KR102427368B1 - Method for recommendating candidate of knowledge graph using semi-structured data and apparatus implementing the same method - Google Patents

Abstract

According to one embodiment of the present invention, a method performed by a computing device includes: a step of extracting column item data identified as a plurality of columns from semi-structured data; and a step of outputting a candidate group of a knowledge graph organized with a plurality of nodes and edges presenting relations between the plurality of nodes by inputting the extracted column item data to a knowledge graph candidate recommendation model. The knowledge graph candidate recommendation model is generated by performing machine learning using learning data including knowledge graph information.

Description

A method for recommending knowledge graph candidates using semi-structured data, and an apparatus for implementing the same

The present invention relates to a knowledge graph candidate recommendation method using semi-structured data, and an apparatus for implementing the same, and more particularly, to a knowledge graph capable of expressing semi-structured data using an artificial intelligence algorithm. It relates to a knowledge graph candidate recommendation method using semi-structured data for generating, and an apparatus for implementing the same.

In the case of public datasets in which data is open and shared, a transformation rule is created by manually defining the relationships or properties included in the data, and the relationships or properties included in the data are converted into LOD (Linked Open Data) and used accordingly. come.

In addition, in the prior art, in order to analyze the relationship included in the data, a person manually defines the relationship or property inherent in the raw data set, writes a transformation rule, builds a knowledge graph, and uses this to search and infer information is being used to

However, since this method relies on manual work, the accuracy and reliability of deriving the relationship or attribute information inherent in the data is low, and it is difficult to use the relationship or attribute information included in the data as learning data or to utilize the previously accumulated information. There is a problem.

Also, when a new data set is input, if the name of a column included in the new data set does not exist on an existing knowledge graph, it is difficult to match a node corresponding to the corresponding column. Moreover, if the nodes are not properly matched, it is difficult to define a node that is highly related to the corresponding column.

Accordingly, in constructing a knowledge graph using a raw data set, a technology capable of generating a knowledge graph by reflecting the relationship inherent in data more accurately is required.

Patent Publication No. 10-2019-0114916 (published on October 10, 2019)

The technical problem to be solved by the present invention is to create a relationship between columns through an AI algorithm using semi-structured data, and based on this, a knowledge graph using semi-structured data that can provide a candidate group of a knowledge graph with high accuracy It is to provide a candidate recommendation method, and an apparatus for implementing the same.

Another technical problem to be solved by the present invention is a knowledge graph candidate recommendation method using semi-structured data that can generate a more sophisticated knowledge graph candidate recommendation model by using relationship information of previously accumulated data as learning data, And to provide a device for implementing this.

Another technical problem to be solved by the present invention is knowledge using semi-structured data that can match the columns included in the new dataset to the nodes and connecting lines of the existing knowledge graph when a new dataset is input. A graph candidate recommendation method and an apparatus for implementing the same are provided.

Another technical problem to be solved by the present invention is a knowledge graph candidate recommendation method using semi-structured data that makes it easy to modify and search the knowledge graph through a visualized user interface when generating a knowledge graph using semi-structured data, And to provide a device for implementing this.

The technical problems of the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the description below.

In order to solve the above technical problem, a method performed by a computing device according to an embodiment of the present invention includes extracting column item data identified as a plurality of columns from semi-structured data, and the and inputting the extracted column item data into a knowledge graph candidate recommendation model to output a candidate group of a knowledge graph consisting of a plurality of nodes and an edge representing a relationship between the plurality of nodes, the knowledge graph The candidate recommendation model is generated by performing machine learning using learning data including knowledge graph information.

As an embodiment, the step of extracting column item data identified as a plurality of columns from the semi-structured data may include changing the name of each column identified in the semi-structured data using synonyms or related word dictionary information. Standardization may be included.

As an embodiment, performing clustering based on the similarity using the column item data includes performing clustering using a dataset name, a column name, and raw data of the column item data. can do.

As an embodiment, the step of inputting the extracted column item data into a knowledge graph candidate recommendation model and outputting a candidate group of a knowledge graph consisting of a plurality of nodes and an edge indicating a relationship between the plurality of nodes includes: , receiving expert recommendation relationship information, generating information about a relationship within the same dataset or a relationship between datasets using the expert recommendation relationship information and the clustering performance result, and the generated same data The method may include generating a candidate group of the knowledge graph by using information about a relationship within a set or a relationship between data sets.

