Method and system for providing conversational ai answer service

Abstract

The present invention relates to a method and a system for providing a conversational artificial intelligence answer service. More specifically, a method for providing a conversational artificial intelligence answer service according to the present invention may comprise the steps of: receiving a user query through a service page; specifying location information from the user query; searching at least one DB for a satellite image related to the user query by using the user query and the location information; performing analysis related to the user query by using the satellite image; generating a prompt by using at least one of the user query, the location information, an analysis result according to the analysis, and the satellite image; processing the prompt as an input of a large-scale language model (LLM); acquiring an answer to the user query from the large-scale language model; and providing the answer to the user query to the service page.

Description

Method and system for providing interactive artificial intelligence answering service

[0001] The present invention relates to a method and system for providing an interactive artificial intelligence answer service.

[0002] Recently, there has been a rapid increase in cases where artificial intelligence, especially deep learning, which extracts data characteristics using deep neural network structures, is achieving outstanding results in various fields such as voice recognition, image recognition, natural language processing, and autonomous driving.

[0003] Along with these advancements in deep learning technology, generative AI technology has recently been attracting attention. More specifically, generative AI models can generate new data in various forms, such as text, images, and voice, from given data. This offers a new level of application potential beyond simply classifying or predicting existing data.

[0004] In other words, as sentences, images, voices, etc. that were previously created by humans can now be automatically generated using generative artificial intelligence models, services using generative artificial intelligence (e.g., ChatGPT) have shown activity and accuracy that differentiate them from existing chatbot services, and are receiving great attention worldwide.

[0005] Meanwhile, advancements in satellite technology are increasingly making satellites increasingly important in areas such as Earth observation, weather forecasting, and communications. Through their sensors and cameras, satellites can collect diverse types of data, including high-definition images and videos, geographic information, and environmental data, providing a wealth of information about the Earth in real time.

[0006] Along with these advancements in satellite technology, there has been a surge in research and attempts to integrate artificial intelligence technology into satellite data and utilize it in various fields. The analysis and processing of the massive amounts of images, videos, and data collected by satellites is crucial, and high-resolution satellite imagery, in particular, is being utilized in diverse fields such as agriculture, disaster management, urban planning, and environmental monitoring.

[0007] Likewise, research is actively underway to incorporate artificial intelligence technology into the satellite field, and there is a need for methods to provide a wider range of satellite-related services by utilizing generative artificial intelligence models.

[0008] The present invention provides a method and system for providing an interactive artificial intelligence answer service that can efficiently analyze satellite data and provide users with information that can be utilized in various fields.

[0009] More specifically, the present invention provides a method and system for providing an interactive artificial intelligence answer service that analyzes satellite data according to a user request and provides information desired by the user through an interactive artificial intelligence chatbot.

[0010] Furthermore, the present invention provides a method and system for providing an interactive artificial intelligence answer service that can provide satellite images collected from satellites to users.

[0011] Furthermore, the present invention provides a method and system for providing an interactive artificial intelligence answer service capable of selling high-resolution satellite images to users.

[0012] In order to solve the problem discussed above, a method for providing an interactive artificial intelligence answer service according to the present invention may include the steps of receiving a user query through a service page, specifying location information from the user query, searching for satellite images related to the user query from at least one DB using the user query and the location information, performing an analysis related to the user query using the satellite image, generating a prompt using at least one of the user query, the location information, an analysis result according to the analysis, and the satellite image, processing the prompt as an input to a large-scale language model (LLM), obtaining an answer to the user query from the large-scale language model, and providing an answer to the user query on the service page.

[0013] Furthermore, the service page may include a first area for receiving the user query and a second area for providing the satellite image searched in response to the user query.

[0014] Furthermore, text-based information related to the user query may be provided in the first area, and image-based information related to the user query may be provided in the second area.

[0015] Furthermore, a map image is displayed in the second area, and when the satellite image is retrieved in response to the user query, the satellite image may overlap at least a portion of the map image.

[0016] Furthermore, the area where the satellite image overlaps in the map image can be specified based on the location information specified for searching the satellite image.

[0017] Furthermore, the method may further include a step of receiving a user input designating a specific area in the satellite image provided in the second area, a step of receiving an additional user query related to the specific area in the first area, and a step of generating an answer to the additional user query using at least one of the satellite image of the specific area, meta information related to the specific area, and the additional user query.

[0018] Furthermore, the step of generating an answer to the additional user query may include the steps of analyzing a user intent of the additional user query, specifying a function corresponding to the user intent, and selecting a specific analysis model that performs the specified function, processing a satellite image of the specific area as an input of the specific analysis model, obtaining an output corresponding to the user intent from the specific analysis model, using the output to generate an answer to the additional user query, and providing the answer to the additional user query to at least one of the first area and the second area.

[0019] Furthermore, the response to the additional user query provided in the second area may include highlighting at least a portion of the satellite image of the specific area provided in the second area.

[0020] Furthermore, the service page may provide at least one tool icon for receiving the user input for the satellite image provided in the second area, and the tool icons may be linked to different functions, a first tool icon among the tool icons may be linked to a first selection function for selecting at least a portion of the satellite image according to a first criterion, and a second tool icon among the tool icons may be linked to a second selection function for selecting at least a portion of the satellite image according to a second criterion different from the first criterion.

[0021] Furthermore, when the user query is received, the method may further include a step of processing the user query as an input to the large-scale language model, selecting a model from the large-scale language model that performs analysis corresponding to the user query, a step of inputting the satellite image as an input to the selected model, and a step of receiving an analysis result for the satellite image corresponding to the user query from the selected model.

[0022] Furthermore, in the step of generating the prompt, the analysis results received from the selected model are included in the prompt, and in the large-scale language model, an answer to the user query can be generated using the analysis results and the satellite image.

[0023] Meanwhile, a method for providing an interactive artificial intelligence answer service according to the present invention further includes a step of receiving a user query through a service page including a first area and a second area, a step of specifying a document to be used for generating an answer to the user query, a step of processing the user query and the specified document as inputs of a large-scale language model, a step of obtaining an answer to the user query from the large-scale language model, and a step of providing the answer to the service page, wherein the answer may be provided in the first area of ​​the service page, and information about the specified document may be provided in the second area of ​​the service page.

[0024] Meanwhile, a system for providing an interactive artificial intelligence answer service according to the present invention includes a memory and at least one processor, wherein the memory and the processor cooperate to receive a user query through a service page, specify location information from the user query, search for satellite images related to the user query from at least one DB using the user query and the location information, perform analysis related to the user query using the satellite image, generate a prompt using at least one of the user query, the location information, an analysis result according to the analysis, and the satellite image, process the prompt as an input of a large-scale language model (LLM), obtain an answer to the user query from the large-scale language model, and provide the answer to the user query on the service page.

[0025] Meanwhile, a program according to the present invention is a program that is executed by one or more processes in an electronic device and can be stored in a computer-readable recording medium, wherein the program may include commands that perform the steps of: receiving a user query through a service page; specifying location information from the user query; searching for satellite images related to the user query from at least one DB using the user query and the location information; performing an analysis related to the user query using the satellite image; generating a prompt using at least one of the user query, the location information, an analysis result according to the analysis, and the satellite image; processing the prompt as an input to a large-scale language model (LLM); obtaining an answer to the user query from the large-scale language model; and providing the answer to the user query on the service page.

[0026] As discussed above, the interactive AI answer service providing method and system according to the present invention can utilize searched documents to generate answers that improve hallucinations and provide them to users. This can contribute to reducing the provision of incorrect information and providing users with highly reliable information.

[0027] Furthermore, the method and system for providing an interactive AI answering service according to the present invention can analyze satellite data related to a user's query and provide a response to the user's inquiry along with visual information. This allows the user to more intuitively confirm the response to the query.

[0028] Furthermore, according to the method and system for providing an interactive artificial intelligence answer service according to the present invention, by comprehensively analyzing various data including satellite images in response to a user's inquiry and providing various information required by the user, the user can utilize the provided information in various fields such as agriculture, environmental monitoring, and urban planning, thereby increasing efficiency across various industries.

[0029] Furthermore, the method and system for providing an interactive AI answering service according to the present invention can provide answers tailored to a user's inquiry by utilizing specific resources at the user's request. In other words, the present invention can provide accurate and reliable information that meets the user's needs without requiring a complex information search process.

[0030] Figure 1 is a conceptual diagram illustrating a conversational artificial intelligence answer service providing system according to the present invention.

[0031] FIG. 2a, FIG. 2b, FIG. 3, FIG. 2a, FIG. 2b, FIG. 3, FIG. 4a, FIG. 4b, FIG. 4c, FIG. 4d, and FIG. 4e are conceptual diagrams for explaining various modes provided in a conversational artificial intelligence answer service providing system according to the present invention.

[0032] Figure 5 is a flowchart illustrating a method for providing an interactive artificial intelligence answer service according to the present invention.

[0033] Figures 6a and 6b are conceptual diagrams for explaining a method for providing an interactive artificial intelligence answer service according to the present invention.

[0034] FIG. 7, FIG. 8, FIG. 9, FIG. 10 and FIG. 11 are conceptual diagrams for explaining a method for selling satellite images and a method for providing paid services according to one embodiment of the present invention.

[0035] Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Regardless of the drawing numbers, identical or similar components will be given the same reference numbers, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" used for components in the following description are assigned or used interchangeably only for the convenience of writing the specification, and do not in themselves have distinct meanings or roles. In addition, when describing the embodiments disclosed in this specification, if it is determined that a specific description of a related known technology may obscure the gist of the embodiments disclosed in this specification, a detailed description thereof will be omitted. In addition, the attached drawings are only intended to facilitate easy understanding of the embodiments disclosed in this specification, and the technical ideas disclosed in this specification are not limited by the attached drawings, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and technical scope of the present invention.

[0036] Terms that include ordinal numbers, such as first, second, etc., may be used to describe various components, but the components are not limited by these terms. These terms are used solely to distinguish one component from another.

[0037] When a component is referred to as being "connected" or "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but that there may be other components intervening. Conversely, when a component is referred to as being "directly connected" or "connected" to another component, it should be understood that there are no other components intervening.

