Method and apparatus for expressing query by using knowledge graph

Abstract

A method for expressing a query by using a knowledge graph, according to an embodiment of the present invention, may comprise the steps of: acquiring query information including a query and a knowledge graph from which the query has been derived; determining location information and role information regarding at least two nodes constituting the query; determining global context information regarding the query on the basis of the location information and the role information; determining local context information regarding each of the at least two nodes on the basis of the relationship between the at least two nodes constituting the query; and determining query expression information by using the location information, the role information, the global context information, and the local context information.

Description

Method and device for expressing queries using knowledge graphs

[0001] The present invention relates to a method and device for expressing a query using a knowledge graph.

[0002] This study is related to the development of a hazardous substance monitoring platform using a color-converting nanoparticle-based AIoT / IoT optical element gas sensor (NO. 2710018181) as part of the research project supported by Sungkyunkwan University with funding from the Ministry of Science and ICT (Government) in 2024.

[0003] In addition, this study is related to the research project Development of Technology to Improve Software Reliability through Identification of Abnormal Open Source and Automatic Application of DevSecOps (NO. 2710007381), which was supported by Sungkyunkwan University with funding from the Ministry of Science and ICT (Government) in 2024.

[0004] In addition, this study is related to the research project for training industrial convergence multimodal generative artificial intelligence talents (NO. 2710008244) supported by Sungkyunkwan University with funding from the Ministry of Science and ICT (government) in 2024.

[0005] In addition, this study is related to the research project “Development of a Stress Visualization and Quantification Durability Evaluation Platform Based on Stimulus-Sensitive Polymer Composites” (NO. 1711181944), which was supported by Sungkyunkwan University with funding from the Ministry of Science and ICT (Government) in 2024.

[0006] In addition, this study is related to the research project Development and Commercialization of AI-based Customized Digital Therapeutic Device for Mood Disorders (NO. H0601241023) supported by the National IT Industry Promotion Agency and funded by the Ministry of Science and ICT (Government) in 2024.

[0007] In addition, this study is related to the research project Development of Brain-Body Interface Technology Using AI-Based Multi-Sensing (NO. 1711196700), which was supported by Sungkyunkwan University with funding from the Ministry of Science and ICT (government) in 2023.

[0008] In addition, this study is related to the Artificial Intelligence Graduate School Support Project (NO. 2710008628) supported by Sungkyunkwan University with funding from the Ministry of Science and ICT (government) in 2024.

[0009] In addition, this study is related to the ICT Talent Development Project (NO. 2710007880) supported by Sungkyunkwan University with funding from the Ministry of Science and ICT (government) in 2024.

[0010] In addition, this study is related to the Deepfake Research Center (NO. 2710008857), a research project supported by Sungkyunkwan University with funding from the Ministry of Science and ICT (government) in 2024.

[0011] In addition, this study is related to the development of core technology for copyrighting datasets for multimodal generative AI models (NO. 2370000050), a research project supported by the Korea Electronics Technology Institute (KET) with funding from the Ministry of Science and ICT (Government) in 2024.

[0012] For reference, this application claims priority to Korean Patent Application No. 10-2023-0193517, filed on December 27, 2023. The entire contents of that application, which serves as the basis for this priority claim, are incorporated herein by reference.

[0013] A knowledge graph is a graph database that contains all the world's vast knowledge information, including encyclopedias, common sense, and the latest news. It is used in various fields such as natural language question-answering systems, recommendation systems, and search engines.

[0014] In order for knowledge graphs to be utilized in the various fields mentioned above, multi-hop logical reasoning technology on knowledge graphs is important, and various methodologies (i.e., geometric (e.g., box, cone)-based methodologies, probability (e.g., beta distribution)-based methodologies) are being studied to perform logical reasoning tasks.

[0015] However, conventional methodologies rely on linear sequential operations within a computational graph and do not consider the context of the input query, so determining the exact geometric boundaries or probability parameters for the query remains an unresolved problem.

[0016] Accordingly, there is a need to develop technologies to improve the effectiveness of multi-hop logical inference techniques.

[0017] (Prior art literature)

[0018] (Patent Document 0001) Korean Patent Publication No. 10-2023-0134798 (September 22, 2023)

[0019] The problem that the present invention seeks to solve is to improve the expressive power of a query by using the location information, role information, global context information, and local context information of nodes included in a query graph, thereby deriving a more accurate answer to a given query.

