KR20250088147A - Method and device for generating learning data for generative artificial intelligence using spatial information and real-time sensor information - Google Patents

Abstract

The present invention relates to a method for generating learning data of generative artificial intelligence using spatial information and real-time sensor information, and a device therefor. The method for generating learning data of generative artificial intelligence, which is performed by a computing device including at least one processor, comprises the steps of: acquiring location-based object data for an object including a preset object; converting metadata and object attribute data specifying the object based on the location-based object data into a text listing format to generate documented data; and generating learning data based on the documented data.

Description

Method and device for generating learning data for generative artificial intelligence using spatial information and real-time sensor information {METHOD AND DEVICE FOR GENERATING LEARNING DATA FOR GENERATIVE ARTIFICIAL INTELLIGENCE USING SPATIAL INFORMATION AND REAL-TIME SENSOR INFORMATION}

The present invention relates to a method for generating learning data for generative artificial intelligence using spatial information and real-time sensor information, and a device therefor. The present invention relates to a method for converting data that does not exist in text form into documented data in text form and providing the data as learning data for generative AI.

The material described in this section merely provides background information for one embodiment of the present invention and does not constitute prior art.

Large-scale artificial intelligence (AI), such as LLM (LLM; large language model), is an artificial intelligence model trained using a very large amount of text data, performs natural language processing tasks, and can be used for various language modeling tasks.

These super-large AIs are the largest and most complex models in the field of artificial intelligence, and they show very high performance in various fields of artificial intelligence, such as natural language processing, image recognition, and speech recognition.

A representative example of a large-scale AI is the GPT (generative pretrained transformer) series developed by Open AI, which has shown outstanding performance in the field of natural language processing and has a high level of language understanding and generation ability through learning using large amounts of text data.

In addition, super-large AI is already being used in various fields. For example, in the field of natural language processing, it can be applied to various applications such as automatic translation, automatic summarization, and question answering. In the field of image analysis, it can be applied to image classification, object recognition, etc., and in the field of speech recognition, it can be used for speech synthesis, speech recognition, etc.

The development of such large-scale AI is largely due to the development of deep learning algorithms and hardware technology. In other words, the development of deep learning algorithms has enabled large-scale AI models to learn more complex patterns, and the development of hardware technology has greatly increased the learning speed of large-scale AI models.

LLM can be trained with a learning dataset consisting of hundreds of billions of sentences containing various web documents and text data such as the Internet, books, newspaper articles, and blogs, and can be used in various applications such as natural language understanding, sentence generation, machine translation, chatbots, and automatic summarization. This deep learning model based on LLM is an artificial intelligence technology that learns the language used by humans and performs various tasks with that language.

Deep learning models based on LLM are representative of tasks that understand sentences written by people through natural language understanding and answer questions based on this, and are receiving great attention in fields such as chatbots, which are artificial intelligence that converse with humans.

Representative deep learning models based on LLM, such as OpenAI's ChatGPT and Google's Bard, are trained based on Internet documents, so their performance is bound to be poor when responding to real-time data that has not been trained in advance. In particular, deep learning models based on LLM have the problem that the accuracy of responses to location-related user queries is significantly low in the case of small and medium-sized businesses and workplaces that open and close approximately 1 million times a year.

Likewise, deep learning models based on LLM have the problem that they cannot answer user queries related to data that is not documented in text form or data that is generated in real time for monitoring, or that the accuracy of the results is significantly low.

In particular, generative AI used in conversational AI services has the problem that learning itself cannot be performed in the case of map data that does not exist in text form on the Internet and real-time data occurring at the present time due to the limitations of language models learned based on data prior to the user entering a question, and as a result, most questions related to location information or environmental information closely related to our lives are either unanswerable or present answers with many errors.

Republic of Korea Patent No. 10-2570178 (2023.08.21)

The present invention has been made in response to the aforementioned background technology, and aims to provide a method for converting location-based object data, such as maps, sensors, weather, and performances, which are not textualized on the Internet, such as the Internet and social networks, into documented data in real time and providing the data as learning data for a generative artificial intelligence model.

However, the problems to be solved in the present invention are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood based on the description below.

