Method and system for providing intelligent response agents based on advanced inference and estimation capabilities

Abstract

The intelligent response agent provision method and system based on advanced inference and estimation functions according to embodiments of the present invention can generate and provide response data to questions related to specialized documents using a deep-learning neural network that realizes a stepwise process for advanced inference and estimation functions.

Description

Technical Field

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

Background Art

[0002] Deep learning research on question answering aims to provide accurate answers to user questions by using artificial intelligence models.

[0003] Generally, a conventional question answering deep learning system is realized as a system that syntactically analyzes a given question, constructs semantic elements as patterns (rules), and extracts responses from a database structured in the same language as the Structured Query Language (SQL).

[0004] However, in the processing of various natural languages, the number of syntactic patterns to be constructed increases geometrically, and the utility of the question answering deep learning system decreases.

[0005] Therefore, recently, a machine reading comprehension (MRC) model, which is a technology in which an artificial intelligence algorithm analyzes questions by itself and finds an optimized answer, has been widely used.

[0006] Specifically, the machine reading comprehension model is a system in which a computer reads and understands a presented text (text for understanding questions such as an article and / or a document) and searches for and presents an answer to the question.

[0007] However, in order to effectively utilize a question-and-answer deep learning system using a machine reading comprehension model, a presentation text must first exist. If a presentation text has not been generated, there is a problem in that it is difficult for the machine reading comprehension model to respond.

[0008] On the other hand, while conventional machine reading comprehension models have been trained over a long period using a variety of previous models and benchmark datasets, most of these training datasets are based on question-and-answer datasets that do not require a deep understanding of the documents.

[0009] Therefore, the development of deep learning-based question-and-answer processes that generate responses to in-depth questions requiring advanced reasoning and estimation is still insufficient.

[0010] In other words, previous research has shown that despite significant improvements in machine reading comprehension capabilities, it has limitations in handling more in-depth question-and-answer sessions that go beyond fact-based, factoid Q&A.

[0011] As a concrete example, the majority of conventional question data consists mainly of fact-checking questions such as "what," "when," "where," and "who." While machine reading models can extract and provide short answers from specific paragraphs related to these questions, they struggle to provide answers to deeper questions such as "how" or "why."

[0012] In other words, conventional technologies have a problem in that, when it comes to in-depth questions that require innovations in knowledge processing and reasoning related to specialized documents (e.g., scientific and technical papers) containing deep technical information, the accuracy and quality of answers from deep learning models are low. [Overview of the Initiative] [Problems that the invention aims to solve]

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

[0014] Furthermore, one embodiment of the present invention aims to realize a method and system for providing generated response data and supporting information in an interconnected manner.

[0015] However, the technical problems that the present invention and its embodiments aim to solve are not limited to those described above, and other technical problems may exist. [Means for solving the problem]

[0016] An embodiment of the present invention provides an intelligent response agent based on advanced reasoning and estimation functions, wherein a response agent application run on at least one processor of a terminal provides an intelligent response agent based on advanced reasoning and estimation functions, comprising the steps of: acquiring question data that identifies a user's question; determining a specialized document for the acquired question data; detecting evidence paragraphs on the determined specialized document that include at least one paragraph that corresponds to the question data; generating response data that identifies an answer to the question based on the detected evidence paragraphs; and providing the generated response data.

[0017] In other aspects, an intelligent response agent provision method based on advanced reasoning and estimation capabilities according to embodiments of the present invention further includes the steps of: acquiring evidence data which is foundational data for generating the response data based on the detected evidence paragraphs; and generating the response data based on the acquired evidence data.

[0018] In other respects, the supporting data includes at least one of the following: primary response data, which includes a direct answer to the question data; explanatory text data, which includes a detailed explanation of the direct answer; and auxiliary information data, which includes background knowledge information related to the question and the answer.

[0019] In other respects, the step of detecting the evidence paragraph includes the step of detecting the evidence paragraph based on a prompt engineering algorithm.

[0020] In other aspects, an intelligent response agent provision method based on advanced reasoning and estimation capabilities according to embodiments of the present invention further includes the steps of inputting the determined expert document and question data into a response agent model that performs a multi-step inference process, and generating the response data based on the response agent model, wherein the multi-step inference process includes an associate 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 other respects, the step of generating the response data based on the response agent model includes the step of performing the related selection process and the rationale generation process, reflecting at least one already stored question data and response data.

[0022] In other respects, the step of providing the generated response data includes the step of providing evidence information relating to the response data, matching the response data with the evidence information, the evidence information including at least two of the following: the evidence data, the evidence paragraph data, the evidence paragraph identification information, the expert document data, the expert document identification information, and the question data.

[0023] In other respects, the step of providing the generated response data further includes the steps of separating the response data into at least one sentence unit, detecting the basis information for each of the separated sentences, and matching the detected basis information for each sentence to the sentence corresponding to each basis information and providing it.

[0024] In other respects, the step of determining a specialized document for the question data includes at least one of the following steps: determining a specialized document for the question data based on at least one specialized document entered by the user; determining a specialized document for the question data based on a specialized document corresponding to at least one specialized document identification piece entered by the user; and detecting at least one specialized document that has a correlation with the question data that exceeds a previously set criterion based on a predetermined deep learning neural network, and determining a specialized document for the question data based on the detected at least one specialized document.

[0025] On the other hand, an intelligent response agent providing system based on advanced inference and estimation functions according to an embodiment of the present invention includes at least one memory storing a response agent application; and at least one processor for reading out the response agent application stored in the memory and providing an intelligent response agent based on advanced inference and estimation functions. The instruction words of the response agent application include steps of obtaining interrogation data for identifying a user's question, determining a specialized document for the obtained interrogation data, detecting an evidence paragraph including at least one text segment associated with the interrogation data on the determined specialized document, generating response data for identifying an answer to the question based on the detected evidence paragraph, and providing the generated response data.

Effect of the Invention

[0026] An intelligent response agent providing method and its system based on advanced inference and estimation functions according to an embodiment of the present invention utilize a deep-learning neural network (Deep-learning Neural Network) that realizes a step-by-step process for advanced inference and estimation functions to generate and provide response data for queries related to specialized documents. Thus, based on advanced inference (Chain-of-thought reasoning) and estimation (Speculation) functions realized in a process similar to human cognitive inference, it is possible to provide response data that is faithful to the question, fact-based (hallucination-controlled), and presents clear evidence for deep queries related to specialized documents.

[0027] In addition, the method for providing an intelligent response agent based on advanced inference and estimation functions according to an embodiment of the present invention and its system provide the generated response data and the basis information supporting the same in mutual cooperation, so that objective evidence materials including the basis for generating the response data can be intuitively and easily confirmed, thereby having the effect of further improving the reliability and validity of the response data.