As an embodiment, the step of inputting the extracted column item data into a knowledge graph candidate recommendation model and outputting a candidate group of a knowledge graph consisting of a plurality of nodes and an edge indicating a relationship between the plurality of nodes includes: , performing tagging on the column item data based on a result of performing the clustering, and generating node recommendation information for the column item data in a supervised learning method using the tagged data. can do.

As an embodiment, the method may further include adding information about a candidate group of the output knowledge graph to the learning data.

As an embodiment, the semi-structured data may be raw data of a type that can be divided into a plurality of columns.

As an embodiment, providing a GUI for visually expressing the candidate group of the knowledge graph to be output, and correcting information about the plurality of nodes and connection lines constituting the candidate group of the knowledge graph according to a user input to the GUI It may further include the step of

As an embodiment, providing a GUI for visually expressing the candidate group of the knowledge graph to be output, and information about the plurality of nodes and connection lines constituting the candidate group of the knowledge graph according to a user input to the GUI. The method may further include providing data search information.

In order to solve the above technical problem, a computer-readable non-transitory recording medium according to an embodiment of the present invention stores a computer program for causing a computer to perform the method.

In order to solve the technical problem, a knowledge graph candidate recommendation apparatus according to an embodiment of the present invention generates learning data using a communication unit communicating with an external server, and knowledge graph information collected from the external server, and A learning unit that generates a knowledge graph candidate recommendation model by learning an AI-based algorithm using data, and extracts column item data identified as a plurality of columns from semi-structured data, and the extracted column item and a knowledge graph candidate recommendation unit configured to input data into the generated knowledge graph candidate recommendation model and generate a candidate group of a knowledge graph including a plurality of nodes and an edge representing a relationship between the plurality of nodes.

As an embodiment, the knowledge graph candidate recommendation unit performs clustering based on similarity using the column item data, and automatically matches the column item data to a node and a connection line having a maximum probability value based on the clustering result. can do.

As an embodiment, the knowledge graph candidate recommendation unit receives the expert's recommendation relationship information, and uses the expert's recommendation relationship information and the clustering performance result to provide information on a relationship within the same dataset or a relationship between datasets. The candidate group of the knowledge graph may be generated by using the generated information on the relationship within the same dataset or the relationship between the datasets.

As an embodiment, the knowledge graph candidate recommendation unit performs tagging on the column item data based on a result of performing the clustering, and performs a supervised learning method on the column item data using the tagged data. Node recommendation information can be generated.

In order to solve the above technical problem, the knowledge graph candidate recommendation apparatus according to an embodiment of the present invention includes one or more processors, a communication interface for communicating with an external device, and a memory for loading a computer program executed by the processor. , and storage for storing the computer program, wherein the computer program includes an operation of receiving information about a knowledge graph candidate recommendation model from a server, an operation of acquiring semi-structured data by a user input, extracting column item data identified as a plurality of columns from the semi-structured data, and inputting the extracted column item data into the knowledge graph candidate recommendation model to obtain a plurality of nodes and relationships between the plurality of nodes and instructions for performing an operation of outputting a candidate group of a knowledge graph composed of an edge indicated, wherein the knowledge graph candidate recommendation model is generated by performing machine learning using learning data including knowledge graph information .

1 is a conceptual diagram according to an embodiment of the present invention.
2 is a block diagram illustrating a configuration of a knowledge graph candidate recommendation apparatus according to an embodiment of the present invention.
3 is a block diagram illustrating a configuration of a knowledge graph candidate recommendation apparatus according to another embodiment of the present invention.
4 to 6 are flowcharts for explaining a knowledge graph candidate recommendation method according to another embodiment of the present invention.
7 is an example of generating a knowledge graph by deriving relationship information between columns of semi-structured data according to some embodiments of the present invention.
8 is an example of matching columns included in a data set according to some embodiments of the present invention to nodes and connecting lines of a knowledge graph.
9 is a hardware configuration diagram of an exemplary computing device that may implement methods according to some embodiments of the present invention.

Hereinafter, preferred embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. Advantages and features of the present disclosure and methods of achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the technical spirit of the present disclosure is not limited to the following embodiments, but may be implemented in various different forms, and only the following embodiments complete the technical spirit of the present disclosure, and in the technical field to which the present disclosure belongs It is provided to fully inform those of ordinary skill in the art of the scope of the present disclosure, and the technical spirit of the present disclosure is only defined by the scope of the claims.

