KR20020048715A - Natural Language Analyzing Apparatus and Method for Controlled Korean Grammar - Google Patents

Abstract

PURPOSE: A system and a method for guiding a standard sentence pattern are provided to guide a user in a standard sentence pattern in real time by consulting the standard Korean sentence pattern. CONSTITUTION: The system includes an input unit(110) through which a user inputs a sentence, a morpheme analyzing unit(120) to divide the inputted sentence into the syllables and to analyze the morphemes composing the syllables, a vocabulary information extracting unit(130) to extract the vocabulary information of the analyzed sentence, a standard sentence pattern rule determining unit(140) to determine standard sentence pattern rules and apply the rules, and an output unit(170) to output a standardized sentence pattern of the inputted sentence depending on the applied standard sentence pattern rules. The standard sentence pattern rule determining unit applies the most proper one among the Korean standard sentence pattern rules loaded and pre-stored in the system. A standard sentence pattern is verified and deducted by using the abbreviation deduction patterns sought by an abbreviation deduction pattern search unit(150).

Description

Standard sentence pattern induction device and its method by Korean standard sentence pattern rules {Natural Language Analyzing Apparatus and Method for Controlled Korean Grammar}

The present invention relates to a standard sentence pattern derivation apparatus and a method according to the Korean standard sentence pattern rules, and more particularly, to a standard system for deriving a standard sentence pattern to a user in real time in an application system conforming to a limited language standard such as a standard sentence pattern. The present invention relates to a sentence-guided apparatus and a method thereof and a computer-readable recording medium having recorded thereon a program for realizing the method.

Standard sentence derivation work using limited grammar has been ongoing in English. The representative one, 'N. E. Fuchs 'and' R. Schatter's proposed "Attempto Controlled English (ACE), <CLAW 96, March 1996>", etc., which proposes the English standard called ACE to prepare the requirements specification.

The biggest problem with language processing systems based on standards is how to drive language generation into standardized types.

In the case of the ACE, 'Paraphrase' was used as a standard draft derivation method. However, since the above rewriting is limited only at the end of a sentence, there is a problem in that it is not possible to give a real-time feedback to a sentence inputter, thereby reducing the productivity of sentence generation.

The present invention proposed to solve the problems as described above, in the application system that conforms to a limited language standard, such as the standard sentence pattern, a standard sentence pattern deriving apparatus and a method and a method for deriving the standard sentence pattern in real time to the user It is an object of the present invention to provide a computer-readable recording medium that records a program for realization.

That is, according to the present invention, in order to create a text according to the standard sentence pattern rule, the user's non-normal and ordinary sentence expressions are replaced with an abbreviated estimation pattern so as not to require the user special knowledge of the standard sentence pattern rule. It is an object of the present invention to provide a standard sentence pattern deriving apparatus for naturally leading to a standard sentence pattern in real time, and a computer readable recording medium recording the method and a program for realizing the method.

1 is a configuration of one embodiment of a standard sentence pattern induction apparatus according to the standard sentence pattern rule according to the present invention.

2 is a flow diagram of one embodiment of a standard sentence derivation process by the standard sentence rules according to the present invention.

3 is an exemplary explanatory diagram of a lexical category list based on a standard sentence pattern according to the present invention;

Figure 4a and 4b is an exemplary explanatory diagram of a sentence structure by the standard sentence pattern rules according to the present invention.

5A and 5B illustrate an example of a progressive syntax analysis process using standard sentence patterns according to the present invention.

6 is an exemplary explanatory diagram of an estimation rule for omitting arguments according to the present invention.

An apparatus of the present invention for achieving the above object is a standard sentence pattern induction apparatus according to the Korean standard sentence pattern rules, the input unit for receiving a sentence by the user; A morpheme analysis unit for analyzing morphemes by dividing sentences received through the input unit into words; A lexical information extraction unit for extracting lexical information of sentences analyzed by morpheme analysis units by the morpheme analysis unit; A standard sentence pattern determination unit for determining and applying a standard sentence pattern rule according to the lexical information extracted through the lexical information extraction unit; And an output unit for outputting a standard sentence pattern verification result for the input sentence according to the standard sentence pattern rule determined and applied by the standard sentence pattern rule determining unit.

