KR20110103580A - Automated classification method of sentential types in korea with morphological analysis and the system using the same - Google Patents

Abstract

The present invention relates to a method for automatically classifying Korean sentence types through morphological analysis and a system using the same. More specifically, the type of Korean sentence is automatically compared by comparing the ending ending, the new ending ending, and the type of communication included in the morpheme. A method of classifying and a system for implementing the same.

Description

AUTOMATTED CLASSIFICATION METHOD OF SENTENTIAL TYPES IN KOREA WITH MORPHOLOGICAL ANALYSIS AND THE SYSTEM USING THE SAME}

The present invention relates to a method for automatically classifying Korean sentence types through morphological analysis and a system using the same. More specifically, the type of Korean sentence is automatically compared by comparing the ending ending, the new ending ending, and the type of communication included in the morpheme. A method of classifying and a system for implementing the same.

A morpheme refers to the smallest unit of grammar units and is the smallest unit of language units that have meaning.

Thus, a morpheme can be said to be a grammatical unit that breaks down further or loses meaning that is related to the previous meaning. For example, splitting a star into stars and columns or splitting into rice and ㄹ have no meaning at all, meaning that columns and rice are meaningless. 'Is 10 or the same as rice, but its meaning is completely unrelated to the stars in the sky, so' star 'is a combination of' rice 'and' ㄹ 'or' ㅂ 'and' ㄹ 'or' ㅂ ' It is not a combination of and 'column'. Thus, the star is the smallest unit that can no longer be broken down by grammatical units, or morphemes. Nose, face, cloud, tree, and sweet potato are also morphemes because they are the smallest semantic unit that can no longer be split. But backpacks are not morphemes. When split into 'books' and 'bags', each of them has a meaning, and the meanings are related to the meaning of a 'bag'. Also, 'books' and 'bags' are morphemes because they can no longer be split. 'Run' is not a morph. When split into 'ju-', '-er', 'ju-' has the same meaning as 'jump' of 'run, jump, run', and '-er' means 'laugh, eat, give'. Because it has the same meaning as '-er'. 'Jump-' and '-er' are morphemes respectively. That is to say, a backpack or a run is made up of morphemes, but they are not morphemes. Some morphemes can appear alone (or even stand alone) in a sentence. There is something that can only appear in combination with morphemes. 'Star', 'book', 'bag', etc. are examples of the former, and 'ju-' and '-er' are examples of the latter. The former is called an independent morpheme and the latter is a dependent morpheme. Dependent morphemes often have connection symbols on the side to which other morphemes combine, such as 'ju-' or '-er', to indicate that they are dependent morphemes. 'Ru' in 'I'm going to run, play, cool' has a connection sign on both sides of it, like '-would-'. The same is true of the '-si-' of 'Come, Come, Ussi'. There are many self-supporting morphemes in Korean nouns and adverbs, but verbs and adjectives can appear in sentences only when their stems and endings are combined, so stems and endings are dependent morphemes, and verbs and adjectives cannot be one morpheme. Other examples of dependent morphemes include prefixes such as barefoot, foot love, and gunsori, and prefixes like barefoot, foot, and military, and bo, cover, width, and proud. And -suffix- ' Also, morphemes are divided into real and formal morphemes according to their loss of meaning. 'The sky is blue' and 'Cheol read fairy tales' are 'real sky' because they represent lexical meanings such as specific objects, actions, and states. This is called a formal morpheme because it attaches to a real morpheme and mainly expresses a relationship or function between words. The roots of the word, formula, admiration, and verb are real morphemes, and investigation, mother, and affixes are formal morphemes. Dependent morphemes can usually be combined with several other morphemes, but only with one morpheme. One example is the 'sol-' of 'sol-gil'. 'Osol' does not combine with any morpheme other than 'Road', where 'Road' is clearly an independent morpheme, and 'Trail' is different from 'Road', so it is obvious that 'Osol-' has a meaning ' Osol- 'also must be separated into separate morphemes. Such a special morpheme that can be combined with only one morpheme is called a specific morpheme. 'Beautiful' of 'beautiful' and 'wearing' of 'good' belong to unusual morphemes.

The most important criteria in determining the type of Korean sentence are the ending ending and the sentence code. For example, a sentence such as “Did you have fun yesterday?” Or “Did you eat lunch?” Can be judged as a question mark because it uses question marks as punctuation marks and '-ni' and '-nya' as endings. Wash your hands. ”,“ Get up, ”and so on, use a period as a punctuation mark and“ -er ”and“-please ”as a closing ending, respectively. In this way, most sentences in Korean have no difficulty in determining their type based on the ending ending and the sentence code alone. However, a sentence containing several specific ending endings can be interpreted in various types even for the same punctuation. A typical example is a sentence written in deconstruction or Haja, and for example, the same sentence, “I eat with Cheolsu” can be interpreted as a completely different sentence depending on the context.

(1) a. “My brother, I'll eat with you today. I eat with Bob. ”

b. “Don't eat now, let's wait. Withdrawal soon ends class. I eat with Bob. ”

c. “I used to eat almost alone, but not now. I eat with Bob. ”

In (1), the sentence “I eat with Cheolsu” is used as a statement in (1a), a petition statement in (1b), and a transcript in (1c). But if you decide to look at the sentence without looking at the context before and after, it will not be easy to determine the type of the sentence, such as "Eat with Bob." In this case, it is difficult to determine the sentence type by using only the punctuation marks and the ending word, because there are several ending endings used in various sentence types such as '-어' shown above.