In adding reference numerals to the components of each drawing, it should be noted that the same components are given the same reference numerals as much as possible even though they are indicated on different drawings. In addition, in describing the present disclosure, if it is determined that a detailed description of a related known configuration or function may obscure the gist of the present disclosure, the detailed description thereof will be omitted.

Unless otherwise defined, all terms (including technical and scientific terms) used herein may be used with the meaning commonly understood by those of ordinary skill in the art to which this disclosure belongs. In addition, terms defined in a commonly used dictionary are not to be interpreted ideally or excessively unless clearly defined in particular. The terminology used herein is for the purpose of describing the embodiments and is not intended to limit the present disclosure. In this specification, the singular also includes the plural, unless specifically stated otherwise in the phrase.

In addition, in describing the components of the present disclosure, terms such as first, second, A, B, (a), (b), etc. may be used. These terms are only for distinguishing the elements from other elements, and the essence, order, or order of the elements are not limited by the terms. When it is described that a component is “connected”, “coupled” or “connected” to another component, the component may be directly connected or connected to the other component, but another component is between each component. It should be understood that elements may be “connected,” “coupled,” or “connected.”

As used herein, "comprises" and/or "comprising" refers to the presence of one or more other components, steps, operations and/or elements mentioned. or addition is not excluded.

Hereinafter, some embodiments of the present disclosure will be described in detail with reference to the accompanying drawings.

1 is a conceptual diagram according to an embodiment of the present invention. Referring to FIG. 1 , a knowledge graph candidate recommendation apparatus 1 according to an embodiment of the present invention performs machine learning based on an artificial intelligence (AI) based algorithm 3 using training data 2 . The knowledge graph candidate group 5 may be output using the knowledge graph candidate recommendation model 31 generated by .

In the illustrated example, the training data 2 is composed of information about the previously built knowledge graph, and the knowledge graph candidate recommendation model 31 is obtained by learning the AI-based algorithm 3 using the training data 2 . can be created. In this case, the AI-based algorithm 3 may be applied to both an unsupervised learning method such as clustering and a supervised learning method such as a neural network. The AI-based algorithm 3 used in the learning process of the learning data 2 is not limited by this embodiment, and various types of algorithms may be used.

When semi-structured data as input information 4 is input to the knowledge graph candidate recommendation model 31 for generation of the candidate group of the knowledge graph, as a result predicted using the knowledge graph candidate recommendation model 31 The candidate group 53 of the knowledge graph may be output. Here, the semi-structured data that is the input information 4 is an intermediate form between structured data such as a database and unstructured data such as free-text, raw data of a type that can be divided into a plurality of columns. As (source data), for example, data having a format such as EXCEL (xls), CSV, rest, open API, HTML, XML, RDF (Resource Description Framework), JSON (JavaScript Object Notation) may be included.

In more detail, semi-structured data, also called semi-structured data or unstructured data, is data in which the format and structure of data can be changed, and is data in a file format that provides structure information of data together with data. Semi-structured data is not structured into rows and columns of a table like structured data, but it is converted into a table-type data schema based on the data structure information contained in the file, and data is mapped to structured data. It is data that can be transformed.

As an embodiment, the knowledge graph candidate recommendation apparatus 1 may extract column item data identified as a plurality of columns from the semi-structured data input as the input information 4 . The knowledge graph candidate recommendation apparatus 1 may input column item data into the knowledge graph candidate recommendation model 31 to output a knowledge graph candidate group including a plurality of nodes and connecting lines representing relationships between nodes.

According to the embodiment of the present invention as described above, a relationship between columns is generated through an AI algorithm using semi-structured data, and based on this, a candidate group of a knowledge graph with high accuracy can be provided. In addition, a more sophisticated knowledge graph candidate recommendation model can be generated by using relationship information of previously accumulated data as learning data.

2 is a block diagram illustrating a configuration of a knowledge graph candidate recommendation apparatus according to an embodiment of the present invention. Referring to FIG. 2 , an apparatus 1 for recommending a knowledge graph candidate according to an embodiment of the present invention includes a learning unit 11 and a knowledge graph candidate recommendation unit 12 , an external server 20 , and a user terminal ( 10) and can be connected through a network. The knowledge graph candidate recommendation apparatus 1 processes a request for generating and analyzing a model for generating a knowledge graph using semi-structured data received from the user terminal 10 and provides the result.

