KR100798752B1 - Apparatus for and method of korean orthography - Google Patents

Abstract

A device and a method for checking Hangul orthography are provided to present a replaceable word spacing nouns when a general rule is not applied to a complex noun, and recognize an idiomatic part-of-speed comparing the multi-word as one checking target while checking Korean orthography. A general dictionary analyzer(10) stores the target word by dividing the inputted word into each morpheme and compares the target word with a general dictionary(12). The general dictionary analyzer stores a corrected replaceable word to a buffer if an error is found the target word, and stores an ID number when a style error is found. A complex noun analyzer(20) memorizes spacing positions between the words if the target word does not follow the rule of a complex noun restriction rule dictionary(22). A multi-word analyzer(30) recognizes the multi-word as one morpheme if the target word is matched with a start word of a multi-word dictionary(32). A meaning/style analyzer(40) generates a rule list by searching a correction rule from a meaning/style correction rule dictionary(42) if the ID number is stored in the target word, and generates/stores the replaceable word by selecting one rule based on the word before/after the target word.

Description

Korean spell checker and test method {APPARATUS FOR AND METHOD OF KOREAN ORTHOGRAPHY}

1-a system for providing corrected word information for the conventional error suspect clause

2-block diagram of a Korean spelling checker in accordance with a preferred embodiment of the present invention.

3 is a block diagram of a semantic analyzer according to a preferred embodiment of the present invention.

4-flowchart of a Korean spelling checking method according to a preferred embodiment of the present invention.

** Description of symbols for the main parts of the drawing **

10: general dictionary analyzer 12: general dictionary

14: Determining Semantic Style Error 20: Complex Noun Analyzer

22: compound noun constraint rule dictionary 30: multiword analyzer

32: multiple word dictionary 40: semantic grammar analyzer

42: Dictionary of semantic correction rules S10: text input step

S20: extract word step S30: general dictionary analysis step

S40: compound noun analysis step S42: constraint rule determination step

S44: Spacing between nouns S50: Multiword analysis step

S52: Multiple word dictionary search step S54: Multiple word alternative word determination step

S60: semantic style analysis step S62: correction rule list step

S64: Step for determining semantic substitute word S70: Step for determining morpheme

S72: calibration routine step S80: display step

S90: Display step S100: Buffer word discrimination step

The present invention relates to a Korean spelling checker and a test method, and more particularly, to a Korean spelling checker and a test method for checking a compound noun, a multiple word clause, and a word having a meaning and stylistic error.

In general, the spell checker narrowly refers to providing information necessary for simple spelling and grammar correction, and to a broader extent, it checks for stylistic errors to support the effective writing of a document or sentence. It provides the terms of punctuation, informs the suitability of punctuation such as punctuation marks, provides learning information about errors, corrected information according to the frequency and intensity of errors, and provides usage examples of words. .

In today's society, many word writers, such as computer-based text editing, desktop publishing (DTP), and electronic forms (CTS), have reduced the time and effort required to create and edit documents, resulting in improved document quality.

However, the retirement of written documents is still dependent on manual work, and the retirement work becomes a bottleneck in the preparation of documents. Misspelled or mistyped materials also make it difficult to retrieve, archive, and process documents. Therefore, the economic and social utility and necessity of automatic retirement systems are increasing day by day.

In general, the spell checking method is a method that checks only the target morpheme by analyzing the sentence to be purified as a morpheme, but the method of checking the spelling only by the morpheme analysis cannot consider a word composed of multiple words as the target of the spell check. There is this.

On the other hand, the prior art related to the spell check is a "spell check method" published by the Republic of Korea Patent Office Publication No. 0048708 dated December 10, 1997 and the Republic of Korea Patent Office filed April 25, 2005 Disclosed is a system for providing an error suspected phrase discrimination and correction information and its implementation method disclosed in Korean Patent No. 01111920.

Looking in detail with respect to the disclosed prior art, the "spell check method" overcomes the difficulty of accurately performing the spelling when the conventional Hangul and foreign languages are mixed, or parentheses are used, the self-reliance during the investigation attached to the independence morpheme A method that examines surveys whose form changes depending on the presence or absence of morphemes. According to this prior art, survey correction can be made according to independent morphemes, but only the spelling of surveys attached to independent morphemes is verified. There is a problem in that compound nouns cannot be spelled out.

Next, FIG. 1 is a block diagram showing a configuration of a conventional error heartbeat determination and correction information providing system and an implementation method thereof.

