and an emotional intimacy of 80%. It can be determined that the higher the parameter value, the more the corresponding item is detected in the conversation text, and the more prominent the characteristics of that item are in the conversation text.

[0127] In one embodiment of the present invention, the parameter derivation step may detect sentence-ending morphemes, titles, and demonstrative vocabulary in the conversation text. For example, by detecting speech styles such as "do it," "please do it," and "I understand" in the sentence-ending morphemes, it may determine whether informal or formal speech is being used, and by detecting titles and demonstrative vocabulary in the conversation text, it may determine the frequency of use of the corresponding titles and demonstrative vocabulary.

[0129] In addition, phonological features, exaggerated expressions, and laughter expressions can be detected in the aforementioned dialogue text. For example, unexpected word combinations or the juxtaposition of contradictory concepts within a sentence can be identified, and linguistic play elements can be extracted by analyzing phonological features such as alliteration, rhyme, and wordplay. Alternatively, the presence and frequency of use of exaggerated expressions or laughter expressions such as "lol" and "haha" can be determined.

[0131] In addition, emotional vocabulary and emotional intensity can be detected in the above dialogue text. For example, the presence and frequency of use of emotional vocabulary in the dialogue text can be identified, and the type and intensity of emotion can be determined accordingly. Based on the emotional vocabulary in the dialogue text, emotion types such as positive, negative, and neutral can be determined, and in detail, emotion types such as happiness, satisfaction, anxiety, and frustration can be identified.

[0133] In addition, language style, emotional synchronization, conversation frequency, and the degree of information exchange can be detected in the above conversation text. For example, the level of matching between the language styles of the counterpart and the user can be determined by identifying their respective language styles, such as a gentle mediator type, a humor-emphasizing type, and an emotion-centered type; the level of emotional synchronization between the counterpart and the user can be determined by identifying their respective emotions based on the third parameter; the conversation frequency regarding how often or how much conversation is conducted can be determined; and the degree of information exchange based on mutual trust, such as exchanging personal information or specific information, can be determined.

[0135] FIG. 6 schematically illustrates the process of performing the final context prompt generation step according to one embodiment of the present invention.

[0137] In summary, the final context prompt generation step can generate a final context prompt by applying the user information, the plurality of parameters, and one or more previous dialogue texts included in the dialogue history information to a preset context prompt template.

[0139] Specifically, the final context prompt generation step may include: a current situation information derivation step that derives current situation information including the age difference, relationship, title, and personality of the counterpart and the user based on the conversation text included in the conversation history information; and a context application step that generates a final context prompt by applying the user information, the current situation information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template.

[0141] More specifically, the current situation information derivation step detects honorifics and titles in the conversation text to derive the age difference, relationship, and title between the other party and the user, and can derive the other party's personality based on conversation pattern indicators including emotional vocabulary, frequency of use of positive and negative words, sentence length, and conversational participation in the conversation text using a text-based personality analysis model.

[0143] As illustrated in FIG. 6, a final context prompt can be generated by applying user information stored in the service server (1000), current situation information derived based on conversation history information, a plurality of parameters derived through the parameter derivation step, and one or more previous conversation texts included in the conversation history information to a pre-set context prompt template.

[0145] In one embodiment of the present invention, current situation information including the age difference, relationship, title, and personality of the other party and the user can be derived based on the conversation text included in the conversation history information through the current situation information derivation step.

[0146] At this time, honorifics and titles can be detected in the conversation text to derive the age difference, relationship, and title between the other party and the user, and a text-based personality analysis model can be used to derive the other party's personality based on conversation pattern indicators including emotional vocabulary, frequency of use of positive and negative words, sentence length, and conversational participation in the conversation text.

[0147] For example, the personality analysis model may correspond to a text-based personality analysis model that quantitatively infers the personality characteristics of the conversation partner and the user according to the Big Five (openness, conscientiousness, extraversion, agreeableness, neuroticism) model based on the conversation text, and the personality analysis model may include a model learned using various linguistic and conversational pattern indicators such as vocabulary diversity, frequency of use of positive and negative words, sentence length, speech act patterns, and conversational participation.

