Method and system for providing dual language model based on license-free data

Abstract

The present invention relates to a method comprising: verifying a license of conversation data; using the conversation data as seed data to perform fine-tuning and additional training on a dual language model; constructing a license-free database based on the trained model; and providing the dual language model based on the constructed license-free database and services related thereto.

Description

Method and System for Providing a License-Free Data-Based Bilingual Model

[0001] The present invention relates to a method and system for providing a license-free data-based bilingual model fine-tuned based on conversational data generated through actual and / or learning.

[0002] With the establishment of development environments utilizing open source software, many open source resources are being utilized even when training Large Language Models (LLMs).

[0003] When utilizing open source, content posted on the internet is typically crawled and utilized through internet searches. However, information obtained through such channels often fails to provide complete license information, and in many cases, the license information is partially damaged or omitted, making it impossible to verify the original license information unless the user carefully reviews the original source.

[0004] Moreover, as AI services based on Large Language Models (LLM) have recently become popularized and are being utilized in a wide variety of fields, a new issue is emerging regarding the disadvantage that if trained with contaminated data that may cause legal problems, they can provide uncertain results by supplying information that has not undergone license review during service delivery.

[0005] Specifically, if a company provides services based on a language model trained with unlicensed, or contaminated, data, disputes with data creators may arise—including copyright infringement under copyright law, violations of the Unfair Competition Prevention Act and the Personal Information Protection Act, civil and criminal liability for trade secret infringement, and infringement of publicity rights—even if used only within an internal environment. Furthermore, if the product is installed and supplied to consumers, disputes with consumers may also occur—including violations of the Consumer Protection Act, failure to notify of defects, and liability for damages due to defects under civil law.

[0006] To this end, data managers must individually verify the license conditions of the sources and sub-information included in each data set; however, given the vast volume of data and its steady growth, it is virtually impossible for an individual to review licenses for all data without error.

[0007] In other words, there is a growing need for the invention of a technology to build a license-free database by automating the license verification of seed data to solve the aforementioned problem.

[0008] The present invention was devised to solve the problems of the prior art as described above, and aims to provide a method and system for providing a license-free data-based bilingual model that generates conversational data based on a cumulative approach to fine-tune and / or train a bilingual model.

[0009] In addition, the present invention aims to provide a method and system for providing a license-free data-based bilingual model that constructs a license-free database by verifying licenses for information used to perform tasks.

[0010] In addition, the present invention aims to provide a method and system for providing a license-free data-based bilingual model that trains a data generation module based on a cross-language instruction tuning method.

[0011] However, the technical problems that the present invention and the embodiments of the present invention aim to solve are not limited to the technical problems described above, and other technical problems may exist.

[0012] A method for providing a bilingual model based on license-free data according to an embodiment of the present invention is a method in which a computing system including memory and a processor provides a bilingual model, comprising: a step of acquiring conversation data based on user input based on a pre-trained bilingual model; a step of verifying a license for the acquired conversation data; a step of generating and storing the conversation data with verified license as seed data; a step of fine-tuning the bilingual model based on the seed data; a step of performing additional learning on the fine-tuned bilingual model based on a plurality of accumulated conversation data; a step of constructing a license-free database based on the learned bilingual model; and a step of providing a bilingual model based on the constructed license-free database and a service using the same.

[0013] Additionally, the step of acquiring the conversation data comprises: acquiring question data including a predetermined task input based on a user interface; providing answer data for the acquired question data based on a pre-trained bilingual model; matching the question data and the answer data into question-answer pairs; and storing at least one question data, answer data, and question-answer pair as conversation data.

[0014] Additionally, the step of storing at least one question data, answer data, and question-answer pair as conversation data includes the step of matching and storing the type of task and the number of turns for the question-answer pair.

[0015] Additionally, the step of verifying the license for the acquired conversation data comprises: a step of performing a search for license-related information regarding the acquired conversation data by linking with a predetermined website and database; a step of acquiring license information, which is information regarding the usage rights of data for at least one of the web page, domain, and file found through the search; and a step of determining the conversation data as either an available license or an unavailable license according to the acquired license information.

[0016] Additionally, the step of fine-tuning the bilingual model includes the step of controlling the instruction module included in the bilingual model to use conversation data determined as an available license as a training data set, the step of controlling the instruction module to exclude conversation data determined as an unavailable license from the training data set, and the step of controlling the answer module included in the bilingual model to generate an answer to an instruction generated from the learned instruction module.

[0017] Additionally, the step of fine-tuning the bilingual model includes the step of matching and storing the generated instructions and answers as question-answer pairs, and the step of repeating the matching and storing of multiple question-answer pairs based on a cumulative approach.

[0018] Additionally, the step of fine-tuning the bilingual model further includes the step of matching conversation data in which a first language is detected to a first general conversation turn, the step of converting the conversation data in which the first language is detected to a second language, the step of obtaining converted answer data converted to the second language based on the answer module for the converted question data converted to the second language, the step of matching the converted question data and the converted answer data to a first converted conversation turn, and the step of fine-tuning the bilingual model by matching the first general conversation turn and the first converted conversation turn.

[0019] In addition, the step of performing additional training on the bilingual model includes the step of performing reinforcement training on the bilingual model based on at least one of the DPO, SFT, and RLHF methods.

[0020] Additionally, the step of constructing the license-free database further includes the step of additionally storing conversation data determined as at least one of an available license and an unavailable license for each of the first language and the second language according to the additional learning progress, and the step of updating the license-free database according to the additional storage.

[0021] Meanwhile, a dual-language model providing system based on license-free data according to an embodiment of the present invention comprises: at least one memory; and at least one processor that reads at least one application stored in the memory and provides a dual-language model; and the instructions of the processor include: a step of acquiring conversation data based on user input based on a pre-trained dual-language model; a step of verifying a license for the acquired conversation data; a step of generating and storing the conversation data with verified license as seed data; a step of fine-tuning the dual-language model based on the seed data; a step of performing additional learning on the fine-tuned dual-language model based on a plurality of accumulated conversation data; a step of constructing a license-free database based on the learned dual-language model; and a step of providing a dual-language model based on the constructed license-free database and a service using the same.

