KR101473982B1 - Knowledge base generating apparatus and knowledge base generating method thereof - Google Patents

Abstract

The present invention relates to a knowledge base establishing apparatus and a method for constructing a knowledge base thereof, and more particularly, to a method and system for constructing a knowledge base, which comprises generating a plurality of unit knowledge bases containing only basic knowledge, independently performing inference, So that the base can be constructed.

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a knowledge base construction apparatus,

The present invention relates to a knowledge base construction technique, and more particularly, to a knowledge base construction apparatus based on an ontology and a method of constructing a knowledge base thereof.

The web search engine, which is a typical form of information retrieval, is based on a keyword-based retrieval technique. It collects documents on the web, indexes frequently occurring words in the document, and searches for words And finds the location of the included document.

Since the keyword-based information retrieval technology is a method of retrieving only documents including the same keyword as the keyword inputted by the user without considering the meaning of the keyword inputted by the user, the recall rate ) Is high, but the precision rate for finding a document that exactly matches the keyword entered by the user is inevitably low.

In order to solve such a problem, a semantic search technique as disclosed in, for example, Korean Patent Laid-Open No. 10-2011-0125872 (November 22, 2011) has emerged. The biggest difference between semantic-based search technology and keyword-based search technology can be found in the structure of the index.

Keyword-based retrieval technology extracts keywords (tokens) from documents through morphological analysis and stores them as indexes of corresponding documents. On the other hand, semantic-based retrieval technology extracts a semantic network from a document, stores it in a knowledge base in a graph structure, and queries the knowledge base in retrieval. Therefore, the search result may be the corresponding document, or it may be the entity described in the document.

Semantic - based retrieval technology has been developed in a wide variety of perspectives, but semantic - based retrieval techniques are reconstructed to represent information in a semantic network, stored in a knowledge base, and queryed.

Semantic networks are one of the ways of expressing knowledge expressed the relationship between concepts and concepts. Therefore, in semantic-based retrieval, relevant information (metadata) is extracted from a web document, and the extracted information (metadata, The process of constructing a knowledge base by expressing it in the form of a semantic network is very important.

In constructing a knowledge base in keyword - based retrieval technology, all knowledge, from extracted explicit metadata to implicit knowledge that can be derived logically, is directly generated and expressed in the knowledge base.

Korean Patent Laid-Open No. 10-2011-0125872 (November 22, 2011)

When constructing a knowledge base in semantic-based retrieval technology, only the extracted metadata corresponding to the basic knowledge can be expressed in the knowledge base, and the task will be simplified if the inference engine logically deduces the derivation knowledge that can be derived from the basic knowledge .

On the other hand, the time required for logical inference increases exponentially with the increase in the amount of basic knowledge on which inference is generally based. In other words, when the amount of basic knowledge doubles, the time required for reasoning increases by an exponent of 2.

Therefore, it is possible to shorten the time of building the final knowledge base by generating several knowledge bases containing only some basic knowledge, performing independent inference, and merging. Further, parallel processing of independent reasoning may provide additional benefits.

The present invention relates to a knowledge base construction apparatus and a knowledge base construction apparatus capable of constructing a plurality of unit knowledge bases including only a part of basic knowledge, independently performing inference, merging and building a final knowledge base, And to provide a method of constructing the same.

It is still another object of the present invention to provide a knowledge base construction apparatus and a knowledge base construction method capable of constructing a knowledge base more easily and easily by constructing a final knowledge base by performing inferences on several unit knowledge bases in parallel and then merging Method.

According to an aspect of the present invention, a knowledge base construction apparatus includes a basic knowledge extraction unit for extracting basic knowledge from web documents; A unit knowledge base generating unit for generating unit knowledge bases based on the basic knowledge extracted by the basic knowledge extracting unit; A unit knowledge base reasoning unit deducing derived knowledge related to each basic knowledge included in the unit knowledge bases generated by the unit knowledge base generating unit; A knowledge base constructing unit for constructing a final knowledge base by merging unit knowledge bases inferred from the derived knowledge related to each basic knowledge by the unit knowledge base reasoning unit; And the like.