The external server 20 may be implemented as a server that provides information on a previously constructed knowledge graph, and information about a plurality of nodes constituting the knowledge graph and a connection line indicating the relationship between the nodes is stored and managed. can

The learning unit 11 is composed of a collection module 111 , a preprocessing module 112 , and a learning module 113 , of which the collection module 111 is a previously constructed knowledge graph provided from the external server 20 . By collecting information about the learning data (2), it is stored in the database.

The preprocessing module 112 reads the learning data 2 stored in the database by the collection module 111 and converts the learning data 2 into a form capable of machine learning.

The learning module 113 performs machine learning based on the AI-based algorithm 3 using the training data 2 that has been pre-processed in the pre-processing module 112, and generates a knowledge graph candidate recommendation model 31 therefrom. . At this time, as the AI-based algorithm 3 used for learning, for example, an unsupervised learning algorithm such as clustering, or a supervised learning algorithm such as a convolutional neural network (CNN), a deep neural network (DNN), and a recurrent neural network (RNN) can be used

The knowledge graph candidate recommendation unit 12 is composed of an input module 121 , a preprocessing module 122 , and a prediction module 123 , and uses semi-structured data input from the user terminal 10 or provided from an external device. Outputs the knowledge graph candidate recommendation result.

In the input module 121 , semi-structured data provided from the user terminal 10 or an external device is input. In this case, the semi-structured data may be prepared as raw data of a type that can be divided into a plurality of columns.

The pre-processing module 122 extracts column item data identified as a plurality of columns from the semi-structured data input from the input module 121 . In this case, the pre-processing module 122 may standardize the name of each column of the extracted column item data using synonym or related word dictionary information. For example, a unified name of 'organization name' may be applied to a case in which the names of each column are different as 'branch', 'institution in charge', and 'source'.

Also, the preprocessing module 122 may define a relationship or attribute between each column of the extracted column item data. For example, the relationship between the 'flooding damage' dataset and the 'heavy rain' dataset may be defined as a 'has_cause' relationship, and the property of the 'heavy rain' dataset may be defined as 'rainfall'.

The preprocessing module 122 converts the column item data extracted from the semi-structured data into a form capable of machine learning. In this case, the conversion process of the column item data is the same as the process performed by the pre-processing module 112 of the learning unit 11 .

The prediction module 123 loads the knowledge graph candidate recommendation model 31 generated by the learning module 113, and inputs the preprocessed column item data into the loaded knowledge graph candidate recommendation model 31 to recommend the knowledge graph candidate. The candidate group 51 of the knowledge graph by the model 31 is output.

As an embodiment, the prediction module 123 may perform clustering based on similarity using column item data, and may perform tagging on column item data based on a result of clustering. In this case, the prediction module 123 may generate node recommendation information for column item data in a supervised learning method using the tagged data. Here, the node recommendation information may include information about a node corresponding to each column of column item data or information about a node to be newly added on the knowledge graph.

As an embodiment, the prediction module 123 may perform clustering based on the similarity using the column item data, and automatically match the column item data to the node and the connection line having the maximum probability value based on the clustering performance result. For example, a node having a maximum probability value is first matched for each column of column item data among a plurality of nodes constituting an existing knowledge graph, and a maximum probability value among connecting lines connected to the matched node Relation information for each column can be created by matching the connecting lines with

As an embodiment, the prediction module 123 may additionally receive relationship information recommended by an expert, and a relationship within the same dataset or a relationship between datasets using the inputted expert recommendation relationship information and a result of clustering A candidate group of the knowledge graph can be generated by generating information about

As an embodiment, the prediction module 123 may generate a GUI for visually expressing the candidate group of the output knowledge graph and provide it to the user terminal 10 . In this case, the prediction module 123 may correct information about a plurality of nodes and connection lines constituting the knowledge graph candidate group according to a user input to the GUI. In addition, when data search for the knowledge graph candidate group is requested according to a user input to the GUI, the prediction module 123 provides data search information based on information about a plurality of nodes and connecting lines constituting the knowledge graph candidate group. can

The knowledge graph candidate recommendation apparatus 1 according to the embodiment of the present invention as described above may be implemented as an apparatus that performs both a learning process and a prediction process for recommending a knowledge graph candidate. Accordingly, the knowledge graph candidate recommendation apparatus 1 may generate a relationship between columns through an AI algorithm using semi-structured data, and provide a candidate group of a knowledge graph with high accuracy based on this. In addition, when generating a knowledge graph using semi-structured data, it is possible to facilitate modification and search of the knowledge graph through a visualized user interface.