[0022] Additionally, a license-free data-based dual-language model providing system according to an embodiment of the present invention comprises: a plurality of neurons configured in an array including at least one register, at least one programmable logic, and at least one input interface; a plurality of synapse circuits storing synapse weights that regulate the connection strength between the plurality of neurons; and at least one routing network that controls the data flow between the plurality of neurons; wherein each of the plurality of neurons further comprises a Field Programmable Gate Array (FPGA) implementation for a predetermined artificial neural network that is connected to at least one other neuron through the routing network to establish a transmission path for the weights.

[0023] Additionally, a license-free data-based bilingual model providing system according to an embodiment of the present invention comprises: a plurality of neurons organized into an array including at least one register, at least one microprocessor, and at least one input; and a plurality of synapse circuits storing synapse weights that regulate the connection strength between the plurality of neurons; wherein each of the plurality of neurons further comprises an Application Specific Integrated Circuit (ASIC) for a predetermined artificial neural network that is connected to at least one other neuron through any one of the plurality of synapse circuits.

[0024] The method and system for providing a license-free data-based bilingual model according to an embodiment of the present invention generates conversational data based on a cumulative approach and fine-tunes and / or trains a bilingual model, thereby providing results that meet given conditions such as task type and the number of multi-turns, which has the effect of increasing applicability and further expanding the scale of the data.

[0025] In addition, the method and system for providing a bilingual model based on license-free data according to an embodiment of the present invention automatically generates license-free data by fine-tuning seed data and builds a license-free database, thereby improving work productivity by performing repetitive tasks on behalf of humans, saving costs by not relying on API calls, and utilizing accumulated data over the long term.

[0026] In addition, the method and system for providing a license-free data-based bilingual model according to an embodiment of the present invention have the effect of providing a higher-performance bilingual model by reducing the gap in results that vary depending on language differences through training a data generation module based on a cross-language instruction tuning method.

[0027] However, the effects obtainable from the present invention are not limited to those mentioned above, and other unmentioned effects can be clearly understood from the description below.

[0028] FIG. 1 illustrates an example of a block diagram of a computing system implementing a license-free data-based bilingual model according to an embodiment of the present invention.

[0029] FIG. 2 illustrates an example of a block diagram of a computing device, which is one of the components of a computing system implementing a license-free data-based bilingual model according to an embodiment of the present invention.

[0030] FIG. 3 illustrates an example of a block diagram in another aspect of a computing device, which is one of the components of a computing system implementing a license-free data-based bilingual model according to an embodiment of the present invention.

[0031] FIG. 4 is a flowchart illustrating a license-free data-based bilingual model according to an embodiment of the present invention.

[0032] FIG. 5 is a conceptual diagram illustrating the entire process of building a license-free database based on seed data and reference documents according to an embodiment of the present invention.

[0033] FIG. 6 is a conceptual diagram illustrating a method for fine-tuning a bilingual model into license-free data according to an embodiment of the present invention.

[0034] Figure 7 is a table showing the performance of a bilingual model in which additional learning was performed according to an embodiment of the present invention.

[0035] The present invention is capable of various modifications and may have various embodiments; therefore, specific embodiments are illustrated in the drawings and described in detail in the detailed description. The effects and features of the present invention, and the methods for achieving them, will become clear by referring to the embodiments described below in detail together with the drawings. However, the present invention is not limited to the embodiments disclosed below but can be implemented in various forms. In the following embodiments, terms such as "first," "second," etc., are used not in a limiting sense but for the purpose of distinguishing one component from another. Furthermore, singular expressions include plural expressions unless the context clearly indicates otherwise. Also, terms such as "include" or "have" mean that the features or components described in the specification exist, and do not preclude the possibility that one or more other features or components may be added. Additionally, in the drawings, the size of components may be exaggerated or reduced for convenience of explanation. For example, the size and thickness of each component shown in the drawings are arbitrarily depicted for convenience of explanation, so the present invention is not necessarily limited to what is depicted.

[0036] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. When describing with reference to the drawings, identical or corresponding components are given the same reference numerals, and redundant descriptions thereof will be omitted.

[0037]

[0038] FIG. 1 illustrates an example of a block diagram of a computing system providing a bilingual model according to an embodiment of the present invention.

[0039] Referring to FIG. 1, a computing system (1000) providing a dual language model according to one embodiment of the present invention includes a user computing device (110), a training computing system (150), and a server computing system (130), and each device and system is connected to communicate through a network (170).

[0040] According to an embodiment of the present invention, 1) a user computing device (110) may provide a bilingual model by using a local or / and external machine learning model (120) or by using a machine learning model (140) provided by a server.

[0041] Additionally, according to another embodiment of the present invention, 2) a server computing system (130) communicating with a user computing device (110) may provide a dual language model to the user computing device (110) on an application or / and the web in response to a request from a user through the user computing device (110).

[0042] In addition, according to another embodiment of the present invention, 3) a user computing device (110) and a server computing system (130) may perform at least a part of the method of providing a bilingual model in conjunction with each other to provide a bilingual model and / or a service based thereon to the user.

[0043]

[0044] Additionally, according to various embodiments of the present invention, a user computing device (110) and / or a server computing system (130) may learn a machine learning model (120 / 140) that is performed in a method of providing a bilingual model through interaction with a training computing system (150) that is communicatedly connected via a network (170). In this case, the training computing system (150) may be separate from the server computing system (130) or may be part of the server computing system (130).

[0045] In some embodiments, the training computing system (150) may be part of the server computing system (130) or part of the user computing device (110).

[0046] In the following description, the dual language model is executed by accessing the server computing system (130) through the user computing device (110), and the dual language model is provided by the server computing system (130) either directly or by using a language model from another server.

[0047] However, it can be understood that cases where part of the process described as being performed in a server computing system (130) is performed in a user computing device (110) are naturally included in the description of the present invention.

[0048] - User Computing Device (110: User Computing Device)

[0049] The user computing device (110) may include all other types of computing devices, such as a smartphone, a mobile phone, a digital broadcasting device, a PDA (personal digital assistants), a PMP (portable multimedia player), a desktop, a wearable device, an embedded computing device and / or a tablet PC.

[0050] Additionally, in the embodiment, the user computing device (110) may further include a predetermined server computing device that provides an environment for providing a dual language model.

[0051] This user computing device (110) includes at least one processor (111) and memory (112).

[0052] Here, the processor (111) may be composed of at least one of a central processing unit (CPU), a graphics processing unit (GPU), ASICs (application specific integrated circuits), DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), controllers, microcontrollers, microprocessors, and / or other electrical units for performing functions, or a plurality of electrically connected processors.