[0148] Therefore, by applying the above current situation information to the above final context prompt, a very specific and in-depth context regarding the conversation between the other party and the user can be provided, and the large language model (3000) can go beyond simply converting text to understand and reflect the subtle social relationships and dynamics between the user and the conversation partner.

[0150] In addition, the above one or more previous conversation texts may include the latest conversation texts within a preset number among the conversation texts included in the conversation history information, and by including the previous conversation texts in the final context prompt, it is possible to maintain the recent context and flexibly respond to the changing conversation flow.

[0152] FIG. 7 schematically illustrates the process of performing the final context prompt generation step according to one embodiment of the present invention.

[0154] Schematically, FIG. 7(a) illustrates a context prompt template, and FIG. 7(b) illustrates a final context prompt.

[0156] Specifically, the final context prompt may include user information including the user's age and gender, current situation information, multiple parameters, and one or more previous conversation texts. Accordingly, the context prompt template may correspond to a pre-configured template to which user information, current situation information, multiple parameters, and previous conversation texts can be applied.

[0158] As illustrated in FIG. 7(a), the context prompt template may include text such as “I am {00} years old and {female / male}. The other party is {00} years older / younger than me and appears to be in a relationship with {00} (me) and {00} (other party), and has a {0000} personality. I call the other party {000}. Hierarchy expression ratio: {000}%, joke frequency: {000}%, personal emotion expression frequency: {000}%, emotional intimacy: {000}%. What the other party said immediately before: {0000}”, and (1) may correspond to a template in which user information, (2) current situation information, (3) multiple parameters, and (4) the text of the previous conversation can be entered.

[0160] As shown in Fig. 7(b), when user information, current situation information, multiple parameters, and one or more previous conversation texts for the user are all collected, the information can be applied to the context prompt template to generate a final context prompt. Therefore, the above final context prompt may include text such as “I am {35} years old and {female}. The other party is {30} years older than me and appears to be in a {nephew} (me) and {aunt} (the other party) relationship, and has a {sensitive} personality. I call the other party {aunt}. Hierarchy expression ratio: {50}%, joke frequency: {20}%, personal emotion expression frequency: {70}%, emotional intimacy: {70}%. What the other party said immediately before: {It’s already dinner time.}”, and (1) may correspond to user information, (2) to current situation information, (3) to multiple parameters, and (4) to a template that can input the previous conversation text.

[0161] Therefore, (1) includes information that the user is 35 years old and female, (2) includes information that the conversation partner is 30 years older than the user, the user and the partner appear to be in a niece-aunt relationship, and the partner has a delicate personality, (3) includes information that among the multiple parameters, the ratio of expression of the superior-subordinate relationship is 50%, the frequency of jokes is 20%, the frequency of expression of personal emotions is 70%, and the emotional intimacy is 70%, and (4) may include information that the words spoken to the partner immediately before are "It's already dinner time."

[0162] Accordingly, the large language model (3000) that receives the above final context prompt can identify the user's age and gender, the age difference between the other party and the user, the relationship, the title and personality, the characteristics of the conversation between the other party and the user, and the words the other party said immediately before, and can output response information that can derive a second text by applying this.

[0164] Preferably, when inputting the final context prompt into the large language model (3000), the final prompt may be input to include the first text to be converted, the final context prompt, and a preset conversion request prompt requesting conversion of the first text, thereby outputting response information, and based on the output response information, the second text converted from the first text may be derived.

[0166] FIG. 8 schematically illustrates the process of performing the second text derivation step according to one embodiment of the present invention.

[0168] In summary, the second text derivation step can derive a second text converted from the first text based on response information output by inputting a final prompt including the first text, the final context prompt, and a preset conversion request prompt into a large language model (3000) outside or inside the service server (1000).

[0169] At this time, the second text is displayed on the conversion interface, and the second text can be transmitted to the messenger app according to the selection input on the user terminal (2000).

[0171] Specifically, the second text derivation step may use an embedding model to derive embedding vectors for each of the nouns and emotion vocabulary included in the first text and the conversation history information, derive an emotion type based on the similarity between a predefined emotion template and the embedding vectors, and generate a conversion request prompt containing text requesting the derivation of response information corresponding to the emotion type, and input it into the large language model (3000).

