KR102053243B1 - Adaptive Knowledge Base Construction Method and System - Google Patents

Abstract

According to an aspect of the present invention, there is provided an adaptive knowledge base construction apparatus comprising: a machine learning engine that generates a learning model based on an input first data set, and analyzes an association relationship with the first data set according to the generated learning model; A rule generator for generating a rule based on the learned model and algorithm; And a semantic rule generator for generating semantic rules using the generated rules and the second data set.

Description

Adaptive Knowledge Base Construction Method and System

The present invention relates to a method and system for constructing an adaptive knowledge base. In particular, the present invention converts a learning result generated through machine learning into a rule, and builds a knowledge base by using semantic techniques. Knowledge base construction method and system thereof.

Recently, research on machine learning and semantic technologies has been actively conducted. Machine learning is a technology that performs data-based learning according to a given purpose, and predicts the information required for a new environment based on the learned results. In general, a learning method may be classified into a tree-based analysis method and an association-based analysis method. A tree-based analysis method generates a rule using node information of a tree constructed as a learning result, and an association-based analysis method generates an association rule by analyzing a pattern of training data.

Semantic technology refers to a technique for creating a rule by a designer having domain knowledge, and using the generated rule, to build a knowledge base by extending-inferring-reusing knowledge.

When generating a rule using machine learning, and generating a result according to a request using the rule, there is a generation cycle, there is a problem that can not recycle the learning results. In the case of semantic technology, the knowledge base is built as a result of expansion and inference based on pre-formed rule-based knowledge before a person with domain knowledge intervenes to change the rules. Therefore, it is necessary to adapt the rules adaptively according to the surrounding environment and to build a knowledge base reflecting the changed rules.

An object of the present invention is to provide an adaptive knowledge base construction method and system for solving the above problems. Specifically, the present invention provides a method and system for constructing an adaptive knowledge base that reflects a dynamically changing environment without human intervention using machine learning and semantic techniques.

The object of the present invention is not limited to the above-mentioned object, and other objects that are not mentioned will be clearly understood by those skilled in the art from the following description.

According to an aspect of the present invention, an adaptive knowledge base construction apparatus generates a learning model based on an input first data set, and analyzes an association relationship with the first data set according to the generated learning model. Machine learning engine; A rule generator for generating a rule based on the learned model and algorithm; And a semantic rule generator for generating semantic rules using the generated rules and the second data set.

According to another aspect of the present invention, there is provided an adaptive knowledge base construction system comprising: a memory in which a program for providing an adaptive knowledge base construction module is stored; And a processor for executing the program, wherein the processor generates a learning model for the input first dataset as the program is executed, and analyzes the correlation of the first dataset according to the generated learning model. Outputting a result, generating a machine learning rule based on the learned model, the learned algorithm and the correlation analysis result, and generating semantic rules using the generated machine learning rule.

According to another aspect of the present invention, an adaptive knowledge base construction method includes (1) machine learning an input first data set; (2) generating a rule based on the learned result; (3) generating a semantic rule using the generated rule and the second data set; And (4) storing the generated semantic rules to build a knowledge base.

According to the present invention, by combining a machine learning technology and semantic technology to build a knowledge base to exclude human intervention, it is effective to achieve optimal analysis and high efficiency.

The present invention can be applied to the field of intelligent services related to IoT technology, analysis related to big data technology, and context awareness service, and can be used as a platform in the field of analysis technology using machine learning.

1 is an exemplary diagram for explaining the configuration of a computer system in which the adaptive knowledge base building method according to the present invention is implemented.
2 is a block diagram of an adaptive knowledge base construction system according to an embodiment of the present invention.
3 is a block diagram of an adaptive knowledge base building system according to another embodiment of the present invention.
4 is an exemplary diagram for explaining a tree learning model based adaptive knowledge base construction method according to an embodiment of the present invention.
5 is an exemplary diagram for explaining an adaptive knowledge base construction method applying the Apriori algorithm according to an embodiment of the present invention.
6 is a block diagram of an adaptive knowledge base building system according to another embodiment of the present invention.
7 is a block diagram of an adaptive knowledge base building system according to another embodiment of the present invention.
8 is a process flow diagram of an adaptive knowledge base building method according to another embodiment of the present invention.

