KR101117298B1 - System and method for parsing korean semantic based on ontology - Google Patents

Abstract

PURPOSE: A system and a method for analyzing the meanings of Korean based on ontology are provided to implement intellectual information retrieval and to provide exact answers by verifying the inner meanings of inputted sentences. CONSTITUTION: A morpheme analyzing part(300) retrieves the morpheme information of an inputted Korean clause by retrieving syllable combined with analyzed part of speech. A clause generating part(500) analyzes clause information and sentence components based on the analyzed morpheme information. An ontology analyzing part(600) establishes the ontology meaning information of an uninflected word as uninflected word ontology. An outputting part(900) the ontology information of uninflected word and predicate as ontology meaning analyzed results.

Description

System and Method for Parsing Korean Semantic Based on Ontology

The present invention relates to a method for analyzing Korean meaning, and more particularly, to an ontology-based Korean semantic analysis system and method capable of semantic-based knowledge information processing to grasp the internal meaning of an input sentence.

The exchange of messages using mobile phone text messages and Internet messengers has become commonplace to many people due to the popularization of mobile devices.

The sentences that make up these messages appear very frequently in abbreviated expressions and spacing errors. Various application softwares based on Korean analysis have a problem in that they cannot perform the expected performance when the sentences including the above error cannot be correctly analyzed.

In the conventional Korean analysis technology, the analysis accuracy of the spacing error sentence is inevitably lowered because only the Korean sentence of the main sentence corresponding to the spelling is analyzed. In addition, the conventional Korean analysis technology has a problem in that the processing speed is slowed and the accuracy of analysis is lowered as the number of syllables without spacing increases.

In the current web environment, the keyword matching method that relies only on form identity is searched only for the excessive information of the same keyword or shape regardless of meaning or content. It was.

For example, the conventional keyword matching method is' boating belly ', body part belly', 'double belly', 'section belly', '(color on paper) belly' in the search result of 'eat belly'. Since the surplus information such as endlessly searched and provided as a search result, the inner meaning of the fruit 'pear' related to 'eat' could not be extracted.

In order to solve this problem, the present invention has an object to grasp the internal meaning of the input statement using the ontology-based Korean semantic analysis system.

Another object of the present invention is to perform a morphological analysis of the spelling and the spacing error sentence in accordance with the spelling using the part-of-speech combination information.

Ontology-based Korean semantic analysis method according to the characteristics of the present invention for achieving the technical problem is to analyze the retrieved parts of speech by searching in the stem dictionary including the word and the verb or the ending dictionary including the investigation or ending Making; Analyzing the morpheme information of the Korean input word by searching the syllables coupled to the analyzed parts of speech from the stem dictionary or the ending dictionary and analyzing the word information using the analyzed morpheme information; Constructing ontology semantic information of the spoken word as a ontology, and constructing an ontology of an ontology based on semantic information in relation to the spoken word; Analyzing the meaning of the verb by connecting the ontology and ontology semantic information of the verb and analyzing ontology information of the verb and the verb through the verb ontology and the verb ontology; And outputting ontology information of the verb and the verb as an ontology semantic analysis result.

The ontology-based Korean semantic analysis system according to the characteristics of the present invention analyzes the retrieved parts of speech by searching the stem syllable to be analyzed in the Korean input word in the stem dictionary including the word and the verb or the ending dictionary including the survey or the ending, A morpheme analysis unit for searching the combined syllables in the stem or dictionary to analyze the morpheme information of the Korean input word; A word generator configured to analyze word information and sentence components using the analyzed morpheme information; Construct the ontology semantic information of the correspondence as a ontology, construct an ontology formed by the ontology according to the semantic relations with the semantic information, connect the ontology with the ontology semantic information of the correspondence Ontology analysis unit for analyzing the meaning; And an output unit for outputting the ontology information of the verb and the verb analyzed by the verb ontology and the verb ontology as a result of the ontology meaning analysis of the verb and the verb.

By the above-described configuration, the present invention has the effect of improving the error sentence and abbreviated expression, unregistered word processing, compound noun processing ability that is not correctly spaced.

The present invention has the effect of enabling intelligent semantic based knowledge information processing to understand and process the meaning of knowledge and information.

According to the present invention, since semantic-based knowledge information processing is possible, the internal meaning of an input statement can be grasped to provide an accurate answer, and at the same time, an upper ontology information and a lower ontology information related to an ontology can be simultaneously provided for intelligent information retrieval. This has a possible effect.

1 is a diagram illustrating a configuration of an ontology-based Korean semantic analysis system according to an exemplary embodiment of the present invention.
2 is a block diagram schematically illustrating an internal configuration of a morpheme analysis unit according to an exemplary embodiment of the present invention.
3 is a diagram illustrating a morpheme analysis method using a Korean semantic analysis system according to an embodiment of the present invention.
4 is a diagram illustrating a morphological analysis result according to an embodiment of the present invention.
5 is a diagram illustrating a word generation result according to an exemplary embodiment of the present invention.
6 is a diagram illustrating a result of constructing a ontology according to an embodiment of the present invention.
7 is a view showing the result of automatic construction of the communication ontology according to an embodiment of the present invention.
8 is a diagram showing a result of constructing an ontology according to an embodiment of the present invention.
9 is a diagram illustrating a result of RDF triple conversion of verb ontology according to an embodiment of the present invention.
FIG. 10 is a diagram illustrating an output result of ontology-based Korean semantic analysis according to an embodiment of the present invention. FIG.

DETAILED DESCRIPTION Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. The present invention may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and like reference numerals designate like parts throughout the specification.

Throughout the specification, when a part is said to "include" a certain component, it means that it can further include other components, without excluding other components unless specifically stated otherwise.

1 is a diagram illustrating a configuration of an ontology-based Korean semantic analysis system according to an exemplary embodiment of the present invention.

Korean semantic analysis system according to an embodiment of the present invention is a user interface 100, dictionary database 200, morpheme analysis unit 300, unregistered noun proper noun combination unit 400, word generation unit 500, ontology The analyzer 600 includes a resource description framework (RDF) triple converter 700, a triple repository system 800, and an output unit 900.

The user interface 100 is an input module for receiving a sentence to be analyzed by a user.