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

Brief Description of Drawings

[0029] [Figure 1] It is a conceptual diagram of an intelligent response agent system according to an embodiment of the present invention. [Figure 2] It is an internal block diagram of a terminal according to an embodiment of the present invention. [Figure 3] It is an example of a diagram for explaining a multi-step inference process of a response deep learning model according to an embodiment of the present invention. [Figure 4] It is an example of a diagram for explaining a multi-step inference process of a response deep learning model according to an embodiment of the present invention. [Figure 5] It is a flowchart for explaining a method for providing an intelligent response agent based on advanced inference and estimation functions according to an embodiment of the present invention. [Figure 6] It is a conceptual diagram for explaining a method for providing an intelligent response agent based on advanced inference and estimation functions according to an embodiment of the present invention. [Figure 7] It is an example of a diagram for explaining a method for obtaining response data for predetermined query data based on a response agent model according to an embodiment of the present invention. [Figure 8]This is an example diagram illustrating a method for providing predetermined response data based on a response output interface according to one embodiment of the present invention. [Figure 9] This is an example diagram illustrating a method for providing predetermined response data based on a response output interface according to one embodiment of the present invention. [Figure 10] This is an example diagram illustrating a method for providing predetermined response data based on a response output interface according to one embodiment of the present invention. [Figure 11] This is an example of a diagram illustrating a method for determining a base document based on supporting information according to one embodiment of the present invention. [Modes for carrying out the invention]

[0030] The present invention can be modified in various ways and has many different 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 how they are achieved, will become clear when you refer to the embodiments described in detail below, along with the drawings. However, the present invention is not limited to the embodiments disclosed below and can be realized in many different forms. In the following embodiments, terms such as "first," "second," etc., are not restrictive and are used to distinguish one component from another. Also, singular expressions include plural expressions unless they have a clearly different meaning in context. Also, terms such as "includes" or "has" 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. Also, the sizes of components may be exaggerated or reduced in the drawings for the sake of illustration. For example, the size and thickness of each component shown in the drawings are arbitrarily shown for the sake of illustration, so the present invention is not necessarily limited to what is shown.

[0031] 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 will be given the same reference numerals, and redundant descriptions thereof will be omitted.

[0032] Figure 1 is a conceptual diagram of an intelligent response agent system according to one embodiment of the present invention.

[0033] Referring to Figure 1, the intelligent response agent system 1000 according to an embodiment of the present invention can realize an intelligent response agent service that acquires and provides response data to questions related to specialized documents by utilizing a deep-learning neural network that enables a stepwise process for advanced inference and estimation functions.

[0034] In one embodiment, the intelligent response agent system 1000 that provides an intelligent response agent service may include a terminal 100, a response agent providing server 200, and a network 300.

[0035] Here, terminal 100 and / or response agent providing server 200 are connected via network 300.

[0036] Here, the network 300 according to this embodiment means a connection structure that enables information exchange between each node, such as the terminal 100 and / or the response agent providing server 200.

[0037] Examples of Network 300 include, but are not limited to, 3GPP (3rd Generation Partnership Project) networks, LTE (Long Term Evolution) networks, WiMAX (World Interoperability for Microwave Access) networks, the 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.

[0038] The terminal 100 and response agent providing server 200 that realize the intelligent response agent system 1000 will be described in detail below with reference to the attached diagrams.

[0039] - Terminal (100:Terminal)

[0040] The terminal 100 according to the embodiment of the present invention may be a predetermined computing device on which a response agent application (hereinafter referred to as "the application") that provides an intelligent response agent service is installed.

[0041] More specifically, from a hardware perspective, terminal 100 may include a mobile type computing device 100-1 and / or a desktop type computing device 100-2, etc., on which an application is installed.

[0042] Here, the mobile computing device 100-1 may be a mobile device on which an application is installed.

[0043] For example, the mobile computing device 100-1 may include a smartphone, a mobile phone, a digital broadcasting device, a PDA (personal digital assistant), a PMP (portable multimedia player), and / or a tablet PC.

[0044] Furthermore, the desktop computing device 100-2 may be a wired / wireless communication-based device on which an application is installed.

[0045] For example, the desktop computing device 100-2 may include a fixed desktop PC, a laptop computer, and / or a personal computer such as an ultrabook.

[0046] Depending on the embodiment, terminal 100 may further include a predetermined server computing device that provides an intelligent response agent service environment.

[0047] Figure 2 is an internal block diagram of terminal 100 according to one embodiment of the present invention.

[0048] On the other hand, referring to Figure 2, from a functional standpoint, 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 this embodiment, the terminal 100 may include these components within a housing.

[0049] More specifically, memory 110 can store application 111.

[0050] Here, application 111 stores one or more of the following: application programs, data, and instruction words, for providing an intelligent response agent service environment.

[0051] In other words, memory 110 stores instructions and data that can be used to generate an intelligent response agent service environment.

[0052] Furthermore, the memory 110 may include a program area and a data area.

[0053] In this embodiment, the program area can be linked between the operating system (OS) that starts the terminal 100 and its functional elements.

[0054] Furthermore, the data area in this embodiment stores data generated by the use of terminal 100.

[0055] Furthermore, the memory 110 may include at least one non-temporary computer-readable storage medium and a temporary computer-readable storage medium.

[0056] For example, memory 110 may be a variety of storage devices such as ROM, EPROM, flash drive, or hard drive, and may also include web storage that performs the storage function of memory 110 over the internet.

[0057] The processor assembly 120 may include at least one processor capable of executing instructions of an application 111 stored in memory 110 in order to perform various tasks for generating an intelligent response agent service environment.

[0058] In one embodiment, the processor assembly 120 can control the overall operation of its components by an application 111 in the memory 110 in order to provide an intelligent response agent service.

[0059] More specifically, the processor assembly 120 may be a system-on-a-chip (SOC) suitable for terminal 100, which includes a central processing unit (CPU) and / or a graphics processing unit (GPU), etc.

[0060] Furthermore, the processor assembly 120 can execute the operating system (OS) and / or application programs stored in the memory 110.

[0061] Furthermore, the processor assembly 120 can control each component mounted on the terminal 100.

[0062] Furthermore, the processor assembly 120 can communicate internally with each component via a system bus, and may include one or more predetermined bus structures, including a local bus.

[0063] Furthermore, the processor assembly 120 can be implemented by including at least one of the following: 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.

[0064] The communication processor 130 may include one or more devices for communicating with external devices. Such a communication processor 130 can communicate via a wireless network.

