KR20110017129A - Apparatus and method for words sense disambiguation using korean wordnet and its program stored recording medium - Google Patents

Abstract

PURPOSE: A device and a method for solving ambiguity of a word using a Korean word meaning network are provided to solve ambiguity of a word by extracting an ambiguity word and a relation word of a word expressed in nearby of ambiguity word, thereby increasing accuracy. CONSTITUTION: A pre-processor(12) selects ambiguity vocabulary about an inputted sentence. A vocabulary extension unit(13) extracts relation word of a word expressed by ambiguity vocabulary and a nearby context nearby by using a Korean vocabulary work network. A statistic information obtain unit(14) obtains statistic information for statistically analyzing a relation between the ambiguity vocabulary and the extracted vocabulary. A selecting unit of a word(15) distinguishes meaning of the ambiguity vocabulary by using the obtained statistic information.

Description

Apparatus and Method for Words Sense Disambiguation Using Korean WordNet and its program stored recording medium}

The present invention utilizes the Korean lexical semantic network to solve the word sense neutrality generated in the semantic analysis stage using the Korean lexical semantic network. Apparatus and method for resolving neutrality of words and a recording medium having recorded thereon a program therefor.

Word Sense Disambiguation (WSD) is an operation of determining the meaning of one or more meaningful vocabulary words (hereinafter, referred to as the "gravity vocabulary") in context and accurately classifying the meanings.

There are two ways to resolve the neutrality of language, depending on the language resources used. One is Knowledge-based WSD based on real-world knowledge bases designed for humans, such as Machine Readable Dictionary (MRD) or thesaurus. Corpus-based WSD using corpus, a collection of sentences used.

The neutrality of knowledge-based words can be resolved by using dictionary semantics of machine-readable dictionaries, using semantic categorization provided by thesaurus or machine-readable dictionaries, and bilingual, depending on the language resources used. It can be distinguished by using word correspondence in a bilingual dictionary.

And neutralization of corpus-based fish is a method using statistical information extracted from large corpus. In this method, the language significance problem is simplified to the statistical classification problem in machine learning and solved by applying various traditional machine learning techniques (instance-based learning, decision tree, Bayesian classifier, etc.).

Resolving the neutrality of words through machine learning uses supervised disambiguation and unsupervised disambiguation depending on whether they use a corpus of words with individual meanings for learning. disambiguation).

In general, supervised neutralization in performance of speech neutrality is higher than unsupervised neutrality and knowledge-based neutralization, but large semantic corpus is required.

In addition, since a classifier must be generated for each vocabulary having neutrality, it is not suitable for eliminating neutrality of large-scale words.

The present invention is to solve the problem of the prior art method of resolving the neutrality of words, the word sense neutrality generated in the semantic analysis step among the various neutralities generated in the analysis process of natural language. It is an object of the present invention to provide an apparatus and method for resolving neutrality of a word that can be solved using a Korean lexical semantic network.

The present invention extracts the relational words of the vocabulary appearing around the neutral vocabulary and the neutral vocabulary, and then determines whether or not there is air between them. The purpose is to provide.

The present invention uses the Korean vocabulary semantic network to realize large-sized neutrality in various application fields such as machine translation. And a computer-readable recording medium having recorded thereon a method for resolving the neutrality of the fish and a method and a program for realizing the method. The purpose is to provide.

The present invention provides a device and method for resolving neutrality of a word to remove an inappropriate meaning by using various kinds of information (selection constraint, compound word formation information, full word, partial word, unit expression) of 'Sejong Electronic Dictionary' It is an object of the present invention to provide a computer-readable recording medium having recorded thereon a program for realizing the method.

The present invention relates to a semantic classification system of the Sejong electronic dictionary used for selection restriction information of the Sejong electronic dictionary with KorLx so that the inappropriate meaning can be removed and the method and a program for realizing the method. Its purpose is to provide a computer readable recording medium having recorded thereon.

In order to achieve the above object, the apparatus for resolving the neutrality of words using the Korean vocabulary meaning network according to the first embodiment of the present invention performs a morphological analysis on the input sentence, resolving part-of-speech neutrality, and selects a neutral vocabulary. Pre-processing unit; Vocabulary expansion unit for extracting the relationship between the neutral vocabulary and the vocabulary appearing in the surrounding context using the Korean vocabulary semantic network; Statistical information for statistically analyzing the relationship between the neutral vocabulary and the extracted vocabulary And obtaining a statistical information acquiring unit; a selecting unit for distinguishing a meaning of the neutral vocabulary by using the obtained statistical information.

The preprocessor may perform part-of-speech tagging on an input sentence using a Korean part-of-speech tagger to remove part-of-speech speech.

The preprocessing unit may perform the reference based on whether the corresponding vocabulary has at least two synonym sets in the Korean vocabulary.

The vocabulary expansion unit may expand the vocabulary by acquiring upper words, lower words, type control, and synonyms of the corresponding vocabulary.

The statistical information acquiring unit may obtain a frequency of a related word of a vocabulary expressed together with a neutral vocabulary and a surrounding context, and each of these air vocabulary frequencies from a statistical dictionary.

According to a first aspect of the present invention, there is provided a method of resolving a word's neutrality using a Korean vocabulary meaning network, comprising: performing part-of-speech tagging when a sentence is input and selecting a neutral vocabulary from the analyzed sentence; Extracting the relational words of the neutral vocabulary and the vocabulary appearing in the surrounding context using the vocabulary meaning network; obtaining statistical information for statistically analyzing the association between the vocabularies; And outputting a sentence in which the significance of the word is resolved by dividing the meaning.

The selection of the neutral vocabulary may be performed based on whether the corresponding vocabulary has at least two synonym sets in the Korean vocabulary network.

The relational word may include a higher word, a lower word, a type control, and a synonym of the corresponding vocabulary.

And the statistical information, characterized in that it comprises a frequency for the relative words of the vocabulary appearing together with the neutral vocabulary and the surrounding context and their respective air vocabulary frequency.

Further, in order to classify the meaning of the corresponding neutral vocabulary by analyzing the relationship between the vocabularies, the Korean relation words of the neutral vocabulary and the Korean relation words of the vocabulary appearing in the context around the neutrality are used.

In addition, in order to distinguish the meaning of the corresponding neutral vocabulary by analyzing the relationship between the vocabularies, the English relation words of the neutral vocabulary and the English relation words of the vocabulary appearing in the context around the neutrality are used.