Therefore, in order to automatically determine the types of sentences that contain such endings, the analysis of morphological and syntactic units is attempted to find clues that play an important role in sentence determination and to reduce the range of possible sentences. Very necessary.

In particular, in a machine translation device for translating Korean into a language of another country, if the type of sentence is not properly judged, awkward interpretation results may occur and the reliability of the machine translation device may be drastically degraded. It is important.

The present invention has been made to solve the above problems, and through a morphological analysis in addition to the ending ending and sentence code, a method of automatically classifying whether the sentence to be interpreted corresponds to any of the statement / review sentence / hearing sentence / admiration sentence / question sentence To provide a system that implements it.

The present invention includes the following technical solutions to achieve the above object.

An input unit for receiving a sentence, a morpheme division unit for dividing the input sentence into morpheme units, a storage unit for classifying and storing the divided morpheme into a subject unit, an adverb unit, and a descriptive unit, and performing a comparative operation on the stored morphemes A morpheme value of the narration part performed by the client terminal includes a calculation unit configured to store an existing morpheme, a database unit classified and stored according to a predetermined criterion, and a type result unit storing a type of a sentence determined by the operation unit. In the method for determining the type of a sentence including a character string belonging to a dissolution terminator database or a dissolution terminator database, the operation unit determines whether the morpheme data value of the predicate phrase includes a character string belonging to the guess / will predicate end database. Judging; Determining, by the operation unit, whether or not the morpheme data value of the descriptor part includes a character string belonging to a zoning word ending database; Determining by the operation unit whether the morpheme data value of the adverb part includes a character string belonging to a negative sentence adverb database or a character string belonging to a negative sentence auxiliary verb database; Determining, by the operation unit, whether the morpheme data value of the predicate language part includes a character string belonging to an operation body language database or a character string belonging to an operation message database; Determining, by the operation unit, whether or not the morpheme data value of the predicate language part includes a character string belonging to a preceding connection ending auxiliary language database; And determining, by the operation unit, whether or not the stemming data value of the subject part includes a character string belonging to a speaker / third party database.

In addition, the guess / will line ending database includes the strings '-'-', '-was-', '-was-', '-the-', '-dun-', '-wo-', The morpheme data value of the predicate word includes any one of '-'-', '-'-', '-'-', '-the-', '-dun-' and '-o'-' If so, further comprising the step of inputting a value corresponding to the written statement in the type result unit.

When the morpheme data value of the predicate idiom includes the character string '-shi-' and the morpheme data value of the narration portion includes the character string of '-shi-', And determining, by the operation unit, whether or not a string includes the character string, and when the morpheme data value of the predicate includes a character string of '-yo', inputting a value corresponding to the plain text in the type result unit. It includes more.

In addition, the negative sentence adverb database includes a string 'not' and 'peg', and the auxiliary verb database includes a string 'not', and the morpheme data values of the adverb part are 'not', 'peg' and ' If the string includes any one of '-', the method further includes the step of inputting a value corresponding to the plain text in the type result section.

The operation idiom database includes a verb indicating an operation, the operation transcript database includes a statement indicating an operation that can be combined with a verb-derived suffix 'ha-', and the morpheme data value of the predicate language is the operation idiom database. And if a character string belonging to the operation message database is not included, inputting a value corresponding to a plain text sentence in the type result unit.

In addition, the preliminary connection ending auxiliary language database is'-a have ','-go ','-even /-should ','-try to ','-say ',' -To see /-, '-can be', '-can't be', '-if you want to-', '-wants to go', '-have ever', '- 'N never', '-break up', '-break up', '-try / to-go', '-walk /-try', '-try', '-try', '-to ',' -Must ',' -shall ',' -shall ', and if the stemming data value of the predicate includes one of the strings belonging to the preceding link end verb database. The step of inputting the value corresponding to the written statement in the type result section.

The speaker / third party database may include a string indicating a speaker or a third party including 'I', 'low', 'we', and 'he', and the stemming data value of the subject may be the speaker / first. In case of belonging to the three-character database, the method may include inputting a value corresponding to the plain text in the type result unit.

The method may further include determining, by the operation unit, whether the morpheme data value of the adverb part includes a character string belonging to the request adverb database, and wherein the morpheme data value of the adverb part includes any one character string of the requested adverb database. If so, further comprising the step of inputting a value corresponding to the statement or the request statement in the type result unit.

The method may further include determining, by the operation unit, whether the morpheme data value of the adverb part includes a character string belonging to the plain text adverb database, wherein the plain text adverb database has a meaning of frequency or a meaning of time or degree. And a character string, and if the morpheme data value of the adverb part includes one of the character strings of the plain text database, inputting a value corresponding to the plain text in the type result unit.

The method may further include determining, by the operation unit, whether or not the morpheme data value includes a character string belonging to the valuation database, wherein the valuation database includes a combination of a message indicating a listener and a valuation or a comma or exclamation point. If the morpheme data value includes a string of any one of the favorable databases, the method may further include inputting a value corresponding to a statement in the type result unit.

And, if the morpheme data value of the predicate phrase includes any one of 'Ju-', 'Bau-', 'Du-', 'Du-' and 'Burry-', the type result section The method may further include inputting a corresponding value.