In addition, the present invention, if the standard sentence pattern determined by the standard sentence pattern determination unit is not directly applied to the input sentence, omitted estimation pattern search unit for searching for the omitted estimation pattern; And a standard sentence pattern verification and estimator for verifying and estimating a standard sentence pattern according to the omitted estimation pattern searched through the abbreviation estimation pattern search unit.

In addition, the method of the present invention, in the standard sentence derivation method applied to the standard sentence derivation apparatus according to the Korean standard sentence pattern rules, when a sentence is input, the input sentence is divided into words to form a morpheme for the separated words A first step of analysis; A second step of extracting lexical information of the analyzed morpheme and determining a standard sentence pattern to be applied according to the extracted lexical information; And a third step of outputting a standard sentence pattern verification result for the separated word, generated by applying the determined standard sentence pattern rule to the extracted lexical information.

In another aspect, the present invention, a standard sentence-type induction system having a processor, when a sentence is input, the first function of analyzing the morpheme for the separated word by dividing the input sentence into a word; A second function of extracting lexical information of the analyzed morpheme and determining a standard sentence pattern to be applied according to the extracted lexical information; And a computer-readable recording medium having recorded thereon a program for realizing a third function of outputting a standard sentence pattern verification result for the separated word, generated by applying the determined standard sentence pattern rule to the extracted lexical information. Characterized in that.

The above objects, features and advantages will become more apparent from the following detailed description taken in conjunction with the accompanying drawings. Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a configuration of an embodiment of a standard sentence pattern induction apparatus according to the standard sentence pattern rule according to the present invention.

As shown in FIG. 1, the standard sentence pattern deriving apparatus according to the Korean standard sentence pattern rule divides a sentence input through the input unit 110 and a sentence received through the input unit 110 by a word by a user. Morphological analysis unit 120 for analyzing the morphemes constituting the separated word, lexical information extraction unit 130 for extracting the lexical information of the sentence analyzed by the morpheme analysis unit 120, the lexical information According to the lexical information extracted by the extraction unit 130, according to the standard sentence pattern determination unit 140 and the standard sentence pattern rule determination unit 140 for determining and applying the standard sentence pattern rule applied according to the standard sentence pattern rule And an output unit 170 for outputting a standardized sentence pattern for a sentence input through the input unit 110.

Here, the standard sentence pattern determination unit 140 selects a suitable one among Korean standard sentence patterns that are stored and loaded in advance in the system to which the standard sentence pattern induction device according to the present invention is applied, according to an input sentence. It performs the function to apply.

In addition, when the standard sentence pattern rule determined by the standard sentence rule determination unit 140 is not directly applied to the input sentence, the elimination estimation pattern search unit 150 and the elimination estimation pattern search unit for searching for the elimination estimation pattern The apparatus further includes a standard sentence pattern verification and estimator 160 for verifying and estimating the standard sentence pattern according to the omitted estimation pattern retrieved through 150.

Detailed functions and operations of the components will be described below.

Figure 2 is a flow diagram of an embodiment of a standard sentence derivation process according to the standard sentence rules according to the present invention.

As shown in FIG. 2, the standard sentence pattern derivation process according to the present invention starts with a sentence being input by a user (220) into a system that is first initialized by loading (210) a standard sentence pattern rule.

In this case, the loaded 210 standard sentence rules are stored in a memory and applied to parsing for standard sentence checks.

The sentence 220 input by the user is analyzed in units of words (230). That is, if one word is input and a blank for confirmation thereof is input, the morpheme analysis and tagging operations are immediately performed in step 231.

Thereafter, if necessary, lexical information extraction is performed (232). The lexical information extraction process (232) is for obtaining descriptive argument composition information for each lexicon from the lexical dictionary. That is, when the predicate "go" is entered, it obtains subcategories of the predicate "--go to-" and applies it to standard sentence checking.

Through the lexical information extraction process 232, a suitable standard sentence pattern for the analyzed word is determined (233).

For further explanation, look at the sentence "I go to the laboratory" illustrated in Figure 5a below, after the first word "I" is analyzed as "I / nc / jc" (512), the analyzed It can be seen that the appropriate standard sentence rules are assigned 513 for each, which is the process of determining 232 standard sentence rules through the lexical information extraction (232).