[0053] In particular, according to the embodiment, this processor (111) may be configured based on a Field Programmable Gate Array (FPGA) implementation and / or an Application Specific Integrated Circuit (ASIC), which is a hardware technology for implementing a certain digital circuit.

[0054] Here, a field programmable gate array (FPGA) can refer to a flexible digital circuit that is programmable according to user needs.

[0055] As an example, a field programmable gate array implementation may include a register that temporarily stores data and controls the flow and timing of signals to maintain intermediate results or state information of operations to support synchronized operation of the FPGA, programmable logic that programs operations within the FPGA to perform specific functions or operations as logic circuits configurable according to user needs, and an input interface that receives signals from external devices or sensors and transmits them to internal circuits as a channel for receiving data from outside the FPGA.

[0056] Through the combination of the above components, a field-programmable gate array implementation can provide flexible and various types of digital circuits.

[0057] Meanwhile, an Application-Specific Integrated Circuit (ASIC) can refer to a custom integrated circuit that is fixedly designed to perform a specific use or function.

[0058] As an example, the application-dedicated integrated circuit may include a register, which is a small memory device for temporarily storing and managing data and supports the rapid processing of ASIC operations by storing intermediate calculation results or state information; a microprocessor, which is a central processing unit that performs control and operations within the ASIC and coordinates the operation of the entire system by performing various operations or generating control signals when necessary; and an input block, which is an interface for receiving data from the outside, which receives data to be processed by the ASIC and transmits it internally, and receives various input data through connections with sensors or external devices.

[0059] Through the combination of the components mentioned above, an application-specific integrated circuit can perform specific purpose tasks in an optimized manner.

[0060] Returning to the point, the memory (112) of the user computing device (110) may include one or more non-transient / transient computer-readable storage media such as RAM, ROM, EEPROM, EPROM, flash memory devices, magnetic disks, and combinations thereof, and may include web storage of a server that performs memory storage functions on the internet. This memory (112) may store data and instructions necessary for the at least one processor (111) to perform the operation of an application to provide a dual language model.

[0061] In one embodiment, the user computing device (110) can perform various deep learnings for providing a bilingual model by linking with a deep-learning neural network.

[0062] Here, the deep learning neural network according to the embodiment may include a Convolutional Neural Network (CNN), R-CNN (Regions with CNN features), Fast R-CNN, Faster R-CNN, Mask R-CNN, etc., and may include any deep learning neural network that includes an algorithm capable of performing the embodiments described below, and the embodiments of the present invention do not limit or restrict such deep learning neural networks themselves.

[0063] At this time, according to the embodiment, the deep learning neural network may be installed directly on the server computing system (130) or operate as a separate device from the server computing system (130) to perform deep learning for the bilingual model.

[0064] Additionally, in one embodiment, the user computing device (110) may store at least one machine learning model (120). For example, the user computing device (110) may be various machine learning models, such as multiple neural networks (e.g., deep neural networks) that provide a bilingual model based on structured / quantitative data, or other types of machine learning models including non-linear models and / or linear models, and may be configured as a combination thereof.

[0065] For example, machine learning models may include linear regression, decision trees, random forests, gradient boosting pre-trained language models or / and deep learning models. And neural networks may include at least one of feed-forward neural networks, recurrent neural networks (e.g., long short-term memory recurrent neural networks), convolutional neural networks or / and other forms of neural networks.

[0066] Specifically, in the embodiment, the user computing device (110) can store an artificial intelligence agent (AI agent) (hereinafter referred to as Agent (AG)) that determines license-free data for an input task while interacting with a given environment (specifically, a web page) and builds a license-free database.

[0067] In an example, an agent (AG) may refer to a computing system that combines question-answer exchanges with a user, fine-tunes a large-scale language model (LLM) (in an example, a bilingual model) using a cumulative approach, and generates license-free data that meets conditions such as task type and / or number of turns.

[0068] Additionally, the user computing device (110) may store a model to be used in each process and a prompt template that serves as the basis for input to the model in order to perform at least part of the process of providing a bilingual model through a large language model (LLM) and / or a large behavior model (LAM).

[0069] For example, the user computing device (110) may store 1) a prompt for generating a query from user input, 2) a prompt for analyzing seed data, 3) a prompt for filtering license-free data, etc.

[0070] That is, in one embodiment, the user computing device (110) may provide a bilingual model based on received data by requesting the execution of some execution steps in a method of providing a bilingual model to a language model of an external server through a prompt, etc.

[0071] In another embodiment, regarding the method of providing a bilingual model requested through a user computing device (110), the server computing system (130) may provide data to the user computing device (110) by providing a bilingual model through at least one machine learning model (140) and a machine learning model of another server.

[0072] Such a user computing device (110) may include at least one input component (121) that detects user input. Specifically, the input component (121) may include a sensor system including an image sensor, a position sensor (IMU), an audio sensor, a distance sensor, a proximity sensor, a contact sensor, etc.

[0073] For example, the user input component (121) may include a touch sensor (e.g., a touch screen or / and a touch pad, etc.) that detects a touch of the user's input medium (e.g., a finger or a stylus), an image sensor that detects the user's motion input, a microphone that detects the user's voice input, a button, a mouse and / or a keyboard, etc.

[0074] Here, the image sensor may include an image processing module. Specifically, the image sensor may process still images or video obtained by an image sensor device (e.g., CMOS or CCD).

[0075] In addition, the image sensor can process a still image or video acquired through the image sensor device using an image recognition process (e.g., OCR, etc.) and / or an image processing module to extract necessary information and transmit the extracted information to a processor.

[0076] Additionally, the input component (121) can receive input from an external controller (e.g., mouse, keyboard, etc.) based on an interface module, and in this case, may include an external output device (e.g., speaker).

[0077] At this time, the interface module may be configured to include at least one of a wired / wireless headset port, an external charger port, a wired / wireless data port, a memory card port, a port for connecting a device equipped with an identification module, an audio I / O (Input / Output) port, a video I / O (Input / Output) port, an earphone port, a power amplifier, an RF circuit, a transceiver, and other communication circuits.

[0078] In addition, the external output device may include a display system that outputs various information related to the bilingual model as a graphic image.

[0079] Such a display system may be implemented by including at least one of a liquid crystal display (LCD), a thin film transistor-liquid crystal display (TFT LCD), an organic light-emitting diode (OLED), a flexible display, a 3D display, and an e-ink display.

[0080] Meanwhile, the user computing device (110) including the above-described components may further perform at least some of the functional operations performed by the server computing system (130) described later.

