Method for providing intelligent response agent based on advanced inference and estimation function, and system therefor

The deep-learning neural network-based intelligent response agent system addresses the challenge of providing accurate responses to in-depth questions by employing a multi-step inference process, generating high-quality response data that is faithful and evidential, thereby improving the handling of complex queries related to professional documents.

WO2024228577A1PCT designated stage expired Publication Date: 2024-11-07LG MANAGEMENT DEV INST CO LTD

Patent Information

Application Number
PCT/KR2024/005992
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2024-01-25
Filing Date
2024-05-03
Publication Date
2024-11-07

AI Technical Summary

Technical Problem

Conventional question-answering deep learning systems face limitations in handling in-depth questions that require advanced inference and estimation, particularly when dealing with professional documents, as they struggle to provide accurate and high-quality responses due to the exponential increase in syntactic patterns across natural languages and the reliance on factoid-based training data sets.

Method used

A deep-learning neural network-based intelligent response agent system that implements a step-by-step process for advanced inference and estimation functions, utilizing a multi-step inference process involving association selection, rationale generation, and systematic composition to generate response data that includes direct answers, explanatory sentences, and background knowledge, while detecting evidence paragraphs through Prompt Engineering algorithms.

Benefits of technology

The system provides faithful, fact-based, and evidential response data for in-depth inquiries, enhancing the reliability and validity of answers by linking generated response data with supporting information, thus overcoming the limitations of conventional models in handling complex queries.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure KR2024005992_07112024_PF_FP_ABST
    Figure KR2024005992_07112024_PF_FP_ABST
Patent Text Reader

Abstract

A method for providing an intelligent response agent based on an advanced inference and estimation function, and a system therefor, according to an embodiment of the present invention, use a deep-learning neural network implementing a stepwise process for an advanced inference and estimation function, and thus can generate and provide response data on a specialized-document-related query.
Need to check novelty before this filing date? Find Prior Art

Description

Method and system for providing an intelligent response agent based on advanced inference and estimation functions

[0001] The present invention relates to a method and system for providing an intelligent response agent based on advanced inference and estimation capabilities. More specifically, the present invention relates to a method and system for generating and providing response data for queries related to specialized documents using a deep-learning neural network that implements a step-by-step process for advanced inference and estimation capabilities.

[0002] Deep learning research on question-answering aims to provide accurate answers to users' questions through artificial intelligence models.

[0003] Typically, conventional question-answering deep learning systems are implemented as systems that parse a given query to construct semantic elements into patterns (rules) and extract responses from a structured database in the same language as the Structured Query Language (SQL).

[0004] However, as the number of syntactic patterns to be built for processing various natural languages ​​increases exponentially, the utility of question-answering deep learning systems is declining.

[0005] Thus, recently, the Machine Reading Comprehension (MRC) model, a technology in which an artificial intelligence algorithm analyzes a query on its own and finds an optimized answer, is being widely used.

[0006] Specifically, a machine reading model is a system in which a computer reads and understands a given text (text for understanding a query, such as a sentence and / or document) and then finds and presents an answer to the query.

[0007] However, in order to effectively utilize a question-answering deep learning system using a machine reading comprehension model, a question must first exist, and if the question is not created, it is difficult to respond using a machine reading comprehension model.

[0008] Meanwhile, conventional machine reading models have been trained for a long time through various previous models and benchmark data sets, but most of the training data sets are based on question-answering data sets that do not require a deep understanding of documents.

[0009] Accordingly, the development of deep learning-based question-answering processes that generate responses to in-depth questions requiring advanced inference and estimation is very limited.

[0010] In other words, previous studies have had limitations in handling more advanced question-answering tasks beyond fact-based factoid QA, despite significantly improving machine reading comprehension answering capabilities.

[0011] As a specific example, most of the existing query data are mainly fact-checking questions such as 'what', 'when', 'where', and 'who', and machine reading comprehension models can extract and provide short answers from specific paragraphs related to these questions, but they have difficulty providing answers to in-depth questions such as 'how' or 'why'.

[0012] That is, according to the conventional technology, there is a problem that the accuracy and quality of the deep learning model's answers are low for in-depth queries that require innovation in knowledge processing and inference related to specialized documents (e.g., scientific and technological papers) containing in-depth technical information.

[0013] One embodiment of the present invention aims to implement a method and system for generating and providing response data for queries related to specialized documents using a deep-learning neural network that implements a step-by-step process for advanced inference and estimation functions.

[0014] In addition, one embodiment of the present invention aims to implement a method and system for providing generated response data and supporting information by linking them with each other.

[0015] However, the technical problems to be solved by the present invention and embodiments of the present invention are not limited to the technical problems described above, and other technical problems may exist.

[0016] A method for providing an intelligent response agent based on an advanced reasoning and estimation function according to an embodiment of the present invention is a method for providing an intelligent response agent based on an advanced reasoning and estimation function by a response agent application executed by at least one processor of a terminal, the method comprising: obtaining query data specifying a user's question; determining a full-text document for the obtained query data; detecting an evidence paragraph including at least one paragraph linked to the query data on the determined full-text document; generating response data specifying an answer to the question based on the detected evidence paragraph; and providing the generated response data.

[0017] In another aspect, a method for providing an intelligent response agent based on an advanced inference and estimation function according to an embodiment of the present invention further includes a step of acquiring basis data, which is base data for generating the response data, based on the detected evidence paragraph, and a step of generating the response data based on the acquired basis data.

[0018] In another aspect, the basis data includes at least one of main answer data including a direct answer to the query data, explanatory sentence data including a detailed explanation of the direct answer, and auxiliary information data including background knowledge information related to the question and the answer.

[0019] In another aspect, the step of detecting the above-mentioned evidence paragraph includes a step of detecting the above-mentioned evidence paragraph based on a prompt engineering algorithm.

[0020] In another aspect, a method for providing an intelligent response agent based on an advanced inference and estimation function according to an embodiment of the present invention further includes a step of inputting the determined specialized document and the query data into a response agent model that executes a multi-step inference process, and a step of generating the response data based on the response agent model, wherein the multi-step inference process includes an associative selection process for detecting the evidence paragraph, a rationale generation process for obtaining the rationale data, and a systematic composition process for generating the response data.

[0021] In another aspect, the step of generating the response data based on the response agent model includes the step of executing the association selection process and the basis generation process by reflecting at least one of previously stored query data and response data.

[0022] In another aspect, the step of providing the generated response data includes a step of providing basis information for the response data by matching the response data with the response data, wherein the basis information includes at least two or more of the basis data, the evidence paragraph data, the evidence paragraph identification information, the professional document data, the professional document identification information, and the query data.

[0023] In another aspect, the step of providing the generated response data further includes the step of dividing the response data into at least one sentence unit, the step of detecting the basis information for each of the divided sentences, and the step of matching the detected basis information for each sentence to a sentence corresponding to each basis information and providing the same.

[0024] In another aspect, the step of determining a specialized document for the query data includes at least one of the steps of: determining a specialized document for the query data based on at least one specialized document input by the user; determining a specialized document for the query data based on a specialized document corresponding to at least one specialized document identification information input by the user; and detecting at least one specialized document having a correlation with the query data exceeding a preset standard based on a predetermined deep learning neural network, and determining a specialized document for the query data based on the detected at least one specialized document.