[0065] More specifically, the communication processor 130 can communicate with the terminal 100, which stores content sources for realizing an intelligent response agent service environment.

[0066] Furthermore, the communication processor 130 can communicate with a variety of user input components, such as controllers that receive user input.

[0067] In this embodiment, the communication processor 130 can send and receive various data related to the intelligent response agent service with other terminals 100 and / or external servers.

[0068] Such a communication processor 130 can wirelessly send and receive data with at least one of a base station, an external terminal 100, or any server on a mobile communication network constructed with communication equipment capable of implementing 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.

[0069] The sensor system 160 may include a variety of sensors, such as an image sensor 161, an IMU (Infrared Unit) 163, an audio sensor 165, a distance sensor, a proximity sensor, and a contact sensor.

[0070] Here, the image sensor 161 can capture images (images and / or videos, etc.) of the physical space surrounding the terminal 100.

[0071] More specifically, the image sensor 161 can capture images of a predetermined physical space using a camera positioned facing outwards from the terminal 100.

[0072] In one embodiment, the image sensor 161 is positioned on the front and / or rear of the terminal 100 to capture images of the physical space on the side in which it is positioned.

[0073] In this embodiment, the image sensor 161 can capture and acquire various images (e.g., images of specialized documents, etc.) related to the intelligent response agent service.

[0074] Such an image sensor 161 may include an image sensor device and an image processing module.

[0075] More specifically, the image sensor 161 can process still images or videos obtained by an image sensor device (e.g., CMOS or CCD).

[0076] Furthermore, the image sensor 161 can use the image processing module to process still images or videos acquired by the image sensor device, extract necessary information, and transmit the extracted information to the processor.

[0077] Such an image sensor 161 may be a camera assembly that includes at least one camera.

[0078] Here, the camera assembly may include a general camera that captures images in the visible light band, and may further include special cameras such as an infrared camera or a stereo camera.

[0079] Furthermore, the image sensor 161 described above can be included in and operated within the terminal 100, depending on the embodiment, or it can be included in an external device (for example, an external server) and operate in conjunction with the aforementioned communication processor 130 and / or interface unit 140.

[0080] The position sensor (IMU) 163 can sense at least one of the motion and acceleration of the terminal 100. For example, it may consist of a combination of various position sensors such as an accelerometer, a gyroscope, and / or a magnetometer.

[0081] Furthermore, the position sensor (IMU) 163 can work in conjunction with the GPS-like position communication processor 130 to recognize spatial information about the physical space around the terminal 100.

[0082] The audio sensor 165 can recognize sounds in the vicinity of the terminal 100.

[0083] More specifically, the audio sensor 165 may include a microphone capable of detecting voice input from a user using the terminal 100.

[0084] In this embodiment, the audio sensor 165 can receive voice data from the user that is necessary for the intelligent response agent service.

[0085] The interface unit 140 can connect the terminal 100 to one or more other devices so that it can communicate with them.

[0086] More specifically, the interface unit 140 may include wired and / or wireless communication devices compatible with one or more different communication protocols.

[0087] Through this interface unit 140, the terminal 100 can be connected to various input / output devices.

[0088] For example, the interface unit 140 can be connected to an audio output device such as a headset port or a speaker to output audio.

[0089] While it has been explained that the audio output device is connected via the interface unit 140 as an example, embodiments in which it is installed inside the terminal 100 are also included.

[0090] Furthermore, for example, the interface unit 140 can be connected to an input device such as a keyboard and / or mouse to acquire user input.

[0091] Such an interface unit 140 may include at least one of the following: 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.

[0092] The input system 150 can sense intelligent response agent services and associated user input (e.g., gestures, voice commands, button activations, or other types of input).

[0093] More specifically, the input system 150 may include predetermined buttons, touch sensors, an image sensor 161 for sensing user motion input, and / or an audio sensor 165 for sensing user voice input.

[0094] Furthermore, the input system 150 can be connected to an external controller via the interface unit 140 to receive user input.

[0095] The display system 170 can output various information related to the intelligent response agent service as a graphic image.

[0096] In one embodiment, the display system 170 can display various user interfaces, images, and / or text for an intelligent response agent service.

[0097] Such displays may include at least one of the following: liquid crystal displays (LCDs), thin-film transistor-liquid crystal displays (TFT LCDs), organic light-emitting diodes (OLEDs), flexible displays, 3D displays, and / or e-ink displays.

[0098] Furthermore, depending on the embodiment, the display system 170 may include a display 171 that outputs an image and a touch sensor 173 that senses user touch input.

[0099] For example, the display 171 can be realized as a touchscreen by forming an interlayer structure with the touch sensor 173 or by being formed as an integrated unit.

[0100] Such a touchscreen can function as a user input unit providing an input interface between the terminal 100 and the user, and can also provide an output interface between the terminal 100 and the user.

[0101] On the other hand, the terminal 100 according to an embodiment of the present invention can perform deep learning related to an intelligent response agent service based on a predetermined deep-learning neural network.

[0102] 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.

[0103] In more detail, in the embodiment, terminal 100 can perform deep learning necessary for an intelligent response agent service in conjunction with at least one deep learning neural network capable of realizing the Response Agent Model according to the embodiment of the present invention.

[0104] In this embodiment, the response agent model can perform deep learning, taking at least one expert document data (e.g., a paper and / or report) and question data as input, and outputting at least one response data based on the input expert document data and question data.

[0105] Alternatively, depending on the embodiment, the terminal 100 may further perform at least some of the functional operations performed by the response agent providing server 200, which will be described later.

[0106] - Response Agent Providing Server (200)

[0107] On the other hand, the response agent providing server 200 according to an embodiment of the present invention can perform a series of processes for providing an intelligent response agent service.

[0108] In more detail, in this embodiment, the response agent providing server 200 can provide an intelligent response agent service by exchanging data with an external device, such as a terminal 100, that is necessary to drive the response agent providing process at the external device.

[0109] More specifically, in the embodiment, the response agent providing server 200 can provide an environment in which the application 111 can run on an external device (in the embodiment, such as a mobile type computing device 100-1 and / or a desktop type computing device 100-2).

[0110] For this purpose, the response agent providing server 200 may include application programs, data, and / or instructions for the operation of application 111, and can send and receive various data based thereon with the external device.

[0111] In addition, the response agent providing server 200 can acquire question data.

[0112] Here, the question data in the embodiment may mean data that identifies a user's question related to a given professional document (e.g., a paper and / or report).

[0113] Furthermore, in this embodiment, the response agent providing server 200 can determine a specialized document for the acquired question data.

