KR101409298B1 - Method of re-preparing lexico-semantic-pattern for korean syntax recognizer - Google Patents

Abstract

The present invention relates to a method for constructing a lexical semantic pattern for recognition of Korean syntax.
The method includes the steps of: (a) defining a lexico-semantic-pattern (LSP) consisting of a morpheme, a syllable, and a word; (b) collecting or generating at least one representative sentence type that the user terminal desires to extract; (c) transforming the representative sentence pattern into an LSP form through morphological processing of the representative sentence pattern; And (d) reconstructing the representative sentence pattern into an LSP by using the representative sentence pattern converted into the LSP form and selecting the object to be changed into the dictionary form.

Description

{METHOD OF RE-PREPARING LEXICO-SEMANTIC-PATTERN FOR KOREAN SYNTAX RECOGNIZER}

The present invention relates to a natural language sentence analysis method, and particularly relates to a method for constructing a lexical semantic pattern for recognizing a Korean phrase.

The language used by humans is called natural language, and the language created by human needs is called artificial language. Natural languages include Korean, English, French, German, Japanese, Chinese and many other languages. Each natural language has occurred naturally, and has evolved naturally over time.

Artificial languages include Esperanto and various programming languages. Since these artificial languages are created by human needs, rather than naturally evolving, only the parts needed by human will and planning are changing little by little.

The artificial language has a definite form because it is created with a specific purpose. Therefore, it is easy to recognize the syntactic structure based on the form of the artificial language. In the case of natural language, although there is a certain form, it occurs naturally for communication between human beings. Since it is not occurred after the formalization like the artificial language, it is possible to infer the grammar which is the form of the language through the language The grammar of the language is established.

The natural language processing is called natural language processing and it is divided into morpheme analysis, tagging with parts (tagging), syntax analysis, semantic analysis, and phonetic analysis step by step. At present, it has been invented up to the level that it can be put to practical use and commercialization in Korean by morphological analysis and part-of-speech.

However, with the development of the Internet, the amount of information is enormous, the creator of information is also diversified, and the number of informal documents becomes much larger than the formalized form of documents, so that the users of the information, We want to extract and use the information in the document. However, it seems that it is almost impossible to parse Korean by the existing natural language processing method. Therefore, natural language processing researchers are attempting to classify and analyze Korean sentence patterns as a method for parsing and apply them in the form of LSP (Lexico Syntactic Pattern).

The inventors of the present invention have completed the present invention after a long effort of studying for the recognition of Korean phrases that can be effectively applied to non-standardized documents.

Accordingly, a problem to be solved by the present invention is to provide a method for constructing a new lexical semantic pattern that can complement a shortage in recognition of the Korean language, which is a natural language, based on natural language processing research. The present invention defines a Lexico-semantic-pattern (LSP) designed to enable a variety of expressions and flexible techniques using morphological, syllabic, and verbal structures, and uses them to recognize the meaning of Korean phrases and sentences And to provide an efficient way to identify them.

On the other hand, other unspecified purposes of the present invention will be further considered within the scope of the following detailed description and easily deduced from the effects thereof.

In order to achieve the above object, the present invention provides:

(a) defining a lexico-semantic-pattern (LSP) consisting of a morpheme, a syllable, and a word;

(b) collecting or generating at least one representative sentence type that the user terminal desires to extract;

(c) transforming the representative sentence pattern into an LSP form through morphological processing of the representative sentence pattern; And

(d) selecting an object to be changed into a dictionary form from the representative sentence pattern converted into the LSP form, and reconstructing the representative sentence pattern into an LSP using the selected sentence pattern, and constructing a lexical semantic pattern for recognizing the Korean sentence do.

In a preferred embodiment of the method for constructing a lexical semantic pattern for recognizing a Korean phrase according to the present invention, it is preferable to further include (e) reconstructing an LSP that assigns a specific vocabulary or expression as a variable in the representative sentence pattern.

Further, in a preferred embodiment of the method for constructing a lexical semantic pattern for recognizing a Korean phrase according to the present invention, the step of extending the LSP of the representative sentence pattern by combining the wildcard operator with the representative sentence pattern converted into the LSP form It is good to do.

Further, in the method of the present invention, elements defined by a basic unit of the lexical semantic pattern may include a word, a morpheme, a syllable, a dictionary, and a variable.