As a technical means for achieving the above technical task, an object of the present invention is to provide a method for generating learning data of a generative artificial intelligence, which is performed by a computing device including at least one processor according to an embodiment of the present invention. The method includes a step of obtaining location-based object data for an object including a preset object; a step of converting metadata and object attribute data specifying the object based on the location-based object data into a text listing format to generate documented data; and a step of generating learning data based on the documented data.

Alternatively, the step of converting metadata and object property data specifying the object based on the location-based object data into a text listing format to generate documented data includes, when the object includes at least one unit object, generating a string by converting metadata and unit object property data specifying the unit object into a text listing format for each unit object, and merging the strings for each unit object to generate documented data.

Alternatively, the location-based object data is at least one of spatial data based on any one of points, polygons or polylines, topographic data based on digital elevation model (DEM) data, sensor data generated through real-time environmental monitoring, public data using an open API (application programming interface), or spatial analysis data based on three-dimensional spatial information.

Alternatively, the metadata includes at least one of coordinate information of a unit object, address information structured by geocoding the coordinate information, absolute altitude information of the coordinate information, and location characteristic information including an average slope, slope direction, or distance to surrounding roads or major facilities for a change area set based on the coordinate information, when the location-based object data includes spatial data based on the point.

Alternatively, the step of converting metadata and object attribute data that specify the object based on the location-based object data into a text listing format and generating documented data is to generate a document string for each unit object by separating the metadata converted into a text listing format and the unit object attribute data with a delimiter.

Alternatively, the metadata includes at least one of address information structured by geocoding based on the central coordinate information of the polygon, and location characteristic information including the area, slope, slope direction or elevation above sea level of the polygon, when the location-based object data includes spatial data based on the polygon.

Alternatively, the metadata includes at least one of coordinate information including the latitude and longitude coordinates of the center point of the polyline and the coordinates of the start and end points, address information structured by geocoding the coordinate information, and location characteristic information including the line length of the polyline, the line slope or the average elevation of the line, or surrounding environment information located in a preset change area based on the coordinate information, when the object data includes spatial data based on the polyline.

Alternatively, the metadata, when the object data includes the topographic data, includes at least one of information from among: center coordinate information of a coordinate group defined based on the unit column, structured address information geocoded based on the center coordinate information, and location characteristic information including elevation, slope, slope aspect or land cover for a change area set based on the center coordinate information.

Alternatively, the metadata may include location information and measurement values of sensors collected at preset time units when the location-based object data includes sensor data.

Alternatively, the sensor is an IoT sensor including one or more of a fine dust sensor, a bridge vibration monitoring sensor, or a traffic volume measurement monitoring sensor.

Alternatively, the step of converting metadata and object attribute data specifying the object based on the location-based object data into a text listing format and generating documented data includes, if the location-based object data includes public data, the step of generating basic learning data based on at least one public API; the step of learning a first language model based on the basic learning data; the step of determining a public API matching a user query using a pre-learned first language model, and then calling the determined public API to receive a result value; and the step of converting the result value into documented data and providing it as a prompt for the first language model.

Alternatively, the metadata may include at least one of spatial extent information including layers and lakes for the object, or visibility information about the ratio of one or more environmental elements based on the visibility analysis results of the object, if the object data includes the spatial analysis data.

Meanwhile, as a technical means for achieving the above-mentioned technical task, an object is to provide a computing device for generating learning data of a generative artificial intelligence according to an embodiment of the present invention. The device includes a processor including at least one core; and a memory including program codes executable by the processor; wherein, when the processor executes the program code, it obtains location-based object data for an object including a preset object, and converts metadata and object attribute data specifying the object based on the location-based object data into a text listing format to generate documented data, and generates learning data based on the documented data.

According to the problem solving means of the present invention described above, the present invention can convert location-based object data such as maps, sensors, weather, and performances that are not textualized on the Internet, such as the Internet and social networks, into documented data in real time and provide the data as learning data for a generative artificial intelligence model, and since the generative artificial intelligence model is re-learned based on the documented data, the generative artificial intelligence model can improve the accuracy of answers to user queries.