[0114] In other words, in this embodiment, the response agent providing server 200 can determine at least one specialized document to be used when generating response data for acquired question data.

[0115] Furthermore, in this embodiment, the response agent providing server 200 can input the determined specialized documents and question data into the response agent model according to the embodiment of the present invention.

[0116] Furthermore, in this embodiment, the response agent providing server 200 can acquire response data for the input question data in conjunction with a response agent model that has been input with specialized documents and question data.

[0117] Here, the response data according to the embodiment may mean data that identifies answers to user questions related to a given professional document (e.g., a paper and / or report).

[0118] In addition, in this embodiment, the response agent providing server 200 can provide the acquired response data in a predetermined manner (for example, by display output).

[0119] Furthermore, in this embodiment, the response agent providing server 200 can perform deep learning necessary for an intelligent response agent service based on a predetermined deep-learning neural network.

[0120] In more detail, in the embodiment, the response agent providing server 200 can perform deep learning necessary for an intelligent response agent service in conjunction with at least one deep learning neural network capable of realizing the response agent model according to the embodiment of the present invention.

[0121] More specifically, in this embodiment, the response agent providing server 200 can read a predetermined deep learning neural network driving program, which has been constructed to perform the deep learning, from the memory module 230.

[0122] The response agent providing server 200 can then perform the deep learning necessary for an intelligent response agent service in accordance with the predetermined deep learning neural network system that has been read out.

[0123] 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.

[0124] Depending on the embodiment, the deep learning neural network may be directly included in the response agent providing server 200, or it may be implemented as a separate device and / or server from the response agent providing server 200.

[0125] In the following explanation, we will describe how the deep learning neural network is implemented within the response agent providing server 200, but we will not be limited to this.

[0126] In addition, in this embodiment, the response agent providing server 200 can store and manage various application programs, commands, and / or data for realizing an intelligent response agent service.

[0127] In one embodiment, the response agent providing server 200 can store and manage at least one question data, specialized document data, response data, supporting information, data processing algorithm, deep learning algorithm, and / or user interface.

[0128] However, the functional operations that the response agent providing server 200 can perform in the embodiment of the present invention are not limited to those described above, and other functional operations can also be performed.

[0129] On the other hand, referring further to Figure 1, in an embodiment, the response agent providing server 200 may be implemented as a predetermined computing device including at least one processor module 210 for data processing, at least one communication module 220 for data exchange with external devices, and at least one or memory module 230 for storing various application programs, data and / or instruction words for providing intelligent response agent services.

[0130] Here, the memory module 230 can store one or more of the following: an operating system (OS), various application programs, data, and instruction words for providing an intelligent response agent service.

[0131] Furthermore, the memory module 230 may include a program area and a data area.

[0132] In this embodiment, the program area can be linked between the operating system (OS) that starts the server and its functional elements.

[0133] Furthermore, the data area according to this embodiment can store data generated by the use of the server.

[0134] Furthermore, the memory module 230 may be a variety of storage devices such as ROM, RAM, EPROM, flash drive, or hard drive, or it may be a web storage that performs the storage function of the memory module 230 over the internet.

[0135] Furthermore, the memory module 230 may be a recording medium that can be attached to or removed from the server.

[0136] On the other hand, the processor module 210 can control the overall operation of each of the aforementioned units in order to realize an intelligent response agent service.

[0137] For more details, the processor module 210 may be a system-on-a-chip (SOC) suitable for a server, which may include a central processing unit (CPU) and / or a graphics processing unit (GPU).

[0138] Furthermore, the processor module 210 can execute the operating system (OS) and / or applications stored in the memory module 230.

[0139] Furthermore, the processor module 210 can control each component installed in the server.

[0140] Furthermore, the processor module 210 can communicate internally with each component via a system bus, and may include one or more predetermined bus structures, including a local bus.

[0141] Furthermore, the processor module 210 can be implemented using at least one of the following: 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.

[0142] In the above description, it has been explained that the response agent providing server 200 according to the embodiment of the present invention performs the functions described above. However, depending on the embodiment, at least a part of the functions performed by the response agent providing server 200 can be performed by an external device (for example, terminal 100), and at least a part of the functions performed by the external device can be further performed by the response agent providing server 200. In short, a variety of embodiments are possible.

[0143] - Response agent model (QALM: Question and Answer Language Model)

[0144] The response agent model (RAM) according to an embodiment of the present invention may be a deep learning model that takes at least one expert document data (e.g., a paper and / or report) and question data as input and outputs at least one response data based on the input expert document data and question data.

[0145] In one embodiment, such a response agent model (RAM), upon input of specific technical documents and questions, can perform deep learning to implement a stepwise process (a multi-step inference process in one embodiment) for advanced inference and estimation capabilities, thereby interpreting the input questions more clearly and providing higher quality, more optimized response data based on this interpretation.

[0146] In this embodiment, the response agent model (RAM) may be a deep learning model (e.g., a large language model, LLM) that has been pre-trained on a training data set containing multiple expert documents in order to perform the aforementioned functional operations.

[0147] Figures 3 and 4 are examples of diagrams illustrating the multi-step inference process of a response deep learning model according to one embodiment of the present invention.

[0148] In more detail, in the embodiment, the response agent model (RAM) can generate response data to predetermined question data based on a multi-step inference process according to the embodiment of the present invention.

[0149] Here, the multi-step inference process according to the embodiment may mean a structured process that uses deep learning to generate response data based on predetermined expert document data and question data.

[0150] Such a multi-step reasoning process may include an Associative Selection Process, a Rationale Generation Process, and a Systematic Composition Process.

[0151] More specifically, in this embodiment, the response agent model (RAM) can 1) determine paragraphs within specialized documents (hereinafter referred to as "evidence paragraphs") that correspond to the question data. (Associative Selection Process)

[0152] Specifically, the Response Agent Model (RAM) can extract at least one evidence paragraph from among the K (K>=1) paragraphs contained in a specialized document, which is a paragraph containing the answer and / or justification for the question data.

[0153] Here, depending on the embodiment, the response agent model (RAM) may also perform the aforementioned related selection process by reflecting at least one or more existing question and answer data (hereinafter referred to as existing question and answer data) that are already stored.

[0154] This allows response deep learning models to generate future response data based on more extensive supporting data compared to conventional methods that extract short-sentence answers to question data.

[0155] In this embodiment, the response agent model (RAM) can perform the aforementioned related selection process based on a prompt engineering algorithm.

[0156] For reference, prompt engineering can be defined as the process of creating and optimizing prompts to obtain desired results from a deep learning model.