The standard sentence rule determination 233 is performed using tagging information of the morpheme.

Subsequently, a standard parsing rule matching (Matching) process, which is a partial parsing process, is performed (240), and if the analyzed 231 morpheme does not conform to the standard sentence formatting rule, an abbreviated estimation pattern search is performed ( 242). In the case of a nested statement or a subsequent sentence, the abbreviated estimation pattern search 242 may estimate the sentence as a sentence or a subsequent clause containing an argument omitted from a nested statement or a preceding clause (243) so that the standard sentence pattern matching operation may be continuously performed. It's a process.

If the analyzed 231 morpheme conforms to the above standard sentence pattern rule (240), or completes the standard sentence pattern estimation through the abbreviated estimation pattern search 242 (243), through the alternate writing output (241). ) Results of the analysis in progress are displayed. Therefore, the sentence inputter can know in real time whether the sentence he is currently generating is suitable for the standardized sentence pattern, and learn a specific standard sentence type by continuous sentence input.

The term information constituting the standard sentence pattern rule is described by way of example in FIG. 3 below, in which the indicator adjective dt, the canonical adjective nu, and the constellation adjective ad all represent the detailed lexical category of the tubular adverb mm.

3 is an exemplary explanatory diagram of a lexical category list based on a standard sentence pattern rule according to the present invention.

As shown in FIG. 3, the list of lexical categories used in the present invention is largely 31, but this is only an arbitrary example. In some cases, the list of categories and names may be different. It is okay.

However, in the present invention, description will be limited to the list and names of the lexical categories shown in FIG. 3.

4A and 4B are diagrams illustrating an example of sentence structure according to a standard sentence pattern rule according to the present invention.

As shown in Figs. 4A and 4B, the basic statements 410 and 411 are composed of one term predicate, two term predicate and three term predicate. The NP1 constituting each basic sentence is constituted by combining the noun phrase _NP and the main investigation jc1 as shown in the section 421 of FIG. 4B. In the same manner, NP2 and NP3 are configured by the target investigation jc2 and the required sub-firing investigation jc3, respectively.

The argument 420 in FIG. 4A refers to the NP1, NP2, and NP3.

The dispute expansion 450 of FIG. 4A is constituted by the dispute expansion 451 of the tubular yarn, the connection investigation, the dispute expansion 452 of the connection adverb, and the dispute expansion 453 of the speed investigation. Here, the argument expansion means an argument including a qualifier rather than an argument constituted by a single NP. That is, in the argument expansion 451 of the tubular sentence, in the case of "this person", the "person" is modified by the tubular sentence "this" to "((/ mm person / nc) / _ NP + / jc1) / NP1". It can be a structural analysis.

Here, when explaining the difference between "NP" and "_NP", NP (471) refers to a compound noun or a single noun, _NP is an extended NP, such as "(/ mm (person / nc) / NP) / _NP" Say.

The standard sentence patterns described in FIGS. 4A and 4B follow a CFG (Context Free Grammar) format and take a multi-tree scheme.

In the notation of the standard sentence syntax, A: (B + [C] D {}…) is an indication that A is a superset of (B + [C] D {}…) and that "[]" is not a mandatory requirement. In addition, "{}" represents constraint information of the component.

In the loading step 210 of the standard sentence pattern rule of FIG. 2, the rule items of FIGS. 4A and 4B are stored in a memory, respectively, and applied to the parsing as the standard sentence pattern rule.

In the tubular article argument expansion 351 of FIG. 4B, "{}" in mm {!. [Dt]} describes constraint information of each rule configuration item. Constraints of standard sentence rules consist of all three types.

"!." Means that the detailed lexical category information, that is, the adjective mm {!. [Dt]} of the adjective argument expansion (451) must be allowed only for the indication adjective dt. ) '! @' Means that "NP + ','" can be used recursively.

In addition, "! ^" Of the argument expansion 453 of the personality investigation means that the noun phrase _NP, which is expressed by the "-of" which is the personality investigation jm, may have only a single NP therein. For example, "Brother's mother and brother" is always "(Brother's mother) and brother" combined with "Brother's mother" and not "(Brother's brother)".