Advantages and features of the present invention, and methods for achieving them will become apparent with reference to the embodiments described below in detail in conjunction with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but will be implemented in various forms, and only the present embodiments are intended to complete the disclosure of the present invention, and the general knowledge in the art to which the present invention pertains. It is provided to fully convey the scope of the invention to those skilled in the art, and the present invention is defined only by the scope of the claims. Meanwhile, the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In this specification, the singular also includes the plural unless specifically stated otherwise in the phrase. As used herein, “comprises” and / or “comprising” refers to a component, step, operation and / or device that is present in one or more other components, steps, operations and / or elements. Or does not exclude additions.

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

1 is an exemplary diagram for explaining a configuration of a computer system in which an adaptive knowledge base construction method according to the present invention is implemented.

Meanwhile, the adaptive knowledge base construction method according to an embodiment of the present invention may be implemented in a computer system or recorded in a recording medium. As shown in FIG. 1, a computer system includes at least one processor 110, a memory 120, a user input device 150, a data communication bus 130, a user output device 160, It may include a reservoir 140. Each of the above components communicates with each other via the data communication bus 130.

The computer system can further include a network interface 170 coupled to the network 180. The processor 110 may be a central processing unit (CPU) or a semiconductor device that processes instructions stored in the memory 120 and / or the storage 140.

The memory 120 and the storage 140 may include various types of volatile or nonvolatile storage media. For example, the memory 120 may include a ROM 123 and a RAM 126.

Therefore, the adaptive knowledge base construction method according to the embodiment of the present invention may be implemented in a computer executable method. When the adaptive knowledge base construction method according to an embodiment of the present invention is performed in a computer device, computer readable instructions may perform the operation method according to the present invention.

Meanwhile, the above-described adaptive knowledge base construction method according to the present invention can be implemented as computer readable codes on a computer readable recording medium. Computer-readable recording media include all kinds of recording media having data stored thereon that can be decrypted by a computer system. For example, there may be a read only memory (ROM), a random access memory (RAM), a magnetic tape, a magnetic disk, a flash memory, an optical data storage device, and the like. The computer readable recording medium can also be distributed over computer systems connected over a computer network, stored and executed as readable code in a distributed fashion.

2 is a block diagram of an adaptive knowledge base construction system according to an embodiment of the present invention.

An adaptive knowledge base building system according to the present invention comprises a machine learning engine; Rule generator; Semantic rule engine; Knowledge base; Preferably, it may further include a rule modeler of a domain expert.

The adaptive knowledge base construction system according to the present invention is based on a knowledge base construction environment based on semantic modeling and a rule base learned based on machine learning to operate a knowledge base construction environment in parallel, or to each knowledge base. The build environment can be operated individually. The case based on semantic modeling and the case based on machine learning can be functionally separated, and can be constructed as a distributed system in consideration of the entire system.

The machine learning engine 210 generates an optimal learning model for the first dataset 260, trains the first dataset using the generated learning model, and relates to the dataset based on the learned results. The analysis results are output as learned model results.

The rule generator 220 generates a rule by regularizing the learned model and the learned algorithm using the output learned model result, and also generates a rule for the analyzed data correlation. The generated rule is passed to the semantic rule engine 230, or a semantic rule is generated from the generated rule and passed to the semantic rule engine 230, or the unstructured data is generated from the generated rule and passed to the semantic rule engine 230. Can be. Typically rule generator 220 generates a generalized rule. However, the semantic rule can be generated directly by applying the API of the semantic rule engine to the rule generator 220. At this time, whether to store the semantic rule that has generated the role of the semantic rule engine 230 in the knowledge base and existing You decide whether to apply the extension to semantic rules.

The adaptive knowledge base building system according to the present invention may be built as a distributed processing system, and the rule generator 220 and the semantic rule engine 230 may be installed in separate servers according to the distributed processing system configuration method. Whether the rule generator 220 generates the semantic rule can usually be appropriately selected according to the load of each server.