3 is a block diagram illustrating a configuration of a knowledge graph candidate recommendation apparatus according to another embodiment of the present invention. Referring to FIG. 3 , an apparatus 9 for recommending a knowledge graph candidate according to an embodiment of the present invention includes a knowledge graph candidate recommendation unit 12 and is connected to a server 8 connected through a network. The server 8 includes the learning unit 11 and may be connected to the external server 20 through a network.

In the illustrated example, the server 8 includes a configuration of the learning unit 11 that generates the knowledge graph candidate recommendation model 31 through machine learning, and the knowledge graph candidate recommendation device 9 is configured in the server 8 . and a configuration of the knowledge graph candidate recommendation unit 12 for generating a knowledge graph candidate group for the input semi-structured data using the generated knowledge graph candidate recommendation model 31 . At this time, the learning unit 11 included in the server 8 and the knowledge graph candidate recommendation unit 12 included in the knowledge graph candidate recommendation device 9 are connected to the knowledge graph candidate recommendation device 1 shown in FIG. 2 . Since the components respectively correspond to the included learning unit 11 and the knowledge graph candidate recommendation unit 12, detailed descriptions of operations performed by each component will be omitted.

The server 8 performs an operation of generating a model for recommending a knowledge graph candidate. The server 8 collects information about the previously built knowledge graph provided from the external server 20 to configure the learning data 2, and uses this to perform machine learning based on the AI algorithm to perform machine learning based on the knowledge graph. A candidate recommendation model 31 is generated.

The knowledge graph candidate recommendation apparatus 9 processes a request for generating a knowledge graph candidate group for semi-structured data input from a user, and provides the result through a screen.

As an embodiment, the knowledge graph candidate recommendation apparatus 9 transmits a request for generating a model for recommending a knowledge graph candidate to the server 8 , and relates to the knowledge graph candidate recommendation model 31 generated by the server 8 . information can be provided. The knowledge graph candidate recommendation apparatus 9 may generate a candidate group of the knowledge graph using semi-structured data input from the user using the knowledge graph candidate recommendation model 31 provided from the server 8, and display it on the screen. .

As described above, the knowledge graph candidate recommendation apparatus 9 according to the embodiment of the present invention performs only a prediction process for generating a candidate group of the knowledge graph, and the learning process for generating a model for prediction is performed through a separate server 8 . can be performed. Accordingly, since the learning process and the prediction process for generating the candidate group of the knowledge graph are performed in different devices, the result can be provided with high performance without delay in generating the candidate group of the knowledge graph.

4 to 6 are flowcharts for explaining a knowledge graph candidate recommendation method according to another embodiment of the present invention.

The knowledge graph candidate recommendation method according to the present embodiment may be executed by the computing device 100 , for example, by the knowledge graph candidate recommendation device 1 . The computing device 100 executing the method according to the present embodiment may be a computing device having an application program execution environment. Note that the description of the subject performing some operations included in the method according to the present embodiment may be omitted, and in this case, the subject is the computing device 100 .

Referring to FIG. 4 , first, in operation S41 , column item data identified as a plurality of columns is extracted from the semi-structured data. Here, the semi-structured data is raw data of a type that can be divided into a plurality of columns, and may be, for example, data in a format such as EXCEL or CSV.

As an embodiment, operation S41 may include an operation of standardizing the name of each column identified in the semi-structured data using synonym or related word dictionary information.

Next, in operation S42, the extracted column item data is input to the knowledge graph candidate recommendation model, and a candidate group of the knowledge graph composed of a plurality of nodes and a connection line representing a relationship between the plurality of nodes is output.

As an embodiment, operation S42 includes an operation of performing clustering based on similarity using column item data, and an operation of automatically matching the column item data to a node and a connection line having a maximum probability value based on the clustering execution result. may include In this case, clustering may be performed using a dataset name, a column name, and raw data of column item data. Accordingly, when a new data set is input, it is possible to match the columns included in the new data set with the nodes and connecting lines of the existing knowledge graph.

As an embodiment, operation S42 includes an operation of receiving the expert's recommendation relationship information, an operation of generating information on a relationship within the same dataset or a relationship between datasets using the expert's recommendation relationship information and a result of clustering; and generating a candidate group of a knowledge graph by using information about a relationship within the same dataset or a relationship between datasets.

