KR101289127B1 - Knowledge managing system - Google Patents

Abstract

A knowledge management system is provided. Knowledge management system according to the present invention, the input analysis unit for analyzing the input expressed in the ontology language and classifies the knowledge declaration and knowledge search; A knowledge base for storing the knowledge declaration if the input is a knowledge declaration as a result of the analysis of the input analysis unit; A query response unit searching for knowledge stored in the knowledge base and outputting a response to the input when the input is a knowledge search as a result of the analysis of the input analysis unit; And a knowledge inference unit that infers a relationship with the concepts stored in the knowledge base with respect to the input.

Knowledge Management, Knowledge Inference, Knowledge Base, Ontology, EOL

Description

Knowledge management system

1 is a block diagram showing the configuration of a knowledge management system according to the present invention;

2 is a diagram illustrating a grammar of an ontology language used in a knowledge management system.

3A and 3B show what is defined by the conventional OWL and the ontology language according to the invention for "airport", respectively.

4 illustrates the grammar used for knowledge retrieval and knowledge declaration.

FIG. 5 shows an example using the grammar of FIG. 4; FIG.

6 is a diagram illustrating an example of an operation performed in the developing unit and the normalizing unit of FIG. 1.

FIG. 7 illustrates an example of a result performed by the normalization unit of FIG. 1. FIG.

The present invention relates to a knowledge management system, and more particularly, to a knowledge management system for providing a simplified frame-based knowledge representation that provides an intuitive interface to a user in a ubiquitous computing environment and processing such knowledge representation.

As the existing web evolves into a semantic web that can be represented by ontology-based metadata, a method in which a person processes a large amount of information can be processed through an automated agent is considered. have.

As a means for representing the semantic web, RDF (Resource Description Language) is used as a standard, and OWL (Web Ontology Language) is currently used to provide a method for merging and inferring ontology by accepting vocabularies that cannot be expressed in RDF. The standard is in progress.

As research on approaches and processing methods for languages that can express ontology is needed, methods for accessing and processing documents expressed in ontology languages are generally RDF, DAML-OIL () using XML data model. DARPA Agent Markup Language-Ontology Inference Layer) expresses semantic information about OWL.

Ontology-based access and processing applications are important to enable semantic-based information access, and many applications of this approach are called automated information processing, information integration, or knowledge management.

Among the ontology-based applications, the logical-based approach is focused on the expressiveness of the model, which does not provide ease of integration with other systems and has a limitation in processing performance.

Due to these problems, successful planning applications based on ontology technologies are not available.

In order to express ontology as an abstracted model, the main problem is the representation and processing of all vocabularies supported by different ontology languages (RDF, DAML-OIL, OWL, etc.).

In order to express conventional ontology as an abstracted model, various methods have been used because one-to-one mapping between words that are dependent on different ontology languages and abstracted models is difficult.

In particular, the model for accessing and expressing the ontology is approached in the form of expressing and extending the basic model using the RDF model, or the models dependent on the ontology language characteristics are applied differently.

In this way, the conventional ontology representation method uses different ontology models to represent semantic information provided by the current semantic web, and reconfigures and approaches duplicate vocabulary according to the model whenever a new ontology language appears. There is also the problem of an inefficient approach that must be created.

As mentioned above, the language most commonly used for expressing ontology and known to the general public is OWL. As OWL is a language proposed by the W3C, it has a wide range of expression and many inference engines using ontologies expressed in OWL have been developed. Not surprisingly, most newly developed ontology inference engines also support OWL. The problem, however, is that the representation of OWL is based on XML. This means that it is difficult for a real user to maintain an intuitive viewpoint in the process of creating an ontology. When you run Internet Explorer right away and go to any site and choose to view the source, the feast of text that appears on the screen is not dizzying, even if you are familiar with the html language. On the other hand, when the ontology, which is the result of formal knowledge expression, is described in XML format and its capacity is several tens of pages, it is not easy to read the contents intuitively and to add new knowledge. It's not easy. Of course, the editor is often provided to solve this problem, but the editor converts the content through the WYSWYG-style website development tool, and the source code is often messed up. Rather than expressing precisely the intention of a, there is a limit to expressing the knowledge within the range provided by the editor.

An object of the present invention is to provide a simplified frame-based knowledge representation that provides an intuitive interface to a user in a ubiquitous computing environment, and to provide a knowledge management system for processing such knowledge representation.