In addition, an input unit for receiving a sentence, a morpheme division unit for dividing the input sentence into morpheme units, a storage unit for classifying and storing the divided morphemes into a subject unit, an adverb unit, and a descriptive unit, and a comparative operation on the stored morphemes The morpheme data value to be performed in a client terminal including an operation unit configured to perform an operation, a database unit classifying and storing an existing morpheme according to a predetermined criterion, and a type result unit storing a type of a sentence determined by the operation unit. In the method for determining the type of a sentence including a character string belonging to a question mark, the morphological data values of the predicates are terminated with '-ni', '-er (yo)', '-rae (yo)', '- Determining, by the operation unit, whether any one of the character strings is included; And a second step of determining, by the operation unit, whether or not the morpheme data value of the predicate word includes an auxiliary word 'main-', wherein the first step includes a string of any one of the ending endings, When the auxiliary step is included in the second step, a value corresponding to the statement is input to the type result unit, and when the second step does not include the auxiliary word, the question corresponds to the type result unit. The method further includes inputting a value.

In addition, an input unit for receiving a sentence, a morpheme division unit for dividing the input sentence into morpheme units, a storage unit for classifying and storing the divided morphemes into a subject unit, an adverb unit, and a descriptive unit, and a comparative operation on the stored morphemes A morpheme of the narration part performed by the client terminal including an operation unit configured to perform an operation, a database unit classifying and storing an existing morpheme according to a predetermined criterion, and a type result unit storing a type of a sentence determined by the operation unit. A method of determining the type of a sentence including a terminating word '-ara' or a string of '-er' in a data value, comprising: storing a verb of an verb in an verb adjective database and indicating a state in an adjective verb database A first step of storing the adjective's statement, and morphological data of the predicate A second step of determining whether the by the string belonging to the company include substantives database to the computing unit and; And a third step of determining, by the operation unit, whether or not the morpheme data value of the predicate word includes the character string belonging to the adjective statement database, wherein in the second step, the morpheme data value of the predicate word belongs to the verb statement database. If a string is included, a value corresponding to a statement is input to the type result unit, and if the stemming data value of the predicate part includes a string belonging to the adjective statement database in the third step, the type result unit corresponds to an exclamation statement. Enter a value.

In addition, an input unit for receiving a sentence, a morpheme division unit for dividing the input sentence into morpheme units, a storage unit for classifying and storing the divided morphemes into a subject unit, an adverb unit, and a descriptive unit, and a comparative operation on the stored morphemes It is a sentence type automatic classification system through morphological analysis including a calculation unit for performing a, a database unit for classifying and storing the existing morphemes according to a predetermined criterion, and a type result unit for storing the type of sentences determined by the operation unit; The operation unit determines whether or not the morpheme data value of the predicate language includes a character string belonging to the guess / will preterm ending database, and if so, inputs a value corresponding to a plain text sentence to the type result unit or the morpheme data value of the predicate language part. Contains a string belonging to this zoning final ending database In this case, the value corresponding to the plain text is input to the type result section, or if the morphological data value of the adverb part includes a string belonging to a negative sentence adverb database or a string belonging to a negative sentence auxiliary verb database. In case of inputting a value corresponding to a plain text sentence in the type result section, or by determining whether the morpheme data value of the predicate word includes a character string belonging to the operation main language database or a character string belonging to the operation message database. , Enter the value corresponding to the statement or the reason statement,

It is determined whether or not the morpheme data value of the predicate language includes a character string belonging to a preceding connection ending auxiliary language database, and if it is not applicable, input a value corresponding to a plain text sentence in the type result part, or the morpheme data value of the subject part is the speaker / It is determined whether to include a character string belonging to a third party database, and if applicable, the value corresponding to the plain text is entered in the type result section.

In addition, an input unit for receiving a morpheme data value divided into morpheme units, a storage unit for classifying and storing the input morpheme into a subject unit, an adverb unit, and a descriptive unit, an operation unit for performing a comparative operation on the stored morphemes, and A sentence type automatic classification system through morpheme analysis comprising a database unit for classifying and storing the morphemes according to a predetermined criterion, and a type result unit for storing the type of sentences determined by the operation unit. It is determined whether or not the stemming data value includes a character string belonging to the guess / will line ending database. If applicable, a value corresponding to the plain text is entered in the type result section, or the stemming data value of the predicate language belongs to the zoned front end database. To determine whether it contains a string In the case of the type result section, the value corresponding to the plain text, or if the morpheme data value of the adverb part contains a character string belonging to the negative sentence adverb database or a character string belonging to the negative sentence auxiliary verb database, if applicable, the type result Input a value corresponding to a plain text statement to the part, or determine whether the morpheme data value of the predicate language includes a character string belonging to an operation body language database or a character string belonging to an action message database and, if applicable, Enter a corresponding value or determine whether the morpheme data value of the predicate language includes a character string belonging to a preceding connection end supplementary database, and if not applicable, enter a value corresponding to the plain text in the type result section; When the morphological data values of the group jueobu to determine whether it contains a character string belonging to the speaker / third-party databases to enter a value corresponding to the declarative type results section.

The present invention improves the reliability of the machine translation apparatus by providing a method of automatically classifying whether the sentence to be interpreted corresponds to a sentence / commentary sentence / hearing statement / interjection sentence / question sentence through morphological analysis in addition to the ending word and the sentence code, There is an effect of providing an optimal translation by preventing the translation of semantically awkward translations. In addition, it is possible to grasp the speaker's basic intention in the process of using the speaker's sentence, thereby improving the accuracy of its application.

1 is a block diagram of the present invention. 2 is a flow chart of the present invention. Figure 3 is a flow chart of the statement / statement / solicitation classification of the present invention. Figure 4 is a flow chart of the statement / statement / solicitation classification step of the present invention. 5 is a flowchart of a question / statement classification step of the present invention. 6 is a flow chart of the exclamation / statement classification step of the present invention.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. However, in describing the present invention, when it is determined that a detailed description of a related known function or configuration may unnecessarily obscure the subject matter of the present invention, the detailed description thereof will be omitted. The terms to be described below are terms set in consideration of functions in the present invention, and may be changed according to a user's intention or custom such as an experimenter and a measurer, and the definitions should be made based on the contents throughout the present specification.