As an embodiment, referring to FIG. 5 , in operation S42, clustering is performed based on similarity using column item data in operation S421, and tagging is performed on the column item data based on a result of performing clustering. It may include operation S422 and operation S423 of generating node recommendation information for the column item data in a supervised learning method using the tagged data.

As an embodiment, the method may further include adding information about a candidate group of a knowledge graph output as a result of performing operation S42 to the learning data. Accordingly, by using the nodes included in the candidate group of the knowledge graph generated using the AI-based algorithm and the relationship information between the nodes as learning data, a more sophisticated knowledge graph candidate recommendation model can be generated later.

As an embodiment, the method includes an operation of providing a GUI for visually expressing a candidate group of a knowledge graph output as a result of performing operation S42, and a plurality of nodes and connecting lines constituting a candidate group of a knowledge graph according to a user input to the GUI It may further include an operation of correcting information about the .

As an embodiment, referring to FIG. 6 , operations S43 and S43 may be performed as additional operations after operation S42 is performed.

In operation S43, a GUI for visually expressing the candidate group of the knowledge graph output as a result of the operation S42 is provided, and in operation S44, information on a plurality of nodes and connection lines constituting the knowledge graph candidate group according to a user input to the GUI Based on the data search information may be provided.

According to the method according to the embodiment of the present invention as described above, a relationship between columns is generated through an AI algorithm using semi-structured data, and based on this, a candidate group of a knowledge graph with high accuracy can be provided.

7 is an example of generating a knowledge graph by deriving relationship information between columns of semi-structured data according to some embodiments of the present invention. Referring to FIG. 7 , the knowledge graph candidate recommendation apparatus 1 identifies a plurality of column items such as a point 83 , an address, and an amount of damage from the semi-structured dataset A 81 , and each identified column item data can be extracted. The knowledge graph candidate recommendation apparatus 1 identifies a plurality of column items, such as area 84, date, rainfall, etc., from the semi-structured dataset B 82 in the same manner as the dataset A 81, and identifies each You can extract the data of column items.

In the illustrated example, the knowledge graph candidate recommendation apparatus 1 performs clustering using the extracted column item data of the dataset A (81) and the column item data of the dataset B (82), and the clustering result dataset A (81) and dataset B (82) may be classified as similar datasets, and relationship information between the two datasets may be generated. In this case, in generating relationship information between datasets, expert recommendation information may be additionally reflected in addition to the clustering result.

Accordingly, when the relationship information between the two datasets is generated, a 'flood damage' node 810 corresponding to the dataset A (81) and a 'heavy rain' node 820 corresponding to the dataset B (82) are generated. and a connection line having a 'Has_Cause' relationship between the flood damage' node 810 and the 'heavy rain' node 820 may be generated.

At this time, the point 83 column of the dataset A 81 and the region 84 column of the dataset B 82 in which the relationship information is generated are determined as similar column items, and the two column items have the same name. tagging may be performed.

8 is an example of matching columns included in a data set according to some embodiments of the present invention to nodes and connecting lines of a knowledge graph. Referring to FIG. 8 , when new data is input while the knowledge graph is generated using an AI-based algorithm, the knowledge graph candidate recommendation apparatus 1 matches the columns included in the new data to the nodes and connecting lines of the knowledge graph. can

In the illustrated example, when a new dataset including a 'medical institution' column 91 is input, the knowledge graph candidate recommendation apparatus 1 displays the 'medical institution' column 91 to a plurality of nodes included in the knowledge graph. ), it is possible to match the 'hospital' node 921 with the maximum probability value of 90%. In addition, the knowledge graph candidate recommendation device 1 selects the 'medical institution' column 91 from the 'hospital-road map' connection line 93 with a maximum probability value of 80% among a plurality of edges included in the knowledge graph. can be matched to Accordingly,

As described above, according to an embodiment of the present invention, a node having a maximum probability value is first matched for each column of column item data among a plurality of nodes constituting an existing knowledge graph, and the matched node is Relation information for each column may be generated by matching the connecting line having the maximum probability value among the connected connecting lines.

9 is a hardware configuration diagram of an exemplary computing device that may implement methods according to some embodiments of the present invention. As shown in FIG. 9 , the computing device 100 loads one or more processors 101 , a bus 107 , a network interface 102 , and a computer program 105 executed by the processor 101 . It may include a memory 103 and a storage 104 for storing the computer program (105). However, only the components related to the embodiment of the present invention are shown in FIG. 9 . Accordingly, those skilled in the art to which the present invention pertains can see that other general-purpose components other than the components shown in FIG. 9 may be further included.