[0038] Singular expressions include plural expressions unless the context clearly indicates otherwise.

[0039] In this application, terms such as “include” or “have” are intended to specify the presence of a feature, number, step, operation, component, part or combination thereof described in the specification, but should be understood not to exclude in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts or combinations thereof.

[0040] The present invention relates to a method and system for providing an interactive AI answer service. The interactive AI answer service providing system according to the present invention generates answers based on generative AI or a large language model (LLM). It may also be referred to as a large language model-based interactive AI answer service providing system. However, for convenience of explanation, it will be referred to as an "answer service providing system" hereinafter.

[0041] The answer service providing system according to the present invention may be a system that, when a user inquiry (or query) related to a satellite and / or satellite image (or image) is received, searches for a document related to the user inquiry from a database (or storage), and, using the searched document and a large-scale language model (LLM), generates and provides an answer that matches the intent of the user's inquiry.

[0042] Here, “satellite image” refers to an image recorded by a detector (or sensor (e.g., image sensor, camera, etc.)) mounted on an artificial satellite, and in a broad sense, may include space photographs CLFC taken by a camera, etc.

[0043] Furthermore, the answer service providing system according to the present invention may be a system that, when a user query related to a satellite and / or satellite image is received, searches a database (or storage) for satellite images related to the user query, and, utilizing the analysis results from the searched satellite images and a large-scale language model, generates and provides an answer tailored to the user's query intent. More specific details regarding this will be described later.

[0044] Meanwhile, the answer service providing system according to the present invention aims to analyze satellite data according to a user request and provide information desired by the user (e.g., information corresponding to the user's inquiry intent) through an interactive artificial intelligence chatbot.

[0045] Hereinafter, we will examine in more detail with the attached drawings. Fig. 1 is a conceptual diagram for explaining an interactive artificial intelligence answer service providing system according to the present invention, and Figs. 2a, 2b, 3, 4a, 4b, 4c, 4d, and 4e are conceptual diagrams for explaining various modes provided in the interactive artificial intelligence answer service providing system according to the present invention. Fig. 5 is a flowchart for explaining an interactive artificial intelligence answer service providing method according to the present invention, and Figs. 6a and 6b are conceptual diagrams for explaining an interactive artificial intelligence answer service providing method according to the present invention. Furthermore, Figs. 7, 8, 9, 10, and 11 are conceptual diagrams for explaining a satellite image selling method and a paid service providing method according to an embodiment of the present invention.

[0046] Meanwhile, as illustrated in FIG. 1, the answer service providing system (100) according to the present invention may include at least one of an input unit (110), an output unit (120), a communication unit (130), a storage unit (140), and a control unit (150).

[0047] Although not illustrated, the answer service providing system (100) according to the present invention may include one or more processors, which may include one or more general-purpose processors and / or one or more special-purpose processors (e.g., digital signal processors, tensor processing units (TPUs), graphics processing units (GPUs), neural network processing units (NPUs), application-specific integrated circuits, application-specific integrated circuits (ASICs), etc.). The one or more processors may be configured to execute instructions stored (or included) in the storage unit (140), computer-readable instructions, and / or other instructions described herein. The answer service providing system and method according to the present invention may enable the memory and at least one processor to cooperate to perform data processing as described below. The processor may perform a series of operations and data processing using data and information stored in the memory. At this time, the memory may be a component of the storage unit (140).

[0048] Meanwhile, the input unit (110) may be configured as a means for data input and may be configured in various types. For example, the input unit (110) may be configured to receive user input. The input unit (110) may be configured to receive user input from a user terminal (10). Here, “receiving input” may mean receiving an input signal (or selection signal) corresponding to the user’s input based on the input being made by the user through the input unit configuration provided in the user terminal (10).

[0049] The input unit (110) may also be referred to as a user interface module. The input unit (110) may include a touch screen, a computer mouse, a keyboard, a keypad, a touch pad, a trackball, a joystick, a voice recognition module, or other similar devices. However, the present invention does not place any limitations on the type of input unit (110). Furthermore, the input unit (110) in the present invention does not necessarily refer to a hardware means, but can be understood as a channel for receiving input from a user.

[0050] Here, the user input may include documents, text, images (or videos), voice, etc. In this case, the answer service providing system (100) may further include a module that converts voice into text.

[0051] Next, the output unit (120) can output information through an output unit configuration (e.g., a display unit, a touch screen, a speaker, etc.) provided in a user terminal (10) linked to the answer service providing system (100) according to the present invention. For example, the output unit (120) can output a page (or service page, 1000) linked to the answer service providing system (100) according to the present invention to the display unit of the user terminal (10). In addition, the output unit (120) does not necessarily mean a hardware means, and can be understood as a passage for outputting results to the user.

[0052] Next, the communication unit (130) may be connected to a user terminal (10), an LLM server (or an artificial intelligence server, 20), a satellite image DB (30), a central server, an external server, a device, and at least one network through a wireless or wired network, and may be configured to receive or transmit overall data and information necessary for the operation of the answer service providing system (100) according to the present invention.

[0053] In this case, the communication unit (130) may include one or more communication modules to enable wireless and / or wired communication between the answer service providing system (100) and the user terminal (10), between the answer service providing system (100) and the artificial intelligence server (20), and between the answer service providing system (100) and the satellite image DB (30). In addition, the communication unit (130) may include one or more communication modules that connect the answer service providing system (100) to one or more networks.

[0054] Here, the user terminal (10) may include at least one of a mobile phone, a smart phone, a notebook computer, a laptop computer, a slate PC, a tablet PC, an ultrabook, a desktop computer, a digital broadcasting terminal, a personal digital assistant (PDA), a portable multimedia player (PMP), a navigation device, and a wearable device (e.g., a smartwatch, a smart glass, a head mounted display (HMD)).

[0055] Furthermore, the communication unit (130) can support various communication methods according to the communication standards of the communicating device.

[0056] For example, the communication unit (130) may be configured to communicate with a communication target using at least one of WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Wi-Fi (Wireless Fidelity) Direct, DLNA (Digital Living Network Alliance), WiBro (Wireless Broadband), WiMAX (World Interoperability for Microwave Access), HSDPA (High Speed ​​Downlink Packet Access), HSUPA (High Speed ​​Uplink Packet Access), LTE (Long Term Evolution), LTE-A (Long Term Evolution-Advanced), 5G (5th Generation Mobile Telecommunication), Bluetooth™, RFID (Radio Frequency Identification), Infrared Data Association (IrDA), UWB (Ultra-Wideband), ZigBee, NFC (Near Field Communication), Wi-Fi Direct, and Wireless USB (Wireless Universal Serial Bus) technologies.

[0057] Meanwhile, the storage unit (140) serves to store various data related to the present invention, and may include one or more non-transitory computer-readable storage media that can be read and / or accessed by at least one of one or more processors.

[0058] The one or more computer-readable storage media may include volatile and / or non-volatile storage components, such as optical, magnetic, organic, or other memory or disk storage devices. In some examples, the storage (140) may be implemented using a single physical device (e.g., a single optical, magnetic, organic, or other memory or disk storage device), while in other examples, the storage (140) may be implemented using two or more physical devices.

[0059] The storage unit (140) may include computer-readable instructions and additional data. The storage unit (140) may include storage necessary to perform at least some of the methods, scenarios, and techniques described herein and / or at least some of the functions of the devices and networks.

[0060] Furthermore, at least a portion of the storage unit (140) may be a cloud storage or a cloud server. At least a portion of data corresponding to user input received from the input unit (110) and learning data may be stored in the storage unit (140).

[0061] That is, it can be understood that the storage unit (140) is sufficient as a space in which information necessary for the operation of the answer service providing system (100) according to the present invention is stored, and there are no restrictions on the physical space.

[0062] Meanwhile, the control unit (150) may perform a role of controlling the overall operation of the answer service providing system (100) related to the present invention. The control unit (150) may process signals, data, information, etc. input or output through the components discussed above, or perform a series of data processing to provide or process appropriate information and functions to the user.

[0063] When a user query related to a satellite and / or satellite image is received, the control unit (150) may retrieve at least one document related to the user query from the storage unit (140) and / or at least one repository and / or document DB, and may generate an answer to the user query using the retrieved document and a large-scale language model (160) and provide the answer to the user. In this case, the control unit (150) may provide the user with at least one document used (or referenced) to generate the answer to the user query, along with the answer.

[0064] Furthermore, when a user query related to a satellite and / or satellite image is received, the control unit (150) may search for at least one satellite image related to the user query from the storage unit (140) and / or at least one repository and / or satellite image DB (30), and may generate an answer to the user query using the analysis result according to the analysis of the searched satellite image and the large-scale language model (160) and provide the answer to the user. In this case, the control unit (150) may provide the user with at least one of the satellite image, location information, and map used (or referred to) to generate the answer to the user query, together with the answer.

[0065] In one embodiment, the analysis results from the satellite image analysis may be related to at least one of the following: satellite image and coordinate search results, satellite orbit prediction results, satellite location information query results, spectral index calculation (or output), and object detection and segmentation results. However, the analysis results from the satellite image analysis in the present invention are not necessarily limited to the examples mentioned above.

[0066] In this regard, the control unit (150) may include at least one model (or module) used to generate a response to a user query (or analyze satellite imagery based on the user query). In the present invention, the model may be configured to perform a specific function and output a result value based on the performance result. Furthermore, in the present invention, the model may also be understood as a "tool." In this case, the tool may include a function or algorithm for performing a specific task.

[0067] In one embodiment, when a user query includes content related to detecting a vehicle included in a satellite image, the control unit (150) may use the first model (170) to detect a vehicle included in the satellite image and provide the user with a response including content related to the number of detected vehicles. The first model (170) may be an object detection model trained to detect at least one object.

[0068] In another embodiment, when a user query includes content related to NDVI (Non-Dispersive Visual Index) analysis of a specific area (or region) included in a satellite image, the control unit (150) may analyze the vegetation index of the specific area using the second model (180) and provide the user with an answer including content related to the analysis result of the vegetation index. The second model (180) may be a spectral index analysis model that has been trained to produce information related to the spectral index analysis result through analysis of the image.

