KR101521862B1 - System and method for classifying patent document - Google Patents

Abstract

The present invention relates to a system for classifying a patent document and a method thereof which comprise: a concept element extract portion for extracting a concept element by analyzing each patent document, and assigning a function attribute to each concept element; a concept structure generation portion for calculating a similarity value of the concept element by grouping the concept element of each patent document by a function attribute, and generating a concept structure including the calculated similarity value of the concept element for each patent document; and a classification portion for obtaining the similarity value of the concept structure between patent documents and classifying the patent document based on the similarity value of the obtained concept structure.

Description

[0001] SYSTEM AND METHOD FOR CLASSIFYING PATENT DOCUMENT [0002]

The present invention relates to a system and method for classifying a patent document, and more particularly, to a system and method for classifying a patent document according to mutual semantic relations between patent documents, And more particularly, to a patent document classification system for classifying patent documents.

With the conclusion of the Korea-US Free Trade Agreement (FTA), patent protection is strengthened by extending the right to intellectual property rights consisting of patents, trademarks, and copyrights, so that interest in patent information directly connected to national industrial competitiveness is demanded more than ever have.

Patent information refers to information related to industrial property rights, including information on patented technology contents and rights, personal information such as applicants and inventors, and other bibliographic information. As the industry becomes more sophisticated, complex, and diversified, a huge amount of patent information is pouring in. In order to survive in a changing industrial society, this information must be properly reflected in corporate management strategies.

Today, patent documents not only formally grant the intellectual property rights of the inventors, but also in the global infinite technology competition environment, the patent DB supports research and development planning of enterprises and establishment of national science and technology policy, Is one of the essential information resources of the world. In addition, as in other fields, the size of patent documents is continuously increasing worldwide, including emerging countries such as China, and there is a growing demand for information levels to be gained through the emergence of new big data analysis. Therefore, accessing demand information quickly and accurately and securing a more in-depth analysis environment is a particularly important task.

In general, the research and analysis of the patent database is conducted by selecting the subject of analysis after reviewing the documents through the selection of the subject word, the composition of the search formula, and the use of the classification code, and then examining the details such as abstract, drawings and claims. Once experience is accumulated and mastery is achieved, it is possible to obtain a certain level of quality based on know-how. However, it is mainly effective when analyzing individual subjects, and it is inevitable Inadequate results, that is, noise removal work, still require a significant amount of intellectual effort and time cost by experts, and technological supplementation and improvement are necessary.

In addition, since the number of classification is significant, and it is necessary to understand the detailed description including the claim for accurate classification, it also gives a considerable stress to the classifier.

Accordingly, there is a need for a method for automatically and accurately classifying patent documents.

Korean Patent No. 1,179,613, entitled " Automatic Classification of Patent Documents Using Frequent Items and Association Rules "

An object of the present invention is to provide a patent document classification system and method for extracting a conceptual structure by analyzing a patent document and classifying the patent document according to mutual semantic relations between patent documents that are obtained through functional attribute analysis and similarity measurement .

According to an aspect of the present invention, there is provided an information processing system including a concept element extracting unit for analyzing each patent document to extract conceptual elements and assigning a function attribute to each conceptual element, A conceptual structure generating unit for generating a conceptual element similarity value by clustering and generating a conceptual structure including the calculated conceptual element similarity value for each patent document, a conceptual structure similarity value between patent documents is obtained, There is provided a patent document classification system including a classification section for classifying a patent document based on the classification result.

The patent document classification system may further include a database of entity names in which the entity names of products and technologies are stored, and a functional property classification database in which rules for classifying functional attributes of entities are set.

The conceptual element extracting unit includes a candidate sentence identification module for identifying a candidate sentence in a predefined area of the patent document with reference to the object name dictionary database, a character string representing a concept element through syntax analysis based on dependency grammar in the identified candidate sentence, And a function attribute assigning module for assigning a function attribute to the extracted character string by referring to the functional classification database, wherein the extracted character string may be a conceptual element.

The conceptual structure generation unit may calculate a similarity value between conceptual element entities for each functional attribute by using a conceptual element clustering module for clustering conceptual elements to which the same functional attributes are assigned in units of patent documents, And a conceptual structure generation module for generating a conceptual structure including similarity calculation module, functional attribute, conceptual elements, and similarity values of conceptual elements in units of patent documents.