FIG. 1 is a block diagram of a computing device according to one embodiment of the present invention.
FIG. 2 is a flowchart illustrating a method for generating learning data of a generative artificial intelligence according to one embodiment of the present disclosure.
FIG. 3 is a diagram illustrating a learning data generation process for spatial data based on points according to one embodiment of the present disclosure.
FIG. 4 is a diagram illustrating a learning data generation process for spatial data based on polygons according to one embodiment of the present disclosure.
FIG. 5 is a diagram illustrating a learning data generation process for spatial data based on a polyline according to one embodiment of the present disclosure.
FIG. 6 is a diagram illustrating a learning data generation process for terrain data according to one embodiment of the present disclosure.
FIG. 7 is a diagram illustrating a learning data generation process for sensor data according to one embodiment of the present disclosure.
FIG. 8 is a diagram illustrating a learning data generation process for public data according to one embodiment of the present disclosure.

Hereinafter, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. The embodiments presented in the present invention are provided so that those skilled in the art can utilize or implement the contents of the present invention. Accordingly, various modifications to the embodiments of the present invention will be apparent to those skilled in the art. That is, the present invention can be implemented in various different forms and is not limited to the embodiments below.

Throughout the specification of the present invention, identical or similar drawing reference numerals refer to identical or similar components. In addition, in order to clearly describe the present invention, drawing reference numerals of parts that are not related to the description of the present invention may be omitted in the drawings.

The term "or" as used herein is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified herein or the context makes clear, "X employs either A or B" should be understood to mean either one of the natural inclusive permutations. For example, unless otherwise specified herein or the context makes clear, "X employs A or B" can be interpreted to mean either X employs A, X employs B, or X employs both A and B.

The term "at least one of A or B" as used in the present invention should be interpreted to refer to all of A, B, and combinations of A and B.

The term "and/or" as used herein should be understood to refer to and include all possible combinations of one or more of the related concepts listed.

The terms "comprises" and/or "comprising" as used in the present invention should be understood to mean that certain features and/or components are present. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, other components, and/or combinations thereof.

Unless otherwise specified in the present invention or unless the context makes it clear that the singular form is intended to be referred to, the singular should generally be construed to include “one or more.”

The term "Nth (N is a natural number)" used in the present invention can be understood as an expression used to distinguish the components of the present invention from each other according to a predetermined standard such as a functional viewpoint, a structural viewpoint, or convenience of explanation. For example, components performing different functional roles in the present invention can be distinguished as a first component or a second component. However, components that are substantially the same within the technical idea of the present invention but should be distinguished for convenience of explanation can also be distinguished as a first component or a second component.

Meanwhile, the term "module" or "unit" used in the present invention may be understood as a term referring to an independent functional unit that processes computing resources, such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, a combination of software and hardware, etc. At this time, the "module" or "unit" may be a unit composed of a single element, or a unit expressed as a combination or set of multiple elements. For example, as a narrow concept, a "module" or "unit" may refer to a hardware element of a computing device or a set thereof, an application program that performs a specific function of software, a processing process implemented through software execution, or a set of instructions for program execution, etc. In addition, as a broad concept, a "module" or "unit" may refer to a computing device itself that constitutes a system, or an application that is executed on a computing device, etc. However, since the above-described concept is only an example, the concept of “module” or “part” may be variously defined within a category understandable to those skilled in the art based on the contents of the present invention.

The explanation of the terms mentioned above is intended to help understanding of the present invention. Therefore, if the terms mentioned above are not explicitly described as matters limiting the content of the present invention, it should be noted that they are not used to limit the technical idea of the content of the present invention.

FIG. 1 is a block diagram of a computing device according to one embodiment of the present invention.

The computing device (100) according to one embodiment of the present invention may be a hardware device or a part of a hardware device that performs comprehensive processing and calculation of data, or may be a software-based computing environment connected to a communication network. For example, the computing device (100) may be a server that performs an intensive data processing function and shares resources, or may be a client that shares resources through interaction with a server. In addition, the computing device (100) may be a cloud system that allows a plurality of servers and clients to interact with each other to comprehensively process data. Since the above description is only one example related to the type of the computing device (100), the type of the computing device (100) may be configured in various ways within a range that can be understood by those skilled in the art based on the contents of the present invention.