[0069] However, the model used to generate answers to user queries in the present invention is not necessarily limited to the aforementioned models, and may include other models in addition to the first model (170) and the second model (180). It goes without saying that the model used to generate answers to user queries in the present invention may be one or more, and may vary in various ways, depending on the case.

[0070] Meanwhile, the answer service providing system (100) according to the present invention can be implemented in various platform forms such as applications, software, and websites. In this specification, for convenience of explanation, the form in which the answer service providing system (100) is implemented is not limited to any one. The answer service providing system (100) according to the present invention can also be referred to as an "interactive artificial intelligence chatbot service providing platform."

[0071] The user discussed above may have a user account pre-registered in the answer service providing system (100) according to the present invention. In this case, the account may be created through a page (or screen) linked to the answer service providing system (100). Alternatively, the account may also be created in at least one other system linked to the answer service providing system (100). However, in this specification, the system where the user account is issued is not separately distinguished, and all accounts that can utilize the various services (or functions) provided by the artificial intelligence chatbot service providing system (100) according to the present invention are referred to as "accounts pre-registered in the answer service providing system (100)."

[0072] Meanwhile, the response service provision system (100) can provide users with multiple modes containing different functions. Here, the multiple modes can include a first mode and a second mode.

[0073] In one area of ​​a service page (1000) linked to a response service provision system (100), a graphic object for selecting one of a plurality of modes from a user terminal (10) may be provided.

[0074] For example, as illustrated in FIG. 5, the control unit (150) can provide (or display or output) a graphic object (312a) for selecting one of a plurality of modes in an area (310) of a service page (1000).

[0075] The control unit (150) can output a GUI (Graphic User Interface) configured to select either the first mode or the second mode based on the selection of the graphic object (312a) from the user terminal (10). In addition, when either the first mode or the second mode is selected by the user, the control unit (150) can control the mode corresponding to the user's selection to be activated.

[0076] In one embodiment, the GUI output may be in the form of a dropdown menu. However, the form of the GUI for selecting a specific mode is not necessarily limited to this, and it is obvious that it may be output in various forms.

[0077] In this case, the activation (or switching) of the first mode and / or the second mode can be understood to be performed based on the user's selection. Furthermore, the activation of the first mode and / or the second mode can also be controlled so that it is activated automatically by the control unit (150) based on the user account's history information. Below, the multiple modes will be examined in more detail.

[0078] First, let's look at the first mode among the multiple modes.

[0079] The first mode (or knowledge mode) may also be named “Knowledhe Mode” and may be a mode configured to retrieve documents related to the user query from the storage (140) and / or at least one repository and / or document DB when a user query related to satellite and / or satellite imagery is received, and to generate an answer to the user query using the retrieved documents and a large-scale language model (160) and provide the answer to the user.

[0080] In one embodiment, the first mode may be configured to search for related content from a document DB when a user needs information or documents related to a specific topic in the field of satellite and / or satellite imagery, and to generate and provide an answer that matches the user's query intent through the searched content.

[0081] That is, in the first mode, the user may be provided with an answer to a user query and at least one document used (or specified) to generate the answer.

[0082] Meanwhile, as illustrated in FIGS. 2a and 2b, the control unit (150) can receive a user query (e.g., “Raw Query”, 200) corresponding to a user input (e.g., “User Input”). Here, the reception of the user query (200) can be received through a service page (1000) linked to the answer service providing system (100).

[0083] Referring to FIG. 3, the service page (1000) may include at least one of a first area (310) that receives a user query and a second area (320) that provides a document searched in response to the user query.

[0084] Specifically, the first area (310) may include at least one of a first sub-area (311) for receiving a user query and a second sub-area (312) for providing (or displaying) a response to the user query. In this case, the first sub-area (311) is an area where user input is made and may also be referred to as an input area.

[0085] The control unit (150) can receive a user query input from a user terminal (10) through a service page (1000). In this case, a first sub-area (311) of the service page (1000) can include a graphic object (311a) linked to a user query receiving function. For example, the control unit (150) can receive a user query corresponding to a user input (e.g., “Tell me about the trend of reused projectiles”, 200) based on the selection of the graphic object (311a) from the user terminal (10).

[0086] In one embodiment, the service page (1000) may further include a third area (330). The third area (330) may be configured to provide information related to a conversation event that occurred in a user account. The third area (330) may include a graphic object corresponding to a conversation list (or list) generated according to a conversation event that occurred in the user account. The control unit (150) may generate a chat list related to the user query (200) based on the user query (200) being received, and provide a graphic object (331) corresponding to the chat list in the third area (330) of the service page (1000). In addition, the control unit (150) may display (or output) the received user query (200) in the second sub-area (312) of the service page (1000).

[0087] Meanwhile, the user query (200) can be rewritten through a large-scale language model (160) provided by the artificial intelligence server (20).

[0088] The control unit (150) can use a user query (200) to generate a prompt (e.g., “Rewrite Prompt Template”, 201) to be input into a large-scale language model (160). The elements (or information) constituting the prompt (201) can be understood to exist and be stored in a storage unit (or memory, 140).

[0089] In one embodiment, the present invention may further include a prompt generation unit (or model) that generates a prompt, but for convenience of explanation, in this specification, it is described that the prompt is generated by the control unit (150) itself.

[0090] The prompt (201) generated by the control unit (150) may include the user query (200) input at the current point in time. At this time, if a conversation (i.e., a previous conversation) exists before the input point of the user query (200), the control unit (150) may generate the prompt (201) by using at least one of the previous conversation and the user query (200) so that the user query (200) is rewritten as an independent input value including the corresponding content from a large-scale language model (160) in consideration of the correlation with the previous conversation. In the following, for the convenience of explanation, it is assumed that the prompt (201) including at least one of the previous conversation and the user query (200) is generated.

[0091] The control unit (150) can process the generated prompt (201) as input to a large-scale language model (160). The large-scale language model (160) can generate a rewritten user query (202) through rewriting the user query (200) so as to confirm (or understand) what intention the user query (200) contains based on the input prompt (201). In this case, the large-scale language model (160) can be understood as an artificial intelligence model provided by an artificial intelligence server (e.g., OpenAI's server, 20).

[0092] Meanwhile, the control unit (150) can process the user query (202) rewritten by the large-scale language model (160) as input to an embedding model (e.g., “Embedding Model”, 210). The embedding model (210) can vectorize the rewritten user query (202) to generate an embedded query vector (e.g., “Embedded Query Vector”, or embedding vector or query vector, 211).

[0093] The control unit (150) can input a query (211) vectorized by the embedding model (210) into a vector DB (e.g., “VectorDB”, or document DB, 220). The control unit (150) can perform a search for at least one document based on the similarity (e.g., cosine similarity) between the embedded query vector (211) and documents existing in the vector DB (220).

[0094] And, the control unit (150) can process the documents (221) searched from the vector DB (220) as input to the document alignment model (or document reordering model, 230). The document alignment model (230) can perform alignment (or reordering) on ​​the searched documents (221). For example, the document alignment model (230) can use a sigmoid function to calculate the similarity between the searched documents (221) and the user's query pairs in order to reorder the searched documents (221), and can express it as a score between 0 and 1. In this case, the document alignment model (230) can filter only the documents whose scores satisfy a preset standard (e.g., 0.7 or higher) and extract the sorted documents (e.g., “Reranked Documents”, or reordered documents, 231).

[0095] The control unit (150) can generate a prompt (e.g., “RAG Prompt Template”, 240) to be input into a large-scale language model (160) using at least one of a rewritten user query (202) and a sorted document (231). For example, the prompt (240) can include at least one of a rewritten user query (202), a sorted document (231), and a previous conversation (or chat) of the user. The elements (or information) constituting the prompt (240) can be understood to exist and be stored in the storage unit (or memory, 140).

[0096] Furthermore, the control unit (150) can process the generated prompt (240) as input to a large-scale language model (160). The large-scale language model (160) can generate an answer (ex: “Output Answer”, 250) to a user query (200) based on the input prompt (240). For example, the answer (250) to the user query (200) can be a text-based answer generated based on an aligned document (231) using the RAG (Retrieval-Augmented Generation) method in the large-scale language model (160).

[0097] An answer (250) to a user query (200) and at least one document (ex: “Output Documents”, 251) used in the process of generating the answer (250) may be transmitted to a backend (ex: “backend”, or backend system, 260) and a frontend (ex: “Front-end”, 270). The backend (260) manages the user’s conversation, the answer (250), the document (251), etc., and may store information related to the user’s conversation history in a DB (ex: “Conversation MongoDB”, 261). In addition, the front end (270) may provide (or display) the answer (250) and the document (251) to the user terminal (10).

[0098] Meanwhile, the control unit (150) can provide a response (250) to a user query (200) through a service page (1000).

[0099] The control unit (150) can provide an answer (250) to a user query (200) in one area of ​​a service page (1000) output to the user terminal (10). For example, as illustrated in FIG. 3, the control unit (150) can provide an answer (ex: “Reusable launch vehicle technology is developing, centered around the United States. SpaceX and Blue Origin are developing large reusable launch vehicles BFR and New Glenn, and the United States is researching small reusable launch vehicles through the XS-1 project. The development of reusable launch vehicles is injecting new vitality into the space industry”, 250).

[0100] In this case, the answer (250) may include at least one of a graphic object (250a) linked to a function that allows copying text included in the answer (250), a graphic object (250b) linked to a function that allows selection of whether to like (or be satisfied with) the answer (250), and a graphic object (250c) linked to a function that allows selection of whether to dislike the answer (250).

[0101] Furthermore, the control unit (150) may provide, together with the answer (250), information about at least one document (251) specific for generating the answer (250) (e.g., a PDF link of the document, a location of a document chunk, a source of the document, etc.). The control unit (150) may provide, in the second area (320) of the service page (1000), information about a plurality of documents (321, 322, 323) specific for generating the answer (250) to the user query (200). Here, the documents (321, 322, 323) provided in the service page (1000) may correspond to documents included in at least one of the searched documents (221) and / or the sorted documents (231).