And in order to distinguish the meaning of the corresponding neutral vocabulary by analyzing the relationship between the vocabulary, it is characterized by using only the upper and lower words above or above the upper and lower words.

In addition, in order to distinguish the meaning of the corresponding neutral vocabulary by analyzing the relationship between the vocabularies, the relation words of the neutral vocabulary and the vocabulary shown in the surrounding context are used.

According to a second aspect of the present invention, there is provided an apparatus for resolving neutrality of a word using a Korean vocabulary meaning network, including: a preprocessor for selecting a neutral vocabulary using a Korean vocabulary meaning network; A first neutrality vocabulary processor for determining and extracting form / meaning / syntax information and selection constraint information and extracting a relational word to expand a vocabulary; an association analysis for selecting a meaning of a neutral vocabulary by analyzing correlations between the extended vocabularies A second step of determining a type of neutral vocabulary based on an air vocabulary, extracting form / meaning / syntax information, selection constraint information, and extracting a relational word based on an analysis result of the correlation analysis unit and extending the vocabulary; And a neutral vocabulary processing unit.

Here, the preprocessor selects a neutral vocabulary based on whether or not it has at least two synonyms in the Korean vocabulary network or based on whether it has at least two headings in the Sejong Electronic Dictionary. Characterized in that.

The first neutral vocabulary processing unit includes a first vocabulary type determination unit for determining whether the neutral vocabulary is a word or a verb, and if the neutral vocabulary is 'word', each type / neutral vocabulary in the Sejong Electronic Dictionary A first information extraction unit for extracting vocabulary appearing in semantic / syntactic information, and extracting the arguments appearing in the selection constraint information for each meaning of the neutral vocabulary in the Sejong electronic dictionary if the neutral vocabulary is 'word', and the Korean vocabulary meaning network (KorLex) is used to extract the relational words of the vocabulary extracted by the first information extraction unit, and KorLex synonyms (synset) corresponding to the arguments extracted by the first information extraction unit using the mapping information of KorLex and Sejong Electronic Dictionary. And extracting the first vocabulary extension unit to expand the vocabulary.

The second neutral vocabulary processing unit may include a second vocabulary type determination unit determining whether the neutral vocabulary is a word or a verb, and if the neutral vocabulary is 'word', each type of neutral vocabulary in the Sejong Electronic Dictionary A second information extraction unit for extracting vocabulary appearing in semantic / syntactic information, and extracting the argument appearing in the optional constraint information for each meaning of the neutral vocabulary in the Sejong electronic dictionary if the neutral vocabulary is 'word', and the Korean vocabulary meaning network (KorLex) is used to extract the relational words of the vocabulary extracted by the second information extraction unit, and KorLex synonyms (synset) corresponding to the arguments extracted by the second information extraction unit using the mapping information of KorLex and Sejong Electronic Dictionary. And extracting the second vocabulary expansion unit to expand the vocabulary.

The relational words may include upper words, lower words, synonyms, and siblings of the neutral vocabulary.

According to a second aspect of the present invention, there is provided a method for resolving neutrality of a word using a Korean vocabulary meaning network, comprising: selecting a neutral word from an input sentence and determining the type of the neutral word; Extracting the vocabulary appearing in the form / meaning / syntactic information and the argument appearing in the optional constraint information according to the type of sexual vocabulary; if the vocabulary appearing in the form / meaning / syntax information is extracted using the Korean vocabulary meaning network Extracting a relational word and extracting a term, and expanding a vocabulary by extracting a synonym synonym of the Korean vocabulary meaning network corresponding to the argument; analyzing a relationship between the extracted vocabulary and resolving the significance of the word Outputting; characterized in that it comprises a.

Here, the selection of the neutral vocabulary is based on whether or not there are at least two synonyms in the Korean vocabulary network, or based on whether or not the semantic vocabulary has at least two heading divisions. It features.

If the neutral vocabulary is 'word', the Sejong electronic dictionary extracts the vocabulary that appears in the form / meaning / syntactic information for each meaning of the neutral vocabulary.If the neutral vocabulary is 'word', each of the neutral vocabulary in the Sejong electronic dictionary It is characterized by extracting arguments appearing in the selection constraint information for each meaning.

In the case of extracting the vocabulary appearing in the form / meaning / syntax information, the vocabulary extension is extracted by extracting the upper words, lower words, synonyms, and sibling words of the extracted vocabulary using the Korean Lexical Meaning Network (KorLex). It is characterized by extending the vocabulary by extracting a synonym of the Korean vocabulary meaning network (KorLex) corresponding to the extracted argument using the Korean vocabulary meaning network (KorLex) and mapping information of the Sejong Electronic Dictionary.

And in the step of analyzing the relationship between the extracted vocabulary, it is characterized by scoring the number of words matched between the extracted neutral vocabulary list and the air vocabulary, and then select the meaning of the neutral vocabulary with a high score.

And if the scores of the semantic vocabulary calculated in the step of analyzing the association between the extracted vocabulary coincide, the type of the neutral vocabulary is determined based on the air vocabulary, the form / meaning / syntactic information and the selection constraint information are extracted and related. It is characterized by repeating the process of extracting words to expand the vocabulary.

The apparatus and method for resolving the neutrality of words using the Korean vocabulary meaning network according to the present invention and a recording medium recording a program therefor have the following effects.

First, it is possible to solve the problem of disambiguation of language in various applications such as machine translation and information retrieval where language importance is a problem, so that a high-performance system can be developed in each application field.

Second, it is possible to improve the accuracy by extracting the related words of the vocabulary appearing around the neutral vocabulary and the neutral vocabulary, and then determining the presence or absence of air between them.

Third, it is advantageous in terms of construction time and cost of the neutrality means of words by applying the method of using relational words when extending the vocabulary appearing around the neutral and neutral words.

Fourth, it is easy to apply the actuality of the application program by using the Korean lexical semantic network.

Fifth, it is possible to improve the accuracy of neutralization by removing inappropriate meanings by using various information (selective pharmaceutical, compound word formation information, whole words, partial words, unit expressions) of the Sejong Electronic Dictionary.

Hereinafter, a preferred embodiment of a device and method for resolving speech neutrality using a Korean vocabulary meaning network according to the present invention and a recording medium recording a program therefor will be described in detail.

The features and advantages of the apparatus and method for resolving speech neutrality using the Korean vocabulary meaning network according to the present invention and the recording medium recording the program therefor will be apparent from the detailed description of each embodiment below.