The embodiments of the present invention have at least the following effects:

It is possible to find out exactly the desired sentence pattern beyond the simple word or combination of words in the unstructured data, and it is possible to extract the desired information accurately in the sentence pattern, and thus it can be utilized more effectively in recognizing and parsing the Korean sentence. In addition, many systems requiring natural language processing such as a speech recognition processor, a query response processor, and an SNS emotion analyzer can analyze Korean sentences more intuitively and efficiently and grasp the meaning.

Even if effects not specifically mentioned in the specification of the present invention are incorporated, the provisional effects expected by the technical features of the present invention are treated as described in the specification of the present invention.

1 is a flowchart schematically showing an entire process of a method for constructing a lexical semantic pattern according to an embodiment of the present invention.
* The accompanying drawings illustrate examples of the present invention in order to facilitate understanding of the technical idea of the present invention, and thus the scope of the present invention is not limited thereto.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description of the present invention, a detailed description of known functions and configurations incorporated herein will be omitted when it may obscure the subject matter of the present invention.

In addition, each block may represent a module, segment, or portion of code that includes one or more executable instructions for executing the specified logical function (s). It should also be noted that in some alternative implementations, the functions mentioned in the blocks may occur out of order. For example, two blocks shown in succession may actually be executed substantially concurrently, or the blocks may sometimes be performed in reverse order according to the corresponding function.

FIG. 1 illustrates a method of constructing a Lexico-semantic-pattern (LSP) according to an embodiment of the present invention. Thus, by constructing the lexical semantic pattern, it provides a means of recognizing the Korean sentence and showing a more useful effect in understanding the meaning of the sentence. In addition, through the above process, the lexical semantic pattern is reconstructed to provide an environment for extracting the desired information in the sentence pattern accurately. This environment is particularly useful for unstructured data where it is difficult to find the desired information simply by word or word combination. It is not simply sending and receiving words, but analyzing sentences through computer devices and understanding the meaning of them, thereby creating a system environment that can process natural language.

The process of FIG. 1 is performed through a Korean syntax recognition apparatus, and the Korean syntax recognition apparatus may be configured by a software module. The device may also be integrated into a speech recognition system, a query response processing system, an SNS emotional analysis system, or a search system. From the viewpoint of the network, the user terminal connects to the system server to execute a series of events such as input, request, and query. The Korean syntax recognition apparatus installed on the server side includes a memory and a processor for storing programs and resources. The software installed in the processor may include means for defining a lexical semantic pattern, means for morpheme processing, means for converting into a lexical semantic pattern, and the like.

First, the definition module of the system defines a statement and an operator necessary for describing the LSP (S100). The " expression " defined in step S100 is a basic unit constituting the LSP, and the expression may include a word, a morpheme, a syllable, a dictionary (category), and a variable. The operator represents the relationship between these expressions, and the statement can be defined as shown in Table 1.

designation Statement meaning Eulogy W -Eyes
- Surface vocabulary separated by white space morpheme {(L; L / P; / P) +? (L; L / P; / P)?} - Expressing the vocabulary and parts of a morpheme
- Expressed in curly brackets ({}) or connected by plus (+)
- Always attach a slash (/) in front of the part of speech Syllable [ C ] - syllable
- Expressed within square brackets ([]) dictionary @ category - record one or more morphemes with the same / similar meaning
- Attach @ before the category name variable & variable = stmt - express the value to extract and the category of its value

Operators representing the relationships between representations can be defined as in Table 2. W is a word, L is a morpheme, P is a part of a morpheme, N is an integer greater than or equal to 0, and stmt is a statement.