[0157] Here, a prompt can refer to structured text that is input to generate an output in a deep learning model.

[0158] Such prompts can guide a deep learning model to search for specific content within the data it has already learned, thereby outputting results that match the objective.

[0159] For example, components of a prompt may include an “Instruction” that identifies the task or guidance the deep learning model is expected to perform; “Context” that identifies external information or additional context that can be used to adjust the deep learning model; “Input Data” that identifies the input or question for which a response is expected; and / or “Output Data” that identifies the type or format of the output.

[0160] In other words, in this embodiment, the response agent model (RAM) can generate structured prompts (hereinafter referred to as "specialized document question prompts") corresponding to predetermined specialized documents and question data based on a prompt engineering algorithm, and can perform a relevant selection process to extract at least one evidence paragraph from the specialized document based on the generated specialized document question prompt.

[0161] Therefore, response deep learning models can extract at least one evidence paragraph based on a clearer interpretation of the question data, such as a more accurate understanding of the questions and the assumptions contained within them.

[0162] In this way, response deep learning models can significantly improve the interpretation performance of question data, and thus directly improve the quality of future response data.

[0163] Furthermore, in the embodiment, the response agent model (RAM) can acquire supporting data for the question data based on the determined evidence paragraphs. (Rationale Generation Process)

[0164] Here, the basis data according to the embodiment may mean the data that serves as the basis for generating response data to question data.

[0165] In other words, in this embodiment, the supporting data may be a collection of various data used when generating the response data.

[0166] More specifically, in the embodiment, the response agent model (RAM) can detect primary response data, which is data containing a direct answer to the question data based on at least one evidence paragraph; explanatory text data, which is data that provides a detailed explanation of the direct answer; and / or auxiliary information data, which is data containing relevant background knowledge.

[0167] The response agent model (RAM) can then obtain the aforementioned supporting data based on the detected data.

[0168] Here, depending on the embodiment, the response agent model (RAM) may also perform the aforementioned evidence generation and selection process by reflecting at least one or more existing question-and-answer data already stored.

[0169] Furthermore, in this embodiment, the response agent model (RAM) can generate response data based on the acquired evidence data. (Systematic Composition Process)

[0170] In the embodiment, the response agent model (RAM) can generate response data by performing data processing based on the acquired evidence data.

[0171] In one embodiment, the response agent model (RAM) can generate response data by performing predetermined data processing (e.g., removing duplicate text) to improve the conciseness and readability of the response based on the underlying data.

[0172] As described above, in the embodiment, the response deep learning model can extract supporting sentences from specialized documents that can answer the question data, generate evidence based on the content of the extracted supporting sentences, and provide response data that provides a more in-depth answer based on the generated evidence.

[0173] In other words, in this embodiment, the response deep learning model can be implemented as a language model specifically for specialized documents that overcomes the limitations of existing general-purpose language models (e.g., OpenAI GPT) in performing in-depth questioning and response tasks based on specialized documents, such as hallucination and ambiguous answers.

[0174] As a result, the expert document sharing platform provision system according to the embodiment of the present invention can provide response data that is faithful to the question, hallucination-controlled, and provides clear evidence to in-depth questions related to expert documents, based on advanced reasoning (chain-of-thought reasoning) and estimation (speculation) functions realized through a process similar to human cognitive reasoning.

[0175] - How to provide an intelligent response agent

[0176] Hereinafter, with reference to the accompanying drawings, a method by which an application 111 executed by at least one processor of terminal 100 according to an embodiment of the present invention generates and provides response data to questions related to specialized documents using a deep-learning neural network that realizes a stepwise process for advanced inference and estimation functions (i.e., a method for realizing an intelligent response agent service).

[0177] In an embodiment of the present invention, at least one processor of the terminal 100 can execute or run at least one application 111 stored in at least one memory 110 in the background.

[0178] In the following embodiments, it will be abbreviated to say that application 111 performs an intelligent response agent provision method based on advanced inference and estimation capabilities, in which at least one processor of terminal 100 operates to execute the instructions of application 111.

[0179] Figure 5 is a flowchart illustrating a method for providing an intelligent response agent based on advanced inference and estimation capabilities according to one embodiment of the present invention, and Figure 6 is a conceptual diagram illustrating a method for providing an intelligent response agent based on advanced inference and estimation capabilities according to one embodiment of the present invention.

[0180] Referring to Figures 5 and 6, in this embodiment, an application 111 running on at least one processor of terminal 100 or operating in the background can acquire question data (S101).

[0181] Here, the question data in the embodiment may mean data that identifies user questions related to a given professional document (e.g., a paper and / or report).

[0182] For example, the question data could be something like, "Please analyze the supporting documents that back up the first content of the first paper," or "Please explain the expected results and effects when combining the first structure of the first paper with a second structure not disclosed in the first paper."

[0183] In this embodiment, application 111 can provide a user interface (hereinafter referred to as the "question input interface") that can input predetermined question data based on an intelligent response agent service platform.

[0184] Furthermore, application 111 can obtain at least one of the aforementioned question data based on user input based on the provided question input interface.

[0185] In addition, in this embodiment, application 111 can determine the expert document for the acquired question data (S103).

[0186] In other words, in this embodiment, application 111 can determine at least one specialized document to be used when generating response data for acquired question data.

[0187] In other words, application 111 can determine at least one expert document (hereinafter referred to as a base document) to input into the response agent model (RAM) according to an embodiment of the present invention in order to generate response data for question data.

[0188] In more detail, as an embodiment, application 111 can determine 1) at least one professional document directly uploaded by the user as the base document on the intelligent response agent service platform.

[0189] In another embodiment, application 111 may determine as a base document at least one specialized document detected on an intelligent response agent service platform based on specialized document identification information (e.g., paper number and / or paper title) entered by the user.

[0190] Specifically, application 111 can determine the base document by retrieving the specialized document corresponding to each of the at least one specialized document identification pieces obtained through user input from memory 110 and / or an external database.

[0191] In another embodiment, application 111 may determine as a base document at least one expert document detected by deep learning based on 3) question data input onto an intelligent response agent service platform.

[0192] In more detail, application 111 can determine the aforementioned base document in conjunction with a deep learning model (hereinafter referred to as the "question-related document detection model") that takes predetermined question data as input and outputs at least one specialized document (hereinafter referred to as the "related document") that has a relationship with the input question data that exceeds a predetermined standard.

[0193] In other words, application 111 can input question data entered on the intelligent response agent service platform into a question-related document detection model, and obtain at least one related document corresponding to the question data from the question-related document detection model that has received the input question data.

[0194] Then, application 111 can determine the base document based on at least one related document obtained.