The present invention relates to resolving the neutrality of the neutral vocabulary by extracting the related words of the neutral vocabulary and the vocabulary appearing around the neutral vocabulary according to the first embodiment and then statistically determining whether there is air between them. The method and a method of removing inappropriate meanings by using various information (selection constraint, compound word formation information, full word, partial word, unit expression) of the 'Sejong Electronic Dictionary' according to the second embodiment.

(First embodiment)

1A and 1B are flowcharts illustrating a configuration diagram of a neutral neutralization device of a fish and a method for neutralization of a fish according to a first embodiment of the present invention.

A first embodiment of the present invention includes a method of using a relational word when extending a vocabulary appearing around a neutral vocabulary and a neutral vocabulary using a Korean vocabulary meaning network to resolve the significance of a word that is represented in a Korean sentence. do.

This solves the problem of cost and construction time use of the supervised neutralization method using a large-scale semantic corpus that shows good performance, and simultaneously solves the performance degradation problem of the non-supervised neutralization method.

According to the first embodiment of the present invention, the apparatus for resolving the neutrality of a word using the Korean vocabulary meaning network uses a relationship between a neutral word and a vocabulary appearing in a surrounding context to distinguish the meaning of the neutral word. Means for distinguishing the meaning of the neutral vocabulary used in a sentence using the related word of the related word and the vocabulary appearing in the surrounding context.

The structure of the neutral neutralizing device for fish according to the first embodiment of the present invention is the same as in FIG. 1A.

First, the apparatus for resolving verbality of speech using the Korean vocabulary meaning network performs a sentence input unit 11 for receiving a sentence for resolving verbality, performing morphological analysis and resolving partiality of the input sentence. Preprocessing unit 12 for selecting sexual vocabulary, vocabulary expansion unit 13 for extracting the relational words of the neutral vocabulary and the surrounding contexts using the Korean vocabulary network, and using the expanded vocabulary A statistical information acquisition unit 14 for acquiring statistical information from the constructed statistical dictionary, a language selection unit 15 for classifying the meaning of the neutral vocabulary using the acquired statistical information, and the significance of the word is eliminated. A sentence output unit 16 for outputting a sentence is included.

The neutralization process of the fish in the neutralization solution for fish according to the first embodiment of the present invention having such a configuration is as follows.

As shown in FIG. 1B, when a sentence is input (S101), the part-of-speech tagging is performed using the Korean part-of-speech tagger (S102) and the neutral vocabulary is selected from the analyzed sentence.

In addition, the relational words (upper words, lower words, siblings, synonyms) between the neutral vocabulary and the surrounding context are extracted using the Korean vocabulary meaning network (S104).

Subsequently, statistical information for statistically analyzing the association between the words is obtained.

In addition, the correlation between the words is analyzed (S106), and the meanings of the corresponding words are classified to output a sentence in which the significance of the word is eliminated.

Here, when analyzing the relationship between the neutral vocabulary and the vocabulary appearing in the surrounding context, the Korean relational words (upper words, lower words, siblings, synonyms) of the neutral vocabulary and the Korean relational words of the vocabulary appearing in the context surrounding the neutrality (parent words) , Subwords, siblings, and synonyms).

And when analyzing the relationship between the neutral vocabulary and the vocabulary appearing in the surrounding context, the English relational words (upper, lower words, siblings, synonyms) of the neutral vocabulary and the English relational words (parent), Subwords, siblings, and synonyms).

And when analyzing the relationship between the neutral vocabulary and the vocabulary that appears in the surrounding context, the upper and lower words of the neutral vocabulary are used, or the upper and lower words are used.

And when analyzing the relationship between the neutral vocabulary and the vocabulary expressed in the surrounding context, the relational word of the neutral vocabulary and the vocabulary expressed in the surrounding context can be used.

The operation of the respective building blocks of the device for resolving the neutrality of words using the Korean vocabulary meaning network according to the present invention will be described in more detail.

First, the preprocessor 102 performs part-of-speech tagging on a sentence input through the sentence input unit 11 using Korean part-of-speech tagging. Part-of-speech tagging is a method of analyzing natural language and removes part-of-speech speech.

After the part-of-speech tagging, the neutral vocabulary is selected from the analyzed sentences. The criterion for selecting a neutral vocabulary is whether the vocabulary has two or more synonyms in the Korean vocabulary network.

The lexical expansion unit 103 extracts the relational words (upper words, lower words, siblings, synonyms) of the neutral vocabulary and the vocabulary appearing in the surrounding context using the Korean vocabulary semantic network. The Korean vocabulary semantic network is built on Princeton University's WordNet as a reference model, and it shows the relationship between upper, lower, sibling and synonyms of a specific vocabulary. In addition, because it uses a synthesized number such as WordNet, it is possible to know a band word of a specific vocabulary.

Since the Korean vocabulary meaning is expressed based on meaning, it expresses different conceptual relations even for neutral vocabulary.

For example, in the case of 'apple', conceptual relationships are expressed by meaning as 'apple 2' (22) and 'apology 2 apple 1' (32). 'Apple 2' means 'fruit of the apple tree' and has 'fruit 1 fruit 1 (21)' as a parent word and 'dolgeumgeum (23)' as a lower word, and various siblings such as 'fall 1'. Has 24). Siblings are vocabulary words that have the same superword as themselves.

On the other hand, 'apology 2 apology 1' means 'acknowledgment of one's fault and forgiveness of forgiveness', which has 'cognition (31)' as a higher word and no lower word (33), Have the same multiple siblings 34.

Since each set of the Korean lexical semantic network has the same set number as WordNet, it is possible to obtain a band word of a specific lexicon using the Korean lexical semantic network and WordNet.

'Apple 2' (22) is connected to 'apple1' (42), the upper word 'fruit 1 fruit 1 (21)' is connected to 'fruit1' (41), 'dunggeum (23)' is connected to 'Crabapple3' Connected. In addition, it is possible to obtain English siblings 44, such as berry1.

In this way, the vocabulary expansion unit 13 expands the vocabulary by obtaining Korean upper words, Korean lower words, Korean siblings, Korean synonyms, English upper words, English lower words, English siblings, and English synonyms of a specific vocabulary.

Next, the statistical information acquisition unit 14 obtains statistical information on the vocabulary expanded by the vocabulary expansion unit 13 from the statistical dictionary that was instrumented. This statistical information is used to statistically analyze the association between the two vocabularies.