[0020] However, the problems to be solved by the present invention are not limited to those mentioned above, and other problems to be solved that are not mentioned can be clearly understood by a person having ordinary skill in the art to which the present invention pertains from the description below.

[0021] A method for expressing a query using a knowledge graph according to one embodiment of the present invention may include: obtaining query information including a query and a knowledge graph from which the query is derived; determining location information and role information regarding at least two nodes constituting the query; determining global context information regarding the query based on the location information and the role information; determining local context information for each of the at least two nodes constituting the query based on a relationship between the at least two nodes; and determining query expression information using the location information, the role information, the global context information, and the local context information.

[0022] Here, the step of determining the location information and role information may include the step of assigning a first unique number to each of the at least two nodes constituting the query by considering the structural locations of the at least two nodes; and the step of determining the location information by referring to the first unique number and a lookup table for each of the at least two nodes.

[0023] In addition, the step of determining the location information and role information may include the step of allocating a second unique number to each of the at least two nodes by considering the types of the at least two nodes constituting the query; and the step of determining the role information by referring to the second unique number and a lookup table for each of the at least two nodes.

[0024] In addition, the step of determining the global context information may include the step of generating a location-role table based on a first unique number included in the location information and a second unique number included in the role information; and the step of determining the global context information based on the location-role table.

[0025] In addition, the step of determining the local context information may include a step of extracting a head node and a tail node by considering the directionality of connection of at least two nodes constituting the query; and a step of determining the local context information based on the head node and the tail node.

[0026] Meanwhile, the method may further include a step of deriving answer information for the query from the knowledge graph based on the query expression information.

[0027] According to another embodiment of the present invention, a query expression device using a knowledge graph comprises: a memory in which a query expression program is stored; and a processor for loading the query expression program from the memory and executing the query expression program, wherein the processor obtains query information including a query and a knowledge graph from which the query is derived, determines location information and role information regarding at least two nodes constituting the query, determines global context information regarding the query based on the location information and the role information, determines local context information for each of the at least two nodes based on a relationship between the at least two nodes constituting the query, and determines query expression information using the location information, the role information, the global context information, and the local context information.

[0028] Here, the processor may assign a first unique number to each of the at least two nodes constituting the query by considering the structural positions of the at least two nodes, and may determine the position information by referring to the first unique number and a lookup table for each of the at least two nodes.

[0029] Additionally, the processor may assign a second unique number to each of the at least two nodes constituting the query by considering the types of the at least two nodes, and may determine the role information by referring to the second unique number and a lookup table for each of the at least two nodes.

[0030] Additionally, the processor can generate a location-role table based on a first unique number included in the location information and a second unique number included in the role information, and determine the global context information based on the location-role table.

[0031] Additionally, the processor can extract a head node and a tail node by considering the directionality of connection of at least two nodes constituting the query, and determine the local context information based on the head node and the tail node.

[0032] Meanwhile, the processor can derive answer information for the query from the knowledge graph based on the query expression information.

[0033] A computer-readable recording medium storing a computer program according to another embodiment of the present invention may include instructions for causing the processor to perform a query expression method using a knowledge graph, the method comprising: obtaining query information including a query and a knowledge graph from which the query is derived; determining location information and role information regarding at least two nodes constituting the query; determining global context information regarding the query based on the location information and the role information; determining local context information regarding each of the at least two nodes based on a relationship between the at least two nodes constituting the query; and determining query expression information using the location information, the role information, the global context information, and the local context information.

[0034] According to an embodiment of the present invention, by determining location information and role information for each of at least two nodes constituting a query graph, it is possible to represent a query graph that takes context into account by reflecting the characteristics and connection relationships between nodes.

[0035] In addition, according to an embodiment of the present invention, by effectively expressing a query by utilizing the global context and local context of the query graph, an effect of being able to derive a more accurate answer to a given query can be achieved.

[0036] In addition, according to an embodiment of the present invention, query expression information can be determined regardless of the type of query expression-based deep learning model.

[0037] FIG. 1 is a block diagram showing a query expression device according to an embodiment of the present invention.