Referring to FIG. 1, a computing device (100) according to one embodiment of the present invention may include a processor (110), a memory (120), and a network unit (130). However, FIG. 1 is only an example, and the computing device (100) may include other configurations for implementing a computing environment. In addition, only some of the configurations disclosed above may be included in the computing device (100).

The processor (110) according to one embodiment of the present invention may be understood as a configuration unit including hardware and/or software for performing computing operations. For example, the processor (110) may read a computer program and perform data processing through a computing function and a control function. The processor (110) for performing such data processing may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application specific integrated circuit (ASIC), or a field programmable gate array (FPGA). Since the type of the processor (110) described above is only an example, the type of the processor (110) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present invention.

The processor (100) can generate response data for a user query on location-based object data using a pre-learned first language model. In addition, the processor (110) can re-learn the first language model and the second language model based on the response data using the first language model.

At this time, the first language model may be a generative AI model, and the second language model may be a large language model (LLM). Accordingly, the second language model may be used in the first language model that generates content based on an input prompt of human language. Accordingly, the processor (110) may generate an agent that responds to a user query by using the first language model based on the second language model.

Here, location-based object data can be any of spatial data based on points, polygons, or polylines, topographic data based on digital elevation model (DEM) data, sensor data for real-time environmental monitoring, public data using an open API (application programming interface), or spatial analysis data based on three-dimensional spatial information.

Location-based object data can include coordinates, text information such as altitude and slope among spatial information, and non-text information that is not converted to text.

The processor (110) can perform an operation representing at least one neural network block included in the first language model and the second language model during the learning process of the neural network model.

The processor (110) can extract text information to be used in the learning data of the first language model using the second language model generated through the above-described learning process, and can generate learning data to be used in the first language model based on the extracted text information. The processor (110) can input a user query into the first language model learned through the above-described process, and generate response data representing an estimated result reflecting knowledge information at the time of the query.

In addition to the examples described above, the types of learning datasets according to language distribution and the outputs of the first language model and the second language model can be configured in various ways within a category understandable to those skilled in the art based on the contents of the present disclosure.

The memory (120) according to one embodiment of the present invention may be understood as a configuration unit including hardware and/or software for storing and managing data processed in the computing device (100). That is, the memory (120) may store any type of data generated or determined by the processor (110) and any type of data received by the network unit (130). For example, the memory (120) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory, a RAM (random access memory), a SRAM (static random access memory), a ROM (read-only memory), an EEPROM (electrically erasable programmable read-only memory), a PROM (programmable read-only memory), a magnetic memory, a magnetic disk, or an optical disk. In addition, the memory (120) may also include a database system that controls and manages data in a predetermined system. The type of memory (120) described above is only an example, and thus the type of memory (120) can be configured in various ways within a range understandable to those skilled in the art based on the contents of the present invention.

The network unit (130) according to one embodiment of the present invention may be understood as a configuration unit that transmits and receives data via any type of known wired and wireless communication system. For example, the network unit (130) may perform data transmission and reception using a wired and wireless communication system such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), fifth generation mobile communication (5G), ultra-wide-band, ZigBee, radio frequency (RF) communication, wireless LAN, wireless fidelity, near field communication (NFC), or Bluetooth. Since the above-described communication systems are only examples, the wired and wireless communication system for data transmission and reception of the network unit (130) may be applied in various ways other than the above-described examples.

FIG. 2 is a flowchart illustrating a method for generating learning data of a generative artificial intelligence according to one embodiment of the present disclosure.

Referring to FIG. 2, a computing device (100) obtains location-based object data including text information or non-text information about an object including a preset object (S100).

The computing device (100) converts metadata and object attribute data that specify an object based on the acquired location-based object data into a text listing format to generate documented data (S200). At this time, the computing device (100) can convert unstructured text into a structured format using a pre-learned second language model.

The computing device (100) generates a string by converting metadata and unit object attribute data that specify each unit object into a text listing format when an object includes at least one unit object, and merges the strings for each unit object to generate documented data.