[0172] Additionally, the second text derivation step derives an emotion type for the first text using an emotion analysis model and generates multiple candidate second texts with different writing styles and vocabulary intensities according to multiple emotion intensities for the emotion type, and the multiple candidate second texts are displayed on the conversion interface, and one of the multiple candidate second texts can be transmitted to the messenger app according to a selection input from the user terminal (2000).

[0174] As illustrated in FIG. 8, a second text can be derived by converting the first text based on the output response information by inputting a final prompt, which includes the first text, the final context prompt, and a preset conversion request prompt, into a large language model (3000). At this time, a plurality of candidate second texts are generated by adjusting the intensity of different styles and vocabulary according to a plurality of emotional intensities of the emotion type of the first text, and the plurality of candidate second texts are displayed on the conversion interface, and one of the plurality of candidate second texts can be transmitted to the messenger app according to a selection input from the user terminal (2000).

[0175] At this time, the above conversion request prompt may include a request to convert the first text based on the input information, such as "Please change the sentence below to be social and considerate based on this information" or "Please change it into a polite and considerate expression suitable for the social context."

[0177] If a user inputs a first text, "Haven't you eaten yet?", through the conversion interface, the service server (1000) can input a final prompt containing text such as "Original text: {Haven't you eaten yet?}. I am {35} years old and {female}. The other person is {30} years older than me and appears to be in a {nephew} (me) and {aunt} (the other person) relationship, and has a {sensitive} personality. I call the other person {aunt}. Hierarchy expression ratio: {50}%, joke frequency: {20}%, personal emotion expression frequency: {70}%, emotional intimacy: {70}%. What the other person said just before: {It's already dinner time.}. Based on this information, please change the sentence below to be social and considerate."

[0178] At this time, the above-mentioned large language model (3000) can derive response information such as, "How is Auntie feeling? You said you were having a hard time because of a cold these days. Did you take good care of your meals?"

[0180] In one embodiment of the present invention, the large language model (3000) can derive response information including three sentences: (1) "Did you eat well?", (2) "How are you feeling? Did you eat well?", and (3) "How are you feeling, Auntie? You said you were having a hard time because of a cold these days. Did you eat well?", and the three sentences can be displayed at once on the conversion interface.

[0181] Accordingly, the user can send a message to the conversation partner by selecting and inputting one of three sentences and sending it to the messenger app. If the user selects and inputs (3) among a plurality of candidate second texts including (1) to (3), the text “How are you feeling, Aunt? You said you were having a hard time because of a cold lately. Did you take good care of your meals?” can be automatically sent to the messenger app of the user terminal (2000). Therefore, the user can conveniently send the converted text to the conversation partner simply by selecting and inputting.

[0183] At this time, the second text derivation step can input the first text into an emotion analysis model to derive an emotion type for the first text, and the large language model (3000) can generate multiple candidate second texts with different styles and vocabulary intensities according to multiple emotion intensities for the corresponding emotion type.

[0184] For example, the above emotion types can be derived in various ways, such as a gentle apology, a strong warning, a celebration, and worry. If the emotion type for the first text is derived as 'worry' through the emotion analysis model, the corresponding emotion type can be input into the large language model (3000) along with the final prompt to derive multiple candidate second texts.

[0185] In this case, the conversion request prompt input to the large language model (3000) may include a request to derive multiple texts with different styles and vocabulary intensities according to the emotional intensity of the corresponding emotion type called 'worry'.

[0186] Accordingly, the plurality of candidate second texts are displayed in the conversion interface, and one of the plurality of candidate second texts can be transmitted to the messenger app according to the selection input at the user terminal (2000).

[0188] In another embodiment of the present invention, the second text derivation step may use an embedding model to derive embedding vectors for each of the nouns and emotion vocabulary included in the first text and the conversation history information, derive an emotion type based on the similarity between a predefined emotion template and the embedding vectors, and generate a conversion request prompt including text requesting to derive response information corresponding to the emotion type, and input it into the large language model (3000).

[0189] For example, if the emotion type derived according to the embedding model and the emotion template corresponds to 'worry', the corresponding emotion type can be input into the large language model (3000) along with the final prompt to derive one second text or multiple candidate second texts.

[0190] At this time, when deriving a second text, the conversion request prompt input into the large language model (3000) may include a request to convert the first text by applying the corresponding emotion type based on the input information, such as, "Please change the sentence below to be social and considerate so that the emotion of 'worry' is included based on this information."