The semantic rule engine 230 is also called an inference engine, and serves to extend the semantic rules. The semantic rule engine 230 generates semantic rules based on the rules and the unstructured data received from the rule generator 220 and stores the generated semantic rules in the knowledge base. In addition, the semantic rule engine 230 determines whether or not the semantic rule received from the rule generator 220 is stored in the knowledge base.

In addition, the semantic rule engine 230 may receive the second dataset 270 and expand the semantic rule stored in the built knowledge base. The second dataset 270 may include a new rule and a new semantic rule manually input.

3 is a block diagram of an adaptive knowledge base building system according to another embodiment of the present invention.

The adaptive knowledge base construction method according to the present invention comprises the steps of: (1) generating a learning model for a first dataset (S310); (2) generating trained model results using the generated learning model; (3) generating a rule using the learned model results (S330); (4) generating a semantic rule using the generated rule (S340); (5) storing the generated semantic rules to build a knowledge base (S350).

In step (1), the machine learning engine 210 learns using the first data set to generate an optimal learning model.

In the step (2), the machine learning engine 210 learns an optimal analysis model and outputs the correlation analysis result for the first data set provided as an input in the inference step.

In step (3), the rule generator 220 rules the learned model and the learned algorithm, and also rules the correlation between the analyzed data using the learned model and the learned algorithm.

In the step (4), the general rule generated by the rule generator 220 is converted into a semantic rule.

In step (5), the transformed semantic rule is merged with previously stored semantic rules to expand the knowledge base.

The method for generating a rule using the machine learning in the step (1) is different according to the algorithm. The present invention provides a method of acquiring the knowledge required for rule generation through machine learning, regenerating the acquired knowledge into a rule, and finally converting the semantic rule into a semantic rule by applying semantic technology. The machine learning method used in step (1) will be described again with reference to FIGS. 4 and 5. However, this is merely exemplary, and rules may be generated by other algorithms or models, and are not intended to limit the scope of the invention.

4 is an exemplary diagram for describing a tree learning model-based adaptive knowledge base construction method according to an exemplary embodiment of the present invention.

4A is an exemplary diagram for explaining a tree-based rule generation method.

4 (a) shows A and C as input data, and learning results B, C-1, and C-2 are generated. For the input data A and C, a tree data structure may be obtained as shown in FIG. Each node can enter data or generate output. Although FIG. 4 (a) shows the simplest example, more complex and various trees can be generated in the form of a binary tree, and control statements can be generated regularly for each node.

For example, the rule generated in FIG. 4 (a) is as follows. If A is greater than the threshold, output "1". Otherwise, if C is yes, print "2". If C is "no", print "3".

The flowchart of FIG. 4 (a) is shown in FIG. 4 (b). The advantage of the tree structure is that you can construct algorithms recursively. Although FIG. 4 (b) is only a flow chart presented to explain FIG. 4 (a), it is possible to generate a complex machine learning model for an input first dataset by repeatedly applying simple rules using a tree structure. There is an advantage.

5 is an exemplary diagram for describing a method of constructing an adaptive knowledge base to which an Apriori algorithm according to an exemplary embodiment of the present invention is applied. In detail, FIG. 5 is an exemplary diagram for describing a method of generating an association analysis result as a rule based on an Apriori algorithm.

The Apriori algorithm is widely known as a machine learning method, but can be applied to generate semantic rules.

The numbers shown in FIG. 5 (a) mean items corresponding to Table 1 below, A, B, C, D, and E denote one transaction, and Table 2 below Represents an item purchased in a transaction.

Item name Item ID soy milk One lettuce 2 diapers 3 wine 4 chard 5 orange juice 6

Deal ID Item list Item ID list A Soy milk, lettuce 1, 2 B Lettuce, diapers 2, 3 C Soy milk, wine 1, 4 D Lettuce, soy milk, diapers, wine 2, 1, 3, 4 E Lettuce, soy milk, diapers 2, 1, 3

Referring to Table 2, the semantic web of Figure 5 (a) can be formed. The formed semantic web may be ordered by semantic rules.

For example, if the bottom box is analyzed, a total of 4 includes 1, 4 includes 2, 4 includes 3, 4 includes 3, and 3 includes 4 It can be seen that. The following table summarizes this.