The semantics of the neutral vocabulary are strongly related to the air vocabulary shown in the surrounding context. Therefore, when a neutral vocabulary appears in a sentence, the meaning of the neutral vocabulary can be distinguished according to what the vocabulary appears in the surrounding context.

In quantitative linguistics, the most common way to measure the co-occurrence of two words is to use hypothesis testing. Hypothesis testing is a statistical analysis that uses statistics to reveal the truth of hypotheses.

To measure the presence or absence of air in two vocabulary using the hypothesis test, the hypothesis to be tested should be set as follows.

의 출현과 어휘

의 출현이 상호독립적이라는 것을 뜻하고, 가설 2는

의 출현이 어휘

의 출현에 종속적이라는 것을 뜻한다. 이때, 가설 1이 가설 검정의 대상이 되는데 이 가설을 귀무가설(null hypothesis)이라고 한다. Hypothesis 1 is

Emergence and vocabulary

Means that the emergence of is mutually independent, hypothesis 2

The emergence of this vocabulary

It is dependent on the appearance of Hypothesis 1 is subject to the hypothesis test, which is called the null hypothesis.

의 출현이 어휘

의 출현에 종속적이라는 것이 된다.Hypothesis 2, on the other hand, is an alternative hypothesis that is accepted when the null hypothesis is rejected. If hypothesis 1 is rejected by the hypothesis test, hypothesis 2 is accepted.

The emergence of this vocabulary

Is dependent on the appearance of.

통계치보다 좀 더 해석이 직관적이다.In the present invention, a hypothesis testing method using likelihood ratio is used among several methods for hypothesis testing. Likelhood ratio is

More intuitive than statistics.

-test 보다 자료 부족 문제에 더 강하다.That is, it shows how much more likely one hypothesis is than another hypothesis. In addition, the hypothesis test using likelihood ratio

Stronger for data shortages than -test

,

를 각각 어휘

,

의 출현 빈도라고 하고

를 두 어휘의 공기 빈도라고 할 때, 확률

는 최대 우도 추정(maximum likelihood estimation)에 의해 다음과 같이 구한다.

,

Each vocabulary

,

Is the frequency of appearance of

When is the air frequency of two vocabularies, the probability

Is obtained by the maximum likelihood estimation as follows.

은 원시말뭉치의 전체 어절 수이다.At this time,

Is the total word count of the primitive corpus.

는 전체 모수 공간상의 maximum likelihood 함수값에 대한 가설에 따라 표현되는 일부 공간(subspace)상의 maximum likelihood 함수값의 비율로써 다음과 같이 표현된다.Likelihood ratio

Is the ratio of the maximum likelihood function value in some subspace expressed according to the hypothesis of the maximum likelihood function value in the overall parameter space.

는 모수 공간

의 한 점(point)이고

는 가설에 따라 표현되는 일부 공간이다. Likelihood ratio의 중요한 특징은

가 점근적으로

분포를 따른다는 것이다. 특히, 이항분포에서 이 점근선은 매우 빠르게 접근하게 된다.At this time,

Parameter space

Is a point of

Is some space represented by the hypothesis. An important feature of the Likelihood ratio

Gradually

It follows the distribution. In particular, in the binomial distribution, this asymptote is very fast.

에서 수학식 4로 변환될 수 있다.Equation 3 is binomial distribution

It can be converted into equation (4) in.

In fact, in hypothesis 1 and hypothesis 2,

이므로 수학식 4는 수학식 5에 의해 다음과 같이 변환된다.

Equation 4 is converted into Equation 5 as follows.

는 자유도가 1인

분포를 따르고 가설 1의 기각 여부를 결정짓는 판단 근거로서

분포를 사용할 수 있다.As above

Has 1 degree of freedom

As a basis for judgment that follows the distribution and determines whether to dismiss hypothesis 1

Distribution can be used.

분포에서 유의수준이 0.005이고 자유도가 1일 때, 임계값(critical value)은 7.88이므로

가 7.88 이상이 되면 가설은 기각되고, 두 어휘는 연관성을 가진 어휘로 판별된다.In other words,

When the significance level in the distribution is 0.005 and the degree of freedom is 1, the critical value is 7.88.

If is equal to or greater than 7.88, the hypothesis is rejected and the two vocabularies are identified as related vocabularies.

In order to calculate the equation (6), the statistical information acquisition unit 14 obtains the frequencies for the relation words of the neutral words and the vocabularies that appear together with the surrounding context and the respective air vocabulary frequencies from the statistical dictionary.

In addition, the word selector 15 calculates Equation 6 based on the statistical information obtained by the statistical information acquirer 14. Equation 6 is to statistically analyze the association between the vocabulary as shown in FIG.

In FIG. 5, the sentence 51 'to eat an apple' includes a neutral vocabulary of 'apple'.

'Apple' can be interpreted in two meanings: 'apple' (52) and 'apology' (53), and the meaning of 'apple' can be distinguished by analyzing the relationship between the semantic relation of the apple and the surrounding context. . In Figure 5, 'eat fruit', 'eat peach', 'eat tangerine' is meaningfully related (54), 'eat cognition', 'eat sake', 'eat thanks', It is irrelevant in the sense of `` to eat apology '' and `` to refuse '' (55).

There is more than one peripheral vocabulary that appears with a neutral vocabulary in sentences that contain a neutral vocabulary.

과 주변 문맥 내 공기 어휘

의

를

라고 할 때,

의 관계어

에 의해

은 다음과 같다고 가정한다.A neutral vocabulary

Air vocabulary in and surrounding context

of

To

When I say

Relative

By

Assume that

Figure 6 is the result of analyzing the "ate one apple" based on the equation (7). The meaning of the Chinese word 'Apple' and the related words 'Han', 'Dog', and 'Eat' appear together in Korean vocabulary, and their statistical information is obtained from a statistical dictionary. Calculate 7. As a result, in the sentence of FIG. 6, the number of words related to 'apple' is three and the number of words related to 'appology' is one. The meaning can be distinguished by 'apple'.

The apparatus and method for resolving the neutrality of words using the Korean vocabulary meaning network according to the first embodiment of the present invention are related words of the vocabulary appearing around the neutral vocabulary and the neutral vocabulary using the Korean vocabulary semantic network. , Subwords, siblings, and synonyms), and then judges whether there is air between them and resolves the neutrality of the words.