[0191] In addition, when deriving multiple candidate second texts, the conversion request prompt input into the large language model (3000) may include a request to derive multiple texts with different styles and vocabulary intensities according to the emotional intensity of the corresponding emotion type called 'worry'.

[0193] Therefore, when deriving a second text by converting a first text through the present invention, not only is the context of the conversation reflected, but the intention of the current conversation, such as an apology, a request, and a compliment, is identified and reflected in the converted language style, thereby providing a second text that matches the user's intention.

[0194] At this time, based on the conversation history information, a second text with an emotional response in actual human conversation applied can be automatically derived by using a large language model (3000) by matching the tone of the language by reflecting the past language usage style, age, and gender of the conversation partner and the user.

[0195] In addition, by deriving multiple candidate second texts that are adjusted in multiple stages, considering not only emotional intensity but also emotional type, a user-customized second text can be provided. That is, the user can receive multiple candidate second texts and select one desired text from among them to automatically send via a messenger app.

[0197] Therefore, the present invention does not provide a single converted second text to the user, but rather generates and provides multiple candidate second texts classified by emotional intensity to the user, allowing the user to select one of them, thereby increasing user satisfaction and usability and providing a user-customized language conversion service beyond simple language conversion.

[0199] FIG. 9 schematically illustrates the process of performing a social language conversion method according to one embodiment of the present invention.

[0201] Schematically, FIG. 9(a) and FIG. 9(b) illustrate the process of receiving a first text from a user terminal (2000) and transmitting a second text to the user terminal (2000).

[0203] Specifically, the first text receiving step can receive the first text input from a conversion interface executed on a user terminal (2000). In one embodiment of the present invention, after the second text is derived through the second text derivation step, the second text is displayed on the conversion interface, and the second text can be transmitted to the messenger app according to a selection input on the user terminal (2000).

[0205] Preferably, the conversion interface may be provided to the user in the form of an app executed on the user terminal (2000), displayed on the user terminal (2000) in the form of a separate app, or displayed on the user terminal (2000) in the form of a widget so as to be used on a messenger app running on the user terminal (2000).

[0207] As illustrated in FIG. 9(a), the L1 and L2 layers correspond to a messenger app running on a user terminal (2000), and the E1 element may correspond to an element that displays the conversion interface according to the selected input. At this time, the conversation history information may include all conversation texts displayed on the L1 layer.

[0208] At this time, the user can proceed with a conversation with the conversation partner through the L1 layer and the L2 layer, and in the conversation history information, (1) to (4) and (7) to (8) correspond to the conversation text of the conversation partner, and (5) to (6) correspond to the conversation text of the user. That is, in the conversation history information, the conversation text of the user and the conversation text of the other party can be classified respectively.

[0210] As shown in Fig. 9(b), when a user selects and inputs the E1 element, a conversion interface corresponding to the L3 layer can be displayed as an overlay on the messenger app in the form of a widget.

[0211] At this time, the L3 layer may include an L4 layer where a user can input a first text to be converted, an E2 element that requests language conversion based on selected input, an L5 layer where the converted second text is displayed, and an E3 element that requests the transmission of the second text to a messenger app.

[0213] In one embodiment of the present invention, when a user wishes to convert language using the social language conversion method of the present invention while conversing with another person through the messenger app, the user terminal (2000) selects and inputs an E1 element to display the L3 layer on the messenger app, and after inputting the first text to be converted into the L4 layer, selects and inputs an E2 element, the converted second text is displayed in the L5 layer, and when the user wishes to transmit the second text to the messenger app, selects and inputs an E3 element, the second text can be transmitted to the L2 layer of the messenger app.

[0215] Accordingly, the user's convenience can be increased by enabling the converted second text to be sent directly to the messenger app of the user terminal (2000), and it can be extended to enable integration with existing smartphone apps, such as by adding a language conversion function through an API for the social language conversion method of the present invention in a messenger app including KakaoTalk, or by providing automatically converted text when text is entered, such as a text auto-completion function on a smartphone.