According to an aspect of the present invention, there is provided a knowledge management system comprising: an input analyzer configured to interpret an input expressed in an ontology language and classify it into a knowledge declaration and a knowledge search; A knowledge base for storing the knowledge declaration if the input is a knowledge declaration as a result of the analysis of the input analysis unit; A query response unit searching for knowledge stored in the knowledge base and outputting a response to the input when the input is a knowledge search as a result of the analysis of the input analysis unit; And a knowledge reasoning unit for inferring a relationship with concepts stored in the knowledge base with respect to the input.

The ontology language, at least one of the atomic concept, the highest concept, the lowest concept, the opposite concept of the existing concept, the logical product between the concepts, the role, the concept of satisfying the number restriction of the role, at least one for a specific role Concepts can be defined, oppositions to complex concepts of multiple concepts, and ORs between concepts.

The knowledge inference unit may include: a deployment unit that develops an input concept into an atomic concept; A normalization unit for deleting and normalizing overlapping concepts among the expanded atomic concepts; And a collision check unit that checks whether a normalized concept collides with a concept stored in the knowledge base.

The normalization unit converts the structure of the developed atomic concept into NNF (Negation Normal Form) by using Demorgan's law, and converts the logical product in the structure of the atomic concept converted into NNF to the NNF-converted atomic concept. Place it below the OR in the structure.

The collision checking unit checks whether a subordinate concept corresponding to a logical product located below the structure of the atomic concept converted into the NNF exists in the knowledge base.

The knowledge base consists of terminological boxes (Tboxes) that describe the relationships between concepts.

In order to fully understand the present invention, operational advantages of the present invention, and objects achieved by the practice of the present invention, reference should be made to the accompanying drawings and the accompanying drawings which illustrate preferred embodiments of the present invention.

Hereinafter, the present invention will be described in detail with reference to the preferred embodiments of the present invention with reference to the accompanying drawings. In each drawing, the same reference numerals denote the same members.

1 is a block diagram showing the configuration of a knowledge management system according to the present invention.

Referring to FIG. 1, the knowledge management system 100 includes an input interpreter 110, a knowledge base 120, a question answering unit 130, and a knowledge reasoning unit 140.

The input analyzer 110 interprets the input expressed in the ontology language from the user. Here, the ontology language used is an intuitive frame-based knowledge ontology expression language, hereinafter referred to as Epistemological Ontology Language (EOL).

The knowledge representation of the EOL according to the present invention supports the range of ALCUN at the DL (Description Logic) definition level, and the knowledge representation that can be expressed as the range of ALCUN is as follows.

first. You can define an atomic concept.

Second, we can define a universal concept.

Third, the bottom concept can be defined.

Fourth, it is possible to define a negation concept of the existing concept.

Fifth, we can define the intersection between the concepts.

Sixth, the definition of a role and the concept of satisfying the number restriction of the role can be defined.

Seventh, one can define the concept of accepting at least one or more for a particular role.

Eighth, it is possible to define opposition to complex concepts of different concepts.

Ninth, we can define a union between existing concepts.

The grammar of the EOL that enables the above expression of knowledge is shown in FIG. 2.

An example of a simple expression of knowledge based on the grammar shown in FIG. 2 is as follows.

Carnivore :: = (Animal (all eat Meat))

This indicates that Carnivore is animal and must eat meat.

Mother :: = (and Woman Parent)

This indicates that Mother is a woman and a parent.

Creature :: = (or (or Animal Plant) Microbe)

This indicates that the creature is an animal or a plant or a microbe.

(not (Person (all hasChild Person)))

This represents the concept of a person who is not a person or who is human but does not have children.

Such expression of knowledge by EOL can be intuitively expressed. For example, when defining "airport", the case of using OWL is illustrated in FIG. 3A, and the case of using EOL is illustrated in FIG. 3B. As shown in the figure, EOL is an intuitive language based on natural language in order to supplement the difficulty of using existing ontology expression languages, and helps general users to define and express knowledge more easily and quickly.

EOL according to the present invention has a representation range of ALCUN. In general, the wider the scope of expression of language, the more complicated the method of reasoning through the knowledge of expression tends to be. The application domain targeted by the knowledge management system according to the present invention is an efficient knowledge management target for low-end devices existing in a ubiquitous computing environment, and thus does not require a high level of knowledge representation. With EOL, users can manage their knowledge so that devices operating in the domain of their choice can perform intelligent operations with the ALCUN representation.