1 is a block diagram of the present invention. The sentence type automatic classification system through morphological analysis according to the present invention is generally installed in the client terminal. An input unit 11 for receiving a sentence, a morpheme division unit 12 for dividing the input sentence into morpheme units, a storage unit 15 for classifying and storing the divided morphemes into a given part, an adverb part, and a descriptive word part, and stored A type result of storing the type of sentence determined by the calculation unit 13, an arithmetic unit 13 performing a comparative operation on a morpheme, a database unit 16 classifying and storing existing morphemes according to a predetermined criterion, and a type of a sentence determined by the arithmetic unit 13 Part 14 is included. In addition, it may further include an output display unit for outputting the type of the determined sentence.

The storage unit may be classified into the subject unit 151, the adverb unit 152, the predicate language unit 153, and the like according to the parts of speech of the sentence. It is also possible to have a sentence code or other classification criteria.

The calculation unit 13 that performs operations using the stored morphemes is the most important part in the present invention. This will be described in detail with reference to FIGS. 2 to 6.

Although the database unit 16 illustrates the dismantling ending mother database 161 and the disassembly ending mother database 162 in FIG. 1, other databases described herein are also possible. This may be understood as a group in which each morpheme is classified by a function or the like in order to determine whether the calculation unit 13 is included in the stored morpheme and its value is included.

The input unit 11 receives a sentence through input means such as a keyboard and a touch screen operated by a user, and provides the sentence to the morpheme division unit 12. However, the present invention classifies the types of sentences by analyzing the divided morphemes, and dividing them into morphemes is not a core idea of the present invention. A device for dividing a sentence into morphemes and the like can be used. For example, Korean Patent Laid-Open No. 10-2008-0091749 discloses a morpheme analysis device and an analysis method in detail.

Further, the present system can also be implemented in a form in which the morpheme division does not exist. In this case, the input unit receives a morpheme data value divided in morpheme units.

As shown in FIG. 2, the present invention has three classification methods. The classification method and system according to the present invention do not work when one sentence type can be determined only by the ending ending and the sentence code.

In Korean sentences, there are several ending endings that are used for each sentence type.

End of Declaration:-b / / b / b / b / b / b-b / b / b

Gamtanhyeong terminated endings: - You / - gumeon / - sucks / - Military / - I / - Yes / - Ara / - Huh / - ㄴgeol

Interrogative closure mother: - I / - Nya / - You / - Want / - b / a - o / - Small / - Caesarea / - ㄹkka / - f Do / - O / - ok / - I / - I / - paper /-

Ending statement terminator - would / - noticed / - Ara / - Huh / - thats / - I / - O / - Please f / - O / - ok / - I / - I / - Ji / - I

Clear type endings:-Well /-Three /-Let 's see /-Oh /-No /-Uh /-Uh /-Ji /-Oh

Of these, '-ah /-no /-ah /-uh /-ji /-jyo', which corresponds to the dismantling / dismantling, is not limited to any sentence but is used in all declarative sentences, interrogations, commands, and solicitations. '-Ara / -ara' is also used for both statements and admiration. In addition, sentences containing some endings used in interrogation may in some cases have a semantically functional function.

As a first embodiment, a description will be given of how to classify sentences of sentences corresponding to '-아 /-아 /-어 /-어 /-지 /-요'. However, the above six termination endings exemplified in the present specification are only examples, and it is obvious that the present invention may be equally applied to the termination endings constituting the dismantling or disassembly.

In order to implement the present invention, the termination mothers corresponding to the dismantling should be stored in the dismantling ending mother database, and the terminating mother used for the dismantling in the dismantling ending mother database.

As shown in Figure 2, in the case of dismantling / disassembly can be used both in the declarative statement / command statement / solicitation statement, the criteria for determining the specific sentence form is as follows.

(2) a. Identify word endings used in descriptors

    b. Identify adverbs and auxiliary words that make negative sentences

    c. Verbs and Verifications Used in Predicates

    d. Identify subjects that match the descriptor

    e. Identify Adverbs That Match Predicates

    f. Confirmation

    g. Identifying auxiliary words used in predicates

Given a sentence, apply the seven criteria of (2a ~ g) to find clues about sentence patterns and narrow the range of possible sentences. However, the order of application of judgment criteria is important because each criterion has a different degree on sentence sentence judgment and there are matters to be considered first.

First, since the criteria of (2a to d) do not contradict each other, it is irrelevant to apply any of the four first. 3 is shown as proceeding in the order of 2a to 2d, but this is for convenience of the flow chart, even if the order is changed, there is no obstacle to the implementation of the present invention.

Also, (2e ~ g) must be applied after (2a ~ d), because (2a ~ d) can be used to determine the sentence type correctly, but (2e ~ g) without going through (2a ~ d) If you apply) first, the sentence pattern may be wrongly determined. Therefore, their order is very important and must be observed.

In relation to 2a, statements and solicitations represent demands for things or actions that have not yet occurred, and therefore they cannot be used in the present tense because they are limited to the present or future tense. Therefore, sentences containing the ending tense endings such as '-er-', '-er-', '-the-' and '-dun-' are not classified as command / request statements. . The word ending ‘-려-’ is also used to express the speaker's will or conjecture, so it cannot be used in command / request statements.