[0195] Here, depending on the embodiment, the question-related document detection model may be implemented as part of the response agent model (RAM), or it may have already been trained based on the training dataset used during the training of the response agent model (RAM) (i.e., a training dataset containing multiple expert documents).

[0196] In the above, the embodiments have been described in the manner described above for the sake of effective explanation, but various embodiments are possible, such as the fact that at least some of the embodiments can be organically combined and operate together depending on the embodiment.

[0197] In more detail, in this embodiment, application 111 can determine a base document based on either 1) a user input base of specialized documents (hereinafter referred to as user input documents) according to user settings, or 2) a user input document and related documents (i.e., specialized documents automatically detected by a deep learning model).

[0198] In other words, application 111 can determine the base document based on at least one expert document directly entered by the user (i.e., user input document) by user selection, or it can determine the base document based on both at least one expert document directly entered by the user (i.e., user input document) and at least one expert document automatically detected by a deep learning model (i.e., related document).

[0199] In this way, application 111 can selectively determine the specialized documents to be used when generating answers to user questions, according to the user's needs.

[0200] This enables application 111 to implement a customized response data provision process that generates answers to relevant questions based on information in the data domain desired by the user, according to the user's needs which vary depending on the situation. For example, if the user wants to obtain answers based only on the specialized documents they are targeting, or if they want to obtain answers by additionally utilizing a variety of specialized documents related to the question in addition to the specialized documents they are already familiar with.

[0201] In addition, in this embodiment, application 111 can input the determined specialized documents and question data into the response agent model (RAM) (S105).

[0202] In other words, in this embodiment, application 111 can input the base document determined as described above and the acquired question data into the response agent model (RAM) according to this embodiment of the present invention.

[0203] In other words, the response agent model (RAM) according to this embodiment may be a deep learning model that takes predetermined specialized document data and question data as input and outputs at least one response data based on the input specialized document data and question data.

[0204] In one embodiment, such a response agent model (RAM), upon input of specific technical documents and questions, can perform deep learning to implement a stepwise process (a multi-step inference process in one embodiment) for advanced inference and estimation capabilities, thereby interpreting the input questions more clearly and providing higher quality, more optimized response data based on this interpretation.

[0205] In addition, in this embodiment, application 111 can obtain response data to the input question data (S107).

[0206] Here, the response data according to the embodiment may mean data that identifies answers to user questions related to a given professional document (e.g., a paper and / or report).

[0207] In other words, response data can be data that provides answers to predetermined question data.

[0208] Figure 7 is an example of a diagram illustrating a method for obtaining response data to predetermined question data based on a response agent model (RAM) according to one embodiment of the present invention.

[0209] For details, as shown in Figure 7, in this embodiment, application 111 can acquire response data for the input question data in conjunction with a response agent model (RAM) into which a base document and question data have been input.

[0210] More specifically, referring further to Figures 3 and 4, in an embodiment, a response agent model (RAM) input with base document and question data can generate response data to the input question data according to a multi-step inference process according to an embodiment of the present invention.

[0211] In other words, the multi-step inference process according to this embodiment may mean a structured process that uses deep learning to generate response data based on predetermined expert document data and question data.

[0212] In some embodiments, such a multi-step reasoning process may include an Associative Selection Process for determining paragraphs in a specialized document that correspond to the question data (i.e., evidence paragraphs), a Rationale Generation Process for obtaining rationale data for the question data based on the determined evidence paragraphs, and a Systematic Composition Process for generating response data based on the obtained rationale data.

[0213] Furthermore, in this embodiment, the response agent model (RAM) that generates response data according to a multi-step inference process can provide the generated response data to application 111.

[0214] Therefore, in this embodiment, application 111 can acquire response data for the input question data by coordinating with the response agent model (RAM).

[0215] Thus, in this embodiment, application 111 can obtain response data for the relevant question data based on a response agent model (RAM) that performs an advanced inference and estimation process to extract supporting sentences that can answer the question data from a specialized document, generate evidence based on the content of the extracted supporting sentences, and generate a deeper answer based on the generated evidence.

[0216] Therefore, Application 111 solves the existing question-answering paradigm of search-rank-select-extract-edit in a single generation process, thereby increasing its efficiency. At the same time, it can provide more detailed, optimized, and high-quality answers based on the premise that the intent of the question is clearly interpreted based on objective data, rather than providing ambiguous answers.

[0217] In addition, in this embodiment, application 111 can provide the acquired response data (S109).

[0218] In more detail, in the embodiment, application 111 can provide at least one response data obtained as described above based on at least one question data.

[0219] Figures 8 to 10 are examples of diagrams illustrating a method for providing predetermined response data based on a response output interface according to one embodiment of the present invention.

[0220] Referring to Figure 8 for more details, in this embodiment, application 111 can provide a response output interface (API) based on an intelligent response agent service platform.

[0221] 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 related data and information in a predetermined manner.

[0222] In addition, in this embodiment, application 111 can display and provide the generated response data (AD) based on the provided response output interface (API).

[0223] In this embodiment, application 111 can provide at least one piece of supporting information corresponding to each provided response data (AD) by matching it with the corresponding response data (AD) and displaying it in conjunction with it.

[0224] Here, the basis information according to the embodiment may mean data and / or information used as supporting material when generating a predetermined response data (AD).

[0225] In embodiments, such evidence information may include specialized document data corresponding to a predetermined response data (AD), specialized document identification information (e.g., article number and / or article title), evidence paragraph data, evidence paragraph identification information (e.g., evidence paragraph identification number / identification code in a pre-established format), and / or evidence data.

[0226] Furthermore, in the embodiment, the supporting information may also further include question data corresponding to predetermined response data (AD).

[0227] More specifically, in the embodiment, application 111 can detect at least one piece of evidence corresponding to each response data (AD).

[0228] In one embodiment, application 111 can detect at least one specialized document data associated with the generation of each response data (AD), specialized document identification information relating to the specialized document data, evidence paragraph data, evidence paragraph identification information relating to the evidence paragraph data, and / or evidence data.

[0229] Furthermore, in this embodiment, application 111 can match the detected evidence information with each response data (AD) and display it in conjunction via the response output interface (API) in a predetermined format (e.g., list, image, and / or text).

[0230] In the following description, we will use as a basis an embodiment in which the supporting information corresponding to each response data (AD) is provided in a list format, but this is merely an example and is not limited to this.

[0231] In one embodiment, application 111 may further include a list interface (LI) in its response output interface (API) that contains at least one piece of rationale information for each response data (AD).