The grammar used for knowledge declaration and knowledge search is shown in FIG. 4. As shown in FIG. 4, the commands used in the present invention are similar to the commands used in the existing inference engine, but it can be seen that the commands are closer to natural language to facilitate intuitive understanding and use. The knowledge definition part following the command follows the grammar defined in EOL (see Figure 2). An example using the grammar shown in FIG. 4 is shown in FIG. 5.

The knowledge base 120 stores the knowledge declaration when the input from the user is a knowledge declaration as a result of the analysis of the input analysis unit 110.

Typically, the knowledge base is divided into Abox (Assertion box) and Tbox (Terminological box). Tbox contains the concept hierarchy, that is, the sentences describing the relationships between the concepts. Abox contains sentences that describe the state of the hierarchy to which individuals belong, ie the relationship between individuals and concepts. For example, the sentence "Every woman is an animal" belongs to Tbox, but the sentence "Sun is a woman" belongs to Abox.

Since the knowledge management system according to the present invention processes only the knowledge representation corresponding to the Tbox, the knowledge base 120 is composed of the Tbox.

The query response unit 130 analyzes the query content INQ transmitted from the input analysis unit 110, extracts a result corresponding to the query word from the knowledge base 120, and outputs an output signal OUTPUT. The question answering unit 130 processes a response to the attributes of a specific concept, concepts having an inclusion relationship among the concepts, and results of the relationships between the concepts.

The knowledge inference unit 140 infers a relationship between concepts stored in the knowledge base 120 with respect to a knowledge declaration or a knowledge search input from a user.

The domain knowledge conceived by the user is stored in the knowledge base 120 through EOL-Lite. However, the user cannot define and declare all knowledge. Whether new knowledge concepts conflict with existing knowledge, or whether there is no undefined, implicit knowledge relationship, is the content that must be automatically inferred as a knowledge management system. The knowledge inference unit 120 uses an extended structural inclusion algorithm to perform an implicit relationship between the knowledge expressed in EOL-Lite and detection of a contradiction relationship between the knowledge. This algorithm performs inference by normalizing and comparing the structure of knowledge. To this end, the knowledge inference unit 140 includes a deployment unit 142, a normalization unit 144, and a collision inspection unit 146.

The deployment unit 142 performs a process of changing the concepts used to define the knowledge into atomic concepts. If the concept itself is an atomic concept, there is no need to expand it, but in the case of a compound concept, the components are all replaced with the atomic concept. Therefore, if there is a component in the definition of a compound concept that is not an atomic concept, the definition of the component is searched and replaced. Because of the binary tree structure, the entire binary tree structure is not broken even if each element is replaced with another binary tree structure. Steps are performed recursively until all components are searched for, replaced, inspected, and all components become atomic concepts.

For example, if the concept of "mother" consists of the concept of "woman" and the role of "one or more children," then the concept of "self" and "person" and "one or more" people Perform the task of changing to.

The normalizer 144 normalizes the relationship of overlapping concepts occurring as a result of the deployment of the deployment unit 142. For example, when the concept of A is represented by B ?? C and the concept of C is represented by B ?? D, when A is unfolded, A is represented by B ?? (B ?? D). do. In this case, since B is duplicated, the normalization unit 144 normalizes the deletion.

As a result of the deployment of the deployment unit 142, a plurality of substitutions occur, the structure of the concept is complicated, there may be a duplicated component, or there may be a component that will occur. Therefore, the normalization unit 144 converts the structure of the concept into NNF (Negation Normal Form) using Demorgan's law. NNF represents a formula in which logical elements are negated only with their basic elements. Demorgan's law is a formula that allows the negative sign to be placed only before the atomic concept.

6 shows an example of operations performed in the deployer and normalizer. In Figure 6, it develops from (a) to (b) and is normalized to (d) via (b) to (c).

In addition, the normalization unit 144 transfers the OR operation in the concept structure in the NNF format to the upper level of the AND operation in the concept structure in the NNF format. The reason for this is to facilitate comparison between concepts in the collision inspection performed by the collision inspection unit 146. If the logical sum is located above the binary tree, the partial binary tree, which is only the logical product, is bounded by the logical sum. This example is illustrated in FIG. 7.