[0025] In another aspect, a system for providing an intelligent response agent based on an advanced reasoning and estimation function according to an embodiment of the present invention includes at least one memory in which a response agent application is stored; and at least one processor for reading out the response agent application stored in the memory to provide an intelligent response agent based on an advanced reasoning and estimation function; wherein the instructions of the response agent application include instructions for performing the steps of: obtaining query data specifying a user's question; determining a full-text document for the obtained query data; detecting an evidence paragraph including at least one paragraph linked to the query data on the determined full-text document; generating response data specifying an answer to the question based on the detected evidence paragraph; and providing the generated response data.

[0026] A method and system for providing an intelligent response agent based on an advanced reasoning and estimation function according to one embodiment of the present invention generates and provides response data for queries related to specialized documents using a deep-learning neural network that implements a step-by-step process for an advanced reasoning and estimation function, thereby having the effect of providing response data that is faithful to the question, based on facts (hallucination-controlled), and presents a clear basis (evidential) for in-depth queries related to specialized documents based on an advanced reasoning (chain-of-thought reasoning) and estimation (speculation) function implemented in a process similar to human cognitive reasoning.

[0027] In addition, the method and system for providing an intelligent response agent based on an advanced inference and estimation function according to one embodiment of the present invention can provide generated response data and supporting evidence in a mutually linked manner, thereby enabling an intuitive and easy confirmation of objective evidence including the basis for generating the response data, thereby further improving the reliability and validity of the response data.

[0028] However, the effects that can be obtained from the present invention are not limited to the effects mentioned above, and other effects that are not mentioned can be clearly understood from the description below.

[0029] FIG. 1 is a conceptual diagram of an intelligent response agent system according to one embodiment of the present invention.

[0030] Figure 2 is an internal block diagram of a terminal according to one embodiment of the present invention.

[0031] FIG. 3 and FIG. 4 are examples of drawings for explaining a multi-step inference process of a response deep learning model according to one embodiment of the present invention.

[0032] FIG. 5 is a flowchart illustrating a method for providing an intelligent response agent based on advanced inference and estimation functions according to one embodiment of the present invention.

[0033] FIG. 6 is a conceptual diagram illustrating a method for providing an intelligent response agent based on advanced inference and estimation functions according to one embodiment of the present invention.

[0034] FIG. 7 is an example of a drawing for explaining a method for obtaining response data for a given query data based on a response agent model according to one embodiment of the present invention.

[0035] FIGS. 8 to 10 are examples of drawings for explaining a method of providing predetermined response data based on a response output interface according to one embodiment of the present invention.

[0036] FIG. 11 is an example of a drawing for explaining a method for determining a base document based on basis information according to one embodiment of the present invention.

[0037] The present invention is capable of various modifications and 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, as well as the methods for achieving them, will become clear with reference to the embodiments described in detail below together with the drawings. However, the present invention is not limited to the embodiments disclosed below and can be implemented in various forms. In the following embodiments, the terms "first," "second," etc. are not used in a limiting sense but are used for the purpose of distinguishing one component from another. Furthermore, the singular expression includes the plural expression unless the context clearly indicates otherwise. Furthermore, terms such as "include" or "have" indicate the presence of a feature or component described in the specification, and do not preemptively exclude the possibility that one or more other features or components may be added. Furthermore, in the drawings, the sizes 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 shown for convenience of explanation, and thus the present invention is not necessarily limited to what is shown.

[0038] 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.

[0039] FIG. 1 is a conceptual diagram of an intelligent response agent system according to one embodiment of the present invention.

[0040] Referring to FIG. 1, an intelligent response agent system (1000) according to an embodiment of the present invention can implement an intelligent response agent service that obtains and provides response data for queries related to specialized documents by using a deep-learning neural network that implements a step-by-step process for advanced inference and estimation functions.

[0041] In an embodiment, an intelligent response agent system (1000) implementing an intelligent response agent service may include a terminal (100), a response agent providing server (200), and a network (300: Network).

[0042] At this time, the terminal (100) and / or the response agent providing server (200) can be connected through a network (300).

[0043] Here, the network (300) according to the embodiment means a connection structure that enables information exchange between each node, such as a terminal (100) and / or a response agent providing server (200).

[0044] Examples of the network (300) include, but are not limited to, a 3rd Generation Partnership Project (3GPP) network, a Long Term Evolution (LTE) network, a World Interoperability for Microwave Access (WIMAX) network, the Internet, a Local Area Network (LAN), a Wireless Local Area Network (Wireless LAN), a Wide Area Network (WAN), a Personal Area Network (PAN), a Bluetooth network, a satellite broadcasting network, an analog broadcasting network, and / or a Digital Multimedia Broadcasting (DMB) network.

[0045] Hereinafter, with reference to the attached drawings, a terminal (100) and a response agent providing server (200) that implement an intelligent response agent system (1000) will be described in detail.

[0046] - Terminal (100)

[0047] A terminal (100) according to an embodiment of the present invention may be a predetermined computing device on which a response agent application (hereinafter, “application”) providing an intelligent response agent service is installed.

[0048] In detail, from a hardware perspective, the terminal (100) may include a mobile type computing device (100-1) and / or a desktop type computing device (100-2) on which an application is installed.

[0049] Here, the mobile type computing device (100-1) may be a mobile device on which an application is installed.

[0050] For example, a mobile type computing device (100-1) may include a smart phone, a mobile phone, a digital broadcasting device, a personal digital assistant (PDA), a portable multimedia player (PMP), and / or a tablet PC.

[0051] Additionally, the desktop type computing device (100-2) may be a wired / wireless communication-based device on which an application is installed.

[0052] For example, a desktop type computing device (100-2) may include a personal computer such as a fixed desktop PC, a laptop computer, and / or an ultrabook.

[0053] According to an embodiment, the terminal (100) may further include a predetermined server computing device that provides an intelligent response agent service environment.

[0054] Figure 2 is an internal block diagram of a terminal (100) according to one embodiment of the present invention.

[0055] Meanwhile, referring to FIG. 2, from a functional perspective, the terminal (100) may include a memory (110), a processor assembly (120), a communication processor (130), an interface unit (140), an input system (150), a sensor system (160), and a display system (170). In an embodiment, the terminal (100) may include the above components within a housing.

[0056] In detail, the memory (110) can store an application (111).

[0057] At this time, the application (111) can store one or more of various application programs, data, and commands for providing an intelligent response agent service environment.

[0058] That is, the memory (110) can store commands and data that can be used to create an intelligent response agent service environment.

[0059] Additionally, the memory (110) may include a program area and a data area.

[0060] Here, the program area according to the embodiment can be linked between the operating system (OS) that boots the terminal (100) and the functional elements.

[0061] Additionally, the data area according to the embodiment may store data generated according to the use of the terminal (100).

[0062] Additionally, the memory (110) may include at least one non-transitory computer-readable storage medium and one or more temporary computer-readable storage medium.