[0232] Furthermore, when application 111 displays each response data (AD) via the response output interface (API), it can match and display each response data (AD) corresponding to the list interface (LI).

[0233] In this embodiment, application 111 can match and display various related information, such as what the supporting documents used to derive each response data (AD) are, what their contents are, and where they are located, together with the corresponding response data (AD).

[0234] This allows Application 111 to clearly and intuitively present objective evidence proving the basis on which each response data (AD) was generated, thereby significantly improving the reliability and validity of the provided response data (AD).

[0235] In this embodiment, application 111 can generate separate lists by classifying each piece of evidence information according to its attributes.

[0236] The application 111 can then match each generated list to the corresponding response data (AD) based on the aforementioned list interface (LI) and provide it.

[0237] As an embodiment, application 111 can generate question data lists (L1), specialized document data lists (L2), evidence paragraph data lists (L3), and / or evidence data lists (L4), etc., by classifying them according to the attributes of each piece of evidence information.

[0238] The application 111 can then match the generated list with the corresponding response data (AD) for each piece of supporting information and display them in conjunction via the response output interface (API).

[0239] In this way, Application 111 can further enhance the readability and recognizability of the materials used as the basis for generating the response data (AD) by categorizing and displaying them by attribute and providing them in a more refined form.

[0240] On the other hand, referring to Figure 9, according to the embodiment, application 111 can display in conjunction the matching of supporting information for each sentence (hereinafter referred to as "response sentence") within each response data (AD).

[0241] In more detail, in this embodiment, application 111 can separate each response data (AD) into at least one response sentence unit.

[0242] Furthermore, application 111 can detect at least one piece of supporting information that was used when generating each separated response sentence.

[0243] Furthermore, in this embodiment, application 111 can match the detected evidence information with the corresponding response text and display it in conjunction via the response output interface (API).

[0244] For example, if the first response data (AD) consists of first to fifth response sentences, application 111 can detect the first supporting information, which is at least one piece of supporting information used when generating the first response sentence; the second supporting information, which is at least one piece of supporting information used when generating the second response sentence; ...; and the fifth supporting information, which is at least one piece of supporting information used when generating the fifth response sentence.

[0245] The application 111 can then match the detected first evidence information with its corresponding first response sentence, match the second evidence information with its corresponding second response sentence, ..., match the fifth evidence information with its corresponding fifth response sentence, and display the results via the response output interface (API).

[0246] In this way, Application 111 can allow users to more thoroughly examine and recognize the evidence supporting the response data (AD) by matching and displaying the supporting documents for each response at a textual level that is even more granular than the response data (AD) level.

[0247] On the other hand, in one embodiment, application 111 can display corresponding supporting information for each response data (AD) and / or each response text if the already set output display conditions are met.

[0248] As an embodiment, the already set display conditions may include: 1) a condition in which the response data (AD) and / or answer text is positioned at the top compared to other response data (AD) and / or other answer texts; 2) a condition in which the response data (AD) and / or answer text is positioned in a predetermined central area within the display area; or 3) a condition in which the response data (AD) and / or answer text is positioned in a user selection area (e.g., a mouseover area).

[0249] In the embodiment, application 111 can select one of the display conditions already set as described above by user input.

[0250] The application 111 can then determine whether the response data (AD) and / or the response text meet the display conditions based on the selected display conditions (hereinafter referred to as the applicable display conditions).

[0251] Furthermore, application 111 can perform the aforementioned linked information display process based on response data (AD) and / or response text that meet the applicable display conditions.

[0252] In other words, in this embodiment, application 111 can determine the real-time gaze status of response data (AD) and / or response text that changes in response to user input (e.g., mouse scrolling), and dynamically select, extract, and display relevant supporting information in conjunction with it.

[0253] This enables application 111 to achieve interactive visualization through interaction with the user to review the response data (AD), thereby allowing the user to recognize the response data (AD) and related background information more easily and intuitively.

[0254] On the other hand, referring to Figure 10, in one embodiment, application 111 can highlight, in a predetermined manner, the supporting information that matches each response data (AD) and / or each response sentence if the already set highlighting conditions are met.

[0255] As an embodiment, the already set display conditions may include: 1) a condition in which the response data (AD) and / or answer text is positioned at the top compared to other response data (AD) and / or other answer texts; 2) a condition in which the response data (AD) and / or answer text is positioned in a predetermined central area within the display area; or 3) a condition in which the response data (AD) and / or answer text is positioned in a user selection area (e.g., a mouseover area).

[0256] In the embodiment, application 111 can select one of the display conditions already set as described above by user input.

[0257] The application 111 can then determine whether the response data (AD) and / or the response text meet the display conditions based on the selected display conditions (hereinafter referred to as the applicable display conditions).

[0258] Furthermore, application 111 can highlight response data (AD) and / or response text that meet the applicable display conditions, along with corresponding supporting information, in a predetermined manner.

[0259] As an embodiment, application 111 can highlight response data (AD) and / or response text that satisfy the applicable display conditions, along with the corresponding supporting information, using bounding boxes.

[0260] This allows application 111 to further improve user readability in the interactive visualization process based on response data (AD).

[0261] Figure 11 is an example of a diagram illustrating a method for determining a base document based on supporting information according to one embodiment of the present invention.

[0262] On the other hand, referring to Figure 11, in one embodiment, application 111 can set at least some of the supporting information provided via the response output interface (API) by user input into a base document (BD).

[0263] More specifically, in this embodiment, application 111 can provide a user interface (hereinafter referred to as the base document (BD) designation interface) that allows the user to select at least a portion of the supporting information displayed via the response output interface (API).

[0264] Furthermore, in this embodiment, application 111 can determine (update) the base document (BD) based on the basis information (C: hereinafter referred to as base basis information) selected by user input based on the provided base document (BD) specification interface.

[0265] In one embodiment, application 111 can determine the base document (BD) by using specialized document data and / or evidence paragraph data selected by user input based on the base document (BD) designation interface as base evidence information (C).

[0266] In this embodiment, application 111 can add base information (C) to an existing base document (BD) based on user input.

[0267] In another embodiment, application 111 can replace an existing base document (BD) using base basis information (C) based on user input.

[0268] Furthermore, in this embodiment, application 111 can execute the processes described above, starting from step S105, based on the updated base document (BD).

[0269] Therefore, application 111 can expand the paradigm of intelligent response agent services while simultaneously improving their usability by effectively supporting newly emerging user needs by providing answers, such as when a user wants to obtain new answers targeting at least some of the specialized documents detected by artificial intelligence in addition to the specialized documents they were already familiar with, or when they want to obtain new answers by further focusing on paragraphs within a particular specialized document.