[0081] -Server Computing System (130: Server Computing System)

[0082] The server computing system (130) can perform a series of processes to provide a bilingual model.

[0083] In detail, in an embodiment, the server computing system (130) can provide the bilingual model by exchanging data necessary to drive a process for providing the bilingual model with the external device, such as the user computing device (110).

[0084] More specifically, in an embodiment, the server computing system (130) can provide an environment in which an application can run on a user computing device (110).

[0085] To this end, the server computing system (130) may include an application program, data and / or instructions, etc. for the application to operate, and may transmit and receive various data based thereon with the external device.

[0086] Additionally, the server computing system (130) includes at least one processor (131) and memory (132). Here, the processor (131) may be composed of at least one or a plurality of electrically connected processors among a central processing unit (CPU), a graphics processing unit (GPU), ASICs (application specific integrated circuits), DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), controllers, microcontrollers, microprocessors, and / or other electrical units for performing functions.

[0087] In particular, depending on the embodiment, such a processor (131) may be configured based on a Field Programmable Gate Array (FPGA) implementation and / or an Application Specific Integrated Circuit (ASIC), which are hardware technologies for implementing a specific digital circuit. A detailed description thereof is omitted by applying the description of the FPGA and ASIC mentioned above.

[0088] And the memory (132) may include one or more non-transient / transient computer-readable storage media such as RAM, ROM, EEPROM, EPROM, flash memory device, magnetic disk, etc. and combinations thereof. This memory (132) may store data and instructions required for prompt templates, machine learning models (140), etc., for the processor (131) to perform tasks through the language model of the server computing system (130) or / and the language model of an external server.

[0089] For example, a server computing system (130) may include a neural network or / and other multi-layer non-linear models as a machine learning model (140). Exemplary neural networks may include a feed-forward neural network, a deep neural network, a recurrent neural network, and a convolutional neural network.

[0090] In one embodiment, the server computing system (130) may be implemented to include at least one computing device. For example, the server computing system (130) may be implemented to operate a plurality of computing devices according to a sequential computing architecture, a parallel computing architecture, or a combination thereof. Additionally, the server computing system (130) may include a plurality of computing devices connected via a network.

[0091] In an embodiment, the server computing system (130) may further include a data store computing system (1000) (hereinafter, data store) which is a storage for continuously storing and managing raw data (e.g., seed data, etc.) that forms the basis of a method (service) for providing a bilingual model. This data store may include various forms of data storage, ranging from file systems to cloud storage.

[0092] For example, a data store may include at least one database among a relational database that uses a structured query language (SQL) to define and manipulate data, a NoSQL database designed for flexibility and scalability to process unstructured and semi-structured data, a data warehouse optimized for querying and analysis by centralizing large volumes of data from multiple sources as a system used for reporting and data analysis, a data warehouse that stores large volumes of raw data in basic formats such as structured data, semi-structured data, and unstructured data, and a local storage device or Network Attached Storage (NAS) that stores data in files in a format generally accessible by a computer operating system.

[0093] - Training Computing System (150: Training Computing System)

[0094] The training computing system (150) includes at least one processor (151) and memory (152). Here, the processor (151) may be composed of at least one or a plurality of electrically connected processors among a central processing unit (CPU), a graphics processing unit (GPU), ASICs (application specific integrated circuits), DSPs (digital signal processors), DSPDs (digital signal processing devices), PLDs (programmable logic devices), FPGAs (field programmable gate arrays), controllers, microcontrollers, microprocessors, and / or other electrical units for performing functions.

[0095] In particular, depending on the embodiment, this processor (151) may be configured based on a Field Programmable Gate Array (FPGA) implementation and / or an Application Specific Integrated Circuit (ASIC), which are hardware technologies for implementing a specific digital circuit. A detailed description thereof is omitted by applying the description of the FPGA and ASIC mentioned above.

[0096] And the memory (152) may include one or more non-transient / transient computer-readable storage media such as RAM, ROM, EEPROM, EPROM, flash memory device, magnetic disk, etc. and combinations thereof. This memory (152) may store data and instructions necessary for the processor (151) to train a machine learning model.

[0097] For example, the training computing system (150) may include a model trainer (160) that trains a machine learning model stored in a user computing device (110) and / or a server computing system (130) using various training or learning techniques, such as back propagation of error.

[0098] For example, the model trainer (160) can perform backpropagation updates to one or more parameters of the machine learning model for the bilingual model based on a defined loss function.

[0099] In some embodiments, performing backpropagation of the error may include performing truncated backpropagation through time. The model trainer (160) may perform a number of generalization techniques (e.g., weight decrement, dropout, knowledge distillation, etc.) to improve the generalization ability of the machine learning model being trained.

[0100] And the model trainer (160) includes computer logic utilized to provide the desired function. The model trainer (160) may be implemented as hardware, firmware and / or software that controls a general-purpose processor. For example, in one embodiment, the model trainer (160) includes a program file stored in a storage device, loaded into memory, and executed by one or more processors. In another embodiment, the model trainer (160) includes one or more sets of computer-executable instructions stored in a tangible computer-readable storage medium, such as a RAM hard disk or an optical or magnetic medium.

[0101] Networks (170) include, but are not limited to, 3GPP (3rd Generation Partnership Project) networks, LTE (Long Term Evolution) networks, WIMAX (World Interoperability for Microwave Access) networks, Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), Bluetooth networks, satellite broadcasting networks, analog broadcasting networks and / or DMB (Digital Multimedia Broadcasting) networks.

[0102] Generally, communication through the network (170) can be performed using any type of wired and / or wireless connection through various communication protocols (e.g., TCP / IP, HTTP, SMTP, FTP), encodings or formats (e.g., HTML, XML), and / or protection schemes (e.g., VPN, Secure HTTP, SSL).

[0103]

[0104] FIG. 2 illustrates an example of a block diagram of a computing device, which is one of the components of a computing system (1000) providing a dual language model according to an embodiment of the present invention.

[0105] Referring to FIG. 2, the computing device (100) included in the user computing device (110), server computing system (130), and training computing system (150) includes a plurality of applications (e.g., applications 1 to N). Each application may include a machine learning library.

[0106] For example, applications may include text messaging applications, virtual keyboard applications, browser applications, chatbot applications, etc.

[0107] In an embodiment, the computing device (100) may include a model trainer (160) for training a machine learning model, and may store and operate the machine learning model to provide a bilingual model for input data.