[0038] FIG. 2 is a block diagram conceptually illustrating the function of a query expression program according to an embodiment of the present invention.

[0039] Figure 3 is a flowchart illustrating a query expression method according to one embodiment of the present invention.

[0040] FIG. 4 is a drawing exemplarily showing assigning a first unique number by considering the structural positions of at least two nodes according to one embodiment of the present invention.

[0041] FIG. 5 is a drawing exemplarily showing assigning a second unique number by considering the types of at least two nodes according to one embodiment of the present invention.

[0042] FIG. 6 is a diagram exemplarily showing determining global context information according to one embodiment of the present invention.

[0043] FIG. 7 is a diagram exemplarily showing determining local context information according to one embodiment of the present invention.

[0044] The advantages and features of the present invention, and the methods for achieving them, will become clearer with reference to the embodiments described in detail below together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. These embodiments are provided solely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the invention, and the present invention is defined solely by the scope of the claims.

[0045] When describing embodiments of the present invention, detailed descriptions of known functions or configurations will be omitted if they are deemed to unnecessarily obscure the gist of the invention. Furthermore, the terms described below are defined in light of their functions in the embodiments of the present invention and may vary depending on the intent or custom of the user or operator. Therefore, their definitions should be based on the overall content of this specification.

[0046] FIG. 1 is a block diagram showing a query expression device according to an embodiment of the present invention.

[0047] Referring to FIG. 1, the query expression device (100) may include a processor (110), an input / output device (120), and a memory (130).

[0048] The processor (110) can control the overall operation of the query expression device (100).

[0049] The processor (110) can receive query information including a query and a knowledge graph derived from the query using an input / output device (120).

[0050] In the present invention, a query is a request to search or extract desired information, and may mean a query graph including nodes and edges.

[0051] Additionally, in the present invention, a knowledge graph may mean a graph database that contains vast knowledge of the real world in the form of a graph among data structures.

[0052] In the present invention, although it has been described that query information including a query and a knowledge graph derived from the query are input through an input / output device (120), it is not limited thereto. That is, depending on the embodiment, the query expression device (100) may include a transceiver (not shown), and the query expression device (100) may receive at least one of the query and the knowledge graph derived from the query using the transceiver (not shown), and at least one of the query and the knowledge graph derived from the query may be generated within the query expression device (100).

[0053] The processor (110) may obtain query information including a query and a knowledge graph from which the query is derived, determine location information and role information about at least two nodes constituting the query, determine global context information about the query based on the location information and the role information, determine local context information about each of at least two nodes based on a relationship between the at least two nodes constituting the query, and determine query expression information using the location information, the role information, the global context information, and the local context information.

[0054] The input / output device (120) may include one or more input devices and / or one or more output devices. For example, the input devices may include a microphone, a keyboard, a mouse, a touch screen, etc., and the output devices may include a display, a speaker, etc.

[0055] The memory (130) can store the query expression program (200) and information necessary for executing the query expression program (200).

[0056] In this specification, a query expression program (200) may mean software that includes commands for receiving query information including a query and a knowledge graph from which the query is derived and determining query expression information.

[0057] In addition, in this specification, the query expression program (200) may mean software that includes commands for receiving query information including a query and a knowledge graph from which the query is derived and deriving answer information for the query from the knowledge graph.

[0058] The processor (110) can load the query expression program (200) and information necessary for executing the query expression program (200) from the memory (130) to execute the query expression program (200).

[0059] The processor (110) can determine query expression information by executing a query expression program (200), using location information, role information, global context information, and local context information determined by inputting query information including a query and a knowledge graph from which the query is derived.

[0060] The functions and / or operations of the query expression program (200) will be examined in detail with reference to Fig. 2.

[0061] FIG. 2 is a block diagram conceptually illustrating the function of a query expression program according to an embodiment of the present invention.

[0062] Referring to FIG. 2, the query expression program (200) may include a query information acquisition unit (210), a node information management unit (220), and a query expression information determination unit (230).

[0063] The query information acquisition unit (210), node information management unit (220), and query expression information determination unit (230) illustrated in FIG. 2 conceptually divide the functions of the query expression program (200) to easily explain the functions of the query expression program (200), but are not limited thereto. According to embodiments, the functions of the query information acquisition unit (210), node information management unit (220), and query expression information determination unit (230) can be merged / separated, and can also be implemented as a series of commands included in a single program.