According to a further aspect of the present invention, the knowledge base constructing apparatus deduces inferred knowledge related to each basic knowledge included in the unit knowledge bases in parallel to a plurality of unit knowledge bases.

The present invention has the effect of constructing a plurality of unit knowledge bases containing only some basic knowledge, performing independent inference, merging and building a final knowledge base, and building a knowledge base easily and quickly.

Furthermore, the present invention has the effect of constructing a knowledge base more easily by constructing a final knowledge base by performing inferences on a plurality of unit knowledge bases in parallel and merging them.

1 is a block diagram showing a configuration of an embodiment of a knowledge base establishing apparatus according to the present invention.
FIG. 2 is a diagram illustrating an example of a knowledge base ontology structure applied to a knowledge base construction apparatus according to the present invention.
3 is a flowchart showing a configuration of an embodiment of a knowledge base establishing method according to the present invention.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Reference will now be made in detail to embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein like reference numerals refer to the like elements throughout.

In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may make the subject matter of the present invention rather unclear.

The terms used throughout the specification of the present invention have been defined in consideration of the functions of the embodiments of the present invention and can be sufficiently modified according to the intentions and customs of the user or operator. It should be based on the contents of.

1 is a block diagram showing a configuration of an embodiment of a knowledge base establishing apparatus according to the present invention. 1, the knowledge base construction apparatus 100 according to this embodiment includes a basic knowledge extraction unit 110, a unit knowledge base generation unit 120, a unit knowledge base reasoning unit 130, And a knowledge base constructing unit 140.

The basic knowledge extraction unit 110 extracts basic knowledge from web documents. Web documents can be collected from the network. Here, explicit knowledge not induced by reasoning is called basic knowledge, and implicit knowledge derived by reasoning is called derived knowledge.

Recently, the Web Ontology Language (OWL) proposed by W3C as the most widely used standard for knowledge base construction consists of class, individual, and property.

A class is also referred to as a concept and is a representative term that can conceptually represent the same set of words. For example, the concept of 'computer' can be sub-concept 'pc', 'workstation', and so on.

An individual is also referred to as an instance (instnce), which is a term not included in the class. For example, 'person' is a class, but 'person name' is an object.

A property is also referred to as a relation and is used to describe the relationship between two terms.

Knowledge is expressed as a set of triples composed of subject, predicate, and object. Then, several triples are connected to form a semantic network.

For example, the knowledge that the actor 'AAA' has produced an unfair transaction in the film is known as' AAA (subject) - Predicate - Objection (object) ',' AAA - instanceOf Transaction - instanceOf (relationship) - movie ".

In a semantic search, when a user inputs a keyword search phrase "AAA directed movie", the search system searches for "AAA" and "direct" and "movie" classes through mapping between user keywords and knowledge base objects Search / extracts a partial semantic network (i.e., a set of triples) containing the instance, and provides an "unreasonable transaction" as a search result.

The knowledge base establishing apparatus 100 according to the present invention is for generating a knowledge base for the semantic based search. The knowledge base constructing apparatus 100 firstly extracts, from the web documents through the basic knowledge extracting unit 110, ) Extract basic knowledge of intellectuals.

A web document can describe various data about a certain object or an object such as a movie or a person using an HTML tag and can use a specific HTML tag for specific data in order to show it to a person at a glance.

For example, a web document providing information about a movie may include information about a movie title, a director, a star actor, a genre, a subject, and the like, and each information may include the following 'meta' tags .