Therefore, in the database unit 16, there is a guess / will premature ending database, which includes'-',' -was- ',' -was- ',' -the- ',' -dun- ','- May be included as an example. In addition, if the morpheme data value of the predicate language part includes any character string of the guess / will preterm ending database, the input sentence has the value of the plain text sentence, and the classification process ends (step 302).

In addition, when the predicate ending '-시-', which indicates respect for a predicate, increases the subject of a sentence, it cannot be a petition sentence that makes both a defect and a listener a subject of a sentence at the same time. However, if there is no closing check 'Yo' indicating exaltation, that is, if the dismantling was used instead of the body, the listener would be lowered, so the statement between the subject and the listener could not be the same. Therefore, this case should be classified as a written comment (step 303).

With regard to 2b, the sentences used in the adverbs ‘안’, ‘pum’ and auxiliary adjunct ‘안-’ can't be ordered / reviewed, so they can be classified as essays. In addition, the auxiliary word 'horse-', which makes negative sentences, is not used in written statements, so it can serve as a basis for narrowing the range of sentence types that can be used for statements and clarification statements.

For example, "take your medicine" can be both a comment and a statement, while "do n't take your medicine" means that the word ending "-si-" is used in the negative "word-" negative statement, “Do n't take your medicine” is a negative statement, which is only possible in writing. Also, "I said do not eat medicine" or "not - 'auxiliary verb, such as" about how late Oh, please be sure "and" end -' The addition in combination with other auxiliary verb bonyongeon + auxiliary verb + auxiliary Statements in verbal form can also be categorized as written statements and statements, respectively.

Therefore, the database 16 has a negative adverb adverb database, which includes '안-' and 'pum-'. There is also a database of auxiliary verbs, which include ‘으-’. However, the present invention may be embodied by including 'an-', 'mop-' or '无-' as an example, in addition to including a string or an auxiliary word for making a negative adverb adverb. If the morpheme data value of the narrator includes the character string of either the negative sentence adverb database or the auxiliary verb database, the input sentence has the value of the plain text sentence and the classification process ends (step 304).

In relation to 2c, statements and solicitation are basically verbal expressions that require the listener to perform an action, so that the adjectives that state the sentence cannot be adjectives that indicate a state, and verbs that indicate action or verbs that indicate an action. It can only be used in verbal form, such as 'study' with the verb-derived suffix '-ha-'. Therefore, it is judged that sentences with other words or verbs cannot be commands / requests.

For example, " animals ", " possible ", and " regret " in "I 'm a small animal ,""I can regret it," and "I regret a lot," are not all statements of action. Neither can be command / clear statement. The phrase “the best way to study ” is used to describe the action of “studying,” but this sentence is not meant to ask the other person to do something but rather to convey something. Therefore, it is hard to see that it has the function of command or solicitation. The predicates of “really good ” and “a cold catches well in winter” use the adjective “good” and the verb “to hang”, which does not represent an action. Of course, in order to figure out whether a statement or a state represents an action, the information used for each word must be included in advance in the dictionary used in the morpheme division unit.

Therefore, the database unit 16 should be provided with an action verbal database in which verbs indicating motion are stored in advance, and an action verb database in which verbs indicating motions that can be combined with the verb-derived suffix 'ha-' are stored in advance. When the morphological data value of the narrator includes a character string of either the operation idiom database or the operation utterance database, it may be determined as a statement or a clarification statement, and if it is not included in all, it may be determined as a plain statement (step 305).

With the exception, some adjectives are sometimes used as commands / hearings with the verb-derived suffix '-er ha-' and certain auxiliary verbs. For example, "Please yeppeohae a lot of new students coming weeks." La thrown "like a lot of you share my picture" is. The adjective 'pretty', which does not represent an action, is used, but it combines with the verb-derived suffix 'ha-' to make a verb that means “pretty”. In addition, as the auxiliary word 'Ju-' is combined, the meaning of the request is added to serve as a statement.

However, such an exception may be solved by additionally determining whether the auxiliary verb 'Ju-' exists in the morpheme data value of the predicate, and in such a case, further determining the statement.

In addition, commands / solicitations are, in principle, sentences that are intended for humans, so if the subject of a sentence comes with a verb that shows its action, if its subject cannot be a person, that is, it is not an action that can be directly performed by a person. The sentence cannot be a command / solicitation statement. However, whether or not a person can be a subject can be solved by storing a case where a person cannot be a subject in advance in a database and comparing them by an operation unit.

Among the supplementary words used in the predicates of sentences, there are a number of other words besides the words `-'and` ʻ-- which make negative sentences, and these provide important clues for sentence classification. For example, the word ‘want-’ used in “I miss you.” Is a representative supplementary verb that is not used in the meaning of command / hearing statement. In particular, even if the same verb is used in the predicate or auxiliary verb in a sentence, the range of possible sentence patterns may vary, and the verb may be a good clue for sentence judgment.

For example, “to do” means “to do something” and can be used in any order / comment, but it means “to deserve”. It is a supplementary term that, in this case, cannot be used semantically in order / solicitation. Since auxiliary verbs are not used independently and are always used together, they are always accompanied by the main verb and the linking ending, so it is usually possible to distinguish between the main verb and the auxiliary verb through the use of the linking ending. In “가야 해요”, the connecting ending ‘-야’ used in ‘가야’ shows that ‘하다’ is an auxiliary word. All of the auxiliary verbs listed below can be seen as preceding verbs, rather than the main verbs, and the sentence type can be judged as a plain text because it is not used in the order / approach.