[0216] In addition, by providing the integration and convenience of an overall workflow that collects conversation history information from a messenger app, converts text through a separate conversion interface or service server (1000), and then transmits it back to the messenger app, the user can conveniently apply the social language conversion method of the present invention when using existing messenger apps such as KakaoTalk, and the present invention can have the scalability to be applied to various existing messenger apps.

[0218] In another embodiment of the present invention, actual user feedback, such as whether the converted second text was actually transmitted to the messenger app and user satisfaction, can be collected to periodically optimize the performance of the parameter calculation model or the large language model (3000). Since it can provide social conversion in various language environments and thus have scalability in the global market, it has various possibilities such as model training and optimization or multilingual support.

[0220] The present invention can provide practical social language conversion by going beyond simply converting language styles, such as written language to spoken language or honorifics to informal language, and by deriving quantified parameters by comprehensively considering social contexts such as age, gender, relationships, and previous conversations, and dynamically reflecting them in the final context prompt. Additionally, by utilizing the excellent conversational context understanding ability of the existing large language model (3000), parameters regarding an individual's language habits and conversational characteristics are derived based on specific social relationships and actual conversation history, thereby enabling customized language conversion.

[0221] In addition, convenience can be provided to users by collecting conversation history information in conjunction with existing messenger apps and easily sharing the converted second text back to the messenger app, and accurate and explainable analysis results can be provided by combining mechanical rules and semantic-based deep learning embedding techniques.

[0223] FIG. 10 illustrates, in an exemplary manner, the internal configuration of a computing device (11000) according to one embodiment of the present invention.

[0225] The service server (1000) mentioned in the description of FIG. 1 may include components of the computing device (11000) illustrated in FIG. 10, which will be described later.

[0227] As illustrated in FIG. 10, the computing device (11000) may include at least one processor (11100), memory (11200), peripheral interface (11300), input / output subsystem (I / O subsystem) (11400), power circuit (11500), and communication circuit (11600).

[0229] Specifically, the memory (11200) may include, for example, high-speed random access memory, magnetic disk, SRAM, DRAM, ROM, flash memory, or non-volatile memory. The memory (11200) may include software modules, instruction sets, or various other data required for the operation of the computing device (11000).

[0230] At this time, access to the memory (11200) from other components, such as the processor (11100) or the peripheral device interface (11300), can be controlled by the processor (11100). The processor (11100) may be composed of a single or multiple units and may include processors in the form of GPUs and TPUs to improve computational processing speed.

[0231] The above peripheral device interface (11300) can connect input and / or output peripheral devices of the computing device (11000) to the processor (11100) and the memory (11200). The processor (11100) can perform various functions for the computing device (11000) and process data by executing a software module or instruction set stored in the memory (11200).

[0232] The input / output subsystem (11400) may connect various input / output peripheral devices to the peripheral device interface (11300). For example, the input / output subsystem (11400) may include a controller for connecting peripheral devices such as a monitor, keyboard, mouse, printer, or, if necessary, a touchscreen or sensor to the peripheral device interface (11300). According to another aspect, the input / output peripheral devices may be connected to the peripheral device interface (11300) without passing through the input / output subsystem (11400).

[0233] The power circuit (11500) may supply power to all or part of the components of the terminal. For example, the power circuit (11500) may include one or more power sources such as a power management system, a battery or alternating current (AC), a charging system, a power failure detection circuit, a power converter or inverter, a power status indicator, or any other components for power generation, management, and distribution.

[0234] The communication circuit (11600) may enable communication with another computing device using at least one external port. Alternatively, as described above, the communication circuit (11600) may enable communication with another computing device by including an RF circuit and transmitting and receiving an RF signal, also known as an electromagnetic signal, as needed.

[0236] The embodiment of FIG. 10 is merely an example of the computing device (11000), and the computing device (11000) may have some components shown in FIG. 10 omitted, additional components not shown in FIG. 10 added, or a configuration or arrangement that combines two or more components. For example, a computing device for a communication terminal in a mobile environment may include a touchscreen or sensors in addition to the components shown in FIG. 10, and the communication circuit (1160) may include a circuit for RF communication of various communication methods (Wi-Fi, 3G, LTE, 5G, 6G, Bluetooth, NFC, Zigbee, etc.). The components that can be included in the computing device (11000) may be implemented as hardware, software, or a combination of both hardware and software, including one or more integrated circuits specialized for signal processing or applications.