<meta name = "Subject" content = "Enter homepage topic" />

<meta name = "Title" content = "Enter name of homepage" />

<meta name = "Description" content = "Enter comment" />

<meta name = "Keywords" content = "Enter Keyword" />

<meta name = "Author" content = "Created name" />

<meta name = "Publisher" content = "Organization / Company Name" />

<meta name = "Other Agent" content = "Web Officer Name" />

<meta name = "Classification" content = "Category Location (Classification)" />

<meta name = "Generator" content = "Generator (Editor)" />

<meta name = "Reply-To (Email)" content = "Enter email address" />

<meta name = "Filename" content = "Enter file name" />

<meta name = "Author-Date (Date)" content = "Date Created" />

<meta name = "Location" content = "Location" />

<meta name = "Distribution" content = "Distributor" />

<meta name = "Copyright" content = "Copyright" />

Accordingly, the basic knowledge extraction unit 110 can mechanically extract basic knowledge using a specific tag such as a 'meta' tag in a web document. The basic knowledge data extracted by the basic knowledge extraction unit 110 may be stored in a relational database in the form of a table.

If a basic knowledge is declared as a primitive class of a Web Ontology Language (OWL) and a derived knowledge is declared as a Defined Class of a Web Ontology Language (OWL) Assuming that the basic knowledge extraction unit 110 extracts properties defined in the basic knowledge and basic knowledge declared as a primitive class from the meta data of the web document,

For example, the basic knowledge extraction unit 110 may include a metadata extraction unit 111, a text analysis unit 112, and a basic knowledge storage unit 113.

The metadata extracting unit 111 extracts metadata from a web document stored in a web document storing DB storing a web document collected from a web site or the like.

The text analysis unit 112 analyzes the unstructured text from the metadata extracted by the metadata extraction unit 111 and extracts basic knowledge that is not an irregular text.

The basic knowledge storage unit 113 checks the redundancy of the basic knowledge extracted by the text analysis unit 112 and stores the non-redundant basic knowledge in the basic knowledge DB.

The unit knowledge base generation unit 120 generates unit knowledge bases based on the basic knowledge extracted by the basic knowledge extraction unit 110, and stores the unit knowledge bases in the unit knowledge base DB. In this case, the unit knowledge base refers to a knowledge base that includes only basic knowledge about some search objects, not the entire search object.

For example, the unit knowledge base generating unit 120 may be implemented to generate 100 unit knowledge bases each containing 10 basic knowledge about 10 web documents, for each of 1000 web documents.

For example, some unit knowledge base including information related to the movie "BBB" starring as actor "AAA" includes "AAA" as an instance of the "person" class according to the knowledge base ontology shown in FIG. 2 And basic knowledge that is an explicit knowledge such as "AAA - Leader - BBB" can be included using the base property "Leader".

From this explicit basic knowledge, it is deduced that the "AAA" derives a derivative knowledge that is an implicit knowledge that it is an instance of "movie actors" and "star actors". The unit knowledge base includes only basic knowledge of explicit knowledge, Knowledge derived knowledge is not included.

On the other hand, another unit knowledge base that includes information related to another movie 'CCC' starring actor 'AAA' may include basic knowledge such as 'AAA - Casting - CCC' .

In the future, these two unit knowledge bases are merged after inferential intellectual derivation knowledge is inferred through inference process, and all knowledge related to 'AAA' is integrated into one knowledge base, .

The unit knowledge base reasoning unit 130 deduces the derived knowledge associated with each basic knowledge included in the unit knowledge bases generated by the unit knowledge base generating unit 120. [ For example, "AAA" by the unit knowledge base reasoning unit 130 from an explicit knowledge base knowledge such as the above-described "AAA-Leading-BBB" is an instance of "Movie Actor", " , The implicit knowledge that is called the knowledge of knowledge.

For example, the basic knowledge, the derivation knowledge related to each basic knowledge and the properties for inference of the derived knowledge are defined, and the basic knowledge is declared as the primitive class of the Web Ontology Language (OWL) Assuming that the derived knowledge is declared as a Defined Class of the Web Ontology Language (OWL), the unit knowledge base inferring unit 130 may classify the attributes for the derived knowledge inference stored in the schema modeling information DB and to derive the derived knowledge related to each basic knowledge included in the unit knowledge bases by referring to the property information.

At this time, the unit knowledge base reasoning unit 130 may be implemented to represent the primitive class instances having the property value among the primitive class instances as instances of the primitive class, and deduce the derivative knowledge related to each basic knowledge have.