[0063] For example, the memory (110) may be a variety of storage devices such as a ROM, EPROM, flash drive, hard drive, etc., and may include web storage that performs the storage function of the memory (110) on the Internet.

[0064] The processor assembly (120) may include at least one processor capable of executing instructions of an application (111) stored in the memory (110) to perform various tasks for creating an intelligent response agent service environment.

[0065] In an embodiment, the processor assembly (120) can control the overall operation of the components through an application (111) of the memory (110) to provide an intelligent response agent service.

[0066] In detail, the processor assembly (120) may be a system on chip (SOC) suitable for a terminal (100) including a central processing unit (CPU) and / or a graphics processing unit (GPU).

[0067] Additionally, the processor assembly (120) can execute an operating system (OS) and / or application programs stored in the memory (110).

[0068] Additionally, the processor assembly (120) can control each component mounted on the terminal (100).

[0069] Additionally, the processor assembly (120) can communicate with each component internally via a system bus and can include one or more predetermined bus structures including a local bus.

[0070] Additionally, the processor assembly (120) may be implemented to include at least one of application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors, and / or other electrical units for performing functions.

[0071] The communication processor (130) may include one or more devices for communicating with external devices. The communication processor (130) may communicate via a wireless network.

[0072] In detail, the communication processor (130) can communicate with a terminal (100) that stores a content source for implementing an intelligent response agent service environment.

[0073] Additionally, the communication processor (130) can communicate with various user input components, such as a controller that receives user input.

[0074] In an embodiment, the communication processor (130) can transmit and receive various data related to the intelligent response agent service to and from another terminal (100) and / or an external server.

[0075] This communication processor (130) can wirelessly transmit and receive data with at least one of a base station, an external terminal (100), and an arbitrary server on a mobile communication network constructed through a communication device capable of performing technical standards or communication methods for mobile communication (e.g., LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), 5G NR (New Radio), WIFI) or short-range communication methods.

[0076] The sensor system (160) may include various sensors such as an image sensor (161), a position sensor (IMU, 163), an audio sensor (165), a distance sensor, a proximity sensor, and a contact sensor.

[0077] Here, the image sensor (161) can capture images (images and / or videos, etc.) of the physical space around the terminal (100).

[0078] In detail, the image sensor (161) can capture a predetermined physical space through a camera positioned toward the outside of the terminal (100).

[0079] As an example, the image sensor (161) can be placed on the front or / and rear of the terminal (100) to capture a physical space in the direction in which it is placed.

[0080] In an embodiment, the image sensor (161) can capture and acquire various images (e.g., professional document images, etc.) related to the intelligent response agent service.

[0081] Such an image sensor (161) may include an image sensor device and an image processing module.

[0082] In detail, the image sensor (161) can process still images or moving images obtained by an image sensor device (e.g., CMOS or CCD).

[0083] In addition, the image sensor (161) can process still images or moving images acquired through the image sensor device using an image processing module to extract necessary information and transmit the extracted information to the processor.

[0084] Such an image sensor (161) may be a camera assembly including at least one camera.

[0085] Here, the camera assembly may include a general camera that captures the visible light band, and may further include a special camera such as an infrared camera or a stereo camera.

[0086] In addition, the image sensor (161) as described above may be included in the terminal (100) and operated according to an embodiment, or may be included in an external device (e.g., an external server, etc.) and operated through linkage based on the communication processor (130) and / or interface unit (140) described above.

[0087] The position sensor (IMU, 163) can detect at least one of the movement and acceleration of the terminal (100). For example, it can be formed by a combination of various position sensors such as an accelerometer, a gyroscope, and / or a magnetometer.

[0088] Additionally, the position sensor (IMU, 163) can recognize spatial information about the physical space around the terminal (100) by working in conjunction with a position communication processor (130), such as the GPS of the communication processor (130).

[0089] The audio sensor (165) can recognize sounds around the terminal (100).

[0090] In detail, the audio sensor (165) may include a microphone capable of detecting voice input from a user using the terminal (100).

[0091] In an embodiment, the audio sensor (165) can receive voice data required for an intelligent response agent service from a user.

[0092] The interface unit (140) can connect the terminal (100) to one or more other devices so that they can communicate with each other.

[0093] In detail, the interface unit (140) may include wired and / or wireless communication devices compatible with one or more different communication protocols.

[0094] Through this interface unit (140), the terminal (100) can be connected to multiple input / output devices.

[0095] For example, the interface unit (140) can be connected to an audio output device such as a headset port or speaker to output audio.

[0096] As an example, the audio output device is described as being connected through the interface unit (140), but an embodiment in which it is installed inside the terminal (100) may also be included.

[0097] Additionally, for example, the interface unit (140) may be connected to an input device such as a keyboard and / or mouse to obtain user input.

[0098] This interface unit (140) 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.

[0099] The input system (150) can detect user input (e.g., gestures, voice commands, button operations, or other types of input) related to the intelligent response agent service.

[0100] In detail, the input system (150) may include a predetermined button, a touch sensor, an image sensor (161) for detecting user motion input, and / or an audio sensor (165) for detecting user voice input.

[0101] Additionally, the input system (150) can be connected to an external controller through the interface unit (140) to receive user input.

[0102] The display system (170) can output various information related to the intelligent response agent service as graphic images.

[0103] As an example, the display system (170) can display various user interfaces, images and / or texts for an intelligent response agent service.

[0104] Such displays may include, but are not limited to, 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 / or an e-ink display.

[0105] Additionally, according to an embodiment, the display system (170) may include a display (171) that outputs an image and a touch sensor (173) that detects a user's touch input.

[0106] For example, the display (171) can be implemented as a touch screen by forming a mutual layer structure or forming an integral structure with the touch sensor (173).

[0107] This touch screen can function as a user input unit that provides an input interface between the terminal (100) and the user, and at the same time, provide an output interface between the terminal (100) and the user.

[0108] Meanwhile, the terminal (100) according to the embodiment of the present invention can perform deep learning related to an intelligent response agent service based on a predetermined deep learning neural network.

[0109] Here, the deep learning neural network according to the embodiment may include, but is not limited to, OpenAI GPT, Instruct GPT, Bi-LSTM (Bidirectional LSTM), LSTM (Long Short-Term Memory models), MLP (Multi-Layer Perceptron), EfficientNet, ResNet, ARIMA (Autoregressive Integrated Moving Average), VAR (Vector Auto Regression), RNN (Recurrent Neural Networks), GRU (Gated Recurrent Unit), GAN (Generative Adversarial Networks), DualStyleGAN, StyleGAN, Graph Convolution Network (GCN), CNN (Convolution Neural Network, CNN), DPSNet (Deep Plane Sweep Network, DPSNet), AGN (Attention Guided Network, AGN), R-CNN (Regions with CNN features), Fast R-CNN, Faster R-CNN, Mask R-CNN, and / or U-Net network.

[0110] In detail, in an embodiment, the terminal (100) can perform deep learning required for an intelligent response agent service by linking with at least one deep learning neural network capable of implementing a response agent model according to an embodiment of the present invention.

[0111] At this time, in the embodiment, the response agent model can perform deep learning by inputting at least one specialized document (e.g., a thesis and / or report, etc.) data and query data, and outputting at least one response data based on the input specialized document data and query data.