[0237] Methods according to embodiments of the present invention may be implemented in the form of program instructions that can be executed through various computing devices and recorded on a computer-readable medium. In particular, the program according to the present embodiment may be configured as a PC-based program or an application dedicated to a mobile terminal. An application to which the present invention is applied may be installed on a user terminal through a file provided by a file distribution system. For example, the file distribution system may include a file transmission unit (not shown) that transmits the file upon a request from the user terminal.

[0239] The device described above may be implemented as a hardware component, a software component, and / or a combination of a hardware component and a software component. For example, the device and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. In addition, other processing configurations, such as parallel processors, are also possible.

[0240] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave in order to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be standardized and stored or executed in a standardized manner on a networked computing device. Software and data may be stored on one or more computer-readable recording media.

[0241] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0243] According to one embodiment of the present invention, when a user inputs a first text into a conversion interface, the user can receive a second text that has been automatically converted, thereby achieving the effect of being able to receive a second text that has been automatically converted.

[0244] According to one embodiment of the present invention, the effect of deriving a second text by transforming a first text by reflecting current situation information derived from conversation history information, a plurality of parameters, and the previous conversation text can be achieved.

[0245] According to one embodiment of the present invention, a parameter derivation model can be used to derive a plurality of parameters that quantify the characteristics of a conversation in the conversation history information by item.

[0246] According to one embodiment of the present invention, the effect of automatically deriving a second text by converting a first text through a large language model can be achieved.

[0247] According to one embodiment of the present invention, the derived second text is transmitted to a messenger app of a user terminal, thereby enabling the user to conveniently use social language conversion in the messenger app.

[0248] According to one embodiment of the present invention, it is possible to achieve the effect of performing social language conversion that comprehensively considers social context and conversational context, going beyond simple text conversion.

[0249] According to one embodiment of the present invention, by reflecting complex social elements such as hierarchical relationships, emotional expressions, and intimacy extracted from the conversation history of an actual messenger app into a first text entered by a user, it is possible to derive a second text that is suitable for the conversation context and natural.

[0250] According to one embodiment of the present invention, by deriving a second text appropriate to the situation that reflects the relationship and emotional state between the other party and the user, it is possible to reduce the possibility of social errors or misunderstandings and to achieve the effect of supporting smooth communication.

[0251] According to one embodiment of the present invention, by providing a plurality of candidate second texts based on emotional intensity or expression style, the effect of providing a user-customized second text can be achieved.

[0252] According to one embodiment of the present invention, since it can provide a social language conversion function in real time by linking with various platforms including messenger apps, it can achieve the effect of providing various scalability and user accessibility.

[0254] Although the embodiments have been described above with reference to limited examples and drawings, those skilled in the art can make various modifications and variations from the description above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents. Therefore, other implementations, other embodiments, and equivalents to the claims below are also within the scope of the claims.