The classification unit includes a concept element similarity degree calculating module that obtains the similarity value of the conceptual elements constituting the conceptual structure between the patent documents, a functional attribute similarity degree calculating module that obtains the functional attribute similarity value between the patent documents, A conceptual structure similarity calculating module for obtaining the conceptual structure similarity value between patent documents using the similarity value, and a classification module for classifying the patent document based on the conceptual structure similarity value between the patent documents.

According to another aspect of the present invention, there is provided a method for analyzing a patent document, comprising the steps of: (a) analyzing each patent document to extract conceptual elements and assigning functional attributes to the conceptual elements; (b) (C) obtaining a concept structure similarity value between patent documents, and calculating a similarity value between the patent documents based on the obtained concept structure similarity value, There is provided a method of classifying a patent document in a patent document classifying system including classifying documents.

The step (a) includes the steps of: identifying a candidate sentence in a predefined area of a patent document with reference to an entity name dictionary database; extracting a character string representing a conceptual element through parsing based on dependency grammar in the identified candidate sentence; And providing a functional attribute to the extracted character string by referring to the functional classification database, wherein the extracted character string may be a conceptual element.

The step (b) includes the steps of: grouping conceptual elements having the same function attributes in units of patent documents, calculating similarity values between conceptual elements in each functional attribute using a predefined similarity calculation model; , A function attribute, conceptual elements, and a similarity value of conceptual elements, in units of patent documents.

The step (c) includes the steps of: obtaining a similarity value of conceptual elements constituting the conceptual structure between patent documents; obtaining a functional property similarity value between patent documents; and using the similarity value or the functional property similarity value of the obtained conceptual elements Obtaining a concept similarity value between the patent documents, and classifying the patent document based on the concept structure similarity value between the patent documents.

According to still another aspect of the present invention, there is provided a system for classifying patent documents, comprising: (a) analyzing each patent document to extract conceptual elements and assigning functional attributes to the respective conceptual elements; (b) (C) generating a conceptual structure similarity value between the patent documents; (c) comparing the conceptual element similarity value between the patent documents, There is provided a computer-readable recording medium containing a program for executing a patent document classification method including classifying a patent document based on the obtained concept structure similarity value.

According to the present invention, a conceptual structure is identified through text mining based on a natural language processing on a patent document, and a patent is classified by analyzing a relation between patent concept structures through functional property and similarity analysis of conceptual elements. It is possible to minimize noise inevitably generated from a one-dimensional approach centered on the keyword (keyword), to reduce intellectual effort and time cost, and to perform in-depth analysis efficiently from a subject point of view.

1 illustrates a system for classifying patent documents according to an embodiment of the present invention.
FIG. 2 is a block diagram schematically showing a configuration of a patent document classification system according to an embodiment of the present invention. FIG.
3 is a diagram showing a configuration of a conceptual element extracting unit shown in FIG.
4 is a diagram showing the configuration of the conceptual structure generation unit shown in FIG.
5 is a view showing a configuration of the classification unit shown in Fig.
6 is a flowchart illustrating a method of classifying a patent document according to an embodiment of the present invention.
FIG. 7 is a flowchart illustrating a method for extracting conceptual elements and functional attributes from a patent document according to an embodiment of the present invention; FIG.
8 is a flow diagram illustrating a method for generating a conceptual structure according to an embodiment of the present invention.

The foregoing and other objects, features, and advantages of the present invention will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which: FIG.

Hereinafter, a patent document classification system and method according to the present invention will be described in detail with reference to the accompanying drawings. The embodiments are provided so that those skilled in the art can easily understand the technical spirit of the present invention, and thus the present invention is not limited thereto. In addition, the matters described in the attached drawings may be different from those actually implemented by the schematic drawings to easily describe the embodiments of the present invention.

In the meantime, each constituent unit described below is only an example for implementing the present invention. Thus, in other implementations of the present invention, other components may be used without departing from the spirit and scope of the present invention. In addition, each component may be implemented solely by hardware or software configuration, but may be implemented by a combination of various hardware and software configurations performing the same function. Also, two or more components may be implemented together by one hardware or software.

Also, the expression " comprising " is intended to merely denote that such elements are present as an expression of " open ", and should not be understood to exclude additional elements.

1 illustrates a system for classifying patent documents according to an embodiment of the present invention.