The dictionary database 200 includes an irregular dictionary 210 for restoring irregular verbs, a stem dictionary 220 for storing the words and verbs, and a mother dictionary 230 for storing the mother and the survey.

The fire rules dictionary includes' illegal rules', 's irregular rules',' le bull rules', 'ㄹ Bull rules',' ㄷ irregular rules', 'le Bull rules',' ㅂ irregular rules', 'bull rules',' bad rules', ' Deformation and restoration of about 10 irregularities, including 'illegal rules'. In addition, the irregular dictionary 210 does not restore the original 'proud' of 'proud' from 'proud son' with 'ㅂ irregularity' to 'proud' with irregular information of 'proud' To speed up the analysis of words without creating unnecessary transform and reconstruction algorithm rules.

The stem dictionary 220 includes a verb (country, ethnicity, oath, contractor, etc.), a verb (such as, under, like, etc.), adverbs (very, etc.), tubular (like, old, back, etc.) Parts of speech information such as; In addition, the stem dictionary 220 according to the parts of speech, quality information, and the number of syllables of the syllables of the left syllable even if a single sentence is searched for in a word to solve the neutrality problem in the grammatical analysis of the error sentence with the incorrect spacing A vocabulary indicating multilingual information and a vocabulary indicating neutrality information are given.

For example, in the case of analyzing 'disease treatment', if the noun 'disease treatment' is searched in the stem dictionary 220 according to the longest matching method, the analysis of 'disease + disease treatment' is performed. In order to prevent such an error analysis, the morpheme analysis unit 300 separates the noun 'therapeutic treatment' having the information and the neutrality information by 'sickness treatment' by syllable like 'sickness + treatment', and then excludes the 'treatment'. If the noun is searched for syllables, and if the noun 'disease' is found, the noun is re-analyzed and stored as a noun, and the left syllable 'disease' is analyzed to perform a correct analysis of 'disease' + 'treatment'.

The ending dictionary 230 includes a check (_, _, _, _, _ to, _, etc.), an assistant (_degree, _ only, _, etc.), the ending (_, _, Etc.), linking endings (such as _go, _ ㄹ, _ ㄴ다 ', etc.), final endings (_ poems, _ all, _op, etc.), prime endings (_ ㄹ, _, _ ,, etc.), etc. Include the same probe and mother. In addition, the parent dictionary 230 includes not only part-of-speech information but also feature information of 'BAN' in morphological analysis in order to indicate a mother and investigation having an incomplete species like 'ㄹ'. That is, the ending dictionary 230 separates the survey and the mother by the longest match from the input word with the information 'BAN' representing the survey and the mother having an incomplete species, such as 'ㄹ' in 'ㄹ'.

The morpheme analysis unit 300 searches the parts of syllables to be analyzed through the stem dictionary 220 or the ending dictionary 230 according to the reverse order of the Korean input words to be analyzed, and analyzes the retrieved parts of speech. The syllables are searched in the stem dictionary 220 or the ending dictionary 230 to perform morphological analysis in the reverse order of the Korean input word.

The non-registered word proper noun combining unit 400 combines proper nouns such as human names after all morphological analysis is performed. In addition, the non-registered word proper noun combining unit 400 performs a 'left to right' check rather than a 'right to left' check used in morphological analysis in order to estimate and combine the unregistered word and the proper noun from the non-spaced error text. By combining English characters, combining three syllable proper nouns, and combining two syllable proper nouns, they are combined according to a specific combination rule by character or syllable.

The word generator 500 generates word information and sentence components using the morpheme information output from the morpheme analyzer 300.

The ontology analyzer 600 includes a verbal ontology analyzer 610, a verbal ontology analyzer 620, and an inference analyzer 630.

The message ontology analysis unit 610 processes not only the classes in the standard properties 'rdfs: subClassOf' and 'owl: equivalentClass' relationships established by the W3C, but also handles the hierarchical and equality relationships of the classes.

Also, the ontology analysis unit 610 receives the morpheme information of the message received from the morphological analysis unit 300 and the ontology 'ONTO' information, and the word information and sentence component information of the word generation unit 500. Form.

The verb ontology analysis unit 620 not only performs property processing in relation to 'owl: equivalentProperty' and 'rdfs: domain', 'rdfs: range', 'rdfs: subPropertyOf', and 'owl: equivalentProperty' which are standard properties established by the W3C. Handles the parent-child relationship, equivalence relationship of (Property), class domain (rdfs: domain) that can have each property as property, and range (rdfs: range) and add-on (ksOnto: AJT) relationship that can be the target of each property. do.

Class domains, ranges, and adjuncts include the subject (SBJ) as the 'domain' in the sentence component information, the object (OBJ) and the bore (CMP) as the 'range', and the adjunct (AJT) as the adjunct (AJT). Ontology attributes are given.

Of the AJTs of a sentence component, the temporal adjunct representing the semantic attribute of time is

'ksOnto: AJTT'; The place adjunct representing the semantic attribute of a place is 'ksOnto: AJTP'; The common adjunct representing the common semantic attribute is 'ksOnto "AJTW'; the source adjunct representing the semantic attribute of the source is 'ksOnto: AJTS'; the destination adjunct representing the semantic attribute of the destination is 'ksOnto: AJTD'; The tool adjunct representing 'ksOnto; AJTI'; the granting adjunct representing the semantic attribute of the grant is 'ksOnto: AJTG'; the comparison adjunct representing the semantic attribute of the comparison is given with ontology information set by itself as 'ksOnto: AJTC'.

The inference analyzer 630 may extend the meaning through the inference of the relationship between the class and the property. In other words, the inference analysis unit 630 is applied to inference rules, such as 'S rdfs: subClassOf 01 + 01 rdfs: subClassOf 02-> S rdfs: subClassOf 02'. For example, RDF triple 'University rdfs: subClassOf School' and 'School rdfs: subClassOf Educational Institution' are deduced from 'University rdfs: subClassOf Educational Institution'. Can extend to

The ontology analyzer 600 stores ontology feature 'ONTO' information indicating an association with the ontology.

For example, the statement 'car' means ONTO information '1132253311' which means 'drinking car', 'TO132 information' which means 'transportation vehicle', and '113229111' which means 'time' as in WWII. Contains 1239212 '.