Item assembly Support {One} 4 {2} 4 {3} 3 {4} 2

In this case, if the minimum support is set to 4, {1} and {2} are classified as a frequent item set, and {3} and {4} are classified as a non-frequent item set. In FIG. 5 (a), the frequent item set is inverted.

Similarly, support can be calculated for a set of items with two elements, which are shown in the table below.

Item assembly Support {1,2} 3 {1,3} 2 {1,4} 2 {2,3} 3 {2,4} One {3,4} One

In this case, if the minimum support is set to 3, {1,2} and {2,3} become frequent item sets, and {1,3}, {1,4}, {2,4}, and {3,4 } Is classified as a set of non-frequent items. In FIG. 5 (a), the frequent item set is inverted.

Similarly, support can be calculated for an item set of three elements, which are shown in the table below.

Item assembly Support {1,2,3} 2 {1,2,4} One {1,3,4} One {2,3,4} One

At this time, if the minimum support is set to 2, {1,2,3} becomes a frequent item set, and {1,3}, {1,2,4}, {1,3,4}, {2,3 , And 4} are classified as non-frequent item sets. In FIG. 5 (a), the frequent item set is inverted.

Similarly, when calculating the support for a set of 4 items, {1,2,3,4} has 1 support, so if the minimum support is 1, it is a frequent item set, and if the minimum support is 2, In other words, it becomes a set of non-frequent items. In FIG. 5 (a), the case where the minimum support degree is 1 is assumed.

In general, the Apriori algorithm is a machine learning method that can prun non-frequent items and proceed to increase computation speed.

5B is an exemplary diagram for describing a method of applying an Apriori algorithm according to an exemplary embodiment of the present invention. Specifically, FIG. 5 (b) shows the result of pruning the nodes below the minimum support in FIG. 5 (a).

When all the operations are performed in the entire tree, the number of operations increases exponentially with the number of items, which causes a problem that the computer cannot keep up with the operation speed. If pruning is performed for nodes below the minimum support and the operation continues only for nodes above the minimum support, semantic rules between nodes can be generated by a small number of operations. Arrows indicated by dotted lines in FIG. 5 (b) indicate nodes to be pointed out.

6 is a block diagram of an adaptive knowledge base building system according to another embodiment of the present invention.

The adaptive knowledge base building system according to the present invention may be configured as a distributed processing system.

An adaptive knowledge base building system according to the present invention includes a machine learning rule generation server 610; Semantic rule generation server 620; And a rule modeler server 630.

Machine learning rule generation server 610 includes a machine learning engine 210; And rule generator 220. The machine learning rule generation server 610 reflects the dynamic environment context through iterative relearning.

The semantic rule generation server 620 extends the knowledge base through semantic inference on the second dataset using semantic rules generated from machine learning rules. In addition, the machine learning engine may transmit the machine learning rule predicted using the second data set as input data to the semantic rule engine, and expand the knowledge base through semantic inference. Description thereof will be described later with reference to FIGS. 7 and 8.

The rule generator 220 converts the generated rule into a semantic rule by the semantic rule engine 240. The semantic rule engine may generate a semantic rule by using a newly input second data set. In this case, the knowledge base may be extended by using semantic inference of the newly input second data set.

The semantic rule may be either an RDF triple tree structure or an OWL structure.

The semantic rule engine 230 may receive the rule model from the rule modeler 240, modify the generated semantic rule, and change a method of converting the machine learning rule into the semantic rule.

The rule modeler server 630 includes a rule modeler for transmitting the rule model input by the domain expert to the semantic rule engine.

Rule Modeler provides a convenient user interface for domain experts to easily enter rule models. It provides a UI for easily creating machine learning rules and semantic rules, and also provides a UI that connects machine learning rules and semantic rules. The rule modeler also provides a user interface for viewing and modifying the relationship between machine learning rules and semantic rules generated by the machine learning engine before storing them in the knowledge base.

7 is a block diagram of an adaptive knowledge base building system according to another embodiment of the present invention.