Operator grammar meaning ; stmt ₁ ; stmt ₂
W ₁ ; W ₂
( L / P ) ₁ ; ( L ) ₂ ; (/ P ) ₃ -OR {} { stmt ₁ } { stmt ₂ } - One qualifier that distinguishes between expressions in a word phrase () () - Qualifiers that distinguish priority and units = & variable = stmt - an operator that specifies the category of values / / P - express the part of morpheme (the value of part of speech is expressed by the value defined in Korean morpheme analyzer) + stmt ₁ + stmt ₂ - Check for morphological bonding to left / right
- Confirm the combination of the rightmost and the leftmost morpheme in the expression on the right and the expression on the right. ^ ^ N _min ~ N _max
N _min < N _max , N _min ≥ 0
^: = ^ 0 ~ ∞, ^ ~ N _max : = ^ 0 ~ N _max
^ N _max : = ^ 0 ~ N _max , ^ N _min ~: = ^ N _min ~ ∞ Replace all W, L , P with N _min ~ N _max
- You can substitute L and P in the case of {} with the qualifier, but not W.
- is used as a wild card form representing one or Eojeol including W text when used in the earlier or later in W. # # N _min ~ N _max
N _min < N _max , N _min ≥ 1
#: = ^ 0 ~ ∞, # ~ N _max : = # 0 ~ N _max
# N _max : = # 0 ~ N _max , # N _min ~: = # N _min ~ ∞ Replace all W , L , P with N _min ~ N _max
- You can substitute L and P in the case of {} with the qualifier, but not W.
- is used as a wild card representing the form of two or more W Eojeol including text when used in the earlier or later in W. ? ? or {?} or stmt ? - Meaning to apply or not to all expressions
- replace all expressions once or implicitly ! ! stmt or! W or! L or! P - Expressions that deny the original meaning of expression or word \ \ character , where
character ∈ {(,), {,}, =, +, *, #, @,?, &,!, \, ~} - literally means the value to be interpreted correctly
- Literal \ is applied to one character * stmt * - Repeating expressions
- The cardinality of * is the same as ^ or # [] POSIX character class
[: alpha:]
[: digit:]
[: lower:]
[: upper:]
PERL character class
[A-Za-z0-9]
[! "# $% &'() * +,. / :; <=>? @ \ ^ _` {|} ~ -] Partial representation of POSIX and PERL character classes $ ^ $, # $, & $ variable - It comes after ^ or #
- Perform post-processing if $ exists

Table 3 shows the priorities of the above-mentioned operators. Specifically, except for the ^, #, and * operators among the operators listed in Table 3, the expression on the left in the combination is defined to take precedence.

ranking Operator Join sequence One () {} \ Left first 2 ! Left first 3 ; Left first 4 = Left first 5 ^ # * Right first 6 + Left first

An example of the expression defined in Table 1 is shown in Table 4.

Example statement meaning Kim, Yeon - Ah; Meaning of the phrase "Kim Yu-na", "Yeon-a" or "Yeon-nim" {Kim Yu-na + / J_; Kim Yeon-ah + / JKG} A morpheme pattern in a word meaning "Kim Yu-na + / J_" or "Kim Yu-na + / JKG" {@ Male; @ animal} Means that one entry in the dictionary "male" and "animal" is assigned It is Kim Yu-na. "Kim Yu-na" means "it is." {& Add1 = @ Address} &Quot; Add1 &quot; is assigned to one of the entries belonging to the dictionary &quot; Address &quot; # +! / EM One or more words, and the part of the last morpheme of the word consisting of two or more morphemes expresses up to the vernacular rather than the ending ending (/ EM) Seoul + # +! / EM In more than one word, more than two morphemes are shown starting from the morpheme "Seoul", and the last morpheme expresses up to the vernacular rather than the ending mother (/ EM) {# +! / EM} It is composed of two or more morphemes, and unless the part of the last morpheme is / EM, ^ 1 ~ 4 Korea Expressing text with one to four random phrases in front of the phrase "Korea" * 3 ~ "A" represents the text that is repeated three or more times, the minimum length of the text is " [0-9] * [2] Represents text where there is no white space between characters, characters between 0 and 9 are repeated 0 to 10 times and ends with 2 [1] * 3 [: digit:] [: digit:] * [4] Minimum length text that matches the expression, eg "1 2 3 4"

Next, one or more representative sentence patterns are collected or generated in the sentences to be extracted by the user (S110). For example, suppose that you have chosen two representative sentence patterns as the representative sentence patterns including the birthplace information of the novelist "other".

Representative Moon Type 1: The other was born in Hamyang-gun, Gyeongsangnam-do.

Representative Moon Type 2: Gyeongnam, Hamyang-gun is a place where a foreigner is born.

Then, the representative sentence pattern is transformed into the LSP form after the morphing process (S120). Analysis of morpheme and selection of morpheme pattern for sentence selected as representative sentence pattern are performed. Through the process, the two representative sentence patterns are expressed in LSP form as follows.

LSP1: <person> + Survey <place> + / Investigation <born-verb> + ^ + mother

LSP2: <place> + survey <person> + survey <born-verb> + tubular mother <place-noun> + ^ + mother

Next, an object to be changed from the representative sentence type converted into the LSP type to the dictionary type is selected, and the representative sentence type is reconfigured into the LSP using the selected target sentence type (S 130). For example, in the above sentence, the symbol <... >, The LSP can be reconfigured by changing the information that the user desires to use only the information in the dictionary information to the dictionary form of the LSP defined in the present invention. In the above example, the objects to be changed to the dictionary form are <born-verb> and <place-noun>.