The message ontology analyzer 610 analyzes the meaning of the message and automatically constructs the message ontology based on the ONTO information. Here, the ONTO information represents ontology semantic information.

The verbal ontology analyzer 620 performs semantic analysis using the verbal ontology of the verbal ontology analyzer 610.

The RDF triple converter 700 converts a plurality of sentences into a set of RDF triples by using ontology information output from the ontology analysis unit 610 and the verbal ontology analyzer 620 received from the ontology analyzer 600. do. Here, the RDF triple expresses unit knowledge and information in three pairs of subject (resource), predicate (property), and object (literal) so that the computer can understand the meaning of knowledge and information. The format is an international standard managed by the World Wide Web Consortium (W3C). Here, three pairs of a subject (resource), a predicate (property), and an object (literal) are called triples.

The triple repository system 800 stores a set of RDF triples received from the RDF triple converter 700 and provides a function of deleting, updating, sorting, and searching a set of RDF triples.

The output unit 900 outputs the morphological analysis result and the word generation result received from the triple repository system 800.

In addition, the output unit 900 outputs ontology semantic analysis results of the message and the ontology information of the input sentence received from the triple repository system 800 from the user interface.

In other words, the output unit 900 outputs the upper ontology information (HONTO), the lower ontology information (LONTO), and the equivalent ontology information (EONTO) containing the internal meaning through the inference analysis of the ontology analyzer 600.

2 is a block diagram schematically illustrating an internal configuration of a morpheme analysis unit according to an exemplary embodiment of the present invention.

The morpheme analysis unit 300 according to an embodiment of the present invention is an adverb analysis unit 310, a connection mother analysis unit 311, an assistant company analysis unit 312, a survey investigation analysis unit 313, and a mother analysis unit 314. , Malleable ending analysis unit 315, front end ending analysis unit 316, designated noun analysis unit 317, verb-derived suffix analysis unit 318, adjective-derived suffix analysis unit 319, verbal auxiliary verb analysis unit 320 , A noun-type derived suffix analyzer 321 and a unregistered word parser 322.

The adverb analyzer 310 includes an adverb analysis module 310a and an adverb combining information analysis module 310b.

The adverb analysis module 310a and the adverb combining information analysis module 310b search for the adverbs in the stem dictionary 220 of the dictionary database unit 200 and separately store the adverbs.

The adverb combining information analysis module 310b searches for a general adverb, an adverb adverb, and an adverb derived suffix in the stem dictionary 220 of the dictionary database 200.

The adverb combining information analysis module 310b analyzes the part-of-speech information located in the left syllable of the retrieved adverbs and transmits the analyzed part-of-speech information to the analytical part-of-speech module.

For example, the adverb analysis module 310a searches for adverb and neutrality information in the stem dictionary 220 from the general adverb 'Hanchang Naie' to 'Ie', and the adverb combining information analysis module 310b retrieves the retrieved neutrality. After the adverb 'Ie' is separated by syllables as in I + e, it is determined whether the noun is searched in the previous syllable except 'e'. When the noun 'age' is searched, the adverb combining information analysis module 310b stores the part of the syllable and the number of syllables of the left syllable of the sub-array survey ('e') after storing the 'e' instead of the adverb as the sub-array survey. In addition, re-analyze adverbs with neutrality information by searching for feature information (which indicates attribute information of parts of speech such as intransitive verb and transitive verb).

The adverb combining information analysis module 310b uses the part-of-speech information and the stored part-of-speech information except the searched adverbs to use the adverbs or change to other parts of speech when an adverb having multilingual information such as 'and' is a connected adverb. You decide.

For example, search for a target search of 'and' in the left syllable except 'and' in 'friends' and 'and', so that 'and' is not a connection adverb. Re-analyze).

The adverb combining information analysis module 310b stores the adverbs according to the parts of the syllable of the left syllable and the parts of the parts except the retrieved adverb derived suffix 'crab' such as the adverb derived suffix 'crab' and stores the left syllable according to the information of the parts of the syllable. Send to the part of speech module that can be analyzed.

The connection end analysis unit 311 includes a connection end analysis module 311a and a connection end connection information analysis module 311b.

The connecting ending analysis module 311a and the connecting ending combining information analysis module 311b search for the connecting ending in the ending dictionary 230 of the dictionary database unit 200 and separately store the connecting ending.

The connection ending analysis module 311a is a connection ending searched as '_bdaman' in 'Study only' is a connection ending with verbal nature, or is a connection ending having a messaging property like '_rang' in 'friends', 'Friend, study' in 'connected mothers' with the nature of the words and words such as '_go' in the search dictionary dictionary 230 to analyze.

The connection ending analysis module 311a separates the connection ending and when the left syllable including the separated connection ending is searched for in the stem dictionary 220, the left syllable including the connecting ending is transmitted to the uncommitted registration analysis module 322a. Analyze and transmit the left syllable including the linking end to the adjective auxiliary verb analysis module 320a when the term is searched for in adverbs, and transmit and analyze the left syllable including the connecting end to the adverb analysis module 310a when searching for adverbs. .

In addition, when the connected syllable analysis module 311a detects the left syllable including the connected ending as an assistant in the ending dictionary 230, transmits and analyzes the left syllable including the connected ending to the assistant analyzing module 312a and searches for the case survey. The left syllable including the linking ending is transmitted to the forensic analysis module 313a, and when it is searched for, the left syllable including the linking ending is transmitted to the ending analysis module 314a for analysis.

After the connection ending combining information analysis module 311b stores the separated connection ending, if the left syllable except the connection ending is searched for in the stem dictionary 220, the remaining syllable except the connecting ending is not registered. And transmits to the adverb analysis module (320a), and transmits the left syllable, except for the connecting ending, to the adverb analysis module (310a). Send and analyze

The connection ending combination information analysis module 311b may accurately analyze error sentences that are not spaced using parts of speech and syllables of the left syllable except the connection ending, and parts of speech information stored on the right side of the connection ending.

The assistant analysis unit 312 includes an assistant analysis module 312a and an assistant combination information analysis module 312b.