[0112] On the other hand, depending on the embodiment, the terminal (100) may further perform at least some of the functional operations performed by the response agent provision server (200) described below.

[0113] - Response Agent Providing Server (200: Response Agent Providing Server)

[0114] Meanwhile, the response agent provision server (200) according to an embodiment of the present invention can perform a series of processes for providing an intelligent response agent service.

[0115] In detail, in the embodiment, the response agent provision server (200) can provide an intelligent response agent service by exchanging data necessary to drive the response agent provision process with an external device, such as a terminal (100).

[0116] In more detail, in an embodiment, the response agent providing server (200) can provide an environment in which an application (111) can operate on an external device (in an embodiment, a mobile type computing device (100-1) and / or a desktop type computing device (100-2) etc.).

[0117] To this end, the response agent providing server (200) may include application programs, data and / or commands for the application (111) to operate, and may transmit and receive various data based thereon with the external device.

[0118] Additionally, in the embodiment, the response agent providing server (200) can obtain query data.

[0119] Here, query data according to the embodiment may mean data specifying a user's question related to a given specialized document (e.g., a thesis and / or report, etc.).

[0120] Additionally, in the embodiment, the response agent providing server (200) can determine a specialized document for the acquired query data.

[0121] That is, in the embodiment, the response agent providing server (200) can determine at least one specialized document to be used when generating response data for acquired query data.

[0122] Additionally, in the embodiment, the response agent providing server (200) can input the determined specialized document and query data into the response agent model according to the embodiment of the present invention.

[0123] In addition, in the embodiment, the response agent providing server (200) can obtain response data for the input query data by linking with the response agent model that has received the professional document and query data.

[0124] Here, response data according to the embodiment may mean data specifying an answer to a user's question related to a given specialized document (e.g., a thesis and / or report, etc.).

[0125] Additionally, in the embodiment, the response agent providing server (200) can provide the acquired response data in a predetermined manner (e.g., display output, etc.).

[0126] Additionally, in the embodiment, the response agent provision server (200) can perform deep learning required for an intelligent response agent service based on a predetermined deep learning neural network.

[0127] In detail, in an embodiment, the response agent provision server (200) can perform deep learning required for an intelligent response agent service by linking with at least one deep learning neural network capable of implementing a response agent model according to an embodiment of the present invention.

[0128] In more detail, in the embodiment, the response agent providing server (200) can read out a predetermined deep learning neural network driving program constructed to perform the above deep learning from the memory module (230).

[0129] And the response agent providing server (200) can perform deep learning required for the intelligent response agent service according to the extracted predetermined deep learning neural network system.

[0130] Here, the deep learning neural network according to the embodiment may include, but is not limited to, OpenAI GPT, Instruct GPT, Bi-LSTM (Bidirectional LSTM), LSTM (Long Short-Term Memory models), MLP (Multi-Layer Perceptron), EfficientNet, ResNet, ARIMA (Autoregressive Integrated Moving Average), VAR (Vector Auto Regression), RNN (Recurrent Neural Networks), GRU (Gated Recurrent Unit), GAN (Generative Adversarial Networks), DualStyleGAN, StyleGAN, Graph Convolution Network (GCN), CNN (Convolution Neural Network, CNN), DPSNet (Deep Plane Sweep Network, DPSNet), AGN (Attention Guided Network, AGN), R-CNN (Regions with CNN features), Fast R-CNN, Faster R-CNN, Mask R-CNN, and / or U-Net network.

[0131] At this time, depending on the embodiment, the deep learning neural network may be directly included in the response agent providing server (200) or implemented as a separate device and / or server from the response agent providing server (200).

[0132] In the following description, the deep learning neural network is described as being implemented by being included in the response agent provision server (200), but is not limited thereto.

[0133] In addition, in the embodiment, the response agent provision server (200) can store and manage various application programs, commands and / or data for implementing an intelligent response agent service.

[0134] As an example, the response agent providing server (200) can store and manage at least one or more query data, specialized document data, response data, basis information, data processing algorithm, deep learning algorithm, and / or user interface.

[0135] However, in the embodiment of the present invention, the functional operations that the response agent provision server (200) can perform are not limited to those described above, and other functional operations can be performed.

[0136] Meanwhile, referring further to FIG. 1, in the embodiment, the response agent provision server (200) may be implemented as a predetermined computing device including at least one processor module (210: Processor Module) for data processing, at least one communication module (220: Communication Module) for data exchange with an external device, and at least one memory module (230: Memory Module) for storing various application programs, data, and / or commands for providing an intelligent response agent service.

[0137] Here, the memory module (230) can store one or more of an operating system (OS), various application programs, data, and commands for providing an intelligent response agent service.

[0138] Additionally, the memory module (230) may include a program area and a data area.

[0139] At this time, the program area according to the embodiment can be linked between the operating system (OS) that boots the server and the functional elements.

[0140] Additionally, the data area according to the embodiment may store data generated according to the use of the server.

[0141] In addition, the memory module (230) may be a variety of storage devices such as ROM, RAM, EPROM, flash drive, hard drive, etc., and may also be web storage that performs the storage function of the memory module (230) on the Internet.

[0142] Additionally, the memory module (230) may be a removable recording medium on the server.

[0143] Meanwhile, the processor module (210) can control the overall operation of each unit described above to implement an intelligent response agent service.

[0144] In detail, the processor module (210) may be a system on chip (SOC) suitable for a server including a central processing unit (CPU) and / or a graphics processing unit (GPU).

[0145] Additionally, the processor module (210) can execute an operating system (OS) and / or application programs stored in the memory module (230).

[0146] Additionally, the processor module (210) can control each component mounted on the server.

[0147] In addition, the processor module (210) can communicate with each component internally via a system bus and can include one or more predetermined bus structures including a local bus.

[0148] Additionally, the processor module (210) may be implemented using at least one of application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, micro-controllers, microprocessors, and / or other electrical units for performing functions.

[0149] In the above description, it has been described that the response agent providing server (200) according to the embodiment of the present invention performs the functional operation as described above, but depending on the embodiment, at least a part of the functional operation performed by the response agent providing server (200) may be performed by an external device (e.g., terminal (100), etc.), and at least a part of the functional operation performed by the external device may be further performed by the response agent providing server (200), and various other embodiments may be possible.

[0150] - Response Agent Model (QALM: Question and Answer Language Model)

[0151] A response agent model (RAM) according to an embodiment of the present invention may be a deep learning model that inputs at least one specialized document (e.g., a thesis and / or report, etc.) data and query data, and outputs at least one response data based on the input specialized document data and query data.

[0152] In an embodiment, such a response agent model (RAM) can, when a specific professional document and query are input, perform deep learning to implement a step-by-step process (in an embodiment, a multi-step inference process) for advanced inference and estimation functions, thereby more clearly interpreting the input query and providing more optimized, high-quality response data based on the same.

[0153] At this time, in the embodiment, the response agent model (RAM) may be a deep learning model (e.g., a large language model (LLM), etc.) pre-trained based on a training data set including a plurality of specialized documents to perform the above-described functional operations.