Referring to FIG. 1, a program module for detecting a suspicious word that detects a word in which a spelling error exists and a word inappropriately used in a context, and providing appropriate correction information, and can be called and used by various applications such as a word processor. In the present invention, the candidate phrase is extracted from the input context text, the candidate phrase is searched in the error information database (2), and when there is information indicating an error phrase, the phrase is an error suspect phrase. An example extraction module for extracting an example of an appropriate article from an error suspect word determination module 1 for determining and a word example database 4 for storing an example of a sentence in which the words stored in the error information database 2 are correctly used ( 3) and a correction word database (6) storing the correction word and a usage example of the sentence using the correction word and the correction word word. While there is an advantage in that it consists of the correction word extraction module (5) which obtains the correction clause in the database (6), while it has the advantage of discriminating and correcting the error clause, the proper substitution for the front and rear words in the case of a semantic grammatical error. There is a problem that can not be presented.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and provides a Korean spelling checker and a method of checking a substitute word between nouns by applying a general rule to a compound noun and applying a space between nouns. There is this.

Another object of the present invention is to provide a Korean spell checker and a method for recognizing an idiomatic part-of-speech consisting of multiple words when checking a spell.

In addition, the present invention provides a Korean spelling checker and method for generating and presenting a substitute word based on a criterion before and after a target word among pre-established correction rules when there is a semantic error in the flow of context. There is another purpose.

In order to achieve the above object, the present invention provides a spelling checker for Korean that provides an alternative word corrected by grammatically dividing the input word into morphemes and storing the target word in a buffer. And a general dictionary analyzer for storing a corrected substitute word in a buffer when there is an error, and storing an identification number if there is a possibility of a stylistic error in comparison with a general dictionary consisting of a terminology dictionary; A compound noun analyzer for storing a space to be spaced between words if the target word is not matched with the rule, compared to a compound noun constraint rule dictionary composed of rules related to compound nouns; A multiple word analyzer that recognizes the majority word as one morpheme if the target word matches the starting phrase of the multiple word dictionary; If an identification number is stored in the target word, a rule list is generated by searching the correction rule indexed from the semantic correction rule rule dictionary, and one of the rule lists is selected based on the front and rear words of the target word. And a semantic parser for generating and storing substitute words.

In addition, the Korean spelling checking method according to the present invention includes a text input step of inputting text to refine the error; A word extraction step of storing a target word obtained by separating the text into a word in a buffer; After analyzing the target phrase in morpheme, compared with the general dictionary consisting of a part-of-speech dictionary, a glossary dictionary, a parent research dictionary, and a terminology dictionary, if there is an error, the corrected substitute word is stored, General dictionary analysis step of storing the identification number; A compound noun analysis step for storing a space to be spaced between words if the target word does not match a compound noun constraint rule dictionary; A multiple word analysis step of recognizing the majority word as one morpheme if the target word matches the starting phrase of the multiple word dictionary; Generate a rule list by retrieving the correction rule indexing the target word in the semantic correction rule dictionary, and selecting a rule from the list of rules based on the front and rear words of the target word to generate a substitute word. A semantic sentence analysis step of storing; If there is no morpheme of the target phrase, a morpheme discrimination step is passed to a correction routine for applying spacing and phoneme substitution, and if there is a morpheme, the process proceeds to a substitute word discrimination step; A substitute word discrimination step, if there is no substitute word for the target word, go to a buffer word clause determination step, and go to a display step if there is a substitute word; A display step of outputting on the screen together with help if there is a substitute word of the target word; And a buffer word determining step of converting the target word to the next word if there is a next word that comes after the target word stored in a buffer, and ending a spell check if there is no stored next word.

Here, in the general dictionary analysis step, it is preferable that the abbreviated target phrase is solved by a morpheme by decompressing the program code.

In the multi-word analysis step, the multi-word phrase consisting of the number of words input by the user is preferably searched in the multi-word dictionary.

The semantic sentence analysis step may include applying a semantic correction rule dictionary in which the target word is converted into various parts of speech and stored in a stylistic rule.

Hereinafter, with reference to the accompanying drawings will be described the present invention.

2 is a block diagram of a Korean spelling checker according to a preferred embodiment of the present invention.

As shown in FIG. 2, the general dictionary analyzer 10, the compound noun analyzer 20, the multiple word analyzer 30, and the semantic style analyzer 40 are included.

In general, since text to be analyzed is input as a signal such as a tap, enter, or space bar corresponding to an input process, and it is already known to generate a target word by dividing the word based on a space, a number, parentheses, etc. The description is omitted.