Referring to FIG. 1, a patent document server 100 and a patent document classification system 200 are connected to each other through a communication network.

The patent providing server 100 provides a patent document to the patent document classification system 200, including a patent database storing patent documents and the like.

The patent document classification system 200 collects patent documents from the patent providing server 100 and analyzes the collected patent documents to classify the patent documents.

That is, the patent document classification system 200 analyzes each patent document collected from the patent providing server 100 to extract conceptual elements, assigns functional attributes to the respective conceptual elements, A similarity value is calculated, and a conceptual structure including the calculated concept element similarity value is generated for each patent document. Then, the patent document classification system 200 obtains the concept structure similarity value between the patent documents, and classifies the patent document based on the concept structure similarity value. Here, the conceptual structure is a set of minimum entities to be established as a patent document, and the minimum entity includes an inventor's own solution to the problem to be solved, that is, a technology, a target object to which the technology is applied, that is, a product. The functional attributes of technology and product are function-oriented search model (Litvin) based on Altshuler (1946) methodology of TRIZ (a problem-solving, analysis and forecasting tool) , 2004)), and defines the combination of the attributes of the action and the object.

The patent document classification system 200 can be implemented through an electronic device capable of communicating with other electronic devices through various communication standards and performing various data processing operations. For example, the patent document classification system 200 may be implemented in the form of a server device, and may be implemented in various electronic devices other than the form of the server device. Also, the patent document classification system 200 may be implemented in the form of a single electronic device, or in a form in which two or more electronic devices are combined.

The patent document classification system 200 will be described in detail with reference to FIG.

Although the patent providing server 100 providing a patent document is described as being outside the patent document classifying system 200, the patent document classifying system 200 may have a database in which patent documents are stored.

FIG. 2 is a block diagram schematically showing a configuration of a patent document classification system according to an embodiment of the present invention. FIG. 3 is a diagram showing the configuration of the conceptual element extraction unit shown in FIG. FIG. 5 is a diagram showing the configuration of the classification unit shown in FIG. 2. FIG.

2, the patent document classification system 200 includes a database 210, an interface unit 220, a conceptual element extraction unit 230, a conceptual structure generation unit 240, a classification unit 250, a control unit 260 ).

The database 210 includes an object name dictionary database 212, a functional property classification database 214, and a conceptual structure database 216.

The entity name dictionary database 212 stores entity names for products and technologies.

The entity name dictionary database 212 may store string entities and attributes identified and extracted from a trademark database, an external disclosure such as wordnet, or a commercial terminology database.

The entity name dictionary database 212 may register, modify, or delete a new entity string based on the text mining result and the entity name identification result value.

The functional attribute classification database 214 stores rules for classifying functional attributes of entities.

The functional attribute classification database 214 includes functional classification matrices composed of TRIZ-based actions and objects (instances) of the objects, and entity data constructed by collecting or identifying the types of the matrices. Here, the action may include move, add, remove, hold, deflect, change, inclusion, operation, etc., Objects may include materials, properties, information / concepts, products, and the like.

The functional attribution classification database 214 includes functional classification matrices as shown in Table 1.

Substance Field Information / Concept (info) Product (artifact) Move m-s m-f m-i m-a Add a-s a-f a-i a-a Remove r-s r-f r-i r-a Hold h-s f-f h-i h-a Deflect d-s d-f d-i d-a Change c-s c-f c-i c-a Configuration( inclusion ) I-s I-f I-i I-a Execution( operation ) o-s o-f o-i o-p

The action-object-based functional classification matrix is based on the TRIZ function oriented search, but can be refined or extended according to the required analysis level, such as the classification matrix.

The concept structure database 216 stores information on the concept structure generated by the concept structure generation unit 240. That is, the conceptual structure analysis database 216 stores conceptual elements (technology, product), conceptual element similarity values, conceptual structure, conceptual structure similarity matrix, and the like.

The interface unit 220 collects the patent document from the patent providing server through the communication network.

The interface unit 220 receives an entity type, a function attribute type, a similarity value, and the like, and retrieves a patent document satisfying the condition.

The conceptual element extracting unit 230 extracts conceptual elements by analyzing each patent document, and assigns functional attributes to each conceptual element. Here, since the conceptual element includes the product and the technology, the conceptual element extraction unit 230 can identify the object for the product and the technology using the pattern analysis of the conceptual element expression from the patent document. At this time, the conceptual element extracting unit 230 can extract the conceptual element from the patent document using the natural language processing method, the mechanical algorithm processing method, and the like.