At this time, the computing device (100) generates a document string for each object by separating the metadata and object attribute data converted into a text listing format with a delimiter such as a comma (,) or special symbol (*, #).

A computing device (100) generates learning data based on documented data and provides the generated learning data to a first language model, thereby enabling the first language model to be trained based on the learning data (S300).

FIG. 3 is a diagram illustrating a learning data generation process for spatial data based on points according to one embodiment of the present disclosure.

There are vectors and rasters as data types that express spatial data. Vector time expresses the real world in the geometric form of points, polylines, and polygons, while raster expresses the real world in the form of grid-shaped pixels (or cells). Points are composed of a combination of latitude and longitude according to most coordinate systems, and are useful when they are too small to be expressed as polygons. For example, points can be used to indicate the location of Seoul on a world map. Polylines are composed of two or more vertices connected by edges, and are used when expressing in the form of natural lines. For example, polylines can be used when indicating subway lines on a map of Seoul. Polygons are composed of three or more vertices in a closed connection form, and are used when expressing the boundaries of a specific area. For example, polygons can be used when indicating the boundaries of Seoul on a map of South Korea.

In the case of spatial data based on points, metadata may include one or more of the following information: coordinate (x, y) information of the unit object, structured address information by geocoding the coordinate information, absolute elevation information of the coordinate information, and location characteristic information including average slope, slope direction, and distance to surrounding roads or major facilities for a preset change area based on the coordinate information.

Here, the unit object can be a point based on the unit sequence of the learning data, and the main facilities can include convenience facilities such as schools, subway stations, and bus stops, educational facilities, medical facilities, restaurants, and financial institutions.

Accordingly, the computing device (100) extracts a unit column reference point (S211), inputs metadata in text form, and generates a string for each unit object by separating the object attribute data previously held by the point information with a delimiter (S212).

The computing device (100) inputs metadata and object attribute data for all unit objects in point-based spatial data to generate a string for each unit object, and then merges the generated strings for each unit object to generate documented data (S213).

The first language model can be trained using training data based on documented data for points. Therefore, when a user query such as "Please tell me a list of Chinese restaurants within 10 minutes from Gangnam Station" is input from a user terminal or application, the first language model can provide response data to the user query through artificial intelligence-based metadata search.

FIG. 4 is a diagram illustrating a learning data generation process for spatial data based on polygons according to one embodiment of the present disclosure.

As illustrated in FIG. 4, in the case of spatial data based on a polygon, the computing device (100) inputs metadata including at least one piece of information from among structured address information, area of the polygon, slope, slope direction, and location characteristic information including the elevation above sea level, by geocoding based on the central coordinate information of the polygon.

The computing device (100) separates metadata for the unit object and unit object attribute information using a delimiter to create a single string, repeats the string creation process for all unit objects in the polygon, and then creates documented data for the polygon.

The first language model can be trained using training data based on documented data about polygons. Therefore, the first language model can provide response data to user queries such as “What is the tallest building in Seoul?”, “What is the largest apartment complex in Busan?”, and “Tell me the land price information of the most expensive land in Korea.”

FIG. 5 is a diagram illustrating a learning data generation process for spatial data based on a polyline according to one embodiment of the present disclosure.

As illustrated in FIG. 5, the computing device (100) inputs metadata including at least one piece of information from among coordinate information including the latitude and longitude coordinates of the center point of the polyline and the coordinates of the start and end points, structured address information obtained by geocoding the coordinate information, the line length of the polyline, the line slope, the average altitude of the line, or location characteristic information including surrounding environment information (distance to the sea, river, etc.) located in a preset change area based on the coordinate information, in the case of spatial data based on a polyline.

The computing device (100) separates metadata for the corresponding unit object and unit object attribute information using a delimiter to create a single string, repeats the string creation process for all unit objects within the polyline, and then creates documented data for the corresponding polyline.

The first language model can be trained using training data based on documented data about polylines. Thus, the first language model can provide response data for user queries such as “What is the longest road name in Seoul?”, “Tell me a route where I can drive while looking at the ocean.”

FIG. 6 is a diagram illustrating a learning data generation process for terrain data according to one embodiment of the present disclosure.