[0064] First, the query information acquisition unit (210) can acquire query information including a query and a knowledge graph from which the query is derived.

[0065] Query information may include a query graph regarding a request sent to a database to search or extract information and information regarding a knowledge graph corresponding to the database.

[0066] Here, the query graph can be configured to include at least two nodes.

[0067] For example, a query graph may include at least one of a variable node and an answer node, as well as an anchor node. Here, an anchor node refers to a core entity or concept that constitutes the query and may refer to a node where the query begins. Furthermore, a variable node may refer to any node (or entity) that satisfies part of the query during the inference process. Furthermore, an answer node may refer to a node that constitutes an answer to the query.

[0068] Additionally, the query graph may include at least one edge representing a relationship between at least two nodes.

[0069] That is, the query graph has a structural form based on at least two nodes and at least one edge.

[0070] Next, the node information management unit (220) can determine location information regarding at least two nodes constituting the query. Here, the location information may refer to information indicating the hierarchical location of the node, taking into account the structural form of the query graph.

[0071] Specifically, the node information management unit (220) can assign a first unique number to each of at least two nodes by considering the structural positions of at least two nodes constituting the query.

[0072] More specifically, the node information management unit (220) can assign the highest first unique number to the node with the largest number of directly and indirectly connected nodes by considering the directionality of the edges among at least two nodes that constitute the query.

[0073] For example, if a query graph is formed by sequentially connecting anchor nodes, variable nodes, and answer nodes, the node information management unit (220) can assign 0 to the anchor node, 1 to the variable node, and 2 to the answer node as the first unique number.

[0074] For another example, if a query graph is formed by sequentially connecting an anchor node, a first variable node, a second variable node, and a correct answer node, the node information management unit (220) may assign 0 to the anchor node, 1 to the first variable node, 2 to the second variable node, and 3 to the correct answer node as the first unique number.

[0075] Additionally, the node information management unit (220) can determine location information by referring to the first unique number and lookup table for each of at least two nodes.

[0076] More specifically, the node information management unit (220) can determine a position embedding representing the position of each of at least two nodes in a latent space by using a first unique number for each of at least two nodes and a lookup table corresponding to the first unique number.

[0077] Additionally, the node information management unit (220) can determine role information regarding at least two nodes constituting a query. Here, the role information may mean information indicating what role a node constituting a query graph plays in the query graph.

[0078] Specifically, the node information management unit (220) can assign a second unique number to each of at least two nodes by considering the types of at least two nodes that constitute the query.

[0079] For example, the node information management unit (220) may assign 0 to the anchor node, 1 to the variable node, and 2 to the correct node as the second unique number.

[0080] Additionally, the node information management unit (220) can determine role information by referring to a second unique number and a lookup table for each of at least two nodes.

[0081] More specifically, the node information management unit (220) can determine a role embedding expressing a role for each of at least two nodes in the latent space by using a second unique number for each of at least two nodes and a lookup table corresponding to the second unique number.

[0082] In this way, by determining location information and role information for at least two nodes, a distinct representation is possible for each knowledge graph, reflecting the characteristics and connection relationships between the nodes.

[0083] However, the first unique number and the second unique number are only examples and may be changed in various ways within the scope that can achieve the purpose of the present invention.

[0084] Meanwhile, the node information management unit (220) can determine global context information regarding the query based on location information and role information.

[0085] Specifically, the node information management unit (220) can create a location-role table based on a first unique number included in the location information and a second unique number included in the role information.

[0086] More specifically, the node information management unit (220) can generate a location-role table representing the entire structure of the knowledge graph by determining the number of nodes corresponding to the first unique number and the number of nodes corresponding to the second unique number.

[0087] Additionally, the node information management unit (220) can determine global context information based on the location-role table.

[0088] More specifically, the node information management unit (220) can determine a multi-hot vector based on the position-role table. In addition, the node information management unit (220) can input the multi-hot vector into a neural network (e.g., a multi-layer perceptron) to determine a global embedding that expresses the structural characteristics of the query graph in the latent space.