<born-verb> is born / is born in verb / is born in verb / verb ... ...

<place-noun> place / noun place / noun area / noun house / noun ... ...

Based on this, the LSP can be reconstructed for the representative sentence pattern as follows.

LSP1: <person> + Survey <place> + / survey @ born-verb + ^ + mother

LSP2: <place> + survey <person> + survey @ born-verb + tubular ending @ place-noun + ^ + ending

In this way, the vocabulary or expression having similar meaning in the representative sentence pattern, and the morpheme constituting the actual grammatical form are grouped together, and each of them is specified in the dictionary, and various sentences corresponding to the representative sentence pattern can be interpreted using the dictionary.

Next, the LSP can be reconstructed by assigning a meaningful specific vocabulary or expression as a variable in the representative sentence pattern (S240). That is, this step is an object recognition step for extracting a specific vocabulary or expression that is meaningful or core in context. When the variable allocation of the LSP for directly extracting the portion corresponding to the information to be obtained and delivering it to the user is applied to the above example, the objects to be allocated are <person> and <place>. The reconstructed LSP is as follows.

LSP1: (& person = person noun) + Survey (& place = nominal noun phrase) + from / survey @ born-verb + ^ +

LSP2: (& place = nominal phrase noun phrases) + survey (& person = person nouns) + survey @ born-verb + tubular endings @ place-noun + ^ +

The information extracted from the representative sentence patterns through variable assignment by performing step S240 is a person = "other number" and place = "Hamyang gun", and many meaningful information can be extracted by using such expression.

Through the above LSP configuration method, information can be extracted from sentences of the same type as the representative sentence pattern. This provides practical assistance in parsing and parsing Korean. However, there is a need for a method that can absorb and utilize more various sentences out of constraints such as syllables, morphemes, and phrases, which are constituent units of LSP. For this purpose, in one embodiment of the present invention, the LSP of the representative sentence type may be extended by combining the wildcard operator with the LSP (S250). The two representative sentence types can be extended as follows by combining wildcard operators.

LSP1: ^ (& person = person noun) + Survey ^ ~ 3 (& place = locative expression noun phrase) + to / search ^ ~ 2 @ born-verb + ^ +

LSP2: ^ (& place = locative noun phrase) + survey ^ ~ 3 (& person = person noun) + survey ^ ~ 2 @ born-verb +

With this extended LSP, it is possible to extract information even in the following sentences.

"The other was born on August 15, 1946, in Hamyang-gun, South Gyeongsang Province."

"Located near the Jirisan National Park, Gyeongsangnam-do is a place where a novelist is born."

Through the above-mentioned embodiments, a method of constructing an LSP for finding a sentence containing the expression or meaning of a person-born place has been described. If you add more LSP to other representative sentence patterns, you can find many sentences corresponding to "someone - born place", and you can use it in various systems.

The scope of protection of the present invention is not limited to the description and the expression of the embodiments explicitly described in the foregoing. It is again to be understood that the scope of protection of the present invention can not be limited by obvious alterations or permutations of the present invention.

Claims (4)

(a) defining a lexico-semantic-pattern (LSP) consisting of a morpheme, a syllable, and a word;
(b) collecting or generating at least one representative sentence type that the user terminal desires to extract;
(c) transforming the representative sentence pattern into an LSP form through morphological processing of the representative sentence pattern; And
(d) selecting an object to be changed into a dictionary form from the representative sentence pattern converted into the LSP form, and reconstructing the representative sentence pattern into an LSP using the selected sentence pattern, and reconstructing a vocabulary semantic pattern for recognizing Korean phrases. The method according to claim 1,
(e) reconstructing an LSP that assigns a meaningful specific vocabulary or expression as a variable in the representative sentence pattern. The method according to claim 1,
And expanding the LSP of the representative sentence pattern by combining the wildcard operator with the representative sentence pattern converted into the LSP form. The method according to claim 1,
A method for reconstructing a lexical semantic pattern for recognition of a Korean phrase, the element being defined as a basic unit constituting the lexical semantic pattern includes a word, a morpheme, a syllable, a dictionary, and a variable.

Images