[0270] In this embodiment, when application 111 updates the base document (BD), it can acquire new question data through user input.

[0271] Furthermore, application 111 can update existing question data based on newly acquired question data.

[0272] In one embodiment, application 111 can add new question data to existing question data based on user input.

[0273] In another embodiment, application 111 can replace existing question data with new question data based on user input.

[0274] Furthermore, in this embodiment, application 111 can execute the processes described above up to step S105 based on the updated question data.

[0275] In this way, application 111 can enhance its convenience by assisting users in easily creating additional questions with a high degree of flexibility even after receiving answers.

[0276] As described above, the intelligent response agent provision method and system based on advanced reasoning and estimation functions according to one embodiment of the present invention utilizes a deep-learning neural network to realize a stepwise process for advanced reasoning and estimation functions to generate and provide response data (AD) to questions related to specialized documents. This enables the provision of response data that is faithful to the question, hallucination-controlled, and provides clear evidence for in-depth questions related to specialized documents, based on advanced reasoning (chain-of-thought reasoning) and estimation (speculation) functions realized through a process similar to human cognitive reasoning.

[0277] Furthermore, the intelligent response agent provision method and system based on advanced inference and estimation functions according to one embodiment of the present invention provides generated response data (AD) and supporting evidence information in an interconnected manner, thereby enabling intuitive and easy verification of objective evidence including the basis for generating the response data (AD), and thereby improving the reliability and validity of the response data (AD).

[0278] On the other hand, the embodiments of the present invention described above can be implemented in the form of program instructions that can be executed by 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., individually or in combination. The program instructions recorded on the computer-readable recording medium may be specifically designed and configured for the present invention, or may be publicly known and usable by those skilled in the field 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 ROMs, RAMs, and flash memory. Examples of program instructions include not only machine code that can be produced by a compiler, but also high-level language code that can be executed by a computer using an interpreter or the like. Hardware devices may be modified into one or more software modules to perform the processing according to the present invention, and vice versa.

[0279] The specific executions described herein are merely embodiments and do not in any way limit the scope of the invention. For the sake of brevity, descriptions of conventional electronic configurations, control systems, software, and other functional aspects of such systems may be omitted. Furthermore, the connections of lines or connecting members between components shown in the drawings are illustrative representations of functional and / or physical or circuit connections and may be replaced or represented as a variety of additional functional, physical, or circuit connections in actual devices. Also, components that are not necessarily required for the application of the invention may not be necessary unless specifically mentioned, such as "essential" or "important."

[0280] Furthermore, while the detailed description of the present invention has been provided with reference to preferred embodiments, a person skilled in the art or someone with ordinary knowledge of the art will understand that the present invention can be modified and altered in various ways, within the scope of the concept and technical domain of the invention as described in the claims. Therefore, the technical scope of the present invention is not limited to what is described in the detailed description of the specification, but must be defined by the claims. [Industrial applicability]

[0281] The present invention relates to a method and system for providing an intelligent response agent based on advanced reasoning and estimation capabilities, and is applicable to the artificial intelligence industry, thus having industrial applicability.

Claims

1. A method for providing an intelligent response agent based on advanced inference and estimation capabilities, wherein a response agent application run on at least one processor of a terminal, The stage of obtaining question data to identify the user's question, The stage of determining a specialized document for the acquired question data, The steps include detecting evidence paragraphs in the determined specialized document that include at least one sentence that corresponds to the question data, The steps include generating response data that identifies the answer to the question based on the detected evidence paragraphs, The step of providing the generated response data includes, A method for providing intelligent response agents based on advanced inference and estimation capabilities.

2. The steps include obtaining the underlying data, which is the basis data for generating the response data based on the detected evidence paragraphs, The step of generating response data based on the acquired basis data further includes, A method for providing an intelligent response agent based on the advanced inference and estimation functions described in claim 1.

3. The aforementioned supporting data includes at least one of the following: primary response data including a direct answer to the question data; explanatory text 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 the advanced inference and estimation functions described in claim 2.

4. The step of detecting the evidence paragraph includes the step of detecting the evidence paragraph based on a prompt engineering algorithm. A method for providing an intelligent response agent based on the advanced inference and estimation functions described in claim 1.

5. The steps include inputting the determined specialized document and the question data into a Response Agent Model that executes a Multi-step Inference Process, The step further includes generating the response data based on the response agent model, The multi-step inference process includes an Associative Selection Process for detecting the evidence paragraph, a Rationale Generation Process for obtaining the evidence data, and a Systematic Composition Process for generating the response data. A method for providing an intelligent response agent based on the advanced inference and estimation functions described in claim 2.

6. The step of generating the response data based on the response agent model includes the step of performing the related selection process and the rationale generation process, reflecting at least one question data and response data already stored. A method for providing an intelligent response agent based on the advanced inference and estimation functions described in claim 5.

7. The step of providing the generated response data includes the step of providing the basis information for the response data by matching it with the response data, The aforementioned basis information includes at least two of the following: the basis data, the evidence paragraph data, the evidence paragraph identification information, the specialized document data, the specialized document identification information, and the question data. A method for providing an intelligent response agent based on the advanced inference and estimation functions described in claim 2.

8. The step of providing the generated response data is, The steps include separating the response data into at least one sentence unit, The step of detecting the aforementioned evidence information for each separated text, The step further includes matching the detected evidence information for each document with the document corresponding to that evidence information and providing it, A method for providing an intelligent response agent based on the advanced inference and estimation functions described in claim 7.

9. The stage of determining the expert document for the aforementioned question data is: A step of determining a specialized document for the question data based on at least one specialized document entered by the user, A step of determining a specialized document for the question data based on a specialized document corresponding to at least one specialized document identification piece entered by the user, The process includes at least one of the following steps: detecting at least one expert document that has a correlation with the question data that exceeds a previously established criterion based on a predetermined deep learning neural network, and determining an expert document for the question data based on the detected at least one expert document. A method for providing an intelligent response agent based on the advanced inference and estimation functions described in claim 1.

10. At least one memory location where the response agent application is stored, The system includes at least one processor that reads response agent applications stored in the memory and provides an intelligent response agent based on advanced inference and estimation capabilities, The command words of the aforementioned response agent application are: The stage of obtaining question data to identify the user's question, The stage of determining a specialized document for the acquired question data, A step of detecting evidence paragraphs in the determined specialized document that include at least one paragraph that corresponds to the question data, The steps include generating response data that identifies the answer to the question based on the detected evidence paragraphs, The step of providing the generated response data includes an instruction word that performs the following steps: An intelligent response agent delivery system based on advanced inference and estimation capabilities.

Images