The assistant analysis module 312a and the assistant combination information analysis module 312b search for and separate the assistants such as '_b, _ degrees, _ only' from the parent dictionary 230 of the dictionary database 200.

The assistant analysis module 312a transmits and analyzes the left syllable including the assistant to the spoken non-registered word analysis module 322a when the left syllable including the assistant is searched for in the stem dictionary 220, and includes the assistant when the verb is found in the verb. The left syllable is transmitted to the verbal auxiliary verb analysis module 320a for analysis. When the search is performed as an adverb, the left syllable is transmitted to the adverb analysis module 310a for analysis.

The assistant combining information analysis module 312b stores and separates the assistant, and when the left syllable except the assistant is searched by the spoken word in the stem dictionary 220, transmits the left syllable except the assistant to the unregistered word analysis module 322a. When the analysis is performed by searching for verbs, the left syllable, except for the auxiliary verb, is transmitted to the verb auxiliary verb analysis module 320a, and when it is detected as an adverb, the left syllable except the auxiliary verb is transmitted to the adverb analysis module 310a for analysis.

Assisted part information analysis module (312b) is part of the syllables and syllables of the left syllables except the assistants and parts of speech information stored in the right syllables of the assistant in order to accurately grasp the meaning of the auxiliary words such as '_b, _do, _man' Analyze together to improve the accuracy of the mother.

Assisted assistant information analysis module (312b) is not only an assistant in a non-sparse error statement, such as 'do' in the 'very good eating', but also among the assistants such as 'chigo' in correct words such as 'chigo' and 'friend'. Re-analyzes the syllables and analyzes them correctly, such as 'galch (+)' and 'connected' and 'friend' (+).

The surveying analysis unit 313 includes a surveying analysis module 313a and a surveying combining information analysis module 313b.

The grading analysis module 313a includes a main survey analysis for processing a survey such as '_a, _, _i' and a target survey analysis for processing a forcing such as '_' and '_' Analyzes sub-trials surveys dealing with cross-checks such as 'to', and tubular surveys analyzes such as '_'.

Checking analysis module 313a transmits the left syllables including the checks to the non-speech word analysis module 322a when the left syllables including the checks are searched for in the stem dictionary 220, and includes the checks if found in the verbs. The left syllable is transmitted to the verbal auxiliary verb analysis module 320a for analysis, and when it is detected as an adverb, the left syllable including the search is transmitted to the adverb analysis module 310a for analysis.

If the left syllable except for the search is retrieved by the spoken word from the stem dictionary 220, the search for combining combining information analysis module 313b stores and separates the search, and then transmits the left syllable except the search to the unregistered word analysis module 322a. When the analysis is performed by searching for verbs, the left syllable except for the interrogation is transmitted to the adjunct auxiliary verb analysis module 320a for analysis. When the search is an adverb, the left syllable except the interrogation is transmitted to the adverb analysis module 310a for analysis.

If the left syllable except for the search is searched as an assistant in the ending dictionary 230, the search for combining combining information analysis module 313b analyzes the left syllable except for the forcing to the assistant analysis module 312a and analyzes the search if it is found as a connected ending. The left syllables are excluded and transmitted to the connected ending analysis module 311a, and when the end words are searched, the left syllables are excluded and transmitted to the ending ending analysis module 314a for analysis.

The mother analyzing unit 314 includes a mother analyzing module 314a and a mother combining information analyzing module 314b.

The mother analyzing module 314a and the mother combining information analyzing module 314b retrieve the mother from the mother dictionary 230 of the dictionary database unit 200 to store and separate the mother.

The ending analysis module 314a performs a final word analysis such as' _, _ b, or _ If you have a speech and verbal property, such as '_, or _ group', you will perform an ending analysis of the word and verbal property.

The ending mother analysis module 314a separates and stores the ending mother, and when the left syllable including the ending mother is searched for in the spoken word in the stem dictionary 220, the ending syllable including the ending mother is sent to the unregistered word analysis module 322a. Transmit and analyze the left syllables containing the ending words to the adverb auxiliary verb analysis module 320a when the words are searched for the adverbs, and transmits and analyze the left syllables containing the ending words to the adverb analysis module 310a when the adverbs are detected. do.

The mother ending information analysis module 314b stores the mother ending and separates the left syllable including the ending mother when the left syllable except the ending mother is searched for by the spoken word in the stem dictionary 220. And transmits to the adverb auxiliary verb analysis module 320a to analyze the left syllable, except for the ending word, if it is searched as a verb and transmits it to the adverb analysis module 310a if it is searched as an adverb. Analyze by

When the ending syllables except the ending words are retrieved as the ending words from the ending dictionary 230, the ending combining information analysis module 314b analyzes the left syllables except the ending words to be transmitted to the ending ending analysis module 316a and adjective-derived suffixes. If searched by the left syllable except the ending is transmitted to the adjective-derived suffix analysis module (319a) for analysis.

The ending combining information analysis module 314b analyzes the part-of-speech information, the number of syllables, and the part-of-speech information stored in the right syllable of the mother ending together to improve the accuracy of the analysis.

Mother ending combined information analysis module (314b), such as 'eat girl', the symptom of the mother's abbreviated '_eun girl' is the 'eat (pronoun) + is (oral-type epilative mother) + thing (cheer) +' (target) As with the survey, the original form '_' is used for restoring the original '_' (word-shaped malleable mother) + one (communication) + (objective investigation).

The prime mother analyzer 315 includes a prime mother analysis module 315a and a prime mother combining information analysis module 315b.

The malleable mother analysis module 315a performs the analysis of the noun-type malleable mothers that handle malleable mothers such as '_ ㅁ, _ 음, _ 기' and the tubular malleable motherland analysis that processes malleable mothers such as '_ ㄹ, _dun' do.

The prime mother analysis module 315a separates and stores the prime mother and then, when the left syllable including the prime mother is retrieved by the spoken word from the stem dictionary 220, the left syllable including the prime mother is transferred to the unregistered word analysis module 322a. Transmit and analyze and transmit the left syllables containing the malleable words to the adverb auxiliary verb analysis module 320a when the words are searched for the adverbs. do.