[0108] Each application of the computing device (100) can communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In one embodiment, each application can communicate with each device component using an API (e.g., a public API). In one embodiment, the API used by each application may be specific to that application.

[0109]

[0110] FIG. 3 illustrates an example of a block diagram in another aspect of a computing device, which is one of the components of a computing system (1000) providing a dual language model according to an embodiment of the present invention.

[0111] Referring to FIG. 3, the computing device (200) includes a plurality of applications (e.g., Application 1 to Application N). Each application can communicate with a central intelligence layer. For example, the applications may include a virtual keyboard application, a browser application, a chatbot application, etc. In one embodiment, each application can communicate with the central intelligence layer (and a model stored therein) using an API (e.g., a common API across all applications).

[0112] And the central intelligence layer may include prompts using a plurality of machine learning models or / and language models. For example, as illustrated in FIG. 3, each machine learning model and at least some thereof may be provided for each application and managed by the central intelligence layer. In another embodiment, two or more applications may share a single machine learning model. For example, in some embodiment, the central intelligence layer may provide a single model for all applications. In some embodiment, the central intelligence layer may be included within the operating system of the computing device (200) or otherwise implemented.

[0113] The central intelligence layer can communicate with the central device data layer. The central device data layer may be a centralized data store for the computing device (200). As illustrated in FIG. 3, the central device data layer can communicate with a number of other components of the computing device, such as, for example, one or more sensors, a context manager, a device state component, and / or additional components. In some embodiments, the central device data layer can communicate with each device component using an API (e.g., a private API).

[0114] The technology described herein may refer to servers, databases, software applications, and other computer-based systems, as well as actions taken and information transmitted to or from said systems. It will be recognized that the inherent flexibility of computer-based systems allows for a wide range of possible configurations, combinations, division of tasks, and functionality between and from components. For example, the processes described herein may be implemented using a single device or component or multiple devices or components operating in combination. Databases and applications may be implemented in a single system or in a distributed system across multiple systems. Distributed components may operate sequentially or in parallel.

[0115] - Method for providing a bilingual model based on license-free data

[0116] Hereinafter, a method for a computing system (1000) according to an embodiment of the present invention to provide a dual language model will be described in detail with reference to the attached FIGS. 4 to 7.

[0117] FIG. 4 is a flowchart illustrating a method for providing a license-free data-based bilingual model according to an embodiment of the present invention.

[0118] Referring to FIG. 4, in an embodiment, the computing system (1000) can acquire conversation data based on user input based on an application of a user computing device (110). (S101)

[0119] To this end, in the embodiment, the user can input question data including a predetermined task (request) based on the user interface implemented in the application.

[0120] In this case, the above-mentioned question data may be data including keywords and / or conditions entered to obtain specific information or data. Typically, it may be in the form of natural language and / or commands requesting specific information from a database, search engine and / or information system, etc.

[0121] In addition, in the embodiment, the computing system (1000) can provide answer data for the input question data based on a pre-trained Bilingual Large Language Model (hereinafter BLLM).

[0122] Here, for reference, a bilingual model (BLLM) may refer to a deep learning model that outputs an appropriate answer (e.g., a natural language response) to a user's question (e.g., a natural language query) written in a first language and / or a second language.

[0123] For example, such a bilingual model (BLLM) is a language model trained on large-scale Korean and English corpora, and can provide response data for both Korean and English.

[0124] In embodiments of the present invention, the deep learning algorithm itself that implements the bilingual model (BLLM) is not limited or restricted, and in embodiments of the present invention, the bilingual model (BLLM) may be implemented based on at least one known deep learning algorithm capable of performing deep learning necessary for the operation of the question-answering function described above.

[0125] In addition, the pre-trained bilingual model (BLLM) can output multiple answer data for multiple question data when refinement is required for the input question data.

[0126] Accordingly, the computing system (1000) can obtain conversation data including at least one question data, answer data and / or question-answer pairs.

[0127] In the embodiment, such conversation data may be stored by matching the task and / or the number of turns (in other words, the number of times a question and answer were exchanged).

[0128] Meanwhile, the answer data included in the above conversation data may be data obtained by crawling from a specific website and / or database to provide an answer to the question, regardless of whether the license has been verified.

[0129] At this time, the above-mentioned answer data is seed data that has not undergone license verification judgment and can be generated through a specific open source call (e.g., Mistral API). Hereinafter, the above-mentioned answer data may be referred to as seed data.

[0130] Additionally, in the embodiment, the computing system (1000) can verify the license for the acquired conversation data. (S103)

[0131] Specifically, in an embodiment, the computing system (1000) can verify the license to comply with the license conditions of the data used according to rules and restrictions on the dataset used when performing a task based on the acquired conversation data.

[0132] More specifically, in an embodiment, a computing system (1000) can train a bilingual model (BLLM) to take at least one answer included in conversation data as input data and provide license suitability (in an embodiment, license information) for said answer as output data.

[0133] Accordingly, in the embodiment, the bilingual model (BLLM) can extract license information of the data constituting the input conversation data.

[0134] To this end, in the embodiment, the bilingual model (BLLM) can perform license-related searching to extract license information by linking with a specific website (e.g., Google, arXiv, etc.) and / or a database.

[0135] In addition, the license information according to the embodiment may refer to information regarding usage rights for data included in web pages, domains, files, etc., retrieved to achieve a task.

[0136] In the embodiments, such license information may include information specified for a plurality of items classified into categories 1 to 4.

[0137] For example, the first category is a data license category, and the first item included in the first category may be the right to modify data and the right to create derivative works, the second item may be the possibility of infringing the original copyright of the output, the third item may be whether rights to the prompt and output are granted, and the fourth item may be whether there is an obligation to notify the data.

[0138] The second category is the personal information and data security category, and the first item included in the second category may be the right to modify data and the right to create derivative works, the second item may be the possibility of infringing the original copyright of the output, the third item may be whether rights to the prompt and output are granted, the fourth item may be whether there is an obligation to notify personal information data, and the fifth item may be whether there is an obligation to notify security data.

[0139] The third category is a data usage period and region category, wherein the first item included in the third category may be a limitation on the data usage period, the second item may be the possibility of revoking the data license grant, the third item may be a limitation on the AI ​​model service period, and the fourth item may be a limitation on the data usage region.