The database unit 16 includes a preliminary link ending supplementary database, and the preliminary linking ending supplementary database includes '-have', '-become', '-also / -must', 'go-to', '-Say-', '--see', '-/-see', '-you can-', '-can't', '-/-want', '-b //-b 'I want to go', '-b never', '-n never', '-be broken', '-to break up', 'to-be-to-be', 'to-be-to-be', '- I'd like to ','-let's', '-let's','-should ','-want ','-want '.

If the morpheme data value of the predicate word includes any one of the character strings belonging to the preliminary connection ending supplementary database, the value corresponding to the plain text is input to the type result part (step 306).

In relation to 2d, the order / solicitation statement is not the command / solicitation statement, because the subject of the sentence includes the other party, so even if the person is the subject, the sentence used by the speaker himself or a third party person is not the order / solicitation statement. "I eat soup" or "he eats soup" cannot be a command / hearing statement because the subject indicating the speaker or third party is obvious.

Therefore, the database unit 16 should be provided with a speaker / third party database in which a string indicating a speaker or a third party including 'I', 'low', 'we', 'he' is stored in advance, If the value belongs to it, a value corresponding to the plain text is input (step 307).

In contrast, when the listener is in the subject, it is somewhat complex. For example, "You are really healthy" is not a problem because "health" is not a statement of movement, so it will already be classified as a commentary in the previous analysis. However, if a verb that represents an action, such as "You do exercise" or "You do exercise really well", is a descriptive phrase, you can think of it as a statement. However, the adjectives "really" and "well" in "You exercise really well" are used only to explain the degree of the "do" action, not to ask the other person to do something. It is right to see. After all, it is clear that if the subject is not the listener but the speaker or the third person, it cannot be a command / request statement, but the listener does not pretend that the statement is a statement. However, it is clear that this is not at least a hearing.

With regard to 2e, commands / solicitations are limited in combination with adverbs because they represent a need for an action that has not yet been executed. For example, a sentence like "Please study tomorrow" doesn't matter, whereas a sentence like "Please study these days" is awkward because the meaning of 'please' and 'now' is in conflict. do. Thus, adverbs that modify the predicates of a sentence can be the basis for distinguishing a command / clear statement from a plain text.

Adverbs commonly used in command / hearing statements are: 1) adverbs used to urgently or eagerly demand actions such as 'come', 'right' and 'quickly', and 2) acts such as 'please', 'please', 'anybody', etc. Adverbs that can be used to eagerly ask for, 3) `` Now '', `` Once '', `` Only once '', `` Only '', `` Some '', etc. These are adverbs that can be stored in the adverb database in advance in the database unit 16, and if the morpheme data value of the adverb part includes any one of the strings, it may be determined as a statement or a request statement (step 401).

1) adverbs indicating the frequency of 'recently', 'fortress', 'currently', 'everyday', 'always', 'usually', etc. 2) 'already', 'just', 'yet', Adverbs that represent times such as 'still', 'still', 'quick' and 'already', 3) 'pretty', 'really', 'very', 'much', 'very', 'true', 'badly' Adverbs that represent the meaning of ',' very ',' nearly ',' too ',' nothing ', etc. 4)' Probably ',' Sure ',' Gear ',' Finally ',' Indefinite ',' Moreover, modal adverbs that modify the entire sentence, such as', ', and so on,' are difficult to use in command / hearing meanings. Therefore, the adverbs may be stored in the database section 16 in the plain text adverb database in advance, and if the morpheme data values of the adverb parts correspond to this, it may be determined as the plain text (step 402).

In relation to 2f, statements or solicitations are used when you want to demand some action from the listener, so the call to the listener is often used. Therefore, it can be used as a basis for sentence sentence determination by checking whether a hit is used. The rage is usually "Father, have dinner." Eat it. ”Appears at the beginning of a sentence with or without a suffix 'Ya', 'Ah', 'Ishi', 'Iye', etc., with a comma or exclamation .

Thus, 1) a combination string of a statement and a comma or exclamation point for the listener, 2) a combination string of a statement and a voting question for the listener, and 3) a hit database with a combination string of the statement and valuation and the comma or exclamation point for the listener. If a morpheme data value is included therein, a value corresponding to the statement may be input (step 403).

With regard to 2g, supplementary verbs combine with the main verbs to add a certain meaning, of which 'ju-', 'bo-', 'du-', 'release-' and 'bury-' are requested respectively. It is often used in statements because it adds meaning to try, hold, prepare, and terminate. For example, a statement that does not reveal any clues other than supplementary verbs, such as “Please go outside” or “Try to exercise every day” is very likely a statement. Accordingly, when any one of these four words is included in the morphological data value of the predicate, it may be determined that the statement corresponds to the statement (step 404).

As a second embodiment of the present invention, in the case of a sentence including a question mark, the process of classifying the question and the statement will be described in detail.

Statements and solicitations are used to request a listener to do something, sometimes in the form of questions when he wants to reduce the burden on the listener to do it or to convey a more polite attitude to the speaker.

For example, in the case of "homework future? Little dash to the line" or "why do not we instead share some homework?" Therefore, it is determined by the operation unit whether the morpheme data value of the predicate word includes one of the ending words' -nee ',' -er (yo), '-rae (yo)', and '-ca (yo)'. If it is satisfactory and the stemming data value of the predicate contains the sub-word 'Note-', the sentence must be entered as a final result (step 503).

Occasionally, the word ending ‘-돼-’ can be used, where ‘- In addition, the auxiliary verbs '-수 수-', '-수 수-' are used to express the meaning of the possibility, or they may be in the form of a negative sentence.