The prime mother combining information analysis module 315b separates and stores the prime mother, and when the left syllable except the prime mother is searched for the term in the stem dictionary 220, the verbal auxiliary verb analysis module 320a When the data is sent to the designated nouns, the left syllables except the malleable words are transmitted to the noun analysis module 317a for analysis.

When the left syllables except the malleable mothers are retrieved from the ending dictionary 230 as prime words, the malleable combining information analysis module 315b transmits the left syllables except the malleable mothers to the line ending analysis module 316a to analyze and adjective derivative suffixes. If searched by the left syllable except for the prime ending is transmitted to the adjective-derived suffix analysis module (319a) for analysis.

The malleable mother analysis module 315a converts the tubular malleable horse 'd' into a target survey in words such as the abbreviation of the oath + d.

The malleable combined information analysis module (315b) excludes the noun-type malleable to solve the difference between the meaning of the word 'fan wind' and 'study wind' in the noun-type malleable words of 'ㅁ, Um, and Ki'. Using the part-of-speech information of the left syllable, the number of syllables, and the part-of-speech information of the right syllable of the noun-type epilator, the word 'wind' in the 'fan wind' means 'wind' in English, and 'wind' in 'study wind'. The word 'wind' which means' Wish 'in English is analyzed as' bara (pron) + ㅁ (noun-type epilogue).

The front end ending analysis unit 316 includes a front end ending analysis module 316a and a front end ending combining information analysis module 316b.

The mother word analysis module 316a analyzes polite mother horse endings that deal with mother endings such as '_op_' and speculative mother ending analysis that handles the final endings such as '_Yo' and the horse endings such as '_Oh'. Analysis of time ending endings that process endings and elevated ending endings that handle ending endings such as '_hours' are performed.

The front end ending analysis module 316a transmits the left syllable including the front end ending to the unregistered word analysis module 322a when the left syllable including the end ending separated by separating the front end ending is searched for in the spoken word in the stem dictionary 220. When the analysis is performed by searching for a verb, the left syllable including the word ending is transmitted to the verb auxiliary verb analysis module 320a for analysis.

The front end ending combining information analysis module 316b stores and separates the end endings, and when the left syllables except the end endings are retrieved by the spoken word from the stem dictionary 220, the left syllables except the ending endings are unregistered words analyzing module 322a. And transmits and analyzes the left syllable, except for the word ending when it is searched for, by the verb adjunct analysis module (320a), and sends the left syllable except for the word ending, to the noun analysis module (317a). Analyze by

When the front end ending combination information analysis module 316b is searched for an assistant in the ending dictionary 230, the left syllable except for the end ending is transmitted to the assistant analysis module 312a to analyze the left syllable except the front end ending. The left syllable excluding the ending ending is transmitted to the connected ending analysis module 311a for analysis, and when it is detected as an adjective derivative suffix, the left syllable excluding the ending ending is sent to the adjective derived suffix analysis module 319a for analysis.

The first word combining information analysis module (316b) is the part of speech and the number of syllables and syllables of the left syllable, except the first ending, in order to accurately grasp the meaning of the first ending "poetry" such as 'It was Gyeongju' and 'Tell me.' By using the part-of-speech information in the right syllable, 'Gyeongju-si (communion) + 2 (specified)) + (time line ending ending) + da (end ending),' word (word) + ha (verb derived suffix) + word (connection ending) ) + Week (verb) + poetry (elevated fresh ending ending) + was (time ending ending ending) + da (ending ending).

The noun analysis unit 317 includes a noun analysis module 317a and a noun combining information analysis module 317b.

The noun analysis module 317a performs affirmative noun analysis that processes nouns such as '_i' and negative noun analysis that handles nouns such as '_no'.

The noun analysis module 317a transmits and analyzes the left syllable including the noun to the unregistered word analysis module 322a when the left syllable including the noun separated from the noun is searched for in the stem cell dictionary 220. If it is searched by the left syllable including the designator sent to the verbal auxiliary verb analysis module 320a for analysis.

The noun combining information analysis module 317b stores and separates the nouns, and when left syllables other than the nouns are retrieved by the spoken word from the stem dictionary 220, the syllable combining information analysis module 317a transmits the left syllables except the nouns to the unregistered word analysis module 322a. After analyzing and searching for an adverb, the left syllable excluding the designated adjective is transmitted to the adverb analysis module 310a for analysis.

If the syllable combining information analysis module 317b is searched for as an assistant in the ending dictionary 230 except for the designator, the syllable combining information analysis module 317b transmits and analyzes the left syllable except for the designator to the assistant analysis module 312a.

The noun combining information analysis module 317b analyzes the part-of-speech information, the number of syllables, and the part-of-speech information stored in the right syllable of the noun, together with the noun.

The verb-derived suffix analyzer 318 includes a verb-derived suffix analysis module 318a and a verb-derived suffix combination information analysis module 318b.

The verb-derived suffix analysis module 318a and the verb-derived suffix combination information analysis module 318b retrieve and store the verb-derived suffixes from the parent dictionary 230 of the dictionary database unit 200.

The verb-derived suffix combination information analysis module 318b not only searches for verbs including verb-derived suffixes, but also analyzes the parts of speech of the left syllable combined with the verb-derived suffixes.

The adjective derived suffix analyzer 319 includes an adjective derived suffix analysis module 319a and an adjective derived suffix combining information analysis module 319b.

The adjective-derived suffix analysis module 319a and the adjective-derived suffix combination information analysis module 319b retrieve and store the adjective-derived suffixes from the ending dictionary 230 of the dictionary database unit 200. In addition, the adjective-derived suffix analysis module 319a and the adjective-derived suffix combination information analysis module 319b may use the irregular dictionary 210 of the dictionary database unit 200 without using a modification rule for prototyping the modified adjective-derived suffix. Use to restore the prototype for irregularities.

The verbal auxiliary verb analysis unit 320 includes a verbal auxiliary verb analysis module 320a and a verbal auxiliary verb combining information analysis module 320b.

The verb auxiliary verb analysis module 320a and the verb auxiliary verb combination information analysis module 320b retrieve and store a verb from the stem dictionary 220 of the dictionary database unit 200.