Unlike FIG. 6, the newly input second dataset may be directly input into the machine learning engine instead of the semantic rule engine. It is also possible to include the machine learning module in the semantic rule engine and process it as shown in FIG. 6, but in order to shorten the machine learning process that normally requires a long time, the second data set is combined with the first data set in the machine learning engine. When inputted and processed at a time in the machine learning process, it is possible to efficiently use computer resources. The adaptive knowledge base building system according to the present invention can be built as a distributed processing system. In general, the machine learning engine is built as a parallel system on a plurality of servers, and the semantic rule engine is installed on a few servers. Inputting datasets into machine learning engines can distribute resources efficiently.

An adaptive knowledge base building system according to an embodiment of the present invention generates a learning model for a first dataset, analyzes an association of the first dataset using the generated learning model, and analyzes a second dataset. A machine learning engine for generating a semantic inference model for, inferring using a second dataset using the generated semantic inference model, and generating a prediction result; A rule generator for generating machine learning rules from the learned model results analyzed by the machine learning engine and generating machine learning rules from the inferred prediction results; And a semantic rule engine that converts the machine learning rules received from the rule generator into semantic rules and stores the converted semantic rules to build a knowledge base.

Preferably, the domain expert may further include a rule modeler for changing the semantic rule generation method.

Not only semantic rules, but also machine learning rules can be extended as needed, machine learning rules themselves are decisive, so machine learning rules need to be changed adaptively in environments where the actual environment is dynamically changing. There is a problem with changing the learning engine. Instead of extending the machine learning rules, the rule modeler changes the way that machine learning rules are translated into semantic rules, helping to choose the appropriate transformation method in a dynamically changing environment. The domain expert is not a person who knows the structure of the adaptive knowledge base construction system according to the present invention, but a person who has a thorough knowledge about the information on the semantic rules, so that the domain expert does not store the generated semantic rules directly in the knowledge base. It allows you to choose whether or not to adopt, and on this basis the rule modeler can work. The rule modeler determines how to convert machine learning rules into semantic rules through a separate learning process, and delivers them to the semantic rule engine. The semantic rule engine can build a knowledge base using the transformation method received from the rule modeler. .

The semantic rule engine builds a knowledge base based on the machine learning rules passed from the rule generator, the second dataset, and the rule models passed from the rule modeler. This process does not proceed immediately, but temporarily stores each data and builds a knowledge base at appropriate intervals. The semantic rule engine has a characteristic that a knowledge base constructed by all of an input order of a first data set, an input order of a second data set, and an order of inputting a rule model is different according to the characteristics of the present invention.

In reality, however, if the input data varies over time, the analysis is also common. For example, when operating a boiler in a house under an Internet of Thing (IoT) environment, the machine learning contents may be different according to the time when the boiler is operated. In other words, the adaptive knowledge base construction system according to the present invention has more robust characteristics in an environment for dynamically converting.

8 is a flowchart illustrating a method of building an adaptive knowledge base according to another embodiment of the present invention.

An adaptive knowledge base construction method, in accordance with some embodiments of the invention, includes generating a semantic inference model for a second dataset; Inferring the semantic inference model generated in the second dataset and generating a prediction result; Generating a machine learning rule from the inferred prediction result; Converting machine learning rules to semantic rules; And storing the transformed semantic rules to build a knowledge base.

In contrast to the description of FIG. 3, in case of FIG. 8, when a knowledge base analyzed by using the first data set as input data is already constructed, the semantic inference is performed by inputting the second data set into the machine learning engine instead of the semantic rule engine. There is a difference in that. When the second dataset is directly input to the semantic rule engine, semantic inference can be performed using the constructed knowledge base. However, since the machine learning engine generally requires a high-end server, the computing power of the machine learning engine is increased. Often it is better. In the case of inputting the second dataset into the machine learning engine, it is easy to apply the learning model including the second dataset, and there is an advantage of rapidly inferring the second dataset using a high specification server. The data used for the reasoning in advance can be confirmed by using the knowledge base built by being input back into the semantic rule engine and expanding the knowledge base.

In the above, the configuration of the present invention has been described in detail with reference to the accompanying drawings, which are merely examples, and those skilled in the art to which the present invention pertains have various modifications and changes within the scope of the technical idea of the present invention. Of course this is possible. Therefore, the protection scope of the present invention should not be limited to the above-described embodiment but should be defined by the description of the claims below.