[0154] FIG. 3 and FIG. 4 are examples of drawings for explaining a multi-step inference process of a response deep learning model according to one embodiment of the present invention.

[0155] In detail, in an embodiment, the response agent model (RAM) can generate response data for a given query data based on a multi-step inference process according to an embodiment of the present invention.

[0156] Here, the multi-step inference process according to the embodiment may mean a structured process that generates response data based on predetermined specialized document data and query data using deep learning.

[0157] This multi-step inference process may include an associative selection process, a rationale generation process, and a systematic composition process.

[0158] In more detail, in the embodiment, the response agent model (RAM) can 1) determine a paragraph (hereinafter, “evidence paragraph”) within a specialized document that is associated with the query data. (Associative Selection Process)

[0159] Specifically, the response agent model (RAM) can extract at least one evidence paragraph, which is a paragraph containing an answer and / or basis for query data, from among K (K>=1) paragraphs included in a specialized document.

[0160] At this time, depending on the embodiment, the response agent model (RAM) may execute the above-described association selection process by reflecting at least one or more existing query data and existing response data (hereinafter, existing query response data) that have been previously stored.

[0161] Through this, the response deep learning model can generate subsequent response data based on more extensive basis data than the conventional method of extracting short answers to query data.

[0162] At this time, the response agent model (RAM) in the embodiment can perform the above-described association selection process based on the prompt engineering algorithm.

[0163] For reference, prompt engineering can refer to the process of creating and optimizing prompts to obtain desired results from a deep learning model.

[0164] Here, a prompt can mean structured text that is input to generate output from a deep learning model.

[0165] These prompts can guide deep learning models to explore specific aspects of their learned data, leading to targeted results.

[0166] For example, components of a prompt may include 'Instruction' that specifies a task or instruction that the deep learning model is desired to perform, 'Context' that specifies external information or additional context that can adjust the deep learning model, 'Input Data' that specifies an input or question for which an answer is sought, and / or 'Output Data' that specifies the type or format of the output.

[0167] That is, in the embodiment, the response agent model (RAM) can generate a structured prompt (hereinafter, a specialized document query prompt) according to a given specialized document and query data based on a prompt engineering algorithm, and can perform an associative selection process for extracting at least one evidence paragraph within the specialized document based on the generated specialized document query prompt.

[0168] Therefore, the response deep learning model can extract at least one paragraph of evidence based on a clearer interpretation of the query data, such as a more accurate understanding of the question and the premises contained within the query data.

[0169] In this way, the response deep learning model can significantly improve the interpretation performance of query data, thereby directly improving the quality of response data generated later.

[0170] In addition, in the embodiment, the response agent model (RAM) can 2) obtain basis data for query data based on the determined evidence paragraph. (Rationale Generation Process)

[0171] Here, the basis data according to the embodiment may mean data that serves as the basis for generating response data for query data.

[0172] That is, in the embodiment, the basis data may be a collection of various data used when generating response data.

[0173] In detail, in an embodiment, the response agent model (RAM) can detect main answer data, which is data that includes a direct answer to the query data based on at least one paragraph of evidence, explanatory sentence data, which is data that explains the direct answer in detail, and / or auxiliary information data, which is data that includes related background knowledge.

[0174] And the response agent model (RAM) can obtain the basis data as described above based on the detected data.

[0175] At this time, depending on the embodiment, the response agent model (RAM) may execute the above-described basis generation selection process by reflecting at least one or more existing question-and-answer data that has been previously stored.

[0176] In addition, in the embodiment, the response agent model (RAM) can generate response data based on the acquired basis data. (Systematic Composition Process)

[0177] In detail, in an embodiment, the response agent model (RAM) can generate response data by performing data processing based on acquired basis data.

[0178] As an example, the response agent model (RAM) can generate response data by performing certain data processing (e.g., removing duplicate text, etc.) to improve the conciseness and readability of the response based on the basis data.

[0179] As described above, in the embodiment, the response deep learning model can extract a paragraph of evidence that can answer the query data from a specialized document, generate evidential evidence based on the content of the extracted paragraph of evidence, and provide response data that provides a more in-depth answer based on the generated evidential evidence.

[0180] That is, in the embodiment, the response deep learning model can be implemented as a language model specialized for specialized documents that resolves the problems of hallucination and vague answers, which are limitations of existing general-purpose language models (e.g., OpenAI GPT, etc.) when performing in-depth query and response tasks based on specialized documents.

[0181] Through this, the system for providing a specialized document sharing platform according to an embodiment of the present invention can provide response data that is faithful, hallucination-controlled, and provides clear evidence for in-depth queries related to specialized documents based on advanced chain-of-thought reasoning and speculation functions implemented through a process similar to human cognitive reasoning.

[0182]

[0183] - Method for providing an intelligent response agent

[0184] Hereinafter, a method for generating and providing response data for a query related to a specialized document using a deep-learning neural network that implements a step-by-step process for advanced inference and estimation functions by an application (111) executed by at least one processor of a terminal (100) according to an embodiment of the present invention (i.e., a method for implementing an intelligent response agent service) will be described in detail with reference to the attached drawings.

[0185] In an embodiment of the present invention, at least one processor of the terminal (100) can execute at least one application (111) stored in at least one memory (110) or operate it in a background state.

[0186] In the following embodiments, the method of providing an intelligent response agent based on advanced inference and estimation functions by having at least one processor of the terminal (100) operate to execute commands of the application (111) is abbreviated as being performed by the application (111).

[0187] FIG. 5 is a flowchart for explaining a method for providing an intelligent response agent based on advanced inference and estimation functions according to one embodiment of the present invention, and FIG. 6 is a conceptual diagram for explaining a method for providing an intelligent response agent based on advanced inference and estimation functions according to one embodiment of the present invention.

[0188] Referring to FIGS. 5 and 6, in an embodiment, an application (111) executed by at least one processor of a terminal (100) or operating in a background state can obtain query data. (S101)

[0189] Here, query data according to the embodiment may mean data that specifies a user's question related to a given specialized document (e.g., a thesis and / or report, etc.).

[0190] For example, query data might be, "Please analyze the supporting data that supports the first content of the first paper." or "Please explain the expected results and expected effects of combining the second structure, not disclosed in the first paper, with the first structure of the first paper."

[0191] In an embodiment, the application (111) may provide a user interface (hereinafter, query input interface) that can input certain query data based on an intelligent response agent service platform.

[0192] And the application (111) can obtain at least one query data as above based on user input based on the provided query input interface.

[0193] Additionally, in the embodiment, the application (111) can determine a specialized document for the acquired query data. (S103)

[0194] That is, in the embodiment, the application (111) can determine at least one specialized document to be used when generating response data for the acquired query data.

[0195] In other words, the application (111) can determine at least one specialized document (hereinafter, base document) to be input into the response agent model (RAM) according to the embodiment of the present invention in order to generate response data for query data.

[0196] In detail, as an example, the application (111) can determine at least one specialized document directly uploaded by a user as a base document on the intelligent response agent service platform.

[0197] In another embodiment, the application (111) may determine at least one specialized document detected based on 2) user-entered specialized document identification information (e.g., paper number and / or paper title, etc.) on the intelligent response agent service platform as a base document.