The verb adjunct analysis module 320a and the verb adjunct combining information analysis module 320b retrieve the irregular dictionary 210 linked to the morphological analysis unit 300 without the rules of circular restoration to restore the prototype and analyze the speed of the word. Maximize

Verbal auxiliary verb analysis module 320a is an auxiliary verb analysis to process verbs such as 'oh, want, not,' and a verb analysis such as 'squeeze, thin' and verbs such as 'squeeze and suck'. We perform bull rule verb analysis and verb abbreviation analysis to process verbs such as 'Wah, Wah'.

Verbal supplementary verb combination information analysis module (320b) is a combination of words such as 'born to live' and 'words' (pron) + ah (connected ending) + flesh (word) + multi (mother) It analyzes by restoring the connection ending 'ah (connection ending)' omitted between the word and the word.

Terminology supplementary verb combination information analysis module 320b stores and separates the longest match, complete the morphological analysis if there are no syllables to analyze in the left syllable except for the verb and unique through the unregistered proper noun combining unit 400 Combine nouns

The noun-derived suffix analyzer 321 includes a noun-derived suffix analysis module 321a and a noun-derived suffix combining information analysis module 321b.

The noun-derived suffix analysis module 321a and the noun-derived suffix combination information analysis module 321b retrieve and store the noun-derived suffixes from the ending dictionary 230 of the dictionary database unit 200. The noun-derived suffix combining information analysis module 321b analyzes the part-of-speech information located in the left syllable of the found noun-derived suffix and transmits the analyzed part-of-speech information to the analytical part-of-speech module.

The spoken non-registered word analysis unit 322 includes a spoken non-registered word analysis module 322a and a spoken combination information analysis module 322b.

The uncommitted word analysis module 322a and the message combination information analysis module 322b search for and store the message in the stem dictionary 220 of the dictionary database unit 200 with the longest match. The unregistered word analysis module 322a and the message combining information analysis module 322b are multilingual information and neutrality information of a message according to the part-of-speech information, syllable number, and feature information of the syllable of the left syllable even if a single word is retrieved in one word. Can be analyzed.

Unregistered Word Analysis Module (322a) analyzes special syllables such as 'dragon, sentence, enemy' and 'nouns' that deal with the words such as 'school, motherland, and nation', as well as 'furniture, light year, dollar, gabby', etc. Performs dependency noun analysis to deal with the problem, rhetorical analysis to process the ninety, ninety, ninety, etc., and unregistered word analysis to process syllables not registered in advance.

The message combining information analysis module 322b stores and separates the longest match, and if there are no syllables to be analyzed in the left syllable except the message, the morpheme analysis is completed and the proper nouns are combined through the unregistered proper noun combining unit 400. Combine.

The message combining information analysis module 322b transmits the left syllable except the message to the adverb analysis module 310a when the left syllable except the message is searched as an adverb in the stem dictionary 220 and analyzes the left syllable except the message. The syllable is transmitted to the unregistered word analysis module 322a for analysis.

The message combining information analysis module 322b analyzes the left syllable excluding the message and transmits the left syllable, excluding the message, to the prime mother analysis module 315a. Transmit the left syllable except the syllable analysis module 313a and analyze it, and if an assistant is found, transmit the left syllable except the message to the assistant analysis module 312a, and analyze the connected syllable. Transmit to module 311a for analysis.

The spoken combination information analysis module 322b may analyze not only compound nouns but also various parts of speech information using parts of speech, syllables, and parts of speech of right syllables except for words.

The first part-of-speech analysis module that analyzes the survey and the mother is the above-mentioned linkage ending analysis module 311a, assistant assistant analysis module 312a, screening analysis module 313a, mother ending analysis module 314a, and malleable mother analysis module 315a. ), A predicate ending analysis module 316a, a noun analysis module 317a, a verb-derived suffix analysis module 318a, an adjective-derived suffix analysis module 319a, and a noun-derived suffix analysis module 321a.

The second part-of-speech analysis module that analyzes the spoken word and the verb includes an adverb analysis module 310a, a verb auxiliary verb analysis module 320a, and a non-spoken word analysis module 322a. In this way, the first part-of-speech analysis module and the second part-of-speech analysis module are combined to function as the part-of-speech analysis module.

The combined information analysis module connected to the first part-of-speech analysis module and the second part-of-speech analysis module collectively functions as a part-of-speech analysis information analysis module.

The part-of-speech information analysis module searches for the left syllables combined with the syllables analyzed by the part-of-speech analysis module in the parent dictionary 230 or the stem dictionary 220 and retrieves the retrieved parts of speech through the first part-of-speech analysis module or the second part-of-speech analysis module. Analyze the left syllable.

The part-of-speech analysis module and the part-of-speech analysis information analysis module search for the syllable to be analyzed in the parent dictionary 230 or the stem dictionary 220 and analyze the retrieved parts of speech in the first part-of-speech analysis module or the second part-of-speech analysis module.

The part-of-speech analysis module searches for the left syllable including the syllable to be analyzed in the stem dictionary 220 and transmits the second part-of-speech analysis module of the corresponding part of speech when a suffix, a verb and an adverb are found. In addition, the part-of-speech analysis module analyzes the first part-of-speech of the syllable to be analyzed by searching for a connecting ending, a mother ending, a dividing word, an auxiliary word, and a syllable ending larger than the syllable to be analyzed in the ending dictionary 230. Send to module

The part-of-speech combining information analysis module is a part of a part that analyzes a part-of-speech that does not change the left syllable when the part of speech is not changed through the stem dictionary 220 or the ending dictionary 230. After the analysis, in case of failure, the parts of speech that can change the morphology are analyzed in the module analyzing the parts of speech that can change the morphology. Here, a part-of-speech that does not change the morphology is, for example, an adjective such as 'Oh, that', an adverb like 'as much as it is', 'a mother' such as 'Daniyo, Laniyo, Eurdu', etc. Such as a connecting ending.

3 is a diagram illustrating a morpheme analysis method using a Korean semantic analysis system according to an embodiment of the present invention.

The user interface 100 receives a Korean word of 'really delicious tea' from the user.