In this way, since the time required to induce a derivative knowledge through inference increases exponentially with the amount of basic knowledge, it is possible to divide the whole knowledge base into several smaller unit knowledge bases, This leads to a reduction in the linear reasoning time overall.

Meanwhile, when the unit knowledge base reasoning unit 130 implements the inference knowledge related to each basic knowledge included in the unit knowledge bases in parallel for inference on a plurality of unit knowledge bases, It can have an additional gain that can further improve.

The knowledge base establishing unit 140 constructs a final knowledge base by merging the unit knowledge bases inferred from the derived knowledge related to each basic knowledge by the unit knowledge base reasoning unit 130. [ The knowledge base construction unit 140 stores the final knowledge base in the knowledge base DB.

For example, if the knowledge base construction unit 140 constructs the unit knowledge bases based on URIs, in which all knowledge objects including class, instance, and property have a unique URI (Uniform Resource Identifier) And may be implemented to merge mechanically.

According to a further aspect of the present invention, the knowledge base establishing apparatus 100 may further include a knowledge base schema modeling unit 150. The knowledge base schema modeling unit 150 defines basic knowledge, properties for derivation knowledge and derived knowledge inference related to each basic knowledge, and generates basic knowledge, derived knowledge related to each basic knowledge, and derived knowledge The schema modeling information defining the property for inference is stored in the schema modeling information DB. The basic knowledge extraction unit 110 extracts the basic knowledge by referring to the schema modeling information stored in the schema modeling information DB.

For example, the knowledge base schema modeling unit 150 declares the basic knowledge as a primitive class of a Web Ontology Language (OWL), and the derived knowledge is defined as a Web Ontology Language (OWL) It can be implemented to be declared as a Defined Class.

According to a further aspect of the present invention, the knowledge base establishing apparatus 100 may further include a web document collecting unit 160. [ The web document collection unit 160 collects web documents from the network.

The present invention models a knowledge base schema (i.e., an ontology) by classifying a basic knowledge and a derived knowledge, generates a plurality of unit knowledge bases including basic knowledge using metadata extracted from a web document, We derive the derived knowledge about the basic knowledge through logical reasoning about the base, and merge them to generate the final knowledge base.

That is, according to the present invention, a plurality of unit knowledge bases having a small size are generated, and derivative knowledge is induced independently of each of the unit knowledge bases, and furthermore, parallel inference is applied, Can be built quickly.

In addition, since the inference engine, that is, the unit knowledge base reasoning unit 130 takes charge of the derivation of the complex derivative knowledge, the complexity of the web document collecting unit 160, which is responsible for the metadata migration, is greatly reduced.

The knowledge base building operation of the knowledge base building apparatus according to the present invention as described above will be described with reference to FIG. 3 is a flowchart showing a configuration of an embodiment of a knowledge base establishing method according to the present invention.

First, in the schema modeling step 310, the knowledge base building apparatus defines basic knowledge, and properties for deriving knowledge and derived knowledge inference related to each basic knowledge. The description of attributes for basic knowledge, derivative knowledge, and inferred knowledge inferences has been described earlier, so redundant explanations are omitted.

Next, in the web document collection step 320, the knowledge base construction apparatus collects web documents from the network. Since we have already explained about the collection of web documents, redundant explanation is omitted.

Next, in the basic knowledge extraction step 330, the knowledge base construction apparatus extracts basic knowledge from web documents. Since we have already explained the basic knowledge extraction, redundant explanation is omitted.

Then, in the unit knowledge base generating step 340, the knowledge base building apparatus generates unit knowledge bases based on the basic knowledge extracted by the basic knowledge extracting step 330. [ Since the unit knowledge base generation process has been described earlier, redundant description is omitted.

Then, in the unit knowledge base reasoning step 350, the knowledge base building device deduces derived knowledge related to each basic knowledge included in the unit knowledge bases generated by the unit knowledge base generating step 340. [ Since we have already explained the derivation of knowledge inference, redundant explanation is omitted.