(Second embodiment)

Hereinafter, an apparatus and method for resolving neutrality of words using the Korean vocabulary meaning network according to the second embodiment of the present invention will be described.

7A and 7B are flowcharts showing the composition of the neutral neutralization device for fish according to the second embodiment of the present invention and a method for neutralization of the fish.

FIG. 8 is a comparison table for resolving neutrality of words using information of the Sejong electronic dictionary, and FIG. 9 is a block diagram of a semantic classification system of the Sejong electronic dictionary.

The method of resolving the neutrality of a word using the Korean vocabulary meaning network according to the second embodiment of the present invention uses various information (selective constraint, compound word formation information, whole word, partial word, unit expression) of 'Sejong Electronic Dictionary'. Includes ways to eliminate neutrality of a word so that inappropriate meanings can be removed.

And optionally extracting relational words (upper words, lower words, sibling words, synonyms) of the vocabulary according to the first embodiment, and determining whether there is air between them and further eliminating the neutrality of the words.

According to a second embodiment of the present invention, the apparatus for resolving the neutrality of words, as shown in FIG. 7A, includes a sentence input unit 71 for receiving a sentence to be resolved, and at least two synonyms in the Korean vocabulary network. The preprocessor 72 selects the neutral vocabulary based on whether or not the set has a synset, and based on whether the sejong electronic dictionary has at least two heading sections, and determines the type of the neutral vocabulary. A first neutrality vocabulary processing unit 700 for extracting meaning / syntactic information and selection constraint information and extracting a relational word to expand the vocabulary, and an association analysis unit for selecting the meaning of the neutral vocabulary by analyzing the relationship between the vocabularies ( 76), a sentence output unit 77 for outputting a vocabulary in which the importance of the word selected by the association analysis unit 76 is eliminated, and a neutrality based on the air vocabulary according to the analysis result of the association analysis unit 76. Sell vocabulary types And a second neutral vocabulary processing unit 710 for extracting form / meaning / syntax information and selection constraint information, extracting relational words, and expanding the vocabulary, and outputting the result to the correlation analysis unit 76. do.

Here, the first neutral vocabulary processing unit 700 determines the first vocabulary type determination unit 73 for determining whether the neutral vocabulary is a word or a verb, and the neutral vocabulary in the Sejong electronic dictionary if the neutral vocabulary is 'word'. A first information extracting unit extracting a vocabulary appearing in form / meaning / syntactic information for each meaning and extracting a topic appearing in the selection constraint information for each meaning of the neutral vocabulary in the Sejong electronic dictionary if the neutral vocabulary is 'term' 74) and extracts the relational words (upper words, lower words, synonyms, siblings) of the vocabulary extracted by the first information extracting unit 74 using the Korean Vocabulary Meaning Network (KorLex) and maps KorLex to the Sejong Electronic Dictionary. The first vocabulary expansion unit 75 extracts a synonym synonym of KorLex corresponding to the argument extracted by the first information extracting unit 74 using the information to expand the vocabulary.

The second neutral vocabulary processing unit 710 may include a second vocabulary type determination unit 78 that determines whether the neutral vocabulary is a word or a verb, and each of the neutral vocabulary words in the Sejong electronic dictionary if the neutral word is a 'word'. A second information extracting unit extracting a vocabulary appearing in form / meaning / syntactic information by meaning, and extracting the argument appearing in the selection constraint information for each meaning of the neutral vocabulary in the Sejong electronic dictionary if the neutral vocabulary is 'word' (79) ) And the related words (upper words, lower words, synonyms, siblings) of the vocabulary extracted by the second information extracting unit 79 using the Korean lexical meaning network (KorLex), and mapping information between KorLex and the Sejong Electronic Dictionary. A second vocabulary expansion unit 80 for extracting synonyms of KorLex corresponding to the arguments extracted by the second information extracting unit 79 to expand the vocabulary is used.

The process for eliminating the neutrality of the fish in the neutral elimination device of the fish according to the second embodiment of the present invention having such a configuration is as follows.

As shown in FIG. 7B, when a sentence is input (S701), the part-of-speech tagging is performed using the Korean part-of-speech tagger (S702) and the neutral vocabulary is selected from the analyzed sentence.

Neutral vocabulary selection proceeds based on whether the Korean vocabulary network has at least two synonyms, and whether the Sejong Electronic Dictionary has at least two heading divisions.

Then, it is determined whether the neutral vocabulary is a verb or a verb (S704).

If the neutral vocabulary is 'word', the vocabulary appearing in the form / meaning / syntax information is extracted for each meaning of the neutral vocabulary in the Sejong electronic dictionary.

If the neutral vocabulary is 'word', then the sejong electronic dictionary extracts the arguments appearing in the selection constraint information for each meaning of the neutral vocabulary (S706).

In this way, when information extraction is performed in step S705, a relational word (upper word, sub-word, synonym, sibling) of the extracted vocabulary is extracted using KorLex, and when information extraction is performed in step S706, KorLex and Sejong electronic dictionary are extracted. A synonym set of KorLex corresponding to the extracted argument is extracted using the mapping information (S707).

Then, the relationship between the vocabulary is analyzed, scored by the number of words matched between the extracted list of neutral vocabulary (an item is a set of subwords of the corresponding synset) and the air vocabulary, and then the meaning of the neutral vocabulary having a high score is selected (S708). ), And outputs a sentence in which the significance of the word is removed (S709).

If the scores of the semantic vocabulary calculated in step S708 coincide with each other, it is determined based on the air vocabulary whether the neutral vocabulary is spoken or used.

If the neutral vocabulary is 'word', Sejong electronic dictionary extracts the vocabulary appearing in the form / meaning / syntactic information for each meaning of the neutral vocabulary (S711).

If the neutral vocabulary is 'word', the Sejong Electronic Dictionary extracts the argument appearing in the selection constraint information for each meaning of the neutral vocabulary (S712).

In this way, when information extraction is performed in step S705, a relational word (upper word, sub-word, synonym, sibling) of the extracted vocabulary is extracted using KorLex, and when information extraction is performed in step S706, KorLex and Sejong electronic dictionary are extracted. A synonym set of KorLex corresponding to the extracted argument is extracted using the mapping information (S713).

Then, the relationship between the vocabulary is analyzed, scored by the number of words matched between the extracted list of neutral vocabulary (an item is a set of subwords of the corresponding synset) and the air vocabulary, and then the meaning of the neutral vocabulary having a high score is selected (S708). ), And outputs a sentence in which the significance of the word is removed (S709).