Referring to FIG. 6, if the location-based object data is terrain data, the computing device (100) sets the preset reference resolution as a unit column (S221). For example, if the computing device (100) sets the reference resolution to 10 cm and configures the unit column based on 1 m, 10 data per unit area (width × height) can be converted into text as one unit column.

The computing device (100) inputs one or more pieces of information as metadata among the center coordinate information of a coordinate group defined based on a set unit column, structured address information geocoded based on the center coordinate information, and location characteristic information including altitude, slope, slope direction, and land cover (mountains, rivers, roads, etc.) for a pre-set change area based on the center coordinate information (S222).

At this time, if the topographic data is contour data, the computing device (100) converts it into DEM data using a known DEM conversion tool and then inputs metadata based on the converted DEM data.

The computing device (100) separates metadata and unit object attribute information for the corresponding unit object using a delimiter to create a single string, repeats the string creation process for all unit objects in the terrain data, and then creates documented data for the corresponding terrain data.

The first language model can be trained using training data based on documented data on terrain data. Thus, the first language model can provide response data for user queries such as "Find coniferous terrain with a slope of 30 degrees or more with a high possibility of landslides."

FIG. 7 is a diagram illustrating a learning data generation process for sensor data according to one embodiment of the present disclosure.

Referring to FIG. 7, in the case of sensor data, the computing device (100) communicates with IoT equipment and standard sensors to collect location information and measurement values of the sensors and inputs them as metadata (S231), and converts the metadata into text to generate documented data at preset time intervals (S232, S233).

At this time, the sensor may be an IoT sensor including at least one of a fine dust sensor for detecting the status of an object, a bridge vibration monitoring sensor, or a traffic volume measurement monitoring sensor.

The computing device (100) embeds documented data in a document file (document A, document B, document C, etc.) (S234) and stores the document file in a database (S235).

For example, the first language model can generate response data in real time using document files stored in a database for user queries such as "Where is the area with the worst fine dust at present?" and "Where is the area with the most traffic congestion in Seoul at present?"

FIG. 8 is a diagram illustrating a learning data generation process for public data according to one embodiment of the present disclosure.

Referring to FIG. 8, if the location-based object data is public data such as weather information (typhoon, rain, weather, etc.) and performance information, the computing device (100) generates basic learning data based on a public API list including at least one public API (S241), and learns a first language model based on the basic learning data (S242).

When a user query is input (S243), the computing device (100) determines a public API matching the user query using a pre-learned first language model (S244), and calls the determined public API to receive a result value (S245, S246).

The computing device (100) converts the received result value into documented data and provides it as a prompt for the first language model (S246). Accordingly, the first language model returns response data to the user based on the prompt input (S247).

For example, the first language model can answer user queries such as “When do you think this typhoon will affect us?” or “What performance is on at the Seoul Arts Center tomorrow?” by searching for information matching the user query in real time.

Meanwhile, if the location-based object data is spatial analysis data based on three-dimensional spatial information, the computing device (100) inputs spatial range information including the floor and lake for the object, and visibility information about the ratio of one or more environmental elements (sky, river, sea, forest, park, road, etc.) within the visible area of the object based on the result of the visibility analysis of the object as metadata.

In this way, the present invention can convert location-based object data such as map information, sensor measurements, weather information, and performance information into documented data in text form and provide it as learning data for a generative artificial intelligence model.

Therefore, the generative AI model can be retrained based on documented data on location-based object data, and can provide accurate answers to user queries that existing generative AI cannot answer. For example, the generative AI model can answer questions about location-related user queries on a map based on accurate address information and coordinate information, such as building area, land, and mutual positional relationships between objects. It can also answer questions about real-time data such as weather, disasters, and sensor monitoring based on real-time collected data.

The various embodiments of the present invention described above can be combined with additional embodiments and can be modified within the scope that can be understood by those skilled in the art in light of the detailed description described above. It should be understood that the embodiments of the present invention are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and likewise, components described as distributed may be implemented in a combined form. Accordingly, all changes or modifications derived from the meaning, scope, and equivalent concept of the claims of the present invention should be interpreted as being included in the scope of the present invention.