The descriptive enlargement 460 of FIG. 4A is comprised of the descriptive enlargement 465 of the adverb of FIG. 4B, and the descriptive enlargement 463 of an auxiliary language. The clause 430 is composed of a tubular clause ETMS 454, a noun clause 472, a quotation clause JQTS 464, an adverb clause 465, a narrative clause 466, and the argument expansion according to the clause is also described in FIG. .

The expansion of the argument of the verse (454), the sentence _S and the syntactic protagonist etm combines to act as a tubular word, and the noun clause 472 is the noun clause by the sentence _S and the noun utter etm or et Will be constructed. In addition, the citation 464 is composed of the sentence _S and a citation search of "say, go". Adverbs 465 are generated by the descriptive checks "like, without, otherwise" and "crab, like,". Finally, the predicate clause 466 is necessary to process a syntax in which two NP1s occur in a sentence, such as "elephant / NP1 long / NP1 long", which expresses a double subject statement.

5A and 5B are diagrams illustrating an example of a gradual syntax analysis process using standard sentence rules according to the present invention.

That is, FIGS. 5A and 5B illustrate an example of a gradual syntax analysis process using a standard sentence pattern according to the present invention, and show a gradual syntax analysis technique and a rewriting method for the sentence “I go to a laboratory”.

In the present invention, a gradual syntax analysis technique means that syntax analysis is performed in parallel with a sentence input process.

First, in the progressive parsing process of FIG. 5A, as described above, the corresponding standard sentence pattern rule 513 is assigned to the first step "I / nc / jc" 511, and "NP1: (_ NP (I / nc)". parallel analysis is performed as in the following.

The first step 530 of FIG. 5B represents the above process in a tree form.

Rewriting refers to the visualization of analysis results.

In the visualization method proposed in the present invention, there are two methods of presenting the analysis result of FIG. 5A and the tree analysis result of FIG. 5B. The two analysis results show the syntax analysis results of each sentence of the sentence, thereby providing the user with the feedback of the gradual syntax analysis so that the user can know the degree of understanding the user input of the current system.

Reference numeral 522 of FIG. 5A represents a result of finding all the arguments and predicates of “I go to the laboratory”. This may be represented by 550 of FIG. 5B.

In this case, as shown in (560) of FIG. 5B, the basic sentence is selected by the arguments NP1, NP3, the descriptor VP, and "ga / pv". In this case, the third clause of S: (NP1 NP2 NP3 VP) is selected. 570 of FIG. 5B shows the final analyzed result.

In addition, even if the position of the subject and the adverb is changed, such as "I go to the research institute", there is no problem in the application of the standard sentence pattern proposed by the present invention.

The decision of NP3 and NP1 itself is independent of the order. In other words, since the arguments and the predicates are only combined in the items that coincide with each other, the order of the arguments is not a problem in the selection of sentence type S: (NP1 NP2 NP3 VP) by the final descriptor VP.

Therefore, the problem of free word order between arguments of Korean sentences can be solved without difficulty in applying this standard sentence pattern rule.

6 is an exemplary explanatory diagram of an estimation rule for skipping an argument according to the present invention.

The estimation rule table for the omission of the argument shown in FIG. 6 is presented as an example of a countermeasure for the frequent omission of the language in Korean.

That is, the estimation rule for the omission of the argument shown in FIG. 6 is used for the omission estimation pattern search 242 and the standard sentence pattern verification and estimation 243 of FIG. 2 described above.

The reason for this omission is that the standard sentence itself must satisfy the argument of the predicate. In other words, there is a problem that the above sentence cannot be seen as a front door because the object of the term "eat" is omitted in "Cheol eats".

In the present invention, there are two ways of dealing with the omission, firstly, when the short sentence has the same non-deterministic argument as in the above example, it is analyzed in the transposition as "E.L./NP1/NP2/VP", and the input to the inputter as it is. To show. In other words, the authenticity of the analysis result is dependent on the confirmation of the user. This method is meaningful to facilitate the natural and gradual analysis of user input.

Second, in the case of compound sentence, the general argument omission is to estimate the argument in the related sentence. In this case, the relational sentence is a nested sentence, and the inner sentence becomes a sentence, and in the case of a subsequent sentence, a preceding clause or a trailing clause.