[0140] The fourth category is a legal risk category, wherein the first item included in the fourth category may be the legality of the data collection process, the second item may be a conflict between data licenses, the third item may be a known dispute regarding the AI ​​model in which the data was used, and the fourth item may be the existence of risks in the license agreement.

[0141] That is, in the embodiment, the bilingual model (BLLM) can extract license information matching the conversation data by performing a search for each item of the first to fourth categories on the acquired conversation data.

[0142] In addition, in the embodiment, the bilingual model (BLLM) can determine license suitability by calculating a risk score for the corresponding conversation data based on the search results of each item included in each category.

[0143] Specifically, the determination of license conformity means that the data can be determined as an available license only if the calculated risk score satisfies the pre-established conditions. Conversely, if the pre-established conditions are not satisfied, the data may be determined as an unavailable license.

[0144] For the sake of convenience of explanation, the license conformity determined above is explained based on the dichotomy of classifying and determining usable licenses and / or unusable licenses.

[0145] That is, in the embodiment, the bilingual model (BLLM) can determine the first conversation data as an available license. If it is determined as an available license, the first conversation data can be stored as a reference document. At this time, the reference document may be various documents proving that there is no license dispute (e.g., papers, etc.).

[0146] In addition, in the embodiment, the bilingual model (BLLM) may determine the first conversation data as having an unusable license. If it is determined to have an unusable license, it may be excluded from the seed data so that it is no longer trained with the first conversation data.

[0147] In this way, in the embodiment, the computing system (1000) can verify the license for the conversation data based on a bilingual model (BLLM) that takes the acquired conversation data as input data and provides license information for the conversation data as output data.

[0148] Additionally, in the embodiment, the computing system (1000) can generate and store licensed conversation data as seed data. (S105)

[0149] In other words, in the embodiment, the computing system (1000) can generate and store only conversation data for which license suitability has been determined to be available license as seed data. Accordingly, the conversation data according to the embodiment is determined to be at least one of seed data and / or reference documents.

[0150] That is, with at least one seed data stored, in the embodiment, the computing system (1000) can build a license-free database by fine-tuning a pre-trained bilingual model (BLLM) using the seed data and / or reference documents as a training set.

[0151] FIG. 5 is a conceptual diagram illustrating the entire process of building a license-free database based on seed data and reference documents according to an embodiment of the present invention.

[0152] Referring to FIG. 5, in an embodiment, a computing system (1000) can build a license-free database (④) based on a more advanced bilingual model (BLLM) by generating seed data and / or reference documents based on conversation data (①), fine-tuning a pre-trained bilingual model (BLLM) with the generated data (②), and performing additional alignment on the fine-tuned bilingual model (BLLM) (③). Based on the license-free database thus built, continuous fine-tuning of the bilingual model (BLLM) can be performed to update the license-free database (⑤).

[0153] In other words, in the embodiment, the computing system (1000) can fine-tune the bilingual model (BLLM) based on the seed data. (S107)

[0154] In detail, in an embodiment, the computing system (1000) can perform a first fine-tuning and / or a second fine-tuning for the dual-language model (BLLM) to provide license-free data-based services based on the seed data.

[0155] Here, fine-tuning may refer to the process of fine-tuning the training process of a bilingual model (BLLM) to meet the data security and / or customer requirements of the customer who intends to use the data.

[0156] First, in an embodiment, the computing system (1000) can train an instruction module and / or a response module included in the dual language model (BLLM) so that the dual language model (BLLM) is fine-tuned for the first time.

[0157] FIG. 6 is a conceptual diagram illustrating a method for fine-tuning a bilingual model into license-free data according to an embodiment of the present invention.

[0158] Referring to FIG. 6, in the embodiment, the bilingual model (BLLM) can generate and provide answer data corresponding to question data based on the data generation module (300).

[0159] To this end, the data generation module (300) of the bilingual model (BLLM) may be composed of an instruction module (310) and / or an answer module (320).

[0160] The instruction module (310) may be a module that generates instructions to find an answer to a question or generates a reverse question for a user's question. The instructions and / or reverse questions generated may be in natural language.

[0161] The answer module (320) may be a module that generates an answer according to instructions generated based on the instruction module.

[0162] In detail, in an embodiment, the computing system (1000) can train the instruction module (310) of the bilingual model (BLLM) with at least one reference document (in an embodiment, seed data verified as available license) stored in a license-free database based on previous turns.

[0163] Accordingly, the instruction module (310) can design and generate a new instruction (400) so that the instruction can lead to the next answer by taking into account the context of the conversation data.

[0164] The instructions (400) generated at this time may be data in a first language and / or a second language that does not rely on an external model, or data that has a higher relevance to the given question than a pre-set standard, or data that clearly exists as a basis.

[0165] Additionally, in an embodiment, the computing system (1000) can train the answer module (320) of the bilingual model (BLLM) based on at least one reference document stored in a license-free database based on previous turns and instructions (400) generated by the instruction module (310).

[0166] Accordingly, the answer module (320) can design and generate an answer (500) to the corresponding instruction (400) by understanding the context of the conversation data based on a predetermined natural language understanding module.

[0167] Additionally, the data generation module (300) can match the generated instructions (400) and answers (500) (hereinafter, question-answer pairs (PR)) into a first conversation turn.

[0168] At this time, in the embodiment, the computing system (1000) can generate a multi-turn conversation by repeating the process of matching, generating, and storing the above-described question-answer pairs (PR) based on a cumulative approach.

[0169] In the embodiments, the cumulative approach may mean a method of considering previous conversation data while processing current conversation data.

[0170] That is, pre-generated question-answer pairs (PRs) can be used as reference documents and / or context for the next turn to support a continuous conversation.

[0171] In this way, in the embodiment, the computing system (1000) can perform a first fine-tuning so that the bilingual model (BLLM) generates a multi-turn conversation with a natural flow.

[0172] Next, in the embodiment, the computing system (1000) can perform a second fine-tuning to provide a bilingual-based service for the bilingual model (BLLM).

[0173] In an embodiment, the bilingual model (BLLM) can match question-answer pairs in the first language to the first general conversation turn for conversation data in which the first language is detected.

[0174] In addition, in the embodiment, the bilingual model (BLLM) can convert the conversation data in which the first language is detected into a second language.

[0175] At this time, as the first language is converted into the second language, the dual language model (BLLM) in the embodiment can set a search country and / or search range corresponding to the second language.