[0102] In this way, the present invention can provide the latest space, satellite, and aviation-related information (e.g., news) according to the user's request through the first mode, and can provide the user with an answer that improves the hallucination phenomenon of the artificial intelligence model by utilizing the searched document.

[0103] Next, let's look at the second mode among the multiple modes.

[0104] The second mode (or map mode) may also be named “Map Mode” and may be a mode configured to retrieve at least one satellite image related to the user query from the storage (140) and / or at least one storage and / or satellite image DB (30) when a user query related to a satellite and / or satellite image is received, and to generate an answer to the user query using the analysis result according to the analysis of the retrieved satellite image and a large-scale language model (160) and provide the answer to the user.

[0105] Unlike the first mode, which provides text-based responses, the second mode can provide answers to user queries in a geographical format or visually display responses to user queries on a map. For example, the second mode can provide satellite imagery of a specific location (e.g., a location requested by the user) or visually represent (or display) information about a specific location.

[0106] That is, in the second mode, the answer to the user's query and at least one of the satellite image, location information, and map used (or specified) to generate the answer may be provided to the user.

[0107] In this regard, referring to FIG. 4, the second mode in the present invention can operate based on the Re-Act (Reasoning Action) + Agent method.

[0108] In one embodiment, the Re-Act + Agent method may include i) a “thought process” that analyzes (or understands) a query (or request) entered by a user and determines which model (or tool) to use to generate (or resolve) an answer to the user query, ii) an “action process” that performs a task necessary to generate an answer to the user query using a specific model determined through the thought process, and iii) an “observation process” that observes (or confirms) the result of the task performed using the specific model. In this case, the thought process may also be understood as a “reasoning (or inference or thinking) process.”

[0109] The Re-Act + Agent method proceeds in the order of "Think → Act → Observe," and, if necessary, the cyclical process can be repeated in the order of "Think → Act → Observe → Think Again." For example, if a specific model determined during the thinking process is not appropriate for generating a response to a user query or additional information is needed, the control unit (150) can return to the thinking process and repeat the cyclical process of determining a new specific model.

[0110] Meanwhile, in the second mode, since the response to the user query must be displayed on a map (or map), the information (or query) requested by the user can be expressed as geometry information (or location information or coordinate information).

[0111] In one embodiment, when a user query requesting satellite images of a specific region is received, the control unit (150) may generate geometry information for the specific region or extract (or specify) geometry information for the specific region using a linked server and / or at least one model.

[0112] Geometry information is transmitted (or delivered) in a specific format for coordinates on a map, and if a satellite image is provided first, geometry information related to the satellite image can be included and input (or delivered) to the answer service providing system (100) together with a user input (or user query).

[0113] Again, referring to FIG. 4, a user query (e.g., “User Input”, 400), satellite imagery (e.g., “Image name”), and geometry information (e.g., “Geometry”) can be used to generate a prompt to be input into a large-scale language model (160). More specifically, the control unit (150) can generate a prompt (e.g., “Re-Act Prompt Template”, 410) to be input into the large-scale language model (160) using at least one of the user query, satellite imagery, and geometry information.

[0114] The prompt (410) may include various information (or elements) required to generate an answer corresponding to the user query (400). For example, the prompt (410) may include at least one of i) a list of available models (411, or tool list) and a description of the role each model plays, ii) the date the prompt (410) is executed, iii) satellite imagery, the name of the satellite imagery, geometry information (or geographic location information), iv) the user's previous conversations (e.g., queries previously entered by the user, information mentioned, requests, etc.), and v) the user's currently entered query (400). However, the information included in the prompt (410) is not necessarily limited thereto, and may include various other information in addition to the examples mentioned.

[0115] The control unit (150) can process the generated prompt (410) as input to a large-scale language model (160). The large-scale language model (160) can determine a method for solving a problem for a user query (400) based on the input prompt (410) and output (or generate) a format (or output data, 420) containing elements (or information) necessary for executing the method.

[0116] The format (420) output from the large-scale language model (160) may include at least one of: i) a first element (e.g., “Thought”) representing a thought process (reasoning) related to thoughts and / or judgments considered to solve the user query (400) based on the user query (400) and the list of available models (411); ii) a second element (e.g., “Action Name”) representing a name for a model specified (or selected) from the large-scale language model (160) to solve the user query (400); and iii) a third element (e.g., “Action Input”) representing input values ​​(or input data) required for the specified model to perform a task.

[0117] In one embodiment, let's assume that a user query (400) includes the content "Detect vehicles from the worldview-2 image taken at Seoul Station in 2023." In this case, the first element may include the content "To resolve the user's request, we need to retrieve the worldview-2 image taken at Seoul Station in 2023 and perform a task to detect vehicles in the worldview-2 image using a model capable of vehicle detection."

[0118] In another embodiment, assume that the user query (400) includes the content “Detect vehicles from the worldview-2 image taken at Seoul Station in 2023.” In this case, the second element may include the name of a specific model (e.g., “SearchSatelliteImageTool”, or a satellite image search model) for performing a search process for a satellite image (e.g., “worldview-2 image”) corresponding to the user query (400), and the name of a specific model (e.g., “VehicleDetectionTool”, or an object detection model) for detecting vehicles in the searched (or specified) satellite image.

[0119] In another embodiment, let's assume that the user query (400) includes the content "Detect vehicles from the worldview-2 image taken of Seoul Station in 2023." In this case, the third element may include input data (e.g., object detection target information (e.g., vehicle), satellite image taken of a specific area (e.g., worldview-2 image), etc.) required for a specific model (e.g., the first model (170)) to perform a specific task (e.g., vehicle detection).

[0120] In this regard, if the name of a specific model included in the second element does not exist in the list of available models (411), the control unit (150) may regenerate the prompt (410) and input it into the large-scale language model (160), thereby causing the large-scale language model (160) to regenerate the format (420). On the other hand, if the name of a specific model included in the second element exists in the list of available models (411), the control unit (150) may input input data included in the third element into the specific model, thereby executing a specific function for generating an answer to the user query (400).

[0121] In this way, the present invention utilizes a large-scale language model (160) to determine which model to use to resolve user input and determine which input values ​​the specified model should use to perform the task. In other words, the present invention can select the optimal model to resolve a user query, thereby performing only the necessary data processing, thereby providing an appropriate response tailored to the user's request.

[0122] Meanwhile, as discussed above, which specific model among the models included in the model list (411) will be used to execute a specific function in order to generate an answer to a user query (400) can be determined based on a specific model name (or second element) from a large-scale language model (160).

[0123] In this case, the input data (e.g., Action Input) required for each model included in the model list (411) may be different. In other words, the input data processed for each model may be different.

[0124] In one embodiment, a satellite image search model (e.g., “SearchSatelliteImageTool”) may search for satellite images in a satellite image database (30) where satellite images are stored and / or at least one storage (e.g., a PostgreSQL database). In this case, the satellite image search model may require processing of input data including at least one of the name of the satellite image, the date of capture, and geographical information.

[0125] In another embodiment, an object detection model (e.g., “VehicleDetectionTool”) may detect at least one object to be detected in satellite images. In this case, the object detection model may require processing input data including information about the object to be detected and at least one satellite image retrieved from the satellite image retrieval model.

[0126] In another embodiment, a satellite image rendering model (e.g., “RenderSatelliteImageTool”) can render satellite images using link information (e.g., Tile JSON URL) provided from a pre-specified server (e.g., Tile Server) or satellite image DB (30). In this case, the size of the image displayed at one time is determined according to the zoom level (or degree) of the map (or map image), and since a wider range is visible when zoomed out, this can be divided into multiple tiles and stored. The server can be configured separately, and the information retrieved from the server can vary depending on the image and map image used to execute a specific function of a specific model.

[0127] Meanwhile, the term for the model list (411) discussed above may also be named as at least one of “function (411)” or “function list (411).” In this case, each of the functions included in the function list (411) may have a preset algorithm that performs a different function (or a specific function). In other words, the expression “function” can be understood to mean processing input data (or input values) according to a preset algorithm and outputting output data (or change values).

[0128] In one embodiment, when a specific function (ex: “SearchCoordinateNaverTool”) is determined from a large-scale language model (160), the control unit (150) can process input data (e.g., geometry information) according to an algorithm preset for the specific function and output output data (e.g., location information).

[0129] Meanwhile, the observation process allows for observing (or monitoring) the results of a specific model executing a specific function. The observation process may vary for each model, and since the observation process is related to user input, it can be used as evidence to determine whether a response appropriate to the user input has been generated.

[0130] For example, the control unit (150) can observe that after the object detection model is executed, a result value such as “A total of OO vehicles were detected in the image at the corresponding location” is output from the object detection model.

[0131] In this case, the results observed through the observation process can be converted into content through post-processing. This content can be used (or reflected) to generate answers to user queries (400).

[0132] The control unit (150) can generate a final result (e.g., “finish answer”) if, through the observation process, it is confirmed that the large-scale language model (160) appropriately generates a name for a specific model and thereby uses a suitable model in response to a user query. The final result is transmitted to the front-end (“front-end”), and can generate an answer to the user query in the form of a description of the results analyzed using the specific model.

[0133] On the other hand, if a suitable result for the user input is not generated, the control unit (150) can proceed with the Re-Act + Agent process again. For example, the control unit (150) can use a large-scale language model (160) to identify a suitable model to resolve the user query, and repeat the process until a suitable result for the user input is generated based on the specific model.

[0134] The control unit (150) can generate a final answer suitable for the user's query based on the output data (e.g., analysis results) output by a specific model, and then transmit the generated final answer to the backend (e.g., “backend”, 430).

[0135] In this regard, referring to FIG. 4e, the relationship between the artificial intelligence server (20), backend (430), front end (440), and a specific server (450, hereinafter, tile server) can be confirmed.

[0136] An artificial intelligence server (20) providing a large-scale language model (160) can process a user query (400), select an appropriate model in response to the user query, and generate an appropriate response to the user query (400). The control unit (150) can transmit various status information generated during this process to the backend (430) in real time via a websocket method. For example, websocket can be understood as a technology that enables real-time two-way communication between a client and a server.