[0198] Specifically, the application (111) can determine a base document by detecting a specialized document corresponding to each of at least one specialized document identification information obtained according to user input from the memory (110) and / or an external database.

[0199] In another embodiment, the application (111) may perform deep learning based on 3) query data input on the intelligent response agent service platform to determine at least one detected specialized document as a base document.

[0200] In detail, the application (111) can determine the base document described above by linking with a deep learning model (hereinafter, a query-related document detection model) that inputs predetermined query data and outputs at least one specialized document (hereinafter, a related document) having a correlation higher than a predetermined standard with the input query data.

[0201] That is, the application (111) can input query data entered on the intelligent response agent service platform into a query-related document detection model, and obtain at least one related document corresponding to the query data from the query-related document detection model that has received the query data.

[0202] And the application (111) can determine a base document based on at least one related document obtained.

[0203] At this time, depending on the embodiment, the query-related document detection model may be implemented by being included in the response agent model (RAM), and may be pre-learned based on the training data set (i.e., the training data set including multiple specialized documents) used in learning the response agent model (RAM).

[0204] In the above, the embodiments have been described separately as described above for effective explanation, but various embodiments are possible, such as at least some of the embodiments may be organically combined and operated depending on the embodiment.

[0205] In detail, in the embodiment, the application (111) may determine a base document based on 1) a user input-based professional document (hereinafter, user input document) or 2) a user input document and related documents (i.e., professional documents automatically detected through a deep learning model) according to user settings.

[0206] In other words, the application (111) can determine a base document based on at least one specialized document directly input by the user (i.e., a user-input document) according to the user's selection, or can determine a base document based on both at least one specialized document directly input by the user (i.e., a user-input document) and at least one specialized document automatically detected through a deep learning model (i.e., a related document).

[0207] In this way, the application (111) can selectively determine the specialized document to be used when generating an answer to a user's question according to the user's needs.

[0208] Through this, the application (111) can implement a customized response data provision process that creates an answer to the question based on information in the data area desired by the user according to the user's needs that vary depending on the situation, such as when the user wants to get an answer based only on the specialized document that the user is targeting, or when the user wants to get an answer by additionally utilizing various specialized documents related to the question in addition to the specialized documents that the user knows.

[0209] Additionally, in the embodiment, the application (111) can input the determined professional document and query data into the response agent model (RAM). (S105)

[0210] That is, in the embodiment, the application (111) can input the base document determined as described above and the acquired query data into the response agent model (RAM) according to the embodiment of the present invention.

[0211] Here, in other words, the response agent model (RAM) according to the embodiment may be a deep learning model that takes certain specialized document data and query data as input and outputs at least one response data based on the input specialized document data and query data.

[0212] In an embodiment, such a response agent model (RAM) can, when a specific professional document and query are input, perform deep learning to implement a step-by-step process (in an embodiment, a multi-step inference process) for advanced inference and estimation functions, thereby more clearly interpreting the input query and providing more optimized, high-quality response data based on the same.

[0213] Additionally, in the embodiment, the application (111) can obtain response data for input query data. (S107)

[0214] Here, response data according to the embodiment may mean data specifying an answer to a user's question related to a given specialized document (e.g., a thesis and / or report, etc.).

[0215] That is, response data may be data that provides an answer to a given query data.

[0216] FIG. 7 is an example of a drawing for explaining a method for obtaining response data for a given query data based on a response agent model (RAM) according to one embodiment of the present invention.

[0217] In detail, referring to FIG. 7, in the embodiment, the application (111) can obtain response data for the input query data by linking with a response agent model (RAM) that has received a base document and query data.

[0218] In more detail, referring to FIGS. 3 and 4, in an embodiment, a response agent model (RAM) that receives a base document and query data can generate response data for the input query data according to a multi-step inference process according to an embodiment of the present invention.

[0219] Here, in other words, the multi-step inference process according to the embodiment may mean a structured process that generates response data based on predetermined specialized document data and query data using deep learning.

[0220] In an embodiment, such a multi-step inference process may include an associative selection process that determines paragraphs (i.e., evidence paragraphs) within a specialized document that are associated with query data, a rationale generation process that obtains rationale data for the query data based on the determined rationale paragraphs, and a systematic composition process that generates response data based on the obtained rationale data.

[0221] Additionally, in the embodiment, a response agent model (RAM) that generates response data according to a multi-step inference process can provide the generated response data to an application (111).

[0222] Thus, in the embodiment, the application (111) can obtain response data for the input query data through linkage with the response agent model (RAM).

[0223] In this way, in the embodiment, the application (111) can obtain response data for the corresponding query data based on a response agent model (RAM) that performs an advanced inference and estimation process that extracts a paragraph of evidence that can answer the query data from a specialized document, creates evidential evidence based on the contents of the extracted paragraph of evidence, and creates a more in-depth answer based on the created evidential evidence.

[0224] Accordingly, the application (111) can increase efficiency by resolving the process of search-ranking-selection-extraction-editing, which is the existing question-and-answer paradigm, into a single process called generation, and can provide a more detailed and optimized high-quality answer based on the premise that the intent of the question is clearly interpreted based on objective data, rather than a vague answer.

[0225] Additionally, in the embodiment, the application (111) can improve the acquired response data. (S109)

[0226] In detail, in an embodiment, the application (111) may provide at least one response data obtained as described above based on at least one query data.

[0227] FIGS. 8 to 10 are examples of drawings for explaining a method of providing predetermined response data based on a response output interface according to one embodiment of the present invention.

[0228] In more detail, referring to FIG. 8, in an embodiment, the application (111) may provide a response output interface (API) based on an intelligent response agent service platform.

[0229] Here, the response output interface (API) according to the embodiment may mean a user interface that visualizes and displays predetermined response data (AD) and various data and information related thereto in a predetermined manner.

[0230] Additionally, in the embodiment, the application (111) can display and provide the generated response data (AD) based on the provided response output interface (API).

[0231] At this time, in the embodiment, the application (111) can provide at least one piece of basis information corresponding to each provided response data (AD) by matching it to the corresponding response data (AD) and displaying it in conjunction with the response data.

[0232] Here, the basis information according to the embodiment may mean data and / or information used as basis data when generating a given response data (AD).

[0233] In an embodiment, such basis information may include specialized document data corresponding to predetermined response data (AD), specialized document identification information (e.g., paper number and / or paper title), evidence paragraph data, evidence paragraph identification information (e.g., evidence paragraph identification number / identification code in a preset format), and / or basis data.

[0234] Additionally, in the embodiment, the basis information may further include query data corresponding to a given response data (AD).

[0235] In detail, in the embodiment, the application (111) can detect at least one piece of evidence corresponding to each response data (AD).

[0236] In an embodiment, the application (111) can detect at least one specialized document data related to the generation of each response data (AD), specialized document identification information for the specialized document data, evidence paragraph data, evidence paragraph identification information for the evidence paragraph data, and / or basis data.

[0237] Additionally, in the embodiment, the application (111) can match the detected basis information to each response data (AD) and display it in a linked manner (e.g., list, image, and / or text, etc.) through an answer output interface (API).