Claims

Claim 1 A conversation context-based social language conversion method using a large language model through a service server comprising one or more processors and one or more memories, comprising: a first text reception step in which user information including the age and gender of a user received from a user terminal is stored in the service server, and a first text input from a conversion interface executed on the user terminal is received; a parameter derivation step in which conversation history information from a messenger app executed on the user terminal is collected, and the conversation history information is input into a parameter derivation model to derive a plurality of parameters that quantify the characteristics of the conversation by item; and a final context prompt generation step in which current situation information including the age difference, relationship, title, and personality of the counterparty and the user is derived based on the conversation text included in the conversation history information, and the user information, the current situation information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information are applied to a preset context prompt template to generate a final context prompt. A second text derivation step of generating a plurality of candidate second texts by analyzing the emotion type of the first text and adjusting the intensity of the style and vocabulary to have different emotional intensities for the emotion type, based on the output response information obtained by inputting a final prompt including the first text, the final context prompt, and a preset conversion request prompt into a large language model outside or inside the service server;The method includes, wherein the plurality of candidate second texts are displayed on the conversion interface, and one of the plurality of candidate second texts is transmitted to the messenger app according to a selection input from the user terminal; the parameter derivation step derives, based on the conversation text, a first parameter for the ratio of expression of hierarchical relationships, a second parameter for joke frequency, a third parameter for the frequency of expression of personal emotions, and a fourth parameter for emotional intimacy; and the final context prompt generation step derives the age difference, relationship, and title between the counterpart and the user by detecting honorifics and titles based on the conversation text included in the conversation history information, and derives current situation information including the counterpart's personality based on conversation pattern indicators including emotional vocabulary and conversation participation of the conversation text using a text-based personality analysis model; A social language conversion method comprising: a step of generating a final context prompt in which a social context is defined by applying the user information, the current situation information, the plurality of parameters, and the previous dialogue text to each of the user information input location, the current situation information input location, the plurality of parameter input location, and the previous dialogue text input location listed in the aforementioned preset context prompt template. Claim 2 In claim 1, the parameter derivation step quantifies the characteristics of each item, including the ratio of expressions of hierarchical relationships, the frequency of jokes, the frequency of expressions of personal emotions, and emotional intimacy, based on the conversation text included in the conversation history information through the parameter derivation model and derives them as parameters, and detects sentence-ending morphemes, titles, and demonstrative vocabulary in the conversation text. A social language conversion method comprising: a first parameter derivation step for deriving a first parameter for the ratio of the expression of the above-mentioned hierarchical relationship; a second parameter derivation step for deriving a second parameter for the frequency of jokes by detecting phonological features, exaggerated expressions, and laughter expressions in the above-mentioned conversation text; a third parameter derivation step for deriving a third parameter for the frequency of the above-mentioned personal emotion expressions by detecting emotional vocabulary and emotional intensity in the above-mentioned conversation text; and a fourth parameter derivation step for deriving a fourth parameter for the above-mentioned emotional intimacy by detecting language style, emotional synchronization, conversation frequency, and degree of information exchange in the above-mentioned conversation text. Claim 3 delete Claim 4 A social language conversion method according to claim 1, wherein the second text derivation step derives an embedding vector for each of the nouns and emotion vocabulary included in the first text and the conversation history information using an embedding model, derives an emotion type based on the similarity between a predefined emotion template and the embedding vector, and generates a conversion request prompt including text requesting the derivation of response information corresponding to the emotion type, and inputs it into the large language model. Claim 5 delete Claim 6 A conversation context-based social language conversion method using a large language model executed on a user terminal comprising one or more processors and one or more memories, wherein user information including the age and gender of the user is stored on the user terminal, and a first text receiving step of receiving a first text input from a conversion interface displayed on the user terminal; a parameter derivation step of collecting conversation history information from a messenger app of the user terminal and inputting the conversation history information into a parameter derivation model to derive a plurality of parameters that quantify the characteristics of the conversation by item; and a final context prompt generation step of deriving current situation information including the age difference, relationship, title, and personality of the counterparty and the user based on the conversation text included in the conversation history information, and generating a final context prompt by applying the user information, the current situation information, the plurality of parameters, and one or more previous conversation texts included in the conversation history information to a preset context prompt template. A second text derivation step of generating a plurality of candidate second texts by analyzing the emotion type of the first text and adjusting the intensity of the style and vocabulary to have different emotional intensities for the emotion type, based on response information output by inputting a final prompt including the first text, the final context prompt, and a preset conversion request prompt into a large language model external or internal to the user terminal;The method includes displaying the plurality of candidate second texts on the conversion interface and transmitting one of the plurality of candidate second texts to the messenger app according to a selection input from the user terminal; the parameter derivation step derives, based on the conversation text, a first parameter for the ratio of expression of hierarchical relationships, a second parameter for joke frequency, a third parameter for the frequency of expression of personal emotions, and a fourth parameter for emotional intimacy; the final context prompt generation step derives the age difference, relationship, and title between the other party and the user by detecting honorifics and titles based on the conversation text included in the conversation history information, and derives current situation information including the other party's personality based on conversation pattern indicators including emotional vocabulary and conversation participation of the conversation text using a text-based personality analysis model; A social language conversion method comprising: a step of generating a final context prompt in which a social context is defined by applying the user information, the current situation information, the plurality of parameters, and the previous dialogue text to each of the user information input location, the current situation information input location, the plurality of parameter input location, and the previous dialogue text input location listed in the aforementioned preset context prompt template.

Images