100 : Computing Device
110 : Processor
120 : Memory
130 : Network Department

Claims (13)

A method for generating learning data for generative artificial intelligence, performed by a computing device including at least one processor,
A step of obtaining location-based object data for an object including a preset object;
A step of converting metadata and object attribute data that specify the object based on the above location-based object data into a text listing format to create documented data; and
A step of generating learning data based on the above documented data;
Including,
method.
In the first paragraph,
The step of converting metadata and object attribute data that specify the object based on the above location-based object data into a text listing format and creating documented data is as follows.
If the above object includes at least one unit object, a string is created by converting metadata and unit object attribute data that specify the unit object for each unit object into a text listing format, and the strings for each unit object are merged to create documented data.
method.
In the first paragraph,
The above location-based object data is,
At least one of spatial data based on any one of points, polygons or polylines, topographic data based on digital elevation model (DEM) data, sensor data generated through real-time environmental monitoring, public data using an open application programming interface (API), or spatial analysis data based on three-dimensional spatial information.
method.
In the third paragraph,
The above metadata is,
In the case where the above location-based object data includes spatial data based on the above point, it includes at least one piece of information from among coordinate information of the unit object, address information structured by geocoding the coordinate information, absolute altitude information of the coordinate information, and location characteristic information including average slope, slope direction, or distance to surrounding roads or major facilities for a change area set based on the coordinate information.
method.
In paragraph 4,
The step of converting metadata and object attribute data that specify the object based on the above location-based object data into a text listing format and creating documented data is as follows.
The metadata converted into a text listing format and the unit object attribute data are separated by a delimiter to create a document string for each unit object.
method.
In the third paragraph,
The above metadata is,
In the case where the above location-based object data includes spatial data based on the polygon, it includes at least one piece of information among structured address information by geocoding based on the central coordinate information of the polygon, and location characteristic information including the area, slope, slope direction or elevation above sea level of the polygon.
method.
In the third paragraph,
The above metadata is,
In the case where the object data includes spatial data based on the polyline, it includes at least one of coordinate information including the latitude and longitude coordinates of the center point of the polyline and the coordinates of the start and end points, address information structured by geocoding the coordinate information, and location characteristic information including the line length of the polyline, the line slope or the average altitude of the line, or the surrounding environment information located in a preset change area based on the coordinate information.
method.
In the third paragraph,
The above metadata is,
In the case where the object data includes the terrain data, the preset standard resolution is set as a unit column, and at least one of the following information is included: center coordinate information of a coordinate group defined based on the unit column, address information structured by geocoding based on the center coordinate information, and location characteristic information including elevation, slope, slope direction or land cover for a preset change area based on the center coordinate information.
method.
In the third paragraph,
The above metadata is,
If the above location-based object data includes the sensor data, it includes the location information and measurement values of the sensor collected in preset time units.
method.
In Article 9,
The above sensor,
An IoT sensor including at least one of a fine dust sensor, a bridge vibration monitoring sensor, or a traffic volume measurement monitoring sensor,
method.
In the third paragraph,
The step of converting metadata and object attribute data that specify the object based on the above location-based object data into a text listing format and creating documented data is as follows.
If the above location-based object data includes public data, a step of generating basic learning data based on at least one public API;
A step of learning a first language model based on the above basic learning data;
A step of determining a public API matching a user query using a pre-trained first language model, and then calling the determined public API to receive a result value; and
Including the step of converting the above result value into documented data and providing it as a prompt of the first language model.
method.
In the third paragraph,
The above metadata is,
If the object data includes the spatial analysis data, at least one of the spatial range information including the layer and lake for the object or the visibility information for the ratio of one or more environmental elements based on the visibility analysis result of the object.
method.
As a computing device that generates learning data for generative artificial intelligence,
a processor comprising at least one core; and
A memory including program codes executable by the processor;
The above processor, upon execution of the above program code,
Obtain location-based object data for objects containing pre-defined objects,
Based on the above location-based object data, metadata and object attribute data that specify the object are converted into text listing format to create documented data.
Generating training data based on the above documented data,
device.

Images