[0238] The following description is based on an example in which the basis information corresponding to each response data (AD) is provided in a list format, but this is only an example and is not limited thereto.

[0239] As an embodiment, the application (111) may further include a list interface (LI: List Interface) including at least one piece of basis information for each response data (AD) in the response output interface (API).

[0240] And, when displaying each response data (AD) through the response output interface (API), the application (111) can further display the list interface (LI) by matching it with each corresponding response data (AD).

[0241] In this way, in the embodiment, the application (111) can display various related information such as what the basis data used to derive each response data (AD) is, what the content is, and where the location is, by matching it with the corresponding response data (AD).

[0242] Through this, the application (111) can clearly and intuitively present objective evidence proving the basis on which each response data (AD) was generated, and thus can significantly improve the reliability and validity of the provided response data (AD).

[0243] At this time, depending on the embodiment, the application (111) can create a separate list by classifying each attribute of the basis information.

[0244] And the application (111) can provide each generated list by matching it to response data (AD) corresponding to each piece of basis information based on the above-described list interface (LI).

[0245] As an example, the application (111) can generate a query data list (L1), a professional document data list (L2), an evidence paragraph data list (L3), and / or a basis data list (L4) by classifying each basis information by attribute.

[0246] And the application (111) can display the generated list by matching it to response data (AD) corresponding to each piece of evidence information through a response output interface (API).

[0247] In this way, the application (111) can further improve readability and recognition thereof by providing the data used as the basis for generating response data (AD) in a more refined form by displaying them according to their properties.

[0248] Meanwhile, referring to FIG. 9, according to an embodiment, the application (111) can display matching basis information in units of sentences (hereinafter, response sentences) within each response data (AD).

[0249] In detail, in the embodiment, the application (111) can divide each response data (AD) into at least one response sentence unit.

[0250] And the application (111) can detect at least one piece of basis information used when generating each separate answer sentence.

[0251] Additionally, in the embodiment, the application (111) can display the detected basis information by matching it to each corresponding answer sentence through a response output interface (API).

[0252] For example, when the first response data (AD) is composed of first to fifth response sentences, the application (111) can detect first basis information, which is at least one piece of basis information used when generating the first response sentence, second basis information, which is at least one piece of basis information used when generating the second response sentence, …, fifth basis information, which is at least one piece of basis information used when generating the fifth response sentence.

[0253] And the application (111) can match the detected first basis information to the corresponding first answer sentence, match the second basis information to the corresponding second answer sentence, …, match the fifth basis information to the corresponding fifth answer sentence, and display them through the answer output interface (API).

[0254] In this way, the application (111) can display and match the data that serves as the basis for each answer at a sentence level that is more detailed than the response data (AD) level, thereby enabling more detailed confirmation and recognition of evidential grounds for the response data (AD).

[0255] On the other hand, according to the embodiment, the application (111) can display the basis information matching each response data (AD) and / or each answer sentence when the preset output display conditions are met.

[0256] In an embodiment, the above-described preset display conditions may include, 1> a condition in which the response data (AD) and / or the answer sentence is located at the top compared to other response data (AD) and / or other answer sentences, 2> a condition in which the response data (AD) and / or the answer sentence is located in a predetermined central area within the display area, or 3> a condition in which the response data (AD) and / or the answer sentence is located in a user selection area (e.g., a mouse over area, etc.).

[0257] In detail, in the embodiment, the application (111) can select any one of the preset display conditions as described above according to user input.

[0258] And the application (111) can determine whether the display conditions of the response data (AD) and / or the answer sentence are satisfied based on the selected display conditions (hereinafter, applied display conditions).

[0259] In addition, the application (111) can perform the above-described basis information linkage display process based on response data (AD) and / or answer sentences that satisfy the applicable display conditions.

[0260] That is, in the embodiment, the application (111) can determine the real-time observation status of response data (AD) and / or answer sentences that change according to user input (e.g., mouse scroll, etc.) and dynamically select and extract related basis information and provide it by displaying it in conjunction.

[0261] Through this, the application (111) can implement interactive visualization through interaction with a user who checks response data (AD), thereby enabling the user to recognize response data (AD) and related basis information more easily and intuitively.

[0262] On the other hand, referring to FIG. 10, according to an embodiment, the application (111) can highlight, in a predetermined manner, the basis information matching each response data (AD) and / or each answer sentence when a preset highlighting condition is satisfied.

[0263] In an embodiment, the above-described preset display conditions may include, 1> a condition in which the response data (AD) and / or the answer sentence is located at the top compared to other response data (AD) and / or other answer sentences, 2> a condition in which the response data (AD) and / or the answer sentence is located in a predetermined central area within the display area, or 3> a condition in which the response data (AD) and / or the answer sentence is located in a user selection area (e.g., a mouse over area, etc.).

[0264] In detail, in the embodiment, the application (111) can select any one of the preset display conditions as described above according to user input.

[0265] And the application (111) can determine whether the display conditions of the response data (AD) and / or the answer sentence are satisfied based on the selected display conditions (hereinafter, applied display conditions).

[0266] In addition, the application (111) can highlight response data (AD) and / or answer sentences that satisfy the applicable display conditions and corresponding basis information in a predetermined manner.

[0267] As an example, the application (111) can perform highlighting by highlighting response data (AD) and / or answer sentences that satisfy the applicable display conditions and corresponding basis information as a bounding box.

[0268] Accordingly, the application (111) can further improve user readability in an interactive visualization process based on response data (AD).

[0269] FIG. 11 is an example of a drawing for explaining a method for determining a base document based on basis information according to one embodiment of the present invention.

[0270] On the other hand, referring to FIG. 11, according to an embodiment, the application (111) may set at least some of the basis information provided through the response output interface (API) as a base document (BD) according to user input.

[0271] In detail, in an embodiment, the application (111) may provide a user interface (hereinafter, a base document (BD) designation interface) that allows selection of at least some of the basis information displayed through the answer output interface (API).

[0272] In addition, in the embodiment, the application (111) can determine (update) a base document (BD) based on selected basis information (C: hereinafter, base basis information) according to user input based on a provided base document (BD) designation interface.

[0273] In an embodiment, the application (111) can determine a base document (BD) using selected specialized document data and / or evidence paragraph data as base basis information (C) based on user input based on a base document (BD) designation interface.

[0274] At this time, as an example, the application (111) can add base basis information (C) to an existing base document (BD) according to user input.

[0275] In another embodiment, the application (111) can replace an existing base document (BD) using base basis information (C) based on user input.

[0276] Additionally, in the embodiment, the application (111) can execute the process below step S105 described above based on the base document (BD) updated as above.

[0277] Accordingly, the application (111) can expand the paradigm of the intelligent response agent service and improve its usability by effectively supporting newly arising user needs as it provides answers, such as cases where the user wants to obtain new answers by targeting at least some of the specialized documents detected by artificial intelligence in addition to the specialized documents he or she knows, or wants to obtain new answers by focusing more on paragraphs within a specific specialized document.

[0278] At this time, according to the embodiment, the application (111) can obtain new query data according to user input when updating the base document (BD).