The apparatus and method for resolving the neutrality of words using the Korean vocabulary meaning network according to the second embodiment of the present invention provide a method for resolving between the vocabulary and the air vocabulary indicated in the form / syntax / meaning information for each semantic word in the dictionary. Contrary to the resolution of the word's importance based on the number of matching vocabulary and the expansion of the words appearing in the form / syntax / meaning information using the Korean vocabulary network. To have a skill.

In addition, by using the dispute information of Sejong electronic dictionary to resolve the significance of the word, mapping the argument with the new set of the Korean vocabulary meaning network eliminates the word's importance based on the number of vocabulary words matching the subwords of the set and the air vocabulary. It includes technology to do.

The Sejong Electronic Dictionary consists of 600,000 items as a result of the '21st Century Sejong Plan' conducted for 10 years from 1998 to 2007. The Sejong electronic dictionary is described in XML format, and its parts are different in structure.

A verb is a key element of a sentence along with a verb, which is divided into common nouns, proper nouns, dependent nouns, classifiers, pronouns, and rhetoric. However, the dictionary of sejong electronic dictionaries used vocabulary belonging to common nouns.

Proverbs play a key role in analyzing the structure of a clause, whether in theoretical linguistics or syntactic analysis. Most verbs determine the key points, such as how many arguments are needed in a section, what signs (such as a cover letter, a bulletin board, etc.) appear, and what kind of meaning the nouns can be placed in each argument position.

For example, the meaning of the object argument plays a decisive role in grasping the meaning of the verb “tada” in the next sentence composed of the verb “tada”, which is a representative neutrality predicate.

A. Bob took the subway.

N. Withdrawal of gayageum.

C. Bob rode coffee.

In other words, the meaning of ride is divided into 'use / rise', 'play', 'preparation', etc., depending on whether the object argument is 'transportation', 'instrument', or 'drink'. do.

Therefore, to correctly understand the meaning of 'ta' and interpret the meaning of the sentence correctly, not only must there be a precise and explicit description of the argument selection constraint according to the meaning of the verb 'ta', that is, the meaning of the argument. You also need to know precisely and explicitly about each meaning.

To this end, the Sejong Electronic Dictionary and the Dictionary of Words and Verbs are composed of the following information.

(Message dictionary)

Isomorphism, derivative formation, compound word formation, adjective combination information, noun combination information, verb combination information, investigation constraints, etc.

(Word dictionary)

Isomorphism, derivative formation, compound word formation, adjective combination information, noun combination information, verb combination information, investigation constraints, etc.

Hereinafter, the neutrality of the fish using information of the Sejong electronic dictionary according to the second embodiment of the present invention will be described in more detail.

Since the neutral vocabulary has different forms / meanings / syntactics information according to the meaning, it is possible to solve the neutrality of the word by using the message information and the verb information of the Sejong electronic dictionary.

FIG. 8 extracts information about a representative Chinese word 'apple' from the Sejong Information Dictionary and compares each information according to the meaning of 'apple'.

For example, to distinguish the meaning of 'apple' in sentences such as 'apple is sweet', 'apple is sour' or 'apple one', 'apple one', etc. This can be done by verifying that 'shall', 'egg', 'dog', etc. appear in the form / meaning / syntactic information of a specific meaning of 'apple'.

In this way, neutrality can be resolved using the shape / meaning / syntactic information of the neutral vocabulary itself, but on the contrary, neutrality can be resolved using the shape / meaning / syntax information of the air vocabulary of the neutral vocabulary. have.

For example, when trying to distinguish the meaning of 'apple' in the sentence 'apple is red', the form / meaning / syntactic of 'red' is based on 'red' rather than on 'apple' information. This can be done by checking if the apple appears in the information.

The disadvantage of this method, however, is that it is based on string comparisons. In other words, the exact match between the neutral vocabulary and the air vocabulary that appears with the neutral vocabulary can cause data shortage problems.

For example, in the sentence “Apple is’ 다 ”, the air vocabulary“ 뻘겋다 ”cannot be identified in the form / meaning / syntax information of“ Apple 1 ”in the Sejong Electronic Dictionary.

However, '뻘겋다' is actually a variation of 'red' and has the same meaning. Therefore, if you expand the vocabulary in the form / meaning / syntactic information of the Sejong electronic dictionary, you can solve the problem of lack of data. In this case, KorLex is used to expand the vocabulary.

The mapping between the Sejong electronic dictionary and KorLex is important for the expansion of the vocabulary in the form / meaning / syntactic information, but it is also important to use the choice constraint information of the dictionary to resolve the significance of the word.

Choice constraints are constraints that occur when two words are in a sentence.

For example, the word 'death' is used only for 'creature' and the word 'death' is used to ensure that there are no semantic contradictions when two vocabulary depend upon it, such as its use only for 'man'. It is a constraint.

In the dictionary of Sejong electronic dictionaries, such selection constraint information is constructed for about 30,000 semantic unit verbs. For example, if you look for `` red '' in the Sejong electronic dictionary,

It can be seen that it has selection constraint information called <selection_constraint arg = "X" tht = "THM"> sphere | phenomena </ selection_constraint>. However, it is impossible to use this information as it is. This is because, in the optional constraint information, arguments of verbs (in the example above, 'concrete' and 'controversy') are represented using the semantic class system of the Sejong Electronic Dictionary.

The Sejong semantic class system used in constructing Sejong electronic dictionaries strictly and precisely divides the semantic domain represented by Korean vocabulary, and uses the vocabulary based on the definition of the semantic domain and the vocabulary shared for each semantic domain. It is a hierarchical system of hierarchical vocabulary and semantics tied together.

That is, as shown in FIG. 9, the semantic domain represented by the Korean vocabulary is divided into five major classes such as <concrete>, <group>, <place>, <abstract object>, and <event>. For example, in the case of <Natural Food> out of <Concrete>, more detailed semantic areas such as <Vegetable>, <Fish>, <Meat>, <Fruit>, <Grain>, <Nuts>, etc. It is divided into three groups with the Sejong semantic classification system.

The Sejong semantic class also has a hierarchical structure with a depth of at least two stories up to seven stories above the top node. In the Sejong Electronic Dictionary, which was released in 2007, 25,458 headings are divided into 35,854 meanings according to the Sejong semantic category. However, this also causes data shortage problems in resolving the neutrality of words based on string matching.