[0089] Meanwhile, the node information management unit (220) can determine local context information for each of at least two nodes based on the relationship between at least two nodes that constitute the query.

[0090] Specifically, the node information management unit (220) can extract a head node and a tail node by considering the directionality in which at least two nodes constituting the query are connected.

[0091] More specifically, the node information management unit (220) can extract a head node and a tail node corresponding to a correlation between at least two nodes constituting a query graph in a knowledge graph.

[0092] Additionally, the node information management unit (220) can determine local context information based on the head node and tail node.

[0093] Specifically, the node information management unit (220) can determine local context information based on edges indicating the correlation between the head node and the tail node.

[0094] More specifically, the node information management unit (220) can determine a local embedding that expresses characteristics of a relationship that specifies a concept or object in a latent space using a head node, a tail node, and edges.

[0095] Next, the query expression information determination unit (230) can determine query expression information using location information, role information, global context information, and local context information.

[0096] Specifically, the query expression information determination unit (230) can determine query expression information by integrating location information, role information, global context information, and local context information.

[0097] More specifically, the query embedding, position embedding, role embedding, global embedding, and local embedding for the query graph itself can be concatenated and fed into a neural network (e.g., a multilayer perceptron) to determine a contextualized query embedding.

[0098] Meanwhile, the query expression information determination unit (230) can derive answer information for the query from the knowledge graph based on the query expression information.

[0099] Specifically, the query expression information determination unit (230) can determine second query expression information for the correct answer node based on first query expression information for the variable node.

[0100] In addition, the query expression information determination unit (230) can determine the node (or entity) with the closest embedding among the contextualized query embedding included in the second query expression information and the embedding for the node (or entity) for the query included in the knowledge graph as the correct answer node (or entity).

[0101] In this way, by effectively expressing a query by utilizing the global context and local context of the query graph, the effect of being able to derive a more accurate answer to a given query can be achieved.

[0102]

[0103] Figure 3 is a flowchart illustrating a query expression method according to one embodiment of the present invention.

[0104] Referring to FIG. 3, the query information acquisition unit (210) can acquire query information including a query and a knowledge graph from which the query is derived (S310).

[0105] Next, the node information management unit (220) can determine location information and role information regarding at least two nodes that constitute the query (S320).

[0106] Additionally, the node information management unit (220) can determine global context information regarding a query based on location information and role information (S330).

[0107] Additionally, the node information management unit (220) can determine local context information for each of at least two nodes based on the relationship between at least two nodes that constitute the query (S340).

[0108] Next, the query expression information determination unit (230) can determine query expression information using location information, role information, global context information, and local context information (S350).

[0109]

[0110] FIG. 4 is a drawing exemplarily showing assigning a first unique number by considering the structural positions of at least two nodes according to one embodiment of the present invention.

[0111] Referring to FIG. 4, for seven types of query graphs (410, 420, 430, 440, 450, 460, 470), the node information management unit (220) can determine location information regarding at least two nodes constituting the query graph.

[0112] Specifically, in the first type of query graph (410), anchor nodes and correct answer nodes are sequentially connected, and the node information management unit (220) can assign 0 to the anchor node and 1 to the correct answer node as the first unique number.

[0113] Additionally, in the second type of query graph (420), anchor nodes, variable nodes, and correct answer nodes are sequentially connected, and the node information management unit (220) can assign 0 to the anchor node, 1 to the variable node, and 2 to the correct answer node as the first unique number.

[0114] In addition, in the third type of query graph (430), an anchor node (431), a first variable node (432), a second variable node (433), and a correct answer node (434) are sequentially connected, and the node information management unit (220) can assign 0 to the anchor node (431), 1 to the first variable node (432), 2 to the second variable node (433), and 3 to the correct answer node (434) as a first unique number.

[0115] In this way, the node information management unit (220) can assign the highest first unique number to the node with the largest number of directly and indirectly connected nodes by considering the directionality of the edges among at least two nodes that constitute the query.

[0116] Additionally, in the fourth type of query graph (440), the correct answer node is connected to the first anchor node and the second anchor node, and the node information management unit (220) can assign 0 to the first anchor node and the second anchor node as the first unique number, and assign 1 to the correct answer node.