When the morpheme analysis unit 300 analyzes the right 'tea' from the 'really tasty tea' which is a spacing error sentence, the search fails from the adverb analysis module 310a to the noun-type derived suffix analysis module 321a. Subsequently, the unregistered word analysis module 322a searches for and stores the left syllable including 'cha' in the stem dictionary 220 of the dictionary database unit 200 as the longest match. The message combining information analysis module 322b searches for the left syllable 'a', except 'difference', as an assistant in the parent dictionary 230, and transmits the left syllable including 'a' to the assistant analysis module 312a.

The assistant analysis module 312a may search for a left syllable including 'a' as the longest match in the parent dictionary 230 of the dictionary database 200, and if the assistant 'a' is searched for an assistant, the left syllable including 'a' may be a noun, Determine if it is an adverb or another part-of-speech.

The assistant analysis module 312a stores the 'A' as an assistant when the 'A' is determined to be an assistant. The assistant combination information analysis module 312b searches for the left syllable 'delicious' except '' as a verb in the stem dictionary 220 and transmits the left syllable including 'delicious' to the verbal auxiliary verb analysis analysis module 320a. .

The verb auxiliary verb analysis module 320a searches for the term 'delicious' for the longest match in the stem dictionary 220 of the dictionary database unit 200. Verb supplementary verb analysis module 320a checks whether the searched word 'delicious' has the information on the feature of the supplementary verb through the stem dictionary 220, and if there is no feature of the supplementary verb is stored in the verb 'delicious'.

The verb auxiliary verb combination information analysis module 320b searches for the left syllable except the verb 'delicious' as an adverb in the stem dictionary 220 and transmits the left syllable including the 'true' to the adverb analysis module 310a.

The adverb analysis module 310a searches for and stores the left syllable including 'really' as the longest match in the stem dictionary 220 of the dictionary database unit 200, and then, if there is no left syllable except the 'really'. Morphological analysis is complete.

As shown in FIG. 4, the error text that is not spaced through this process shows the result of morphological analysis as 'really / delicious + / tea'.

As shown in FIG. 4, the morpheme analysis unit 300 interoperates with the ontology analyzer 600 and the triple repository system 800 to show 'ONTO' information corresponding to the message together with the morpheme analysis result.

The non-registered word proper noun combining unit 400 combines proper nouns such as human names after all morphological analysis is performed. In addition, the non-registered word proper noun combining unit 400 performs a 'left to right' check rather than a 'right to left' check used in morphological analysis in order to estimate and combine the unregistered word and the proper noun from the non-spaced error text. By combining English characters, combining three syllable proper nouns, and combining two syllable proper nouns, they are combined according to a specific combination rule by character or syllable.

As shown in FIG. 5, the word generation unit 500 receives morpheme information from the unregistered noun proper noun combining unit 400, generates words, sentence components using the received morpheme information, and the ontology analyzer 600. ), In conjunction with the triple repository system 800 generates 'ONTO' information of the message and verbs. Here, a word means a sentence component that is separated by a space in a sentence spelled correctly, and according to the part of speech character, NN, verb, VNP, affirmative verb, MM, and adverb. , Interpolation (IC) and conjunction (CONJ).

The correspondence ontology analyzer 610 processes the upper and lower relations and the equality relations of classes, such as 'tea rdfs: subClassOf beverages' (teas belong to beverages) in association with the inference analyzer 630. In other words, the correspondence ontology analyzer 610 constructs upper ontology information such as tea-related favorite products and beverages and lower ontology information such as beverages, honey tea, green tea, plum tea, and herbal tea as the correspondence ontology.

The correspondence ontology analyzer 610 converts the constructed ontology into an RDF triple through the RDF triple converter 700 and stores the generated ontology in the triple repository system 800.

If the sentence to analyze the sentence ontology analysis unit 610, such as 'Samsung + tea', 'Hyundai + tea', 'Kia + tea' in the word generation unit 500, the 'cha' is Each ontology class corresponds to the ONTO information '1132253311' which means' drinking car ', the' ONT information '113229111' which means' transportation vehicle 'and the ONTO information' 1239212 'which means' time' like World War II. Auto build.

In other words, as shown in Figure 7, 'Samsung + tea', 'Hyundai + tea', 'Kia + tea' is not automatically received, the ontology class means 'tea drinking' is automatically built. In addition, although not shown in FIG. 7, ontology classes respectively meaning 'transport vehicle' and 'time' are also automatically constructed.

The verb ontology analysis unit 620 works in conjunction with the inference analyzer 630 to clearly express the exact semantic relationship between the verb and the verb, the subject of the verb, the object of the verb and the range of the bore. In other words, the semantic relations that AJT should have internally are converted into RDF triple form by constructing each ontology according to 2900 semantic information (concrete, non-concrete, animal, plant, food, drink, fruit, etc.). And stored in triple repository system 800 (shown in FIG. 8).

As shown in FIG. 9, the RDF triple converter 700 converts the spoken ontology in the verb ontology analyzer 620 into a set of RDF triples. Similarly, the RDF triple converter 700 converts the spoken ontology in the message ontology analyzer 610 into a set of RDF triples.

The verbal ontology analyzer 620 connects domain, range, and AJT attribute information values with the 'rdf: ID' or 'owl: equivalentClass' values of the verbal ontology analyzer 610. Analyze the meaning of the statement.

For example, the expression 'delicious' can be implied as well as expressing the parent-child relationship and property of a property, such as 'eat' delicious 'rdfs: subProperty0f' ('delicious' belongs to 'eat'). The semantic attribute of the domain can be variously expressed as shown in FIG. 8.

The verbal ontology analysis unit 620 is connected to the 'tea' of the communication ontology in FIG. 6 to receive various upper and lower meaning information of the 'tea' so as to grasp the meaning that the 'tea' is a sub-information of food as a symbol rather than a machine. do.

Therefore, the meaning attribute of 'tea', which is a domain containing 'delicious' in 'delicious tea', is 'drinking tea (1132253311)' rather than 'tea (113229111)' meaning 'transportation vehicle'.

As illustrated in FIG. 10, the output unit 900 outputs ontology semantic analysis results of the message and the ontology information of the input sentence received from the triple repository system 800 from the user interface.

The output unit 900 outputs the upper ontology information (HONTO), the lower ontology information (LONTO), and the equivalent ontology information (EONTO) containing internal meanings through the inference analysis of the ontology analyzer 600.