On the other hand, KorLex, which maps to the Sejong semantic class, contains 90,134 synonym sets and 102,358 word meanings for nouns, with a massive hierarchy with a depth of up to 17 levels. Korean noun vocabulary. Therefore, by mapping the semantic class system of the Sejong electronic dictionary and KorLex, it is possible to maximize the utilization of the selection constraint information of the Sejong electronic dictionary.

The method for resolving the neutrality of a word using the Korean lexical meaning network according to the present invention described above can be embodied as computer readable codes on a computer readable recording medium.

For example, in a computer-readable recording medium, when a sentence is input, a part-of-speech tagging is performed and a neutral vocabulary is selected from the analyzed sentence. Extracting a relational word of a word, acquiring statistical information for statistically analyzing the relationship between the vocabularies, and analyzing the relationship between the words to classify the meanings of the vocabulary, The program for executing the outputting step is recorded.

In addition, the computer-readable recording medium may include selecting the neutral vocabulary from the input sentence and determining the type of the neutral vocabulary, the vocabulary appearing in the form / meaning / syntax information according to the type of the neutral vocabulary, and the selection constraint. Extracting a term appearing in the information, and extracting a related term of the extracted vocabulary using a Korean vocabulary network in case of extracting a vocabulary appearing in the form / meaning / syntactic information; A program for executing a step of extracting a synonym synonym of a vocabulary meaning network, and outputting a sentence in which the word significance is resolved by analyzing the association between the extracted vocabulary is recorded.

The program recorded on the computer-readable recording medium is not limited to the above description, but may include other contents for the execution of the method according to the embodiment of the bit allocation method in the multi-input / output communication system of the present invention. Of course.

Computer-readable recording media include all kinds of recording devices that store data that can be read by a computer system. Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disks, optical data storage devices, and the like, which are also implemented in the form of carrier waves (for example, transmission over the Internet). Include. The computer readable recording medium can also be distributed over network coupled computer systems so that the computer readable code is stored and executed in a distributed fashion.

It will be apparent to those skilled in the art that various modifications and variations can be made in the present invention without departing from the spirit or scope of the invention.

Therefore, the technical scope of the present invention should not be limited to the contents described in the embodiments, but should be defined by the claims.

1A and 1B are flowcharts showing the construction of the neutralization device of a fish and the method for neutralization of a fish according to the first embodiment of the present invention.

Figure 2 is an example of the expansion of the Korean vocabulary of 'apple 2' extracted from the Korean vocabulary meaning network used in the present invention

3 is an example of expanding the Korean vocabulary of 'apple 1' extracted from the Korean vocabulary meaning network used in the present invention

4 is an example of expanding the English vocabulary of 'apple 1' extracted from the Korean vocabulary meaning network used in the present invention

5 is an exemplary view showing the relationship between each meaning of the Chinese-language vocabulary and the vocabulary appearing together in order to solve the neutrality of the language to which the present invention is applied.

6 is a diagram showing whether there is air between a neutral vocabulary and a vocabulary that appear together in a surrounding context by using a relational word extracted from the Korean vocabulary meaning network according to the present invention.

7A and 7B are flowcharts showing the construction of the neutralization device of a fish and the method for neutralization of a fish according to the second embodiment of the present invention.

8 is a comparison table for solving the neutrality of the fish using the information of the Sejong electronic dictionary

9 is a block diagram of a semantic classification system of the Sejong electronic dictionary

Explanation of symbols for the main parts of the drawings

11. Sentence input section 12. Preprocessor

13. Vocabulary expansion unit 14. Statistical information acquisition unit

15. Selection of words 16. Sentence output

Claims (26)