The processor 101 controls the overall operation of each component of the computing device 100 . The processor 101 includes at least one of a central processing unit (CPU), a micro processor unit (MPU), a micro controller unit (MCU), a graphic processing unit (GPU), or any type of processor well known in the art. may be included. In addition, the processor 101 may perform an operation on at least one application or program for executing the method/operation according to various embodiments of the present disclosure. The computing device 100 may include one or more processors.

The memory 103 stores various data, commands and/or information. Memory 103 may load one or more programs 105 from storage 104 to execute methods/operations according to various embodiments of the present invention. For example, if the computer program 105 is loaded into the memory 103 , logic (or modules) may be implemented on the memory 103 . An example of the memory 103 may be a RAM, but is not limited thereto.

The bus 107 provides communication functions between the components of the computing device 100 . The bus 107 may be implemented as various types of buses, such as an address bus, a data bus, and a control bus.

The network interface 102 supports wired/wireless Internet communication of the computing device 100 . The network interface 102 may support various communication methods other than Internet communication. To this end, the network interface 102 may be configured to include a communication module well known in the art.

The storage 104 may store one or more computer programs 105 non-transitory. The storage 104 may include a non-volatile memory such as a flash memory, a hard disk, a removable disk, or any type of computer-readable recording medium well known in the art.

The computer program 105 may include one or more instructions in which methods/operations according to various embodiments of the present invention are implemented. When the computer program 105 is loaded into the memory 103 , the processor 101 may execute the one or more instructions to perform methods/operations according to various embodiments of the present invention.

So far, various embodiments of the present invention and effects according to the embodiments have been described with reference to FIGS. 1 to 9 . Effects according to the technical spirit of the present invention are not limited to the above-mentioned effects, and other effects not mentioned will be clearly understood by those skilled in the art from the following description.

The technical idea of the present invention described so far may be embodied as computer-readable codes on a computer-readable medium. The computer-readable recording medium may be, for example, a removable recording medium (CD, DVD, Blu-ray disk, USB storage device, removable hard disk) or a fixed recording medium (ROM, RAM, computer-equipped hard disk). can The computer program recorded in the computer-readable recording medium may be transmitted to another computing device through a network such as the Internet and installed in the other computing device, thereby being used in the other computing device.

In the above, even if all the components constituting the embodiment of the present invention are described as being combined or operating in combination, the technical spirit of the present invention is not necessarily limited to this embodiment. That is, within the scope of the object of the present invention, all the components may operate by selectively combining one or more.

Although acts are shown in a specific order in the drawings, it should not be understood that the acts must be performed in the specific order or sequential order shown, or that all shown acts must be performed to obtain a desired result. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of the various components in the embodiments described above should not be construed as necessarily requiring such separation, and the program components and systems described may generally be integrated together into a single software product or packaged into multiple software products. It should be understood that there is

Although embodiments of the present invention have been described above with reference to the accompanying drawings, those of ordinary skill in the art to which the present invention pertains can practice the present invention in other specific forms without changing the technical spirit or essential features. can understand that there is Therefore, it should be understood that the embodiments described above are illustrative in all respects and not restrictive. The protection scope of the present invention should be interpreted by the claims below, and all technical ideas within the equivalent range should be interpreted as being included in the scope of the technical ideas defined by the present invention.

Claims (16)