[0137] Next, the control unit (150) can transmit the status information transmitted to the backend (430) to the frontend (440). The frontend (440) can also be understood as a user interface (or user interface module) and can visually display information or results requested by a user. For example, the frontend (440) can display satellite images on a map image or display searched information.

[0138] Here, the front end (440) can communicate directly with the tile server (450). The tile server (450) can provide information related to tile images to be displayed on the map. For example, when a user zooms in or out or moves the map, the tile server (450) can provide the front end (440) with a tile image corresponding to the corresponding location.

[0139] Furthermore, at least one of the user query (400) and / or the response to the user query and / or the user's conversation content and commands may be stored in the storage (140) and / or at least one repository (e.g., MongoDB (460)) in association with the user account. This may be utilized to understand the conversation context in the future or to efficiently store and retrieve the conversation content.

[0140] Below, a method for providing a response service according to the present invention will be described in more detail, assuming that the second mode has been selected by the user.

[0141] First, in the present invention, a process of receiving a user query can be performed through a service page (S510, see FIG. 5).

[0142] As illustrated in FIG. 6a, the control unit (150) can provide a service page (1000) linked to the answer service provision system (100) to the user terminal (10).

[0143] A first area (610) of a service page (1000) with the second mode activated may provide text-based information related to a user query (e.g., a user query, a response to a user query, etc.). Additionally, a second area (620) different from the first area (610) may provide image-based information related to the user query (e.g., a map, satellite image, etc.).

[0144] In this regard, the service page (1000) may include at least one of a first area (610) for receiving a user query and a second area (620) for displaying a map image (621).

[0145] Specifically, the first area (610) may include at least one of a first sub-area (611) for receiving a user query and a second sub-area (612) for providing (or displaying) a response to the user query. In this case, the first sub-area (611) is an area where user input is made and may also be referred to as an input area.

[0146] The control unit (150) can receive a user query input from a user terminal (10) through a service page (1000). In this case, a first sub-area (611) of the service page (1000) can include a graphic object (611a) linked to a user query receiving function. For example, the control unit (150) can receive a user query corresponding to a user input (ex: “Detect vehicles at Pangyo Station from the worldview-2 image taken at Pangyo Station in 2021”, 612a) based on the selection of the graphic object (611a) from the user terminal (10).

[0147] Meanwhile, the control unit (150) can process the received user query (612a) as input to a large-scale language model (160).

[0148] Specifically, the control unit (150) can generate a prompt (e.g., Re-Act Prompt, or first prompt) to be input into the large-scale language model (160) using a user query (612a) (see FIGS. 4a to 4d).

[0149] Hereinafter, a more detailed explanation will be provided with reference to FIGS. 4a to 4d.

[0150] As discussed above, a prompt to be input into a large-scale language model (160) may include various elements necessary to generate an answer corresponding to a user query (612a). For example, the prompt may include at least one of: i) a list of available models and a description of the role each model plays; ii) the date the prompt is executed; iii) a satellite image, a name of the satellite image, geometry information (or geographic location information); iv) the user's previous conversation (e.g., a query previously entered by the user, information mentioned, a request, etc.); and v) a query currently entered by the user (612a).

[0151] The control unit (150) processes the generated prompt as input to a large-scale language model (160), and can select a model that performs analysis corresponding to a user query from the large-scale language model (160).

[0152] Specifically, the large-scale language model (160) can determine how to solve a problem for a user query (612a) based on an input prompt and output output data including various elements necessary to execute the decision.

[0153] The output data output from the large-scale language model (160) may include at least one of: i) a first element (e.g., “Thought”) representing a thought process (reasoning) related to thoughts and / or judgments considered to solve the user query (612a) based on the user query (612a) and the list of available models; ii) a second element (e.g., “Action Name”) representing a name for a specific model from the large-scale language model (160) to solve the user query (612a); and iii) a third element (e.g., “Action Input”) representing input data required for the specific model to perform a specific function.

[0154] In one embodiment, assume that a user query (612a) includes the content, “Detect vehicles in Pangyo Station from the worldview-2 image taken at Pangyo Station in 2021.” In this case, the first element may include the content, “To resolve the user’s request, we need to retrieve the worldview-2 image taken at Pangyo Station in 2021 and perform a task of detecting vehicles in the worldview-2 image using a model capable of vehicle detection.”

[0155] Next, the second element output together with the first element may include the name of a specific model (ex: “SearchSatelliteImageTool”, or satellite image search model) for performing a search process for a satellite image (ex: “worldview-2 image”) corresponding to a user query (612a), and the name of a specific model (ex: “VehicleDetectionTool”, or object detection model) for detecting a vehicle in the searched (or specific) satellite image. For convenience of explanation, the model specified for searching a satellite image is named “satellite image search model”, and the model specified for detecting a vehicle is named “first model (170)”.

[0156] Furthermore, the third element output together with the first element and the second element may include input data (e.g., name of satellite image, shooting date, shooting satellite, location information, geometry information, etc.) required for a specific satellite image model to perform a specific function (satellite image search) and input data (e.g., target information to be the object detection target (e.g., vehicle), satellite image searched from the satellite image search model (e.g., worldview-2 image)) required for the first model (170) to perform a specific function (object detection).

[0157] Meanwhile, in the present invention, a process of specifying location information from a user query can be performed (S520, see FIG. 5).

[0158] As discussed above, the large-scale language model (160) can understand the content included in the user query (612a) based on the input prompt, and based on the result of the understanding, output the analysis result for the user intent of the user query (612a) (e.g., “To resolve the user’s request, we need to search the worldview-2 image taken at Pangyo Station in 2021 and perform the task of detecting a vehicle in the worldview-2 image using a model capable of detecting a vehicle.”, or the first element).

[0159] The control unit (150) can use the output data of the large-scale language model (160) to specify location information (or coordinate information or coordinate values) from a user query (612a).

[0160] Specifically, the control unit (150) can extract (or specify) information necessary for specifying location information based on the analysis results of the user's intent. Examples of information necessary for specifying location information include at least one of: i) a place name, ii) a location name, and iii) an address. However, the information necessary for specifying location information in the present invention is not necessarily limited to the examples mentioned above, and any information that can be utilized to specify location information may be included.

[0161] For example, the control unit (150) can extract “Pangyo Station” corresponding to the place name among the information required to specify the location information from the analysis result of the user’s intention (ex: “In order to resolve the user’s request, the worldview-2 image taken at Pangyo Station in 2021 must be searched, and a vehicle detection model must be used to detect the vehicle in the worldview-2 image.”, or the first element).

[0162] Furthermore, the control unit (150) can specify the location information from the extracted information based on the information required to specify the location information.

[0163] At this time, if information required to specify location information is extracted, specifying the location information can be made possible by the control unit (150) itself. In addition, the control unit (150) can i) specify (or extract) location information (e.g., latitude and longitude) by using an external server linked with the answer service provision system (100) and / or a model (e.g., “SearchCoordinateNaverTool”, or location information analysis model) provided by the external server.

[0164] In one embodiment, the control unit (150) can specify location information (e.g., latitude: 37.3947, longitude: 127.1104) for a place name (e.g., “Pangyo Station”) extracted from a user query (612a) through a map API provided by an external server linked to the answer service providing system (100) or link information (URL) of a website linked to the map.

[0165] In another embodiment, the control unit (150) can use a location information analysis model to specify location information (e.g., latitude: 37.3947, longitude: 127.1104) for a place name (e.g., “Pangyo Station”) extracted from a user query (612a).

[0166] However, in the process discussed above, the user intention is analyzed through a large-scale language model (160), but in the present invention, the user intention can be analyzed by the control unit (150) itself.

[0167] Meanwhile, the control unit (150) can generate geometry information using specific location information. Here, “generating geometry information” can also be understood as converting (or expressing) specific location information into geometry information.

[0168] Geometry can refer to the method of representing spatial data in a Geographic Information System (GIS). Geometric information is used to define spatial characteristics such as location, shape, size, and direction in maps and geographic data, and can be expressed in various forms such as points, lines, and polygons.

[0169] In this case, the conversion of geometry information can be performed either i) by the control unit (150) itself, or ii) by using a model (or module) configured to convert location information (or coordinate information or coordinate values) into geometry information. For convenience of explanation, the above-described cases will be described below without distinction.

[0170] The control unit (150) can convert location information (e.g., latitude: 37.3947, longitude: 127.1104) for a place name extracted from a user query (612a) into geometry information using a geometry information generation (or conversion) model.

[0171] Furthermore, the control unit (150) can process the location information for the above place name as input to a geometry information generation model. In one embodiment, the geometry information generation model can convert the input location information into geometry information having at least one of a point and / or a polygon shape.

[0172] Meanwhile, in the present invention, a process of searching for satellite images related to a user query from at least one DB using user query and location information can be performed (S530, see FIG. 5).

[0173] As discussed above, the large-scale language model (160) may determine which model to use to search for satellite images related to a user query (612a). For example, the control unit (150) may utilize a satellite image search model to perform a satellite image search process based on a second element output from the large-scale language model (160).

[0174] That is, the satellite image search model may be a model selected from a large-scale language model (160) to perform a search process for satellite images (ex: “worldview-2 images”) corresponding to a user query (612a).

[0175] The control unit (150) can specify input data required for the satellite image search model to perform the satellite image search process and process it as input to the satellite image search model. In this case, the input data to be input to the satellite image search model can be specified based on a third element output from a large-scale language model (160). For example, the control unit (150) can input at least one of the shooting date (e.g., “2021”), the name of the satellite image (e.g., “worldview-2 image”), specified location information, and converted geometry information as input data to the satellite image model.

[0176] The satellite image search model can retrieve satellite images related to a user query (612a) from at least one repository and / or database based on input data. For example, the satellite image search model can retrieve satellite images related to the satellite image capture date (e.g., “2021”), the name of the satellite image (e.g., “worldview-2 image”), specific location information, and transformed geometry information from the satellite image database (30).

[0177] Meanwhile, as an embodiment different from the one discussed above, the control unit (150) can search for satellite images related to a user query (612a) using specific location information.

[0178] Specifically, the control unit (150) can search for satellite images related to a user query (612a) using the converted geometry information and at least one storage and / or DB. In this case, the DB used for satellite image search can store various satellite images, and each satellite image can include spatial information (i.e., geometry information (e.g., location and range)) regarding which region it covers.