As a third embodiment of the present invention, '-ara' and '-er' are representative terminating endings used in a statement, which, when combined with a verb representing an action such as "eat", "go!" It has a meaning, but when combined with adjectives that indicate status, such as "cold" and "like a child!" In this case, therefore, we need to see whether the combined word is a verb or an adjective.

Therefore, after storing the verb of the verb indicating the action in the verb verbal database, storing the adjective of the adjective indicating the state in the adjective verb database, the operation unit determines whether the stemming data value of the predicate contains a character string belonging to the verb verb database. After determining by the operation unit whether or not the morpheme data value of the predicate contains a character string belonging to the adjective statement database, and if the morpheme data value of the predicate part contains a character string belonging to the verb declaration database, If the value is input (step 602), and if the morpheme data value of the predicate language includes a character string belonging to the adjective statement database, the value corresponding to the exclamation sentence is input (step 603).

Hereinafter, a method and a system for automatically classifying Korean sentence types through morpheme analysis according to the present invention have been described in detail. Through the above method, the present invention provides an effect of automatically classifying Korean character types, thereby increasing the reliability and accuracy of the machine translation apparatus.

The present invention is not limited to the scope of the embodiments by the above embodiments, all having the technical spirit of the present invention can be seen to fall within the scope of the present invention, the present invention is the scope of the claims by the claims Note that is determined.

10: client terminal, 11: input unit, 12: morphological division unit, 13: calculation unit, 14: type result unit, 15: storage unit, 16: database unit

Claims (15)