Meanwhile, the conceptual element extracting unit 230 extracts a threshold value for an object name collection execution count, a number of processed document units, a discrimination rate per number of processed document units and a target document area (e.g., title, abstract, The detailed description, the drawings, the claims), and the like. In this case, the conceptual element extracting unit 230 selects an arbitrary document set for the specified number of documents according to the environment variable setting information, and loads the specified document area. Then, the conceptual element extracting unit 230 identifies or extracts the entity name from the specified document region by loading the entity name dictionary and the pattern recognition rule defined for each entity type. The conceptual element extracting unit 230 executes the object name collecting process for the designated number of times by referring to the number of documents and the discrimination rate that are not identified or recognized, and stops when the minimum discrimination rate is not satisfied. The administrator can query the unidentified document, register a new object name or recognition rule, and re-execute the document set that does not meet the minimum recognition rate.

Referring to FIG. 3, the conceptual element extracting unit 230 includes a candidate sentence identifying module 232, a character extracting module 234, and a function attribute assigning module 236.

The candidate sentence identification module 232 references the entity name dictionary database 212 to identify candidate sentences in predefined areas of the patent document.

The candidate sentence identification module 232 identifies the candidate sentence by referring to the object name dictionary database 212 for the designated document area. For example, a candidate sentence including a character string registered in the object name dictionary is analyzed by analyzing the title, abstract, detailed description, first paragraph of the claim, and the like. At this time, the candidate sentence identification module 232 may identify the candidate sentence by designating the whole or an arbitrary range according to the document structure.

The candidate sentence identification module 232 can sequentially identify the candidate sentence by loading the patent document from the patent providing server based on the preset environment variable setting information. At this time, you can specify a new environment variable value. Then, the candidate sentence identification module 232 identifies the entity name in the document area designated according to the environment parameter setting information or the recognized sentence as the candidate sentence.

The string extracting module 234 extracts a character string representing a concept element from the candidate sentence identified by the candidate sentence identifying module 232 through parsing based on the dependency grammar. At this time, the extracted character string may be a conceptual element, and the character string extraction module 234 may extract a character string using various methods such as text mining, natural language processing, and morphological analysis. That is, the string extracting module 234 identifies the independent noun phrase of the main sentence by referring to the entity name dictionary database 212 and the stop word dictionary (not shown) for identification and recognition of the concept element (entity name) Identify or recognize entity names through analysis. At this time, syntax analysis can utilize open source such as Stanford Parser which supports dependency analysis.

That is, the string extracting module 234 analyzes the syntactic structure based on the dependent grammar of the candidate sentence to identify the noun phrase of the order chapter and the head word of the noun phrase or recognize the entity name. For example, in the case of a descriptive noun phrase such as 'anti-piracy system for protecting distributed software applications from unauthorized use', 'anti-piracy system' can be identified as a product by analyzing syntax based on dependency grammar as follows.

(ROOT (NP ( NP ( JJ anti - piracy ) ( NN system)) (PP (IN for ) (S (VP (VBG protecting) (NP (VBN distributed) (NN software) (NNS applications)) (PP (IN from) (NP (JJ unauthorized) (NN use))) )))))))

As another example, in the case of simple nouns such as' solid polymer electrolyte membrane ',' ROOT (NP ( NP ( JJ solid ) ( NN polymer)) (NP (NN electrolyte ) (NN membrane )) )) can be identified as an individual.

The function attribute assigning module 236 refers to the function attribute classification database 214 and gives a function attribute to the string extracted by the string extracting module 234. [

The function attribute assignment module 236 analyzes the syntax structure with reference to the function attribute classification database 214 and identifies the function attribute of the corresponding object.

For example, in the case of 'a system for reporting ( add ) security information (information) relating to a mobile device', the following syntax analysis result based on dependence grammar can be obtained.

(ROOT (NP (NP (DT a) (NN system)) (PP (IN for) (S (VP (VBG reporting ) (NP ( NP ( NN security ) ( NN information)) (VP (VBG relating ) (PP (TO to) (NP (DT a) (JJ mobile) (NN device)))))))) (..)))