[0179] In the present invention, there may be one or more DBs used for satellite image search as needed (or in some cases), but for the convenience of explanation, the DBs used for satellite image search will not be distinguished separately and will be described as “satellite image DB (30).”

[0180] Specifically, the control unit (150) can search for satellite images related to a user query by comparing the converted geometry information with the geometry information of satellite images stored in the satellite image DB (30) based on the converted geometry information.

[0181] In one embodiment, the control unit (150) can use PostGIS (e.g., a spatial query function of PostGIS or a PostGIS function, etc.) capable of storing, managing, and analyzing spatial data to search for at least one satellite image corresponding to the converted geometry information among a plurality of satellite images stored in the satellite image DB (30).

[0182] In another embodiment, the control unit (150) can search for at least one satellite image that includes location information (or coordinate information or coordinate value) of a place name (ex: “Pangyo Station”) included in a user query (612a) among a plurality of satellite images stored in a satellite image DB (30) using PostGIS.

[0183] In this way, in the present invention, searching for satellite images related to a user query can be performed either by a satellite image search model or by the control unit (150) itself.

[0184] Meanwhile, when multiple satellite images related to a user query (612a) are retrieved from the satellite image DB, the control unit (150) can specify one of the multiple satellite images based on various criteria preset in the answer service provision system (100). For example, the various criteria may include at least one of i) the most recent date (or time), ii) a user's request (or selection), iii) a weight including specific location information, and iv) a weight including a user's area of ​​interest (or designation). However, the criteria for specifying a satellite image in the present invention are not necessarily limited to the examples mentioned, and may further include various criteria in addition to the examples mentioned.

[0185] In one embodiment, if the preset criterion is “most recent date,” the control unit (150) can analyze metadata matched to each of the searched satellite images to identify the most recently captured satellite image as the satellite image related to the user query (612a).

[0186] Furthermore, in the present invention, a process of performing analysis related to a user query using satellite images can be performed (S540, see FIG. 5).

[0187] The control unit (150) can perform analysis corresponding to a user query using the searched (or specified) satellite image.

[0188] The control unit (150) can input satellite images searched in relation to the user query (612a) as input to a specific model to perform analysis corresponding to the user query (612a).

[0189] As discussed above, which model to use to perform analysis corresponding to a user query (612a) can be determined by the large-scale language model (160). More specifically, the control unit (150) can utilize the first model (170) to perform an object detection process corresponding to the user query (612a) based on the second element output from the large-scale language model (160).

[0190] That is, the first model (10) may be a model selected from a large-scale language model (160) to perform a process of detecting an object that is a target of detection (e.g., a vehicle) from a satellite image (e.g., “worldview-2 image”) searched in response to a user query (612a).

[0191] The control unit (150) can specify the input data required for the first model (170) to perform an object detection process corresponding to the user query (612a) and process it as input to the first model (170). In this case, the input data to be input to the first model (170) can be specified based on the third element output from the large-scale language model (160). For example, the control unit (150) can input at least one of the searched satellite image (e.g., “worldview-2 image”) and the target information to be detected (e.g., “vehicle”) as input data to the first model (170).

[0192] The first model (170) can detect objects to be detected from retrieved satellite images based on input data. For example, the first model (170) can detect vehicles from the input satellite images and output analysis results (e.g., "A total of 614 vehicles were detected") that include information related to the number of detected vehicles (e.g., 614 vehicles).

[0193] In another embodiment, the first model (170) may perform an object detection task when a satellite image is rendered on a map image (621) of a service page (1000), as the case may be. In this case, the control unit (150) may render a satellite image searched in response to a user query (612a) on the map image (621) and control the first model (170) to detect a target object in the satellite image.

[0194] Furthermore, the control unit (150) may receive, from the selected model, analysis results for satellite images corresponding to the user query (612a). For example, the control unit (150) may receive, from the first model (170), analysis results for satellite images corresponding to the user query (612a) (e.g., “There are a total of 614 vehicles detected.”).

[0195] Meanwhile, in the present invention, a process of generating a prompt using at least one of a user query, location information, analysis results according to analysis, and satellite imagery may be performed (S550, see FIG. 5).

[0196] The control unit (150) may generate a prompt to be input into the large-scale language model (160) using at least one of a user query (612a), specified location information, an analysis result received from the first model (170), and a retrieved satellite image. For example, the prompt may include at least one of a user query (612a), specified location information (e.g., latitude: 37.3947, longitude: 127.1104), output data of the first model (170) (e.g., “There are a total of 614 vehicles detected.”), and a satellite image (or retrieved satellite image) used for analysis by the first model (170).

[0197] And, in the present invention, a process of processing a prompt as input to a large-scale language model can be performed (S560, see FIG. 5).

[0198] The control unit (150) can process the above prompt as input to a large-scale language model (160). The large-scale language model (160) can generate an answer to a user query (612a) (ex: “Red, green, and blue of worldview-2_20211106_wydku8vn were rendered. The worldview-2 satellite image taken on November 6, 2021 was rendered in red, green, and blue. In addition, a total of 614 vehicles were detected in the image at the corresponding location.”, 612b) based on the input prompt, using the analysis result of the first model (170) and the retrieved satellite image (see FIG. 6a).

[0199] Furthermore, in the present invention, a process of obtaining an answer to a user query from a large-scale language model can be performed (S570, see FIG. 5).

[0200] The control unit (150) can receive an answer to a user query (612a) from a large-scale language model (160) (ex: “Red, green, and blue of worldview-2_20211106_wydku8vn were rendered. The worldview-2 satellite image taken on November 6, 2021 was rendered in red, green, and blue. In addition, a total of 614 vehicles were detected in the image at the corresponding location.”, 612b).

[0201] Meanwhile, in the present invention, a process of providing an answer to a user query on a service page may be performed (S580, see FIG. 5).

[0202] The control unit (150) can provide a response (612b) to a user query (612a) and at least one satellite image (622) searched in response to the user query (612a) to the service page (1000).

[0203] Text-based information related to a user query (612a) may be provided in a first area (610) of a service page (1000), and image-based information related to a user query (612a) may be provided in a second area (620).

[0204] Accordingly, the control unit (150) can provide an answer generated from a large-scale language model (160) in the first area (610) of the service page (1000) (ex: “Red, green, and blue of worldview-2_20211106_wydku8vn were rendered. The worldview-2 satellite image taken on November 6, 2021 was rendered in red, green, and blue. In addition, a total of 614 vehicles were detected in the image at the corresponding location.”, 612b).

[0205] In addition, the control unit (150) may provide a satellite image (622) searched in response to a user query in the second area (620) of the service page (1000). More specifically, the control unit (150) may provide (or display) the satellite image (622) by overlapping it with at least a portion of the map image (621) displayed in the second area (620) based on the satellite image searched in response to the user query (612a).

[0206] In this case, the area where the satellite image (622) overlaps the map image (621) can be specified based on location information specified for searching the satellite image.

[0207] In one embodiment, the control unit (150) may specify an area corresponding to the specified location information among a plurality of areas included in the map image (621) based on the fact that the specified location information corresponds to “latitude: 37.3947, longitude: 127.1104.” The control unit (150) may display a satellite image (622) by overlapping (or rendering) it in the specified area.

[0208] In another embodiment, the control unit (150) may overlap a satellite image (622) on a map image (621) based on information about a specific place name (e.g., “Pangyo Station”) extracted from a user query (612a). The control unit (150) may overlap and display the satellite image (622) on a specific area corresponding to a specific place name among a plurality of areas included in the map image (621).

[0209] Furthermore, the satellite image (622) may include and be provided the analysis results of the satellite image (622) of the first model (170). For example, the satellite image (622) may be provided together with information about an object (e.g., a vehicle) detected by the first model (170) through the object detection analysis results.

[0210] Meanwhile, the control unit (150) can receive a user input designating a specific area in the satellite image provided in the second area (620).

[0211] In the present invention, user input for satellite images is not necessarily limited to specifying a specific area. In addition to specifying a specific area in a satellite image, the present invention can provide a user environment that allows input related to satellite images provided in the second area (620) through input for the entire satellite image or text query input for the satellite image. However, for convenience of explanation, the present specification assumes that user input for a specific area has been received.

[0212] The service page (1000) may be provided with at least one tool icon (or graphic object) for receiving user input for satellite images provided in the second area (620). For example, as illustrated in FIG. 6b, the control unit (150) may provide at least one of a first tool icon (622a) and a second tool icon (622b) in the second area (620) of the service page (1000).

[0213] In the present invention, the first tool icon (622a) and the second tool icon (622b) may exist linked to different functions.

[0214] The first tool icon (622a) may be linked to a first selection function that selects at least a portion of a satellite image (631) based on a first criterion. For example, the first selection function of the first tool icon (622a) may enable a user to select a square range centered on a selected area (or point) in the satellite image (631), and may enable rotation of the selected range.

[0215] The second tool icon (622b) may be linked to a second selection function that selects at least a portion of the satellite image (631) based on a second criterion different from the first criterion. For example, the second selection function of the second tool icon (622b) may allow the user to specify a square rectangular range centered (or based) on an area (or point) selected in the satellite image (631), and rotation about the specified range may be impossible.

[0216] The control unit (150) can receive a user input for designating a specific area in a satellite image (622) using any one of a plurality of tool icons (622a, 622b). For example, as illustrated in FIG. 6b, assume that a second tool icon (622b) is selected from a user terminal (10). The control unit (150) can receive a user input for designating a specific area (632) among a plurality of areas included in a satellite image (631) from the user terminal (10) in which the second tool icon (622b) is activated.

[0217] Meanwhile, the control unit (150) can receive additional user queries related to a specific area (632).

[0218] Specifically, the control unit (150) can receive an additional user query (ex: “Calculate the NDVI index of the area”, 632a) related to a specific area (632) in the first area (610) of the service page (1000).

[0219] Here, if an additional user query (632a) is received, a response to the additional user query (632a) can be provided based on the Re-Act + Agent method discussed above. Since more specific details regarding this have been described above, a brief description will be provided below.