[0176] In addition, in the embodiment, the bilingual model (BLLM) can provide answer data (hereinafter, converted answer data) converted into a second language based on the data generation module (300) of the first fine-tuned bilingual model (BLLM) for question data converted into a second language (hereinafter, converted question data).

[0177] In other words, the above-mentioned converted response data and the response data of the conversation data in which the first language was detected may differ. This is because the scope of information search based on each language and the resulting search results differ. For example, results such as papers in the original language may be exposed only when searched in a specific language.

[0178] Additionally, in the embodiment, the bilingual model (BLLM) can match the transformed question data and / or transformed answer data to a first transformed conversation turn (i.e., a question-answer pair in a second language).

[0179] In addition, similar to the first fine-tuning described above, the bilingual model (BLLM) in the embodiment can generate a multi-turn conversation in the second language by repeating the process of generating and saving the matched first conversion conversation turn.

[0180] Additionally, in the embodiment, the bilingual model (BLLM) can train the data generation module (300) by matching the first general conversation turn and / or the first transformation conversation turn. For example, the data generation module (300) can be trained based on a cross-language instruction tuning method that transfers the effect to the second language after fine-tuning for the first language.

[0181] Accordingly, in an embodiment, the computing system (1000) can generate answer data by combining search results of the first language and the second language and providing the generated answer data in the input language, even if a task is input in one of the first language and the second language.

[0182] In other words, by performing a second fine-tuning to provide bilingual-based services, the gap in results that varies depending on language differences is reduced, thereby having the effect of providing a higher-performance bilingual model.

[0183] Additionally, in the embodiment, the computing system (1000) can perform additional training on the fine-tuned bilingual model (BLLM) based on accumulated multiple conversation data. (S109)

[0184] Here, additional learning for the above-mentioned bilingual model (BLLM) may refer to reinforcement learning for data alignment (e.g., DPO, SFT, RLHF methods).

[0185] In the embodiment, the computing system (1000) can rank and sort multiple answer data according to a predetermined standard during the first and second fine-tuning.

[0186] Additionally, in the embodiment, the computing system (1000) can train a bilingual model (BLLM) to select a higher predetermined answer data among the sorted answer data.

[0187] Accordingly, in the embodiment, the computing system (1000) can improve the linguistic and / or technical performance of the bilingual model (BLLM).

[0188] Figure 7 is a table showing the performance of a bilingual model (BLLM) that has undergone additional learning according to an embodiment of the present invention.

[0189] Referring to FIG. 7, in an embodiment, a computing system (1000) can verify the linguistic and / or technical performance of a bilingual model (BLLM) by performing machine translation and multi-task benchmark tests on the bilingual model (BLLM) and calculating a benchmark score (MT-bench Score).

[0190] The multiple benchmark test according to the embodiment may be a test that comprehensively evaluates the conversational and reasoning abilities of a large language model.

[0191] In detail, in the embodiment, the computing system (1000) may perform additional learning to improve the performance of the corresponding detailed task based on the linguistic and / or technical performance verified based on multiple benchmark tests. For example, additional learning may be performed on a given detailed task by extending SFT (supervised fine-tuning) based on more conversational data and applying RLHF (reinforcement learning).

[0192] For example, the computing system (1000) can perform additional learning to improve the writing performance of the Korean model (700) because the performance of the Korean model (700) in the detailed task of writing was measured to be lower than that of the English model (600). In addition, since the performance of non-verbal tasks was measured to be lower than that of linguistic tasks in both the English model (600) and the Korean model (700), similar additional learning can be performed.

[0193] That is, as additional learning is performed, the computing system (1000) in the embodiment improves the ability to understand and generate conversations based on conversational data (in the embodiment, conversational data) and fine-tunes the existing model, thereby having the effect of creating a model specialized for a given task.

[0194] In this way, through the reinforcement learning described above, the computing system (1000) in the embodiment can provide a higher performance bilingual model (BLLM).

[0195] Additionally, in an embodiment, the computing system (1000) can build a license-free database based on a learned bilingual model (BLLM). (S111)

[0196] In detail, in an embodiment, the computing system (1000) can build a license-free database by storing conversation data for which the license has been verified by a first language and / or a second language.

[0197] At this time, the conversation data with the verified license mentioned above may be conversation data with matched license information.

[0198] In the embodiment, the license-free database may include data generated internally or collected under available licenses, rather than data originating from an external agent.

[0199] Additionally, conversation data determined to be an unusable license may also be stored in the aforementioned license-free database. Such unusable licenses are not stored in the license-free database but may be stored by matching them as unusable for exclusion from use and / or non-exposure.

[0200] In addition, in the embodiment, the computing system (1000) can perform additional fine-tuning by increasing the data size of the license-free database through continuous conversation data collection.

[0201] Accordingly, the data included in the above-mentioned license-free database can be updated at predetermined intervals.

[0202] Additionally, in an embodiment, the computing system (1000) may provide a bilingual model based on a constructed license-free database and / or a service using the same. (S113)

[0203] In an embodiment, the computing system (1000) can execute a license-free based service by setting a search range according to user input.

[0204] At this time, the search range set above may differ from the language of the detected user input. For example, the user input may be a first language, but the set search range may be a second language.

[0205] Additionally, in the embodiment, the computing system (1000) can acquire a task included in the user input according to the set content. Additionally, it can determine a work type for the acquired task.

[0206] At this time, by referring to the task type matched with the conversation data stored in the license-free database, an answer according to the determined task type can be provided.

[0207] Accordingly, in the embodiment, the computing system (1000) can provide a service according to a task type determined through a predetermined multi-turn conversation.

[0208]

[0209] In summary, the method and system for providing a license-free data-based bilingual model according to an embodiment of the present invention generate conversational data based on a cumulative approach and fine-tune and / or train a bilingual model, thereby providing results that meet given conditions such as task type and the number of multi-turns, which has the effect of increasing applicability and further expanding the scale of the data.

[0210] In addition, the method and system for providing a bilingual model based on license-free data according to an embodiment of the present invention automatically generates license-free data by fine-tuning seed data and builds a license-free database, thereby improving work productivity by performing repetitive tasks on behalf of humans, saving costs by not relying on API calls, and utilizing accumulated data over the long term.

[0211] The embodiments according to the present invention described above may be implemented in the form of program instructions that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., either individually or in combination. The program instructions recorded on the computer-readable recording medium may be those specifically designed and configured for the present invention or 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. Hardware devices may be modified into one or more software modules to perform processing according to the present invention, and vice versa.