[0117] In addition, in the seventh type of query graph (470), the correct answer node is connected with the first anchor node and the variable node, and the variable node is sequentially connected to the second anchor node, and the node information management unit (220) can assign 0 to the second anchor node, 1 to the first anchor node and the variable node, and 2 to the correct answer node as the first unique number.

[0118]

[0119] FIG. 5 is a drawing exemplarily showing assigning a second unique number by considering the types of at least two nodes according to one embodiment of the present invention.

[0120] Referring to FIG. 5, the node information management unit (220) can assign a second unique number to each of at least two nodes by considering the types of at least two nodes constituting the query graph.

[0121] For example, the node information management unit (220) may assign 0 to the anchor node, 1 to the variable node, and 2 to the correct node as the second unique number.

[0122] Through this, unlike the conventional technology where the expression for the second node sequentially connected to the first node was the same regardless of the type of node, a distinct expression can be made possible for each knowledge graph by reflecting the characteristics and connection relationships between nodes.

[0123]

[0124] FIG. 6 is a diagram exemplarily showing determining global context information according to one embodiment of the present invention.

[0125] Referring to FIG. 6, the node information management unit (220) can generate a location-role table (602) based on a first unique number and a second unique number assigned to each of the first anchor node, the second anchor node, the variable node, and the answer node constituting the query graph (601). Here, the location-role table (602) can be generated based on the number of nodes corresponding to the first unique number and the number of nodes corresponding to the second unique number.

[0126] Additionally, the node information management unit (220) can input a multi-hot vector determined based on a location-role table into a multilayer perceptron to determine a global embedding (603) that expresses the structural characteristics of the query graph in the latent space.

[0127]

[0128] FIG. 7 is a diagram exemplarily showing determining local context information according to one embodiment of the present invention.

[0129] Figure 7 shows a query graph containing queries regarding the languages ​​used by Asian countries that hosted the Olympic Games.

[0130] Referring to FIG. 7, the node information management unit (220) can determine a local embedding that expresses characteristics of a relationship that specifies a concept or object in a latent space using a head node, a tail node, and edges.

[0131] Specifically, the node information management unit (220) can extract a head node and a tail node corresponding to a correlation between at least two nodes constituting a query graph in a knowledge graph.

[0132] In addition, when determining local context information for a variable node (700), the node information management unit (220) can extract a tail node (701) that is correlated with hosting the Olympics in the knowledge graph, a tail node (702) that is correlated with being located in Asia, and a head node (703) that is correlated with using a language.

[0133] Additionally, the node information management unit (220) can determine a local embedding for the variable node (700) by calculating an embedding for each of the tail nodes (701, 702) and the head node (703).

[0134] For example, local embedding can be expressed as in the following mathematical expression (1).

[0135]

[0136]

[0137] Here, can mean local embedding, may mean the embedding for the tail nodes (701, 702), may mean an embedding for the head node (703).

[0138]

[0139] The combination of each block of the block diagram and each step of the flowchart attached to the present invention may be performed by computer program instructions. These computer program instructions may be installed in an encoding processor of a general-purpose computer, a special-purpose computer, or other programmable data processing equipment, so that the instructions executed by the encoding processor of the computer or other programmable data processing equipment create a means for performing the functions described in each block of the block diagram or each step of the flowchart. These computer program instructions may also be stored in a computer-available or computer-readable memory that can direct a computer or other programmable data processing equipment to implement the functions in a specific manner, so that the instructions stored in the computer-available or computer-readable memory can also produce an article of manufacture that includes an instruction means for performing the functions described in each block of the block diagram or each step of the flowchart. Since the computer program instructions can also be installed on a computer or other programmable data processing device, a series of operational steps are performed on the computer or other programmable data processing device to create a computer-executable process, and the instructions that cause the computer or other programmable data processing device to perform the steps for executing the functions described in each block of the block diagram and each step of the flowchart can also provide steps for executing the functions described in each block of the block diagram and each step of the flowchart.

[0140] Additionally, each block or step may represent a module, segment, or portion of code that includes one or more executable instructions for performing a specific logical function(s). It should also be noted that in some alternative embodiments, the functions mentioned in the blocks or steps may occur out of order. For example, two blocks or steps depicted in succession may actually be performed substantially concurrently, or the blocks or steps may sometimes be performed in reverse order, depending on the functionality they perform.