In the knowledge graph candidate recommendation method using semi-structured data performed by a computing device,
extracting column item data identified by a plurality of columns from semi-structured data; and
The extracted column item data is input to the knowledge graph candidate recommendation model, and candidate nodes and candidate edges of the knowledge graph matched to the extracted column item data using the output data of the knowledge graph candidate recommendation model ) comprising the step of displaying information about,
Displaying information on candidate nodes and candidate connection lines of the knowledge graph comprises:
performing clustering based on similarity using the column item data; and
and automatically matching the column item data to a node and a connecting line having a maximum probability value based on the clustering performance result,
The knowledge graph candidate recommendation model is generated by performing machine learning using learning data including knowledge graph information,
A knowledge graph candidate recommendation method using semi-structured data. The method of claim 1,
The step of extracting column item data identified as a plurality of columns from the semi-structured data,
Normalizing the name of each column identified in the semi-structured data using synonym or related word dictionary information;
A knowledge graph candidate recommendation method using semi-structured data. delete The method of claim 1,
The step of performing clustering based on the similarity using the column item data includes:
Comprising performing clustering using a dataset name, a column name, and raw data of the column item data,
A knowledge graph candidate recommendation method using semi-structured data. The method of claim 1,
The step of inputting the extracted column item data into a knowledge graph candidate recommendation model and outputting a candidate group of a knowledge graph consisting of a plurality of nodes and an edge indicating a relationship between the plurality of nodes comprises:
receiving an input of expert recommendation relationship information;
generating information on a relationship within the same dataset or a relationship between datasets using the expert's recommendation relationship information and the clustering performance result; and
and generating a candidate group of the knowledge graph by using information about a relationship within the generated same dataset or a relationship between datasets.
A knowledge graph candidate recommendation method using semi-structured data. The method of claim 1,
The step of inputting the extracted column item data into a knowledge graph candidate recommendation model and outputting a candidate group of a knowledge graph consisting of a plurality of nodes and an edge indicating a relationship between the plurality of nodes comprises:
performing tagging on the column item data based on a result of performing the clustering; and
generating node recommendation information for the column item data in a supervised learning method using the tagged data;
A knowledge graph candidate recommendation method using semi-structured data. The method of claim 1,
Further comprising the step of adding information about the candidate group of the output knowledge graph to the learning data,
A knowledge graph candidate recommendation method using semi-structured data. The method of claim 1,
The semi-structured data is raw data of a type that can be divided into a plurality of columns,
A knowledge graph candidate recommendation method using semi-structured data. The method of claim 1,
providing a GUI for visually expressing the candidate group of the output knowledge graph; and
Further comprising the step of modifying information about the plurality of nodes and connection lines constituting the knowledge graph candidate group according to a user input to the GUI,
A knowledge graph candidate recommendation method using semi-structured data. The method of claim 1,
providing a GUI for visually expressing the candidate group of the output knowledge graph; and
Further comprising the step of providing data search information based on information on the plurality of nodes and connection lines constituting the knowledge graph candidate group according to a user input to the GUI,
A knowledge graph candidate recommendation method using semi-structured data. A computer program for causing a computer to perform the method of any one of claims 1 to 10 is stored therein,
A computer-readable non-transitory recording medium. a communication unit for communicating with an external server;
a learning unit that generates learning data using the knowledge graph information collected from the external server, and generates a knowledge graph candidate recommendation model by learning an AI-based algorithm using the learning data; and
Extracting column item data identified as a plurality of columns from semi-structured data, inputting the extracted column item data into the generated knowledge graph candidate recommendation model, and output data of the knowledge graph candidate recommendation model A knowledge graph candidate recommendation unit for displaying information on a knowledge graph candidate node and a candidate edge matching the extracted column item data using
The knowledge graph candidate recommendation unit,
performing clustering based on similarity using the column item data, and automatically matching the column item data to a node and a connection line having a maximum probability value based on the clustering result;
Knowledge Graph candidate recommendation device. delete 13. The method of claim 12,
The knowledge graph candidate recommendation unit,
Receives expert recommendation relationship information, generates information about a relationship within the same dataset or a relationship between datasets using the expert recommendation relationship information and the clustering performance result, and generates the generated internal relationship within the same dataset or generating a candidate group of the knowledge graph by using information about a relationship between data sets,
Knowledge Graph candidate recommendation device. 13. The method of claim 12,
The knowledge graph candidate recommendation unit,
performing tagging on the column item data based on a result of performing the clustering, and generating node recommendation information for the column item data in a supervised learning method using the tagged data,
Knowledge Graph candidate recommendation device. one or more processors;
a communication interface for communicating with an external device;
a memory for loading a computer program executed by the processor; and
a storage for storing the computer program;
The computer program is
Receiving information about the knowledge graph candidate recommendation model from the server;
obtaining semi-structured data by user input;
extracting column item data identified by a plurality of columns from the semi-structured data;
The extracted column item data is input to the knowledge graph candidate recommendation model, and knowledge graph candidate nodes and candidate edges matching the extracted column item data using the output data of the knowledge graph candidate recommendation model are used. ) to display information about, and
Clustering is performed based on the similarity using the column item data,
and instructions for automatically matching the column item data to a node and a connecting line having a maximum probability value based on the clustering execution result,
The knowledge graph candidate recommendation model is generated by performing machine learning using learning data including knowledge graph information,
Knowledge Graph candidate recommendation device.

Images