[0212] The specific embodiments described in this invention are examples and do not limit the scope of the invention in any way. For the sake of brevity of the specification, descriptions of prior electronic configurations, control systems, software, and other functional aspects of said systems may be omitted. Additionally, the connections of lines or connecting members between components shown in the drawings are illustrative of functional connections and / or physical or circuit connections, and may be replaced or additionally represented as various functional connections, physical connections, or circuit connections in actual devices. Furthermore, unless specifically stated as “essential,” “importantly,” etc., a component may not be strictly necessary for the application of the invention.

[0213] Furthermore, although the detailed description of the present invention has been explained with reference to preferred embodiments of the invention, those skilled in the art or those with ordinary knowledge in the relevant technical field will understand that various modifications and changes can be made to the invention without departing from the spirit and technical scope of the invention as set forth in the claims below. Accordingly, the technical scope of the present invention should not be limited to the contents described in the detailed description of the specification, but should be determined by the claims.

[0214] The mode for carrying out the invention is the same as the best mode for carrying out the invention described above.

[0215] The present invention can be utilized in the AI ​​model development industry, which generates and sells bilingual models and training data without license restrictions. Since the language model developed through the present invention can be applied to various service industries such as multilingual translation and chatbots and can be commercially repeated, it has industrial applicability.

Claims

1. A method for a computing system including memory and a processor to provide a bilingual model, A step of acquiring conversational data based on user input based on a pre-trained bilingual model; A step of verifying the license for the above-mentioned acquired conversation data; A step of generating and storing the conversation data with the above-mentioned verified license as seed data; A step of fine-tuning the bilingual model based on the seed data above; A step of conducting additional training on the fine-tuned bilingual model based on the accumulated multiple conversation data; A step of constructing a license-free database based on the above-mentioned learned bilingual model; and A step comprising providing a bilingual model based on the above-described license-free database and a service using the same Method for providing a license-free data-based bilingual model.

2. In Paragraph 1, The step of acquiring the above conversation data is, A step of obtaining question data including a predetermined task entered based on a user interface, and A step of providing answer data for the acquired question data based on the above-mentioned pre-trained bilingual model, and A step of matching the above question data and the above answer data into question-answer pairs, and A step comprising storing at least one question data, answer data, and question-answer pairs as conversation data Method for providing a license-free data-based bilingual model.

3. In Paragraph 2, The step of storing at least one question data, answer data, and question-answer pairs as conversation data is: A step comprising matching and storing the type of task and the number of turns for the above question-answer pairs. Method for providing a license-free data-based bilingual model.

4. In Paragraph 1, The step of verifying the license for the above-mentioned acquired conversation data is, A step of performing license-related searching for the acquired conversation data by linking with a predetermined website and database, and A step of obtaining license information, which is information regarding the right to use data for at least one of the web page, domain, and file searched through the above search, and A step comprising determining the conversation data as either an available license or an unavailable license based on the acquired license information. Method for providing a license-free data-based bilingual model.

5. In Paragraph 4, The step of fine-tuning the above-mentioned bilingual model is, A step of controlling an instruction module included in the above-mentioned bilingual model to use conversation data determined by the above-mentioned available license as a training data set, and A step of controlling the instruction module to exclude conversation data determined to be unusable licenses from the training data set, and A step comprising controlling an answer module included in the above-described bilingual model to generate an answer to an instruction generated from the above-described learned instruction module. Method for providing a license-free data-based bilingual model.

6. In Paragraph 5, The step of fine-tuning the above-mentioned bilingual model is, The step of matching and storing the generated instructions and answers as question-answer pairs, and A step comprising repeating the matching and storage of multiple above-mentioned question-answer pairs based on a cumulative approach Method for providing a license-free data-based bilingual model.

7. In Paragraph 6, The step of fine-tuning the above-mentioned bilingual model is, A step of matching conversation data in which a first language is detected to a first general conversation turn, and The step of converting conversation data in which the first language is detected into a second language, and A step of obtaining converted answer data converted into the second language based on the answer module for the converted question data converted into the second language, and A step of matching the above-mentioned conversion question data and the above-mentioned conversion answer data to a first conversion conversation turn, and The method further includes the step of fine-tuning the bilingual model by matching the first general conversation turn and the first transformation conversation turn. Method for providing a license-free data-based bilingual model.

8. In Paragraph 1, The step of conducting additional training on the above-mentioned bilingual model is, A step comprising performing reinforcement learning on the bilingual model based on at least one of the DPO, SFT, and RLHF methods. Method for providing a license-free data-based bilingual model.

9. In Paragraph 7, The step of building the above license-free database is, A step of additionally storing conversation data determined as at least one of an available license and an unavailable license for each of the first language and the second language according to the additional learning progress above, and A step further comprising updating the license-free database according to the additional storage above Method for providing a license-free data-based bilingual model.

10. At least one memory; and It includes at least one processor that reads at least one application stored in the memory and provides a bilingual model; The instructions of the above processor are, A step of acquiring conversational data based on user input based on a pre-trained bilingual model, and A step of verifying the license for the above-mentioned acquired conversation data, and The step of generating and storing the conversation data with the above-mentioned license verified as seed data, and A step of fine-tuning the bilingual model based on the seed data, and A step of conducting additional training on the fine-tuned bilingual model based on the accumulated multiple conversation data, and A step of constructing a license-free database based on the above-mentioned learned bilingual model and A step comprising providing a bilingual model based on the above-described license-free database and a service using the same License-free data-based bilingual model provision system.

11. In Paragraph 10, A plurality of neurons comprising an array including at least one register, at least one programmable logic, and at least one input interface; a plurality of synapse circuits storing synapse weights that regulate the connection strength between the plurality of neurons; and at least one routing network that controls the data flow between the plurality of neurons; Each of the plurality of neurons further comprises a Field Programmable Gate Array (FPGA) implementation for a predetermined artificial neural network, wherein each neuron is connected to at least one other neuron through the routing network to establish a transmission path for the weights. License-free data-based bilingual model provision system.

12. In Paragraph 10, A plurality of neurons organized into an array comprising at least one register, at least one microprocessor, and at least one input; and a plurality of synapse circuits storing synapse weights that regulate the connection strength between the plurality of neurons; comprising Each of the plurality of neurons further comprises an Application Specific Integrated Circuit (ASIC) for a predetermined artificial neural network, which is connected to at least one other neuron through any one of the plurality of synaptic circuits. License-free data-based bilingual model provision system.

Images