An input unit for receiving a sentence, a morpheme division unit for dividing the input sentence into morpheme units, a storage unit for classifying and storing the divided morpheme into a subject unit, an adverb unit, and a descriptive unit, and performing a comparative operation on the stored morphemes A morpheme value of the narration part performed by the client terminal includes a calculation unit configured to store an existing morpheme, a database unit classified and stored according to a predetermined criterion, and a type result unit storing a type of a sentence determined by the operation unit. In the method of determining the type of a sentence including a character string belonging to the dismantling ending database or the dismantling ending database,
Determining, by the operation unit, whether or not the morpheme data value of the predicate language part includes a character string belonging to a guess / will preterm ending database;
Determining, by the operation unit, whether or not the morpheme data value of the descriptor part includes a character string belonging to a zoning word ending database;
Determining by the operation unit whether the morpheme data value of the adverb part includes a character string belonging to a negative sentence adverb database or a character string belonging to a negative sentence auxiliary verb database;
Determining, by the operation unit, whether the morpheme data value of the predicate language part includes a character string belonging to an operation body language database or a character string belonging to an operation message database;
Determining, by the operation unit, whether or not the morpheme data value of the predicate language part includes a character string belonging to a preceding connection ending auxiliary language database;
Determining by the operation unit whether the stemming data value of the subject part includes a character string belonging to a speaker / third party database,
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
The guess / will premature end database includes the strings '-'-', '-was-', '-was-', '-the-', '-dun-', '-wo-',
The morpheme data value of the predicate word includes any one of '-'-', '-'-', '-'-', '-the-', '-dun-' and '-o'-' If so, further comprising the step of inputting a value corresponding to the written statement in the type result unit,
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
The zoning word ending database includes the string '-shi-',
Determining, by the operation unit, whether or not the morpheme data value of the predicate constituent part includes the character string of '-when-',
If the morpheme data value of the predicate contains a string of '-yo', further comprising the step of inputting a value corresponding to the plain text in the type result unit;
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
The negative sentence adverb database includes the strings 'not' and 'peg', and the auxiliary verb database includes the string 'not',
If the morpheme data value of the adverb part includes any one of the strings 'not', 'not', 'not', the method may further include inputting a value corresponding to the plain text in the type result unit.
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
The action idiom database includes a verb indicating an action, the action statement database includes a statement indicating an action that can be combined with a verb derived suffix 'ha-',
If the morpheme data value of the predicate language does not include a character string belonging to the operation idiom database and the operation dictation database, inputting a value corresponding to a plain text to the type result unit;
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
The preliminary connection ending supplementary database is'-ah ','-go ','-even /-should go ','-try to go ','-say ','-see ','- /-See, ','-can be ','-can't ','-if you want to-','-wants to be / '-' No, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no, no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no no , '-Must', '-do', '-if',
If the morpheme data value of the predicate language includes any one of character strings belonging to the preceding linked ending auxiliary language database, inputting a value corresponding to a plain text sentence in the type result unit;
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
The speaker / third party database includes a string indicating a speaker or a third party including 'I', 'low', 'us', 'he',
If the morphological data value of the subject part belongs to the speaker / third party database, inputting a value corresponding to a plain text sentence in the type result unit;
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
Determining whether the morpheme data value of the adverb part includes a character string belonging to a request adverb database;
If the morpheme data value of the adverb part includes a string of any one of the requested adverb databases, further comprising: inputting a value corresponding to a statement or a request statement in the type result unit;
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
Determining whether the morpheme data value of the adverb part includes a character string belonging to a plain text adverb database;
The test sentence adverb database includes a string having a meaning of frequency or a meaning of time or degree.
If the morpheme data value of the adverb part includes a string of any one of the plain text databases, further comprising: entering a value corresponding to the plain text text in the type result section;
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
Determining, by the operation unit, whether or not the morpheme data value includes a character string belonging to an attractive database;
The rapport database includes a combination of a statement indicating a listener and a rapport or comma or exclamation point,
If the stemming data value includes a string of any one of the favorable databases, further comprising: inputting a value corresponding to a statement in the type result unit;
Automatic Classification of Korean Sentence Types through Morphological Analysis. The method according to claim 1,
When the morpheme data value of the predicate phrase includes any one of 'Ju-', 'Bo-', 'Du-', 'Lu-' and 'Burry-', the type result part corresponds to a statement. Further comprising the step of entering a value,
Automatic Classification of Korean Sentence Types through Morphological Analysis. An input unit for receiving a sentence, a morpheme division unit for dividing the input sentence into morpheme units, a storage unit for classifying and storing the divided morpheme into a subject unit, an adverb unit, and a descriptive unit, and performing a comparative operation on the stored morphemes A question mark on the morpheme data value, which is performed in a client terminal, comprising: an operation unit configured to classify and store an existing morpheme according to a predetermined criterion; and a type result unit that stores a type of a sentence determined by the operation unit. In the method of determining the type of a sentence containing a string belonging to,
The calculation unit may determine whether the morpheme data value of the predicate language part includes any one of a terminal '-nee', '-er (yo)', '-rae (yo)', and '-ka (yo)'. Judging by;
And a second step of judging by said calculating section whether or not the morpheme data value of said predicate word contains the character string of a supplementary word 'main-',
When the first step includes the string of any one of the ending ending, and when the second step includes the auxiliary word, a value corresponding to the statement is input to the type result section, and in the second step If it does not include an auxiliary verb, further comprising the step of inputting a value corresponding to the question in the type result section
Automatic Classification of Korean Sentence Types through Morphological Analysis. An input unit for receiving a sentence, a morpheme division unit for dividing the input sentence into morpheme units, a storage unit for classifying and storing the divided morpheme into a subject unit, an adverb unit, and a descriptive unit, and performing a comparative operation on the stored morphemes A morpheme data value of the narration part performed by the client terminal including an operation unit configured to classify and store an existing morpheme according to a predetermined criterion, and a type result unit which stores a type of a sentence determined by the operation unit. In the method of determining the type of a sentence including a string of ending ending '-ara' or '-어',
A first step of storing the verb of the verb indicating the action in the verb verb database and storing the adjective of the adjective indicating the status in the adjective verb database;
A second step of determining by the operation unit whether the morpheme data value of the descriptor part includes a character string belonging to the verb verbal database;
And a third step of judging by said operation unit whether or not the morpheme data value of said predicate language part contains a character string belonging to said adjective statement database,
In the second step, if the morpheme data value of the predicate part includes a character string belonging to the verb verbal database, a value corresponding to a statement is input to the type result part. In the case of including a character string belonging to the adjective statement database, the value corresponding to the exclamation sentence is entered in the type result section.
Automatic Classification of Korean Sentence Types through Morphological Analysis. An input unit for receiving a sentence, a morpheme division unit for dividing the input sentence into morpheme units, a storage unit for classifying and storing the divided morpheme into a subject unit, an adverb unit, and a descriptive unit, and performing a comparative operation on the stored morphemes It is an automatic sentence type classification system through morphological analysis, including a calculation unit, a database unit for classifying and storing the existing morphemes according to a predetermined criterion, and a type result unit for storing the type of sentences determined by the operation unit,
The operation unit determines whether the morpheme data value of the predicate language includes a character string belonging to the guess / will preterm ending database, and if so, inputs a value corresponding to the plain text to the type result unit,
It is determined whether or not the stemming data value of the predicate language includes a character string belonging to a zoning word ending database, and if so, inputs a value corresponding to a plain text sentence in the type result section.
It is determined whether the morpheme data value of the adverb part includes a character string belonging to a negative sentence adverb database or a character string belonging to a negative sentence auxiliary verb database.
It is determined whether or not the morpheme data value of the predicate language includes a character string belonging to an operation idiom database or a character string belonging to an operation dictation database.
It is determined whether or not the morpheme data value of the predicate language includes a character string belonging to a preceding connection ending supplementary database, and if not applicable, enter a value corresponding to a plain text sentence in the type result section.
It is determined whether the stemming data value of the subject part includes a character string belonging to a speaker / third party database, and if applicable, inputting a value corresponding to a plain text sentence into the type result part.
Korean sentence type automatic classification system through morphological analysis. An input unit for receiving a morpheme data value divided into morpheme units, a storage unit classifying and storing the input morpheme into a subject unit, an adverb unit, and a descriptive unit, an operation unit performing a comparative operation on the stored morphemes, and an existing morpheme It is a sentence type automatic classification system through morphological analysis including a database unit for classifying and storing according to a predetermined criterion, and a type result unit for storing the type of the sentence determined by the operation unit,
The operation unit determines whether the morpheme data value of the predicate language includes a character string belonging to the guess / will preterm ending database, and if so, inputs a value corresponding to the plain text to the type result unit,
It is determined whether or not the stemming data value of the predicate language includes a character string belonging to a zoning word ending database, and if so, inputs a value corresponding to a plain text sentence in the type result section.
It is determined whether the morpheme data value of the adverb part includes a character string belonging to a negative sentence adverb database or a character string belonging to a negative sentence auxiliary verb database.
It is determined whether or not the morpheme data value of the predicate language includes a character string belonging to an operation idiom database or a character string belonging to an operation dictation database.
It is determined whether or not the morpheme data value of the predicate language includes a character string belonging to a preceding connection ending supplementary database, and if not applicable, enter a value corresponding to a plain text sentence in the type result section.
It is determined whether the stemming data value of the subject part includes a character string belonging to a speaker / third party database, and if applicable, inputting a value corresponding to a plain text sentence into the type result part.
Korean sentence type automatic classification system through morphological analysis.

Images