In this case, the function attribute assigning module 236 refers to the functional attribute classification matrix to provide 'reporting' and 'security information' respectively, and 'reporting security information' as the information type, Attribute and can assign a functional attribute to the type of provision-information (ai).

The function attribute assignment module 236 stores the object string and the function attribute in the concept structure database 216 together with the corresponding patent document information.

The conceptual structure generation unit 240 generates conceptual element similarity values by grouping the conceptual elements of each patent document by functional attributes, and generates conceptual structures including the calculated conceptual element similarity values for each patent document. That is, the conceptual structure generation unit 240 calculates the semantic similarities between the entities extracted by the conceptual element extraction unit 230, and calculates a similarity matrix between conceptual structures for patent documents including entities having the same functional attributes . In other words, the conceptual structure generation unit 240 classifies product type entities or technology type entities by type, and measures the similarity according to the same function attribute type in the action-object classification matrix.

The conceptual structure generation unit 240 selects a calculation model for object similarity measurement. At this time, the environment variable for the similarity threshold can be specified and changed.

The conceptual structure generation unit 240 calculates and stores the similarity value for the object string stored in the conceptual structure database 216. Then, the conceptual structure generation unit 240 generates a conceptual structure, that is, a similarity matrix between entities of document units, and stores the generated similarity matrices in the conceptual structure database 216.

Referring to FIG. 4, the conceptual structure generation unit 240 includes a conceptual element grouping module 242, a similarity calculation module 244, and a conceptual structure generation module 246.

The conceptual element cluster module 242 groups conceptual elements having the same function attribute in units of patent documents.

The similarity calculation module 244 calculates a similarity value between the concept element entities for each functional attribute using the previously defined similarity calculation analysis model. The concept element is an entity having attribute values. The similarity measure between two entities is based on analysis of dependecy tree which is separated into head word and expression part, lemmatisation of each word, Can be calculated. The degree of similarity between two object strings a and b is defined as the sum of object similarity and functional attribute similarity. Each similarity degree is calculated as a representative DamerauLevenshtein distance among the minimum editing distance scale of the following equation by referring to the syntax structure attribute and the functional attribute.

That is, the similarity degree of the object can be obtained by using Equations (1) to (4).

Here, LD means Damerau Levenshtein distance, i.e., the minimum editing distance of two strings, and means that the similarity degree between the strings a and b is calculated by calculating the number of times of editing of insertion, deletion, and replacement necessary for the two strings to be the same string.

C (S) is the entity class of the string S , H (S) is the entry word of the string S , E ' is the entity set, and F is a set of words that have no meaningful discriminative power in expressing an object as an abbreviation or monolingual. Each expression considers the case where the heading identified according to the syntactic structure analysis belongs to F, and the case of the same or different heading words.

The editing distance d can be obtained using Equation (5).

However, even if the same editing distance is used, the degree of similarity varies depending on the length of a character string. Therefore, the length of the character string is normalized as shown in Equation 6

The conceptual structure generation module 246 generates a conceptual structure including functional attributes, conceptual elements, and similarity values of conceptual elements in units of patent documents.

The classification unit 250 obtains conceptual structure similarity values between patent documents, and classifies patent documents based on the obtained conceptual structure similarity values.

5, the classification unit 250 includes a concept element similarity degree calculation module 252, a functional property similarity degree calculation module 254, a conceptual structure similarity degree calculation module 256, a classification module 258 ).

The conceptual element similarity calculation module 252 obtains a similarity value of conceptual elements constituting the conceptual structure between patent documents. The method for calculating the similarity value of the conceptual elements by the conceptual element similarity degree calculation module 252 is the same as the method for calculating the conceptual element similarity value by the similarity degree calculation module, and a description thereof will be omitted.

The functional property similarity calculating module 254 obtains a functional property similarity value between patent documents. At this time, the functional attribute similarity degree calculation module 254 calculates the similarity degree of the target object having the same action attribute using Equations (7) to (10).

Here, C (S) is the object class of the character string S , H (S) is the entry word of the character string S , O is the target object set of the functional attribute, F is a word having no meaning discriminating power Means a set. Each expression considers the case where the heading identified according to the syntactic structure analysis belongs to F, and the case of the same or different heading words.