[0220] The control unit (150) may generate a prompt (e.g., Re-Act Prompt) to be input into the large-scale language model (160) using the additional user query (632a). Here, the control unit (150) may generate the prompt using at least one of the user query (612a) and the additional user query (632a) corresponding to the previous conversation, so that the additional user query (632a) is analyzed as an independent input value including the corresponding content from the large-scale language model (160) by considering the correlation with the previous conversation.

[0221] In addition, the control unit (150) can process the generated prompt as input to a large-scale language model (160). The large-scale language model (160) can analyze the user intent of the additional user query (632a) based on the input prompt, specify a function corresponding to the user intent, and select a specific analysis model that performs the specified function.

[0222] The large-scale language model (160) determines that the user intention is related to calculating the NDVI index of a specific area (632) based on the fact that the additional user query (632a) includes the content “calculate the NDVI index of the area”, and based on the determination result, the large-scale language model (160) can specify (or determine) a satellite image search model for searching satellite images for the specific area (632) and a second model (180) that performs a specific function (e.g., calculating the NDVI index) corresponding to the additional user query (632a).

[0223] The control unit (150) can search for satellite images for a specific area (632) from the satellite image DB (30) using a satellite image search model determined from a large-scale language model (160), and can specify satellite images that satisfy preset criteria among the searched satellite images.

[0224] Additionally, the control unit (150) can process a specific satellite image as input to a second model (180) corresponding to a specific analysis model. For example, the second model (180) can analyze a satellite image of a specific area (632) and output an analysis result including an NDVI index of the specific area (632).

[0225] Furthermore, the control unit (150) can obtain an output (or output data) corresponding to the user intention of the additional user query (632a) from the second model (180).

[0226] Meanwhile, the control unit (150) can generate an answer to an additional user query (ex: “NDVI of worldview-2_20211106_wydku8vn was rendered. After detecting a total of 614 vehicles in the corresponding area of ​​the WorldView-2 satellite image taken on November 6, 2021, the NDVI index was calculated”, 632a) by using at least one of a satellite image of a specific area (632), meta information related to the specific area (632), and an additional user query (632a) (see FIG. 6b).

[0227] Here, examples of meta information related to a specific area (632) may include at least one of location information, geometry information, climate data, land use status, and environmental data of the specific area (632). In this case, the meta information may be understood as having been collected by the control unit (150) from the storage unit (140) and / or a predetermined storage (or DB).

[0228] Furthermore, the control unit (150) can provide a response to an additional user query (632a) to at least one of the first area and the second area.

[0229] As illustrated in FIG. 6b, the control unit (150) can provide a text response (632b) to an additional user query (632a) in the first area (610) and provide a satellite image (631) for a specific area (632) in the second area (620).

[0230] In this case, the answer to the additional user query (632a) provided in the second area (620) may be displayed with at least a portion highlighted in the satellite image (631) of the specific area (632) provided in the second area (620).

[0231] That is, the control unit (150) can highlight at least a portion of a satellite image (631) of a specific area (632) and display it on the second area (620).

[0232] Furthermore, satellite images for a specific area (632) may be provided with the analysis results of the second model (180). For example, satellite images for a specific area (632) may be provided with information on the NDVI index analyzed through the NDVI analysis results of the second model (180).

[0233] Meanwhile, in the second mode, in addition to the functions discussed above, various functions can be provided to users.

[0234] In the second mode, users may be provided with functions related to at least one of the following: satellite image and coordinate search, satellite orbit prediction (or query), satellite location information query, spectral index calculation (or output), object detection and segmentation results, and satellite image purchase (or sale). However, the functions provided in the second mode are not necessarily limited to the examples mentioned, and various other functions may be included.

[0235] In this case, the answer service providing system (100) according to the present invention may include at least one model for performing the various functions provided in the second mode. However, the process for providing the various functions may also be performed by the control unit (150) itself. For convenience of explanation, the following description will assume that the various functions are performed by the control unit (150) itself, without separately distinguishing between the entities providing them.

[0236] In one embodiment, as illustrated in FIG. 7, the control unit (150) may receive a user query (e.g., “Tell me the current location of sentinel-2a”, 711a) requesting location information of a specific satellite through the service page (1000). The control unit (150) may specify location information of a specific satellite related to the user query (711a) and provide a response (e.g., “The current location of sentinel-2a is latitude -8.062, longitude -75.039”, 711b) including the specified location information to the first area (710). In addition, the control unit (150) may display an object (721a) of a specific satellite by overlapping it on a map image (721) provided in the second area (720). In this case, the location of the object (721a) of the specific satellite displayed on the map image (721) may be determined based on the location information of the specific satellite.

[0237] In another embodiment, the control unit (150) may receive a user query (ex: “Disseminate the orbit of the sentinel-2a satellite until June 24, 2024”, 712a) requesting a satellite orbit prediction of a specific satellite through the service page (1000). The control unit (150) may predict the satellite orbit of the specific satellite related to the user query (712a) and provide a response (ex: “The orbit of the sentinel-2a satellite has been propagated until June 24, 2024”, 712b) including information about the predicted orbit to the first area (710). In addition, the control unit (150) may display the predicted orbit information of the specific satellite by overlapping it on a map image (721) provided to the second area (720).

[0238] Additionally, the second mode can provide a user environment that allows users to purchase satellite images they wish to purchase.

[0239] In one embodiment, as illustrated in FIG. 8, the control unit (150) may receive a user query (e.g., “I want to purchase a satellite image”, 811a) requesting the purchase of a specific satellite image through the service page (1000). Based on the reception of the user query (811a), the control unit (150) may provide an answer related to the user query (811a) (e.g., “Would you like to purchase a WorldView-2 satellite image of Pangyo Station taken on November 6, 2021? If you purchase the satellite image, a high-quality satellite image taken by the first satellite will be provided...”, 811b) in the first area (810), and may provide a graphic object (822) linked to a specific satellite image (821) that the user has requested to purchase and a purchase request reception function (or a payment function for purchase) of the specific satellite image (821) in the second area (820). A specific satellite image (821) may be a satellite image searched in response to a user query, or a satellite image specified based on information entered by the user regarding a satellite image they wish to purchase. The control unit (150) may sell a specific satellite image (821) to the user based on the selection of the graphic object (822).

[0240] In this case, the satellite image rendered on the service page (1000) may be a low-quality satellite image, while the satellite image sold to users may be a high-quality satellite image. In other words, the satellite image sold to users may be understood to have improved resolution and / or image quality compared to the satellite image rendered on the service page (1000).

[0241] In another embodiment, as illustrated in FIG. 9, the control unit (150) may receive a user query requesting the purchase of a specific satellite image (ex: “I want to purchase all the satellite images shown so far”, 911a) through the service page (1000). The control unit (150) may provide, based on the user query (911a) received, a response related to the user query (911a) in the first area (910) (ex: “Would you like to purchase both the WorldView-2 satellite image of Pangyo Station taken on November 6, 2021 and the satellite image for 9.7046° N, 38.2528° E taken on October 9, 2022? There is a discount if you purchase two together!”, 911b), and may provide, in the second area (920), a graphic object (822) linked to a purchase request receiving function of a plurality of satellite images (921, 922) that the user requested to purchase. The plurality of satellite images (921, 922) may be specific based on the user's conversation history (e.g., satellite images retrieved based on previous conversations, current conversations, etc.), and the control unit (150) may sell the plurality of satellite images (921, 922) to the user based on the selection of the graphic object (923). In this case, specific benefits may be provided depending on whether the user purchases all of the plurality of satellite images (921, 922) or only one.

[0242] Meanwhile, in the second mode, various services can be provided according to the Euro Plan subscription.

[0243] The control unit (150) can provide answers to user queries based on available models.

[0244] In one embodiment, the control unit (150) may generate an answer to a user query using a model-specific result, or, if the specified model is included in the list of models available to the user account, using the output of the specified model.

[0245] On the other hand, in another embodiment, as illustrated in FIG. 10, the control unit (150) may provide, in the first area (1010), a model-specific result for generating an answer to a user query (e.g., “Analyze the NDWI of the area”, 1011a), a response requesting payment related to a paid plan subscription if the specified model is not included in the list of models available to the user account (e.g., “An NDWI analysis model is used to analyze the NDWI of the area, and this requires an additional paid service subscription.”, 1011b), and a graphic object (1012a) associated with the paid plan subscription function. In this case, the paid plan subscription may be associated with a subscription to use a specific model that performs a specific function (e.g., NDWI analysis). At this time, if a paid plan is paid for in the user account based on the selection of the graphic object (1012a), the control unit (150) may provide the user with NDWI analysis results for a specific area (1022) analyzed using a specific model (NDWI analysis model). However, the subscription to the paid plan is not necessarily limited to a subscription for using a specific model, and it is obvious that it may be varied in various ways depending on the purpose of the present invention.

[0246] Meanwhile, in the present invention, various services (or personalized services) related to the present invention can be provided based on various information (or history information) stored in connection with a user account.

[0247] The control unit (150) can perform clustering on various pieces of information stored in connection with a user account based on various criteria. Here, clustering may mean grouping data with similar characteristics into a single cluster (or cluster or group), and separating data with different characteristics into different clusters.

[0248] In the present invention, the criteria for clustering can be set in various ways. For example, when performing clustering on satellite images, the various criteria may include at least one of: i) location information matched to the satellite image, ii) the capture date matched to the satellite image, and iii) attributes of the satellite image (e.g., satellite images based on vegetation index analysis results, satellite images including orbital information, etc.).

[0249] In one embodiment, as illustrated in FIG. 11, the control unit (150) may perform clustering on satellite images matched to user accounts. In this case, the first group (1110) and the third group (1130) may include satellite images having the same location information (or coordinate information). Alternatively, the second group (1120) may include satellite images based on satellite image attributes (e.g., satellite images based on vegetation index analysis results).

[0250] At this time, a graphic object linked to a function for purchasing satellite images included in each group may be provided in one area of ​​the service page (1000). For example, the control unit (150) may sell all satellite images included in the first group (1110) to the user based on the selection of the first graphic object (1110a) from among the plurality of graphic objects (1110a, 1110b) included in one area of ​​the service page (1000) from the user terminal (10). As another example, when the selection of the second graphic object (1110b) occurs, the control unit (150) may sell at least one satellite image included in the first group (1110) to the user according to the user's selection.