[0141] The above description is merely an illustrative illustration of the technical idea of ​​the present invention, and those skilled in the art will appreciate that various modifications and variations can be made without departing from the essential quality of the present invention. Therefore, the embodiments disclosed in the present invention are intended to illustrate, rather than limit, the technical idea of ​​the present invention, and the scope of the technical idea of ​​the present invention is not limited by these embodiments. The scope of protection of the present invention should be interpreted by the following claims, and all technical ideas within a scope equivalent thereto should be interpreted as being included in the scope of the rights of the present invention.

Claims

1. A query expression method using a knowledge graph, A step of obtaining query information including a query and a knowledge graph from which the query is derived; A step of determining location information and role information regarding at least two nodes constituting the above query; A step of determining global context information regarding the query based on the location information and the role information; A step of determining local context information for each of at least two nodes based on the relationship between the at least two nodes constituting the above query; and A step of determining query expression information using the above location information, the above role information, the above global context information, and the above local context information. How to express a query.

2. In paragraph 1, The steps of determining the above location information and role information are: A step of assigning a first unique number to each of at least two nodes constituting the above query, taking into account the structural positions of the at least two nodes; and A step of determining the location information by referring to a first unique number and a lookup table for each of the at least two nodes. How to express a query.

3. In paragraph 1, The steps of determining the above location information and role information are: A step of assigning a second unique number to each of said at least two nodes, taking into account the types of said at least two nodes constituting said query; and A step of determining said role information by referring to a second unique number and a lookup table for each of said at least two nodes. How to express a query.

4. In paragraph 1, The step of determining the above global context information is: A step of generating a location-role table based on a first unique number included in the location information and a second unique number included in the role information; and A step of determining the global context information based on the above location-role table is included. How to express a query.

5. In paragraph 1, The step of determining the above local context information is: A step of extracting a head node and a tail node by considering the directionality of connection of at least two nodes constituting the above query; and A step of determining the local context information based on the head node and the tail node. How to express a query.

6. In paragraph 1, Further comprising a step of deriving correct answer information for the query from the knowledge graph based on the query expression information. How to express a query.

7. As a query expression device using a knowledge graph, Memory where the query expression program is stored; and A processor for loading the query expression program from the memory and executing the query expression program, The above processor, Obtain query information including a query and a knowledge graph from which the query is derived, Determine location information and role information about at least two nodes that constitute the above query, Determine global context information about the query based on the above location information and the above role information, Determine local context information for each of at least two nodes based on the relationship between the at least two nodes constituting the above query, Determining query expression information using the above location information, the above role information, the above global context information, and the above local context information. A query expression device.

8. In paragraph 7, The above processor, Assigning a first unique number to each of at least two nodes constituting the above query, taking into account the structural positions of the at least two nodes, Determining the location information by referring to the first unique number and lookup table for each of the at least two nodes. A query expression device.

9. In paragraph 7, The above processor, Considering the types of at least two nodes that constitute the above query, assign a second unique number to each of the at least two nodes, Determining the role information by referring to the second unique number and lookup table for each of the at least two nodes. A query expression device.

10. In paragraph 7, The above processor, Generate a location-role table based on the first unique number included in the above location information and the second unique number included in the above role information, Determining the global context information based on the above location-role table A query expression device.

11. In paragraph 7, The above processor, Extract the head node and tail node by considering the directionality of connection of at least two nodes composing the above query, Determining the local context information based on the head node and the tail node A query expression device.

12. In paragraph 7, The above processor, Based on the above query expression information, the correct answer information for the above query is derived from the above knowledge graph. A query expression device.

13. A non-transitory computer-readable recording medium storing a computer program, The above computer program, when executed by a processor, A step of obtaining query information including a query and a knowledge graph from which the query is derived; A step of determining location information and role information regarding at least two nodes constituting the above query; A step of determining global context information regarding the query based on the location information and the role information; A step of determining local context information for each of at least two nodes based on the relationship between the at least two nodes constituting the above query; and A step of determining query expression information using the above location information, the above role information, the above global context information, and the above local context information. A method of expressing a query using a knowledge graph, comprising a command for causing the processor to perform the method. A non-transitory computer-readable recording medium.

Images