[0279] And the application (111) can update existing query data based on the acquired new query data.

[0280] As an example, the application (111) can add new query data to existing query data based on user input.

[0281] In another embodiment, the application (111) may replace existing query data with new query data based on user input.

[0282] Additionally, in the embodiment, the application (111) can execute the process below step S105 described above based on the updated query data.

[0283] In this way, the application (111) can enhance convenience by allowing the user to easily write additional questions with a high degree of freedom even after receiving an answer.

[0284] Above, the method for providing an intelligent response agent based on an advanced reasoning and estimation function according to one embodiment of the present invention and the system thereof generate and provide response data (AD) for a query related to a specialized document using a deep-learning neural network that implements a step-by-step process for an advanced reasoning and estimation function, thereby having the effect of providing response data (AD) that is faithful to the question, fact-based (hallucination-controlled), and presents a clear basis (evidential) for an in-depth query related to a specialized document based on an advanced reasoning (Chain-of-thought reasoning) and estimation (Speculation) function implemented in a process similar to human cognitive reasoning.

[0285] In addition, the method and system for providing an intelligent response agent based on an advanced inference and estimation function according to one embodiment of the present invention can provide generated response data (AD) and supporting evidence in a mutually linked manner, thereby enabling an intuitive and easy confirmation of objective evidence including the basis for generating the response data (AD), thereby further improving the reliability and validity of the response data (AD).

[0286] Meanwhile, the embodiments of the present invention described above may be implemented in the form of program commands that can be executed through various computer components and recorded on a computer-readable recording medium. The computer-readable recording medium may include program commands, data files, data structures, etc., either singly or in combination. The program commands recorded on the computer-readable recording medium may be specially designed and configured for the present invention or may be 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 commands, such as ROMs, RAMs, and flash memories. Examples of program commands include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. Hardware devices may be changed into one or more software modules to perform processing according to the present invention, and vice versa.

[0287] The specific implementations described in the present invention are exemplary embodiments and do not limit the scope of the present invention in any way. For the sake of brevity, descriptions of conventional electronic components, control systems, software, and other functional aspects of the systems may be omitted. In addition, the lines connecting or connecting members between components illustrated in the drawings are merely representative of functional connections and / or physical or circuit connections, and may be replaced or represented as various additional functional connections, physical connections, or circuit connections in an actual device. In addition, unless specifically mentioned as “essential,” “important,” etc., a component may not be absolutely necessary for the application of the present invention.

[0288] Although the detailed description of the present invention has been described with reference to preferred embodiments of the present invention, it will be understood by those skilled in the art or having ordinary knowledge in the art that various modifications and changes can be made to the present invention without departing from the spirit and technical scope of the present 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 defined by the claims.

[0289] The present invention relates to a method and system for providing an intelligent response agent based on advanced inference and estimation functions, and is applicable to the artificial intelligence industry, and thus has industrial applicability.

Claims

1. A method for providing an intelligent response agent based on advanced inference and estimation functions, wherein a response agent application executed by at least one processor of a terminal comprises: A step of obtaining query data that specifies a user's question; A step of determining a specialized document for the above-obtained query data; A step of detecting an evidence paragraph including at least one paragraph linked to the query data on the determined professional document; A step of generating response data specifying an answer to the question based on the detected evidence paragraph; and comprising a step of providing the generated response data; A method for providing an intelligent response agent based on advanced inference and estimation functions.

2. In paragraph 1, A step of acquiring basis data, which is the base data for generating the response data based on the above-detected evidence paragraph, Further comprising a step of generating the response data based on the acquired basis data. A method for providing an intelligent response agent based on advanced inference and estimation functions.

3. In paragraph 2, The above supporting data is, At least one of main answer data including a direct answer to the above query data, explanatory sentence data including a detailed explanation of the direct answer, and auxiliary information data including background knowledge information related to the question and the answer. A method for providing an intelligent response agent based on advanced inference and estimation functions.

4. In paragraph 1, The step of detecting the above evidence paragraph is: A step of detecting the above-mentioned evidence paragraph based on a prompt engineering algorithm is included. A method for providing an intelligent response agent based on advanced inference and estimation functions.

5. In paragraph 2, A step of inputting the above-determined professional document and the above-determined query data into a response agent model that executes a multi-step inference process, Further comprising a step of generating the response data based on the response agent model, The above multi-step inference process is, It includes an associative selection process for detecting the above-mentioned evidence paragraph, a rationale generation process for obtaining the above-mentioned evidence data, and a systematic composition process for generating the above-mentioned response data. A method for providing an intelligent response agent based on advanced inference and estimation functions.

6. In paragraph 5, The step of generating the response data based on the above response agent model is: A step of executing the associated selection process and the basis generation process by reflecting at least one of the previously stored query data and response data. A method for providing an intelligent response agent based on advanced inference and estimation functions.

7. In paragraph 2, The step of providing the generated response data is as follows: A step of providing basis information for the above response data by matching it to the above response data is included, The above information is, Containing at least two of the above-mentioned basis data, the above-mentioned evidence paragraph data, the above-mentioned evidence paragraph identification information, the above-mentioned professional document data, the above-mentioned professional document identification information, and the above-mentioned query data. A method for providing an intelligent response agent based on advanced inference and estimation functions.

8. In paragraph 7, The step of providing the generated response data is as follows: A step of dividing the above response data into at least one sentence unit, A step of detecting the above-mentioned basis information for each separated sentence, It further includes a step of providing the basis information for each detected sentence by matching it to the sentence corresponding to each basis information. A method for providing an intelligent response agent based on advanced inference and estimation functions.

9. In paragraph 1, The step of determining the specialized document for the above query data is: A step of determining a specialized document for the query data based on at least one specialized document entered by the user; A step of determining a specialized document for the query data based on a specialized document corresponding to at least one specialized document identification information entered by the user; A method comprising: detecting at least one specialized document having a correlation higher than a preset standard with the query data based on a predetermined deep learning neural network; and determining a specialized document for the query data based on the detected at least one specialized document. A method for providing an intelligent response agent based on advanced inference and estimation functions.

10. At least one memory in which the response agent application is stored; and At least one processor for providing an intelligent response agent based on advanced reasoning and estimation functions by reading out the response agent application stored in the memory; The command of the above response agent application is: A step for obtaining query data that specifies the user's question, A step of determining a specialized document for the above-mentioned acquired query data, A step of detecting an evidence paragraph including at least one paragraph linked to the query data on the above-determined professional document, A step of generating response data that specifies an answer to the question based on the detected evidence paragraph, A command that performs a step of providing the generated response data A system that provides intelligent response agents based on advanced inference and estimation capabilities.

Citation Information

Patent Citations

  • Indexing documents for information retrieval

    KR101319628B1

  • System for Answering and the Method thereof

    KR1020150129134A

  • System for analyzing process performance of shipyards

    KR1020240044003A

  • Apparatus for Intellectual Learning Service and Driving Method Thereof

    KR102282798B1

  • System for Providing Semantic Analysis Finding and Analyzing Sentence Meaning

    KR102439321B1

Cited By

  • Resource dynamic scheduling method and system based on multi-dimensional feature fusion and task aggregation

    CN122798122A