[0251] Meanwhile, in addition to the first tool icon (622a) and the second tool icon (622b) as shown in FIG. 6b, the present invention may further include a third tool icon.

[0252] In one embodiment, as illustrated in FIG. 12, the control unit (150) may provide a third tool icon (1210) linked to a specific function on the service page (1000). The third tool icon (1210) may be associated with a specific function that allows the user to display specific coordinates on a map that the user wishes to include in the query, in parallel with the method of extracting a specific region from a user query and searching for coordinates. This specific function may be used when searching for satellite images containing specific coordinates.

[0253] When a third tool icon (1210) is selected from the user terminal (10), the control unit (150) can display a graphic object (1212) linked to the third tool icon (1210) on at least a portion of the map image (1211). When the third tool icon (1210) is selected, the graphic object (1212) can be displayed on the map image (1211) in a first visual appearance (e.g., in the form of a blue circle). In addition, when the graphic object (1212) is displayed on the map, the corresponding geometry information can be stored in the answer service providing system (100, or front end).

[0254] As another example, when a specific area (or location) is selected on a map image (1211) from a user terminal (10) while the graphic object (1212) is not activated (or selected), the control unit (150) may display the graphic object (1212) on the map image (1211) with a second visual appearance (e.g., a blue-bordered circle on a white background).

[0255] Meanwhile, the present invention can provide various functions related to answers of a large-scale language model (160).

[0256] In one embodiment, as illustrated in FIG. 13, upon receiving a user query requesting a list of satellite images including a specific location (e.g., “Show me a list of satellite images including the location”, 1301), the control unit (150) may provide a response (1302) including a list of satellite images related to the location corresponding to the user query (1301) on the service page (1000). Here, the specific location may be an area designated by the user through at least one tool icon. In addition, the control unit (150) may render the most recently captured satellite image (1312) on the map image (1311) while providing the response (1302). Through this, the user may set a specific area of ​​interest, utilize the object detection function, or render other satellite images on the map and continue a conversation with the chatbot (or large-scale language model).

[0257] In another embodiment, the control unit (150) may provide, in an area of ​​the service page (1000), a graphic object (1320) linked to a function for selecting a chatbot model that provides an answer to a query. When the graphic object (1320) is selected from the user terminal (10), the control unit (150) may provide a list (1321) including a plurality of chatbot models (1321a, 1321b, 1321c), and the user may select at least one of the different chatbot models (1321a, 1321b, 1321c) included in the list (1321). In this case, the control unit (150) may utilize a specific model selected by the user to perform an answer to a query input by the user.

[0258] As discussed above, the interactive AI answer service providing method and system according to the present invention can utilize searched documents to generate answers that improve hallucinations and provide them to users. This can contribute to reducing the provision of incorrect information and providing users with highly reliable information.

[0259] Furthermore, the method and system for providing an interactive AI answering service according to the present invention can analyze satellite data related to a user's query and provide a response to the user's inquiry along with visual information. This allows the user to more intuitively confirm the response to the query.

[0260] Furthermore, according to the method and system for providing an interactive artificial intelligence answer service according to the present invention, by comprehensively analyzing various data including satellite images in response to a user's inquiry and providing various information required by the user, the user can utilize the provided information in various fields such as agriculture, environmental monitoring, and urban planning, thereby increasing efficiency across various industries.

[0261] Furthermore, the method and system for providing an interactive AI answering service according to the present invention can provide answers tailored to a user's inquiry by utilizing specific resources at the user's request. In other words, the present invention can provide accurate and reliable information that meets the user's needs without requiring a complex information search process.

[0262] Meanwhile, the present invention discussed above can be implemented as a program that is executed by one or more processes on a computer and can be stored on a medium (or recording medium) that can be read by the computer.

[0263] Furthermore, the present invention discussed above can be implemented as computer-readable code or instructions on a program-recorded medium. In other words, the present invention can be provided in the form of a program.

[0264] Meanwhile, computer-readable media include all types of recording devices that store data that can be read by a computer system. Examples of computer-readable media include hard disk drives (HDDs), solid-state disk drives (SSDs), silicon disk drives (SDDs), ROMs, RAMs, CD-ROMs, magnetic tapes, floppy disks, and optical data storage devices.

[0265] Furthermore, the computer-readable medium may include a storage device and may be a server or cloud storage device accessible via communication. In this case, the computer may download the program according to the present invention from the server or cloud storage device via wired or wireless communication.

[0266] Furthermore, in the present invention, the computer described above is an electronic device equipped with a processor, i.e., a CPU (Central Processing Unit), and there is no particular limitation on its type.

[0267] Meanwhile, the above detailed description should not be construed as limiting in any respect and should be considered illustrative. The scope of the present invention should be determined by a reasonable interpretation of the appended claims, and all modifications within the equivalent scope of the present invention are intended to be included within the scope of the present invention.

Claims

1. Step of receiving user queries through the service page; A step of specifying location information from the above user query; A step of searching for satellite images related to the user query from at least one DB using the user query and the location information; A step of performing analysis related to the user query using the satellite image; A step of generating a prompt using at least one of the user query, the location information, the analysis result according to the analysis, and the satellite image; A step of processing the above prompt as input to a large-scale language model (LLM); A step of obtaining an answer to the user query from the large-scale language model; and A method for providing an interactive artificial intelligence answer service, characterized in that it includes a step of providing an answer to the user query on the service page.

2. In paragraph 1, The above service page is, A first area receiving the above user query and A method for providing an interactive artificial intelligence answer service, characterized in that it includes a second area that provides the satellite image searched in response to the user query.

3. In paragraph 2, In the first area, text-based information related to the user query is provided, A method for providing an interactive artificial intelligence answer service, characterized in that in the second area, image-based information related to the user query is provided.

4. In paragraph 3, In the second area above, a map image is displayed, When the satellite image is retrieved in response to the user query above, A method for providing an interactive artificial intelligence answer service, characterized in that the satellite image overlaps at least a portion of the map image.

5. In paragraph 4, A method for providing an interactive artificial intelligence answer service, characterized in that the area where the satellite image overlaps in the above map image is specified based on the location information specified for searching the satellite image.

6. In paragraph 2, A step of receiving a user input designating a specific area in the satellite image provided in the second area; A step of receiving an additional user query related to the specific area in the first area; A method for providing an interactive artificial intelligence answer service, characterized in that it further comprises a step of generating an answer to the additional user query by using at least one of a satellite image of the specific area, meta information related to the specific area, and the additional user query.

7. In paragraph 6, The steps for generating answers to the above additional user queries are: A step of analyzing the user intent of the above additional user query; A step of specifying a function corresponding to the user's intention and selecting a specific analysis model that performs the specified function; A step of processing satellite images of the specific area as input to the specific analysis model; A step of obtaining an output corresponding to the user intention from the specific analysis model; A step of generating an answer to the additional user query using the above output; and A method for providing an interactive artificial intelligence answer service, characterized in that it comprises a step of providing an answer to the additional user query to at least one of the first area and the second area.

8. In paragraph 7, The answers to the additional user queries provided in the second area above are: A method for providing an interactive artificial intelligence answer service, characterized in that it includes highlighting at least a portion of the satellite image of the specific area provided in the second area.

9. In paragraph 6, The above service page provides at least one tool icon for receiving the user input for the satellite image provided in the second area, The above tool icons are linked to different functions, The first tool icon among the above tool icons is linked to a first selection function for selecting at least a portion of the satellite image according to a first criterion, A method for providing an interactive artificial intelligence answer service, characterized in that the second tool icon among the above tool icons is linked to a second selection function that selects at least a portion of the satellite image according to a second criterion different from the first criterion.

10. In paragraph 1, When the user query is received, a step of processing the user query as input to the large-scale language model, and selecting a model that performs analysis corresponding to the user query from the large-scale language model; A step of inputting the satellite image as input of the selected model; A method for providing an interactive artificial intelligence answer service, characterized in that it further comprises a step of receiving an analysis result for the satellite image corresponding to the user query from the selected model.

11. In paragraph 10, In the step of generating the above prompt, Include the analysis results received from the above selected model in the above prompt, In the above large-scale language model, A method for providing an interactive artificial intelligence answer service, characterized in that it generates an answer to the user query using the analysis results and the satellite image.

12. A step of receiving a user query through a service page including a first area and a second area; A step of specifying a document to be used to generate an answer to the above user query; A step of processing the user query and the specified document as input to a large-scale language model; A step of obtaining an answer to the user query from the large-scale language model; In the above service page, further comprising a step of providing the answer, A method for providing an interactive artificial intelligence answer service, characterized in that the answer is provided in the first area of ​​the service page, and information about the specific document is provided in the second area of ​​the service page.

13. In the interactive artificial intelligence response service provision system, The system comprises a memory and at least one processor, The above memory and the above processor cooperate to: Through the service page, we receive user queries, From the above user query, specify the location information, Using the user query and the location information, search for satellite images related to the user query from at least one DB, Using the above satellite image, analysis related to the user query is performed, Generate a prompt using at least one of the user query, the location information, the analysis result according to the analysis, and the satellite image, Treat the above prompt as input to a large-scale language model (LLM), From the above large-scale language model, an answer to the user query is obtained, An interactive artificial intelligence answer service providing system characterized by providing an answer to the user query on the above service page.

14. A program that is executed by one or more processes in an electronic device and stored in a computer-readable recording medium, Step of receiving user queries through the service page; A step of specifying location information from the above user query; A step of searching for satellite images related to the user query from at least one DB using the user query and the location information; A step of performing analysis related to the user query using the satellite image; A step of generating a prompt using at least one of the user query, the location information, the analysis result according to the analysis, and the satellite image; A step of processing the above prompt as input to a large-scale language model (LLM); A step of obtaining an answer to the user query from the large-scale language model; and A program stored on a computer-readable recording medium, characterized in that it includes commands for performing a step of providing an answer to the user query on the service page.

Images