The conceptual structure similarity calculation module 256 uses the similarity value of the conceptual elements obtained in the conceptual element similarity calculation module 252 or the functional property similarity value obtained in the functional property similarity calculation module 254 to calculate the conceptual structure similarity value between patent documents I ask. At this time, the conceptual structure similarity degree calculation module 256 can be obtained by subdividing each object and the attribute of the function or by the total sum of the two concept element similarities.

The conceptual structure similarity degree calculating module 256 obtains the conceptual structure similarity value between patent documents using Equation (11).

Where N _E (P) is the number of individuals included in patent P, N _E (P _i , P _j ) is the number of patents P _i And P _j is the number of pairs of entities with a similarity value of 1 between the product and technology entities. Similarly, N _F (P) is the number of functional properties of the object of the patent P, N _F (P _i, P _j) are patent P _i And the number of pairs in which the similarity value between the functional attributes of the product and technology entities included in P _j is 1. If the two similarity values are each 1, the conceptual structure of the two patents can be said to be the same.

The classification module 258 classifies patent documents based on the concept structure similarity value between patent documents. That is, the classification module 258 can classify patent documents having a similarity value equal to or greater than a preset constant value into documents having the same or similar conceptual structure.

The conceptual element extracting unit 230, the conceptual structure generating unit 240, and the classifying unit 250 may each be implemented by a processor or the like necessary for executing a program on a computing device. The conceptual element extracting unit 230, the conceptual structure generating unit 240, and the classifying unit 250 may be physically independent of each other, or may be implemented in a functional manner in one processor have.

The control unit 260 controls the operations of various components of the database 210, the interface unit 220, the conceptual element extracting unit 230, the conceptual structure generating unit 240, and the classifying unit 250.

The control unit 260 may include at least one computing device, which may be a general purpose central processing unit (CPU), programmable device elements (CPLDs, FPGAs) that are suitably implemented for a particular purpose, Device (ASIC) or a microcontroller chip.

These components that may be included in the patent document classification system 200 may be implemented by hardware, software, or a combination thereof, and two or more components may be simultaneously implemented by one hardware or software.

6 is a flowchart illustrating a method of classifying a patent document according to an embodiment of the present invention.

Referring to FIG. 6, the patent document classification system extracts conceptual elements by analyzing each patent document, and assigns functional attributes to the conceptual elements (S602). A detailed description of step S602 will be made with reference to FIG.

The patent document classification system calculates a concept element value of each patent document and generates a concept structure for each patent document (S604). A detailed description of step S604 will be made with reference to FIG.

Then, the patent document classification system obtains the concept structure similarity value between the patent documents, and classifies the patent document based on the obtained concept structure similarity value (S606). That is, the patent document classification system obtains similarity values of conceptual elements constituting the conceptual structure between patent documents and functional attribute similarity values between patent documents. Then, the patent document classification system obtains the concept structure similarity value between the patent documents using the similarity value or the functional property similarity value of the obtained concept elements, and classifies the patent document based on the obtained concept structure similarity value between the patent documents.

FIG. 7 is a flowchart illustrating a method for extracting conceptual elements and functional attributes from a patent document according to an embodiment of the present invention.

Referring to FIG. 7, the patent document classification system refers to the object name dictionary database to identify a candidate sentence in a predefined area of a patent document (S702).

Then, the patent document classification system extracts a character string representing the concept element through parsing based on the dependency grammar in the identified candidate sentence (S704).

The patent document classification system refers to the functional classification database and gives a functional attribute to the extracted character string (S706). Thereafter, the patent document classification system stores conceptual elements, functional attributes, and the like in units of patent documents.

Figure 8 is a flow diagram illustrating a method for generating a conceptual structure according to an embodiment of the present invention.

Referring to FIG. 8, the patent document classification system groups conceptual elements having the same function attributes in units of patent documents (S802).

Then, the patent document classification system calculates a similarity value between the conceptual element entities according to each functional attribute using the previously defined similarity calculation model (S804), and includes similarity values of the functional attributes, conceptual elements, and conceptual elements Is generated in units of patent documents (S804).

Such a patent document classification method can be written in a program, and the codes and code segments constituting the program can be easily deduced by programmers in the field. In addition, the program relating to the patent document classification method is stored in an information storage medium (Readable Media) readable by an electronic apparatus, and can be read and executed by an electronic apparatus.