100: computer system
110: processor
120: memory
123: ROM
126: RAM
130: data communication bus
140: storage
150: user input device
160: user output device
170: network interface
180: network
210: machine learning engine
220: rule generator
230: semantic rule engine
240: rule modeler
250: domain expert
260: first data set
270: second dataset
280: Knowledge Base
S310: Learning Model Generation Steps
S320: Generate trained model results
S330: generating machine learning rule;
S340: generating semantic rule;
S350: building a knowledge base;
S360: Semantic inference using the second dataset;
S370: expanding the knowledge base;
610: Machine learning rule generation server
620: semantic rule generation server
630: Rule Modeler server
S810: predictive model generation step;
S820: generating a prediction result;
S830: generating rules of the prediction result;
S840: Semantic Rule Conversion Step;
S850: expanding the knowledge base;

Claims (16)

A machine learning engine that generates a learning model based on the input first data set and analyzes an association relationship with the first data set according to the generated learning model;
A rule generator for generating a rule based on the learned model and algorithm; And
A semantic rule generator for generating semantic rules using the generated rules and the second data set;
Adaptive knowledge base building system comprising a.
The method of claim 1,
The semantic rule generator,
Generating rules by any of tree-based rule generation algorithms and apriori algorithms
Adaptive knowledge base building system.
The method of claim 1,
Further comprising a rule modeler for changing the semantic rule generation method by the domain expert
Adaptive knowledge base building system.
The method of claim 1,
The semantic rule generator,
Periodically expanding the knowledge base using temporarily stored machine learning rules, the second dataset, and rule models
Adaptive knowledge base building system.
In the adaptive knowledge base construction system,
A memory storing a program for providing an adaptive knowledge base building module; And
Including a processor for executing the program,
As the processor executes the program,
Generate a learning model for the input first dataset, output the correlation analysis result of the first data set according to the generated training model, and machine learning based on the learned model, the learned algorithm and the correlation analysis result. Generating a rule, and generating semantic rules using the generated machine learning rule
Adaptive knowledge base building system.
The method of claim 5,
The processor,
Generating semantic rules using rules generated by either RDF or OWL and the second data set
Adaptive knowledge base building system.
The method of claim 5,
The processor,
Generating a learning model and an inference model for the first dataset and the second dataset
Adaptive knowledge base building system.
The method of claim 5,
The processor,
Generating semantic rules using a rule model input by a domain expert
Adaptive knowledge base building system.
(1) machine learning an input first data set by a machine learning engine;
(2) the rule generator generating a rule based on the learned result;
(3) generating a semantic rule using the rule generated by the rule generator and the second data set; And
(4) constructing a knowledge base by storing the generated semantic rules by the semantic rule engine;
Adaptive knowledge base construction method comprising a.
The method of claim 9,
Providing, by a rule modeler, a model for generating a rule by a domain expert to generate the semantic rule;
Adaptive knowledge base construction method further comprising.
The method of claim 9,
Generating the semantic rule,
The rule generator generating a semantic rule using a rule generated by one of RDF and OWL and a second data set
Adaptive knowledge base construction method.
The method of claim 9,
Generating the semantic rule,
The rule generator generating a rule by any one of a tree-based rule generation algorithm and an apriori algorithm
Adaptive knowledge base construction method.
The method of claim 12,
The apriori algorithm,
The rule generator pruning the segmentation stage of each node according to the degree of support
How to build an adaptive knowledge base.
An adaptive knowledge base construction method for constructing a knowledge base using a machine learning engine and a semantic rule engine,
Generating a semantic inference model for the second dataset;
Inferring the semantic inference model generated in the second dataset and generating a prediction result;
Generating a machine learning rule from the inferred prediction result;
Converting machine learning rules to semantic rules; And
Storing the transformed semantic rules to expand the knowledge base;
Adaptive knowledge base construction method comprising a.
The method of claim 14,
Converting the machine learning rule into semantic rule,
Rules model provided by domain experts to determine how to translate machine learning rules into semantic rules
Adaptive knowledge base construction method.
The method of claim 14,
Converting the machine learning rule into semantic rule,
Generating semantic rules by either a tree structure model or an apriori algorithm
Adaptive knowledge base construction method.

Images