A preprocessor configured to perform morphological analysis and resolving part-of-speech neutrality on input sentences and to select a neutral vocabulary; A vocabulary expansion unit for extracting relational words of a neutral vocabulary and a vocabulary appearing in a surrounding context using a Korean vocabulary semantic network; A statistical information obtaining unit obtaining statistical information for statistically analyzing the association between the neutral vocabulary and the extracted vocabulary; And a language selection unit for classifying the meaning of the neutral vocabulary using the obtained statistical information. The method of claim 1, wherein the pretreatment unit, An apparatus for resolving disambiguation using a Korean lexical semantic network, characterized in that part-of-speech tagging is performed by using a Korean part-of-speech tagging on an input sentence. The method of claim 1, wherein the pretreatment unit, Apparatus for resolving speech vocabulary using the Korean vocabulary meaning network, wherein the vocabulary is performed based on whether the vocabulary has at least two synonym sets in the Korean vocabulary network for the selection of the neutral vocabulary. . The method of claim 1, wherein the lexical expansion unit, Apparatus for solving the neutrality of speech using the Korean vocabulary meaning network, characterized by expanding the vocabulary by acquiring upper words, lower words, type control, and synonyms of the vocabulary. The method of claim 1, wherein the statistical information obtaining unit, An apparatus for resolving speech ambiguity using a Korean lexical semantic network, characterized in that the frequency of the relative words of the neutral vocabulary and the surrounding context and the respective air vocabulary frequencies are obtained from a statistical dictionary. Performing part-of-speech tagging when a sentence is input and selecting a neutral vocabulary from the analyzed sentence; Extracting a relational word between a neutral vocabulary and a vocabulary appearing in a surrounding context using a Korean vocabulary meaning network; Obtaining statistical information for statistically analyzing associations between vocabularies; And analyzing the relationship between the vocabulary and outputting a sentence in which the meaning of the vocabulary has been resolved by dividing the meaning of the vocabulary. The method of claim 6, wherein the selection of the neutral vocabulary is performed based on whether the corresponding vocabulary has at least two synonym sets in the Korean vocabulary. How to resolve neutrality of fish. The method of claim 6, wherein the relationship word, Method for solving the neutrality of the language using the Korean vocabulary meaning network, characterized in that it includes a higher word, a lower word, a type control, a synonym. The method of claim 6, wherein the statistical information, A method for resolving a word's neutrality using a Korean lexical semantic network, comprising a frequency of a related word of a vocabulary expressed together with a neutral vocabulary and a surrounding context. According to claim 6, In order to distinguish the meaning of the corresponding neutral vocabulary by analyzing the relationship between the vocabulary, A method for resolving neutrality of words using the Korean lexical semantic network, characterized by using the Korean relational words of the neutral vocabulary and the Korean relational words of the vocabulary shown in the context around the neutrality. According to claim 6, In order to distinguish the meaning of the corresponding neutral vocabulary by analyzing the relationship between the vocabulary, A method for resolving neutrality of words using a Korean lexical semantic network, using the English relational words of a neutral vocabulary and the English relational words of a vocabulary expressed in a context surrounding neutrality. According to claim 6, In order to distinguish the meaning of the corresponding neutral vocabulary by analyzing the relationship between the vocabulary, A method of resolving neutrality using a Korean vocabulary meaning network, which uses only the upper and lower words above or below the neutral vocabulary. 7. The Korean language of claim 6, wherein the relational words of the neutral words and the vocabulary appearing in the surrounding context are used from the relational words of the relational words to distinguish the meaning of the corresponding neutral vocabulary by analyzing the relation between the vocabularies. A method for resolving neutrality of words using lexical semantics. A recording medium having recorded thereon a program for resolving ambiguity of a word using a Korean lexical meaning network, wherein the method of any one of claims 6 to 13 is read by a computer and executable program code. A preprocessing unit for selecting a neutral vocabulary using a Korean vocabulary meaning network; A first neutral vocabulary processing unit for determining a type of neutral vocabulary, extracting form / meaning / syntax information, selection constraint information, and extracting a related word to expand the vocabulary; An association analysis unit that selects a meaning of a neutral vocabulary by analyzing the relations between the extended vocabularies; A second degree of repetition of determining a type of neutral vocabulary based on an air vocabulary, extracting form / meaning / syntactic information, selection constraint information, and extracting relational words to expand the vocabulary according to the analysis result of the correlation analysis unit; And a vocabulary processing unit. The method of claim 15, wherein the preprocessing unit, Korean vocabulary characterized by selecting neutral vocabulary on the basis of whether it has at least two synonyms in the Korean vocabulary network or whether it has at least two heading divisions in the Sejong Electronic Dictionary Absolute neutralization device of words using semantic network. The method of claim 15, wherein the first neutral vocabulary processing unit, A first vocabulary type determination unit for determining whether the neutral vocabulary is a verb or a verb, If the neutral vocabulary is 'word', the Sejong Electronic Dictionary extracts the vocabulary that appears in the form / meaning / syntactic information for each meaning of the neutral vocabulary. If the neutral vocabulary is 'word', each of the neutral vocabulary is A first information extracting unit for extracting the arguments appearing in the selection constraint information for each meaning; KorLex is a synonym for the vocabulary extracted from the first information extractor using the Korean Vocabulary Meaning Network (KorLex) and corresponding to the argument extracted from the first information extractor using the mapping information of KorLex and Sejong Electronic Dictionary. And a first vocabulary expansion unit for extracting a set to extend the vocabulary. The method of claim 15, wherein the second neutral vocabulary processing unit, A second vocabulary type determination unit that determines whether the neutral vocabulary is a verb or a verb, If the neutral vocabulary is 'word', the Sejong electronic dictionary extracts the vocabulary that appears in the form / meaning / syntax information for each meaning of the neutral vocabulary. If the neutral vocabulary is 'word', each meaning of the neutral vocabulary in the Sejong electronic dictionary A second information extracting unit for extracting the arguments appearing in the selection constraint information for each; KorLex is synonymous with the vocabulary extracted from the second information extractor using the Korean Vocabulary Meaning Network (KorLex) and the argument extracted from the second information extractor using the mapping information of KorLex and Sejong Electronic Dictionary. And a second vocabulary expansion unit for expanding a vocabulary by extracting a synset. 19. The apparatus of claim 17 or 18, wherein the relational word includes an upper word, a lower word, a synonym, and a sibling of a neutral vocabulary. Selecting a neutral vocabulary from the input sentence and determining the type of the neutral vocabulary; Extracting the vocabulary appearing in the form / meaning / syntax information and the argument appearing in the selection constraint information according to the type of neutral vocabulary; In the case of extracting the vocabulary appearing in the form / meaning / syntactic information, the relational word of the extracted vocabulary is extracted using the Korean vocabulary meaning network. Expanding the vocabulary; And analyzing the relationship between the extracted vocabulary and outputting a sentence in which the significance of the word has been eliminated. 21. The method of claim 20, wherein the neutral vocabulary selection is based on whether or not there are at least two synonyms in the Korean vocabulary network, or whether there are at least two heading divisions in the Sejong Electronic Dictionary. Method for solving the neutrality of the language using the Korean vocabulary meaning network, characterized in that proceeding to. 21. The method of claim 20, wherein if the neutral vocabulary is 'word', the vocabulary appearing in the form / meaning / syntax information is extracted for each meaning of the neutral vocabulary in the Sejong electronic dictionary. If the neutral vocabulary is 'word', the method for solving the neutrality of the language using the Korean lexical semantic network, characterized in that the sejong electronic dictionary extracts the arguments that appear in the selection constraint information for each meaning of the neutral vocabulary. 23. The method of claim 22, wherein in the case of extracting the vocabulary appearing in the form / meaning / syntax information, the vocabulary expansion is performed by extracting the upper words, lower words, synonyms, and sibling words of the extracted vocabulary using Korean Lexical Meaning Network (KorLex). In the case of extracting a term, it is possible to extend the vocabulary by extracting a synonym of the Korean term vocabulary (KorLex) corresponding to the extracted term using the Korean Lexical Meaning Network (KorLex) and mapping information of Sejong Electronic Dictionary. A method for resolving neutrality of words using Korean vocabulary meaning. The method of claim 20, wherein in the step of analyzing the association between the extracted vocabularies, A method for solving the neutrality of a word using the Korean lexical meaning network, comprising: selecting a meaning of a neutral word having a high score after scoring the number of words matching the extracted neutral word list and the air word. The method of claim 20, wherein the types of neutral vocabulary are determined based on the air vocabulary, and the type / meaning / syntactic information and selection constraints are determined based on the air vocabulary if the scores of the neutral vocabulary calculated in the step of analyzing the relation between the extracted vocabulary coincide. A method of resolving word neutrality using a Korean lexical semantic network, characterized by repeating a process of extracting information and extracting a related word to expand a vocabulary. A recording medium having recorded thereon a program for resolving ambiguity using a Korean lexical meaning network, wherein the method according to any one of claims 20 to 25 is read by a computer and recorded as executable program code.

Images