6, 610, 620, and 630 show an estimation example for NP1, NP2, and NP3, respectively, and 640 shows an estimation example for NP1 and NP2, respectively.

In 670 and 680, example sentences for estimating the following sentences are described.

Although the present invention described above is limited to input and output of a character, the current speech recognition technology receives a voice signal, converts it to a character, and performs internal operations. It is obvious that there is no problem in applying and implementing the input / output system.

As described above, the method of the present invention may be implemented as a program and stored in a recording medium (CD-ROM, RAM, ROM, floppy disk, hard disk, magneto-optical disk, etc.) in a computer-readable form.

The present invention described above is capable of various substitutions, modifications, and changes without departing from the spirit of the present invention for those skilled in the art to which the present invention pertains, and the above-described embodiments and accompanying It is not limited to the drawing.

As described above, the present invention has an effect of deriving a standard sentence pattern through a real-time response by enabling dynamic analysis of word units using a gradual syntax analysis technique according to a standard sentence pattern rule.

In addition, the present invention, by solving the free word order and abbreviation phenomena frequently in the Korean standard sentence rules and argument estimating method, it is possible to make the standard sentence pattern derivation process for the user's input sentence more natural.

Claims (8)

In the standard sentence design apparatus according to the Korean standard sentence pattern rules, An input unit for receiving a sentence by a user; A morpheme analysis unit for analyzing morphemes by dividing sentences received through the input unit into words; A lexical information extraction unit for extracting lexical information of sentences analyzed by morpheme analysis units by the morpheme analysis unit; A standard sentence pattern determination unit for determining and applying a standard sentence pattern rule according to the lexical information extracted through the lexical information extraction unit; And An output unit for outputting a standard sentence pattern verification result for the input sentence according to the standard sentence pattern rule determined and applied by the standard sentence pattern rule determining unit Standard sentence pattern induction apparatus according to the Korean standard sentence pattern rules, including. The method of claim 1, An omission estimation pattern search unit for searching for an omission estimation pattern when the standard sentence pattern rule determined by the standard sentence rule determination unit is not directly applied to the input sentence; And Standard sentence pattern verification and estimator for verifying and estimating a standard sentence pattern according to the omitted estimation pattern retrieved through the elimination estimation pattern search unit Standard sentence pattern induction apparatus according to the Korean standard sentence pattern rules further comprising. The method according to claim 1 or 2, The sentence sentence input device according to the Korean standard sentence pattern rule, characterized in that the input and output of the sentence includes the input and output by the character and the input and output by the voice. In the standard sentence derivation method applied to the standard sentence deduction device according to the Korean standard sentence pattern rules, A first step of dividing the input sentence into a word and analyzing a morpheme of the divided word when a sentence is input; A second step of extracting lexical information of the analyzed morpheme and determining a standard sentence pattern to be applied according to the extracted lexical information; And A third step of outputting a standard sentence pattern verification result for the separated word, generated by applying the determined standard sentence pattern rule to the extracted lexical information; Standard sentence pattern derivation method according to the Korean standard sentence pattern rules, including. The method of claim 4, wherein A fourth step of verifying and estimating the standard sentence pattern by searching for an omitted estimation pattern if the determined standard sentence rule is not directly applied to the extracted lexical information; Standard sentence pattern derivation method according to the Korean standard sentence pattern rules further comprising. The method according to claim 4 or 5, Input and output of the sentence, the input and output by the character and the input and output by the voice, characterized in that the standard sentence pattern derivation method according to the Korean standard sentence pattern rules. In a standard sentence guidance system with a processor, A first function of dividing the input sentence into a word and analyzing a morpheme of the divided word when a sentence is input; A second function of extracting lexical information of the analyzed morpheme and determining a standard sentence pattern to be applied according to the extracted lexical information; And A third function of outputting a standard sentence pattern verification result for the separated word, generated by applying the determined standard sentence pattern rule to the extracted lexical information A computer-readable recording medium having recorded thereon a program for realizing this. The method of claim 7, wherein A fourth function of verifying and estimating a standard sentence pattern through a search for an omitted estimation pattern if the determined standard sentence rule is not directly applied to the extracted lexical information; A computer-readable recording medium that records a program for further realization.