The embodiments of the present invention described above are not implemented only by the apparatus and / or method, but may be implemented through a program for realizing functions corresponding to the configuration of the embodiment of the present invention, a recording medium on which the program is recorded And such an embodiment can be easily implemented by those skilled in the art from the description of the embodiments described above.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments, It belongs to the scope of right.

Claims (14)

Analyzing the retrieved part-of-speech by searching the stem syllable including the word and the verb or the ending dictionary including the survey or the ending syllable of the Korean input word;
Analyzing the morpheme information of the Korean input word by searching the syllables coupled to the analyzed parts of speech from the stem dictionary or the ending dictionary and analyzing the word information using the analyzed morpheme information;
Constructing ontology semantic information of the spoken word as a ontology, and constructing an ontology of an ontology based on semantic information in relation to the spoken word;
Analyzing the meaning of the verb by connecting the ontology and ontology semantic information of the verb and analyzing ontology information of the verb and the verb through the verb ontology and the verb ontology; And
Outputting ontology information of the message and the verb as an ontology semantic analysis result;
Ontology-based Korean semantic analysis method comprising a. The method of claim 1,
Analyzing ontology information of the message and the verb,
If the term is not found in the analyzed word information, receiving a plurality of ontology semantic information related to the word from the spoken ontology; And
Constructing each ontology class corresponding to each ontology semantic information in a first correspondence combined with the correspondence;
Ontology-based Korean semantic analysis method further comprising. The method of claim 1,
Analyzing the word information,
The ontology-based Korean semantic analysis method of claim 11, wherein the morpheme information and the word information include ontology meaning information of the message and ontology meaning information of the word. The method of claim 1,
The building of the verb ontology,
Converting the communication ontology and the verb ontology into a set of resource description framework (RDF) triples and storing the ontology ontology
Ontology-based Korean semantic analysis method further comprising. The method of claim 1,
The outputting of the ontology semantic analysis result may include:
Outputting the upper ontology information (HONTO), the lower ontology information (LONTO), and the equivalent ontology information (EONTO) containing the internal meaning through the inference analysis of the verb and the verb.
Ontology-based Korean semantic analysis method comprising a. The method of claim 1,
Analyzing the word information,
Analyzing the retrieved parts of speech by searching the stem syllable or the ending dictionary for the syllable to be analyzed in the reverse order of the Korean input word to be analyzed; And
Morphological analysis of the left syllable in the reverse order of the Korean input word by searching for the left syllable coupled to the analyzed part of speech in the stem dictionary or the ending dictionary
Ontology-based Korean semantic analysis method comprising a. The method of claim 6,
Analyzing the word information,
When the left syllable is a part-of-speech that does not change the phonetic form through the stem dictionary or the mother dictionary, after analyzing the parts of speech that do not change the form of the left syllable, the left syllable is formed through the stem dictionary or the mother dictionary. In the case of changeable part-of-speech, analyzing parts of speech that can be changed
Ontology-based Korean semantic analysis method comprising a. Analyze the retrieved parts of speech by searching the stem syllables of the phrases and idioms or the dictionary containing the words or words of the Korean input word, and search the syllables bound to the analyzed parts of speech in the stem or dictionary. A morpheme analysis unit configured to analyze morpheme information of the Korean input word;
A word generator configured to analyze word information and sentence components using the analyzed morpheme information;
Construct the ontology semantic information of the correspondence as a ontology, construct an ontology formed by the ontology according to the semantic relations with the semantic information, connect the ontology with the ontology semantic information of the correspondence Ontology analysis unit for analyzing the meaning; And
An output unit for outputting the ontology information of the verb and the verb as analyzed by the ontology meaning of the verb and the verb as analyzed through the verb ontology and the verb ontology
Ontology-based Korean semantic analysis system comprising a. The method of claim 8,
The ontology analyzer,
A coronation ontology analyzer configured to construct the coronary ontology by forming a correlation between upper ontology information and lower ontology information on the ontology semantic information of the correlator; And
A verb ontology analysis unit for constructing the ontology by constructing semantic relations of domains, ranges, and additional words (AJTs) of the verbs according to semantic information.
Ontology-based Korean semantic analysis system comprising a. 10. The method of claim 9,
When the ontology semantic information is searched for in the statement or the word from the analyzed word information, the ontology analyzer analyzes the ontology information of the message using the verb ontology analyzer, and uses the verb ontology analyzer to detect the ontology. Ontology-based Korean semantic analysis system, characterized in that to analyze the ontology information. The method of claim 8,
An RDF triple conversion unit for converting the communication ontology and the verb ontology into a set of Resource Description Framework (RDF) triples; And
A triple repository system for storing the converted RDF triple set and providing functions for deleting, updating, sorting, and searching the RDF triple set.
Ontology-based Korean semantic analysis system, characterized in that it further comprises. The method of claim 8,
The morpheme analysis unit,
The syllable to be analyzed according to the reverse order of the Korean input word to be analyzed is analyzed in the first part-of-speech analysis module analyzing the survey and the ending when the syllable to be analyzed is searched in the ending dictionary, and analyzing the word and the verb when searching in the stem dictionary A part-of-speech analysis module for analyzing in the part-of-speech analysis module; And
A part-of-speech that analyzes the left syllable through the first part-of-speech analysis module or the second part-of-speech analysis module by searching for the left syllable coupled to the syllable analyzed by the part-of-speech analysis module in the stem dictionary or the mother dictionary Join information analysis module
Ontology-based Korean semantic analysis system comprising a. The method of claim 12,
The part-of-speech analysis information analysis module searches for the left syllable except the syllable to be analyzed as a specific part-of-speech in the stem or dictionary and analyzes the specific part-of-speech in the left syllable including the syllable to be analyzed. Or morphological analysis by transmitting to the second part-of-speech analysis module. The method of claim 12,
The part-of-speech analysis module analyzes the survey or ending in the first part-of-speech analysis module according to the reverse order of the Korean input word to be analyzed, and then, if the survey or ending is not the first part of speech, proclaims and verbs. Ontology-based Korean semantic analysis system characterized in that the analysis in the two parts of speech analysis module.

Images