Next, in the knowledge base construction step 360, the knowledge base construction apparatus constructs the final knowledge base by merging the unit knowledge bases inferred from the derived knowledge related to each basic knowledge by the unit knowledge base inference step 350. Since we have already explained the merging of unit knowledge bases, redundant explanation is omitted.

Therefore, according to the present invention, a plurality of unit knowledge bases including only some basic knowledge can be generated, independently inferred, and merged to construct a final knowledge base, thereby making it possible to construct a knowledge base easily and quickly.

Furthermore, the present invention can construct a knowledge base more easily by constructing a final knowledge base by performing inferences for a plurality of unit knowledge bases in parallel and merging them, Can be achieved.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it will be understood by those of ordinary skill in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. .

INDUSTRIAL APPLICABILITY The present invention is industrially applicable in the field of knowledge base construction technology and its application technology field.

100: knowledge base establishing apparatus 110: basic knowledge extracting unit
111: Metadata extraction unit 112: Text analysis unit
113: basic knowledge storage unit 120: unit knowledge base generation unit
130: unit knowledge base reasoning unit 140: knowledge base building unit
150: Schema modeling unit 160: Web document collecting unit

Claims (12)

A basic knowledge extraction unit for extracting basic knowledge from web documents;
A unit knowledge base generating unit for generating unit knowledge bases based on the basic knowledge extracted by the basic knowledge extracting unit;
A unit knowledge base reasoning unit deducing derived knowledge related to each basic knowledge included in the unit knowledge bases generated by the unit knowledge base generating unit;
A knowledge base constructing unit for constructing a final knowledge base by merging unit knowledge bases inferred from the derived knowledge related to each basic knowledge by the unit knowledge base reasoning unit;
Wherein the knowledge base construction apparatus comprises: The method according to claim 1,
Knowledge base building device:
A knowledge base schema modeling unit that defines basic properties, a class property for derivation knowledge related to each basic knowledge and inferred knowledge inference;
Further comprising the steps of: 3. The method of claim 2,
Unit knowledge base reasoning part:
And derives derived knowledge related to each basic knowledge included in the unit knowledge bases by referring to a class property for inference knowledge inference. 3. The method of claim 2,
Knowledge base schema modeling part:
The basic knowledge is defined as a primitive class of a Web Ontology Language (OWL), and the derived knowledge is declared as a Defined Class of a Web Ontology Language (OWL) And the knowledge base construction apparatus. 5. The method of claim 4,
Unit knowledge base reasoning part:
Wherein a primitive class instance having a specific class property value is represented as an instance of a primitive class to deduce derived knowledge related to each basic knowledge. 6. The method of claim 5,
Basic knowledge extraction unit:
Wherein a class property defined in a basic knowledge and basic knowledge declared as a primitive class is extracted from meta data of a web document. The method according to claim 1,
Unit knowledge base reasoning part:
Wherein inferring knowledge related to each basic knowledge included in unit knowledge bases is inferred in parallel to a plurality of unit knowledge bases inferring. The method according to claim 1,
Knowledge base building part:
It is possible to mechanically merge unit knowledge bases based on URIs in which all knowledge entities including class, instance, and class properties have a unique URI (Uniform Resource Identifier) And the knowledge base construction apparatus. 9. The method according to any one of claims 1 to 8,
Knowledge base building device:
A web document collection unit for collecting web documents from a network;
Further comprising the steps of: A basic knowledge extraction step of extracting basic knowledge from web documents;
A unit knowledge base generation step of generating unit knowledge bases based on the basic knowledge extracted by the basic knowledge extraction step;
A unit knowledge base inferring step of deriving the derived knowledge related to each basic knowledge included in the unit knowledge bases generated by the unit knowledge base generating step;
A knowledge base constructing step of constructing a final knowledge base by merging unit knowledge bases inferred from the derived knowledge related to each basic knowledge by unit knowledge base inference step;
Wherein the knowledge base comprises a plurality of knowledge bases. delete delete

Images