The collision check unit 146 checks whether the normalized concept collides with the concept stored in the knowledge base 120. In other words, the collision check unit 146 verifies whether the normalized concept exists in the knowledge base 120.

In the embodiment of the present invention, the collision check unit 146 checks whether or not the sub-concept corresponding to the logical product located below the structure of the concept in the NNF format exists in the knowledge base 120. In the binary concept tree shown in FIG. 7, it is considered to be divided into several subtrees connected only by logical products. The reason for this is that since the bundles connected by logical products marked 'and group' are connected by logical sum, the concept exists even if there is a concept including only one of the bundles among the concepts in the knowledge base 120. There is no process of comparing objects like the Tableau algorithm does. This comparison is equivalent to comparing existing 'and Group's to existing concepts in a normalized format.

The extended structural inclusion algorithm performed in the knowledge inference unit 140 combines the advantages of the structural inclusion algorithm using only TBox and the concept substitution method of the Tableau algorithm that can infer a higher expression power than the structural inclusion algorithm. . In order to know if this algorithm is practical, the completeness and soundness of the algorithm must be verified. The key to this algorithm is the normalization and comparison of binary concept trees, and since there is no object comparison process, which is a factor in improving time complexity in the Tableau algorithm, all work does not deviate from the already proven binary tree operations. Thus, this algorithm is also complete. And the robustness proceeded by comparing logical process with actual execution result. The logical process can be verified because the Tableau algorithm does not perform ABox comparisons, or object comparisons, but instead implements the process of substituting and comparing concepts through their own binary tree operations.

Although a preferred embodiment of the present invention has been described in detail above, those skilled in the art to which the present invention pertains can make various changes without departing from the spirit and scope of the invention as defined in the appended claims. It will be appreciated that modifications or variations may be made. Accordingly, modifications of the embodiments of the present invention will not depart from the scope of the present invention.

According to the present invention, the scope of the knowledge expression managed by EOL is ALCUN, which is weaker than that of the existing ontology language, but in general, the low-level devices operating in the ubiquitous computing environment use the knowledge used for intelligent behavior. There is no shortage of expression. Accordingly, the knowledge management system according to the present invention using a fast algorithm that compares the structure of concepts in a limited domain called ALCUN to find a relationship can be efficiently applied and used for knowledge management of low specification devices in a ubiquitous computing environment.

In addition, by providing a simplified frame-based knowledge representation and processing engine that provides an intuitive interface to the user, it provides a specialized knowledge of the limited computing capabilities of devices in the ubiquitous computing environment where general users can freely receive computing services. Optimal effects can be expected when applied to fields that can be modified quickly and conveniently, such as intelligent cameras and intelligent TVs in ubiquitous computing environments.

Claims (6)

An input interpreter that interprets the input expressed in the ontology language and classifies it into a knowledge declaration and a knowledge search; A knowledge base for storing the knowledge declaration if the input is a knowledge declaration as a result of the analysis of the input analysis unit; A query response unit searching for knowledge stored in the knowledge base and outputting a response to the input when the input is a knowledge search as a result of the analysis of the input analysis unit; And A knowledge reasoning unit for inferring a relationship with concepts stored in the knowledge base for the input; The knowledge reasoning unit, An expansion unit for expanding the input concept into an atomic concept; A normalization unit for deleting and normalizing overlapping concepts among the expanded atomic concepts; And And a collision check unit for checking whether a normalized concept collides with a concept stored in the knowledge base. The method of claim 1, wherein the ontology language is, Atomic concepts, top concepts, lowest concepts, opposite concepts, logical products between concepts, roles, concepts that satisfy the number restriction of those roles, concepts that accept at least one or more for a particular role, A knowledge management system, characterized by defining oppositions to a complex concept of a plurality of concepts, and a logical sum between the concepts. delete The method of claim 1, wherein the normalization unit, The structure of the developed atomic concept is transformed into NNF (Negation Normal Form) using Demorgan's law, and the logical product in the structure of the atomic concept converted into NNF is the logical sum in the structure of atomic concept converted into NNF. Knowledge management system, characterized in that located below. The method of claim 4, wherein the collision detection unit, And checking whether or not the subordinate concept corresponding to the logical product located under the structure of the atomic concept converted into the NNF exists in the knowledge base. The method of claim 1, wherein the knowledge base, Knowledge management system, consisting of a terminological box (Tbox) describing the relationships between concepts.

Images