Thus, those skilled in the art will appreciate that the present invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. It is therefore to be understood that the embodiments described above are to be considered in all respects only as illustrative and not restrictive. The scope of the present invention is defined by the appended claims rather than the detailed description and all changes or modifications derived from the meaning and scope of the claims and their equivalents are to be construed as being included within the scope of the present invention do.

100: Patent providing server 200: Patent document classification system
210: Database 220: Interface Unit
230: Conceptual element extracting unit 240: Conceptual structure generating unit
250: Classification unit 260:

Claims (10)

A conceptual element extracting unit for extracting conceptual elements by analyzing each patent document, and assigning a functional attribute to each conceptual element;
A conceptual structure generation unit for generating conceptual element similarity values by clustering conceptual elements of each patent document by functional attributes and generating conceptual structures including the calculated conceptual element similarity values for each patent document;
A classifying unit for classifying the patent document based on the concept structure similarity value obtained between the patent documents and the concept structure similarity value between the patent documents;
A patent document classification system.
The method according to claim 1,
Product, and technology. Dictionary database;
And a function attribute classification database in which rules for classifying functional attributes of entities are set.
The method according to claim 1,
The conceptual element extracting unit may include a candidate sentence identifying module for identifying a candidate sentence in a predefined area of the patent document by referring to the object name dictionary database;
A character string extraction module for extracting a character string representing a concept element through syntax analysis based on a dependency grammar in the identified candidate sentence;
And a functional attribute assigning module for referring to the functional classification database and assigning a functional attribute to the extracted character string,
Wherein the extracted character string is a conceptual element.
The method according to claim 1,
The conceptual structure generation unit,
A conceptual element cluster module which clusters conceptual elements to which the same functional attributes are assigned in units of patent documents;
A similarity calculation module for calculating a similarity value between the conceptual element entities for each functional attribute using a previously defined similarity calculation model;
And a conceptual structure generation module for generating a conceptual structure including functional attributes, conceptual elements, and similarity values of conceptual elements in units of patent documents.
The method according to claim 1,
Wherein,
A concept element similarity calculating module for obtaining a similarity value of concept elements constituting a conceptual structure between patent documents;
A functional attribute similarity calculation module for obtaining a functional attribute similarity value between patent documents;
A conceptual structure similarity calculation module for obtaining a conceptual structure similarity value between patent documents using the similarity value or the functional property similarity value of the obtained concept elements;
And a classification module for classifying the patent document based on the conceptual structure similarity value between the patent documents.
(a) analyzing each patent document to extract conceptual elements, and assigning functional attributes to each conceptual element;
(b) calculating conceptual element similarity values by clustering conceptual elements of each patent document by functional attributes, and generating a conceptual structure including the calculated conceptual element similarity values for each patent document; And
(c) classifying the patent document based on the concept structure similarity value between the patent documents and the concept structure similarity value;
The method comprising the steps of:
The method according to claim 6,
The step (a)
Identifying a candidate sentence in a predefined area of the patent document by referring to the object name dictionary database;
Extracting a character string representing a concept element through parsing based on a dependency grammar in the identified candidate sentence;
And providing a functional attribute to the extracted character string by referring to the functional classification database,
Wherein the extracted character string is a conceptual element.
The method according to claim 6,
The step (b)
Clustering conceptual elements to which the same functional attributes are assigned in units of patent documents;
Calculating a similarity value between the conceptual element entities for each functional attribute using a previously defined similarity calculation model;
And generating a conceptual structure including a feature attribute, conceptual elements, and similarity values of the conceptual elements in units of a patent document.
The method according to claim 6,
The step (c)
Obtaining similarity values of conceptual elements constituting a conceptual structure between patent documents;
Obtaining a functional property similarity value between patent documents;
Obtaining a concept structure similarity value between patent documents using the similarity value or the functional property similarity value of the obtained concept elements;
And classifying the patent document based on the concept structure similarity value between the patent documents.
When implemented by a patent document classification system,
(a) analyzing each patent document to extract conceptual elements, and assigning functional attributes to each conceptual element;
(b) calculating conceptual element similarity values by clustering conceptual elements of each patent document by functional attributes, and generating a conceptual structure including the calculated conceptual element similarity values for each patent document; And
(c) obtaining a conceptual similarity value between patent documents, and classifying the patent document based on the obtained conceptual structure similarity value.

Images