The general dictionary analyzer 10 stores the target word in which the input text is morphologically analyzed in a buffer, and refers to the general dictionary 12 such as a part-of-speech dictionary, a misuse dictionary, a ending dictionary, and a terminology dictionary. As a result, the target word is checked for an error, and if there is an error, a general word is generated and stored with reference to the general dictionary 12.

After storing a specific number for classification according to the parts of speech in the target word, the parts of speech are grouped by property to store a group number. The group number is for use in the compound noun analyzer 20 to be described later. An embodiment of a specific grouping is classified into a 'Hada noun' group and an 'animal' group, and the like is stored in each of the compound noun constraint rules to be described later. It can be applied in advance. This grouping is not limited to the above embodiments, and various embodiments are possible, such as using a specific dictionary.

Also, in order to determine whether the semantic analyzer 40 to be described later is applied, the semantic correction rule checks whether a target word is searched by an index, and if there is an error in the semantic sentence, an identification number is added to the target word and stored.

The compound noun analyzer 20 stores a space for spacing between nouns if the target phrase does not have a compound noun through a compound noun constraint rule dictionary 22 that determines the possibility of a compound noun according to the morphological analysis result. In the display step S90 to be described later, a substitute word containing a space is generated and stored. If the target phrase is a correct compound noun by the compound noun constraint rule dictionary 22, the next step is performed without spacing between nouns.

For example, a word such as 'processing' can be a compound noun with 'processing (Hada noun) + process'. However, in the case of 'elephant process', the rule is restricted to 'no elephant (animal) + process' so that it can't be a compound noun. In the display step S90 to be described later, a substitute word is generated.

The multiple word analyzer 30 processes a foreign person's name, idiomatic notation, a spacing error, or other difficult multiple word phrases as one test object. In order to discriminate such idiomatic majority words, the multiple word dictionary 32 is used to search whether the target word is the start word of the multiple word dictionary 32. At this time, if a majority word is found, the majority word is stored in a buffer as a substitute word, the display proceeds to step S90, the substitute word and the help are output, and if there is no substitute word related to the semantic sentence analysis step S60. Move.

When the analysis of each word through the multiple word analyzer 30 is not the ones that can not be analyzed when several words are gathered to have a single meaning, the accuracy of the morpheme analysis can be increased by the analysis by the multiple words. have. For example, in the case of 'Osama bin Laden', the multi-word dictionary 32 is searched through the multi-word dictionary 32, which indexes the beginning phrase of 'Osama', and corresponds to the corresponding 'Osama bin'. If it finds a multiword clause, it is treated as a morpheme and does not indicate an error when checking spelling. In addition, as an example of the correction through the multiple word misuse, the multi-word dictionary 32 is searched with 'Michael', which is the beginning word of 'Michael Jordan', and a multi-word word related to 'Michael Jordan' is generated as a substitute for the buffer. After storing in the display, go to the display step (S90), and replaces the word 'Michael Jordan' with the related help. As such, the multiword analyzer 30 can correct the error word, so that the reproducibility of the spell check (probability of finding an error word in a given corpus / example: there are 10 errors in a sentence. If the dog is caught, the recall rate is 90%).

The semantic style analyzer 40 searches the semantic correction rule dictionary 42 to create a rule list, and selects a rule that best matches the relationship between the front and rear words of the target word based on the target word. Check and correct natural or literal errors. A more detailed description thereof will be described with reference to FIG. 3.

3 is a block diagram of a semantic analyzer 40 according to a preferred embodiment of the present invention.

As shown in FIG. 3, in the present semantic stylus analyzer 40, if there is a possibility of meaning and stylistic error, the presence or absence of an identification number stored in the target word is determined. The rule is searched for in the semantic correction rule dictionary to create a rule list to be applied. In this case, if the identification number is not present, the semantic sentence analyzer 40 is not executed, thus saving time of the spell check.

In addition, it is possible to apply the partial sentence analysis technique to the created rule list, and the partial sentence analysis technique uses the target phrase as an index even if an adverb exists before or after the target word or if the target word is converted to a different part of speech. If there are rules, it means creating a list of rules to extend the rules.

In the rule list created by the rule expansion, the front and rear words of the target word are finally compared to determine the most relevant rule that can correct the error of the target word that can be applied, select one rule, and then Generate a substitute.

Referring to a specific example through the general dictionary analyzer 10 and the semantic style analyzer 40, in the case of 'the ugly alley', when the morpheme analysis is performed in the general dictionary analyzer 10, the morphological analysis is performed. And then decompose it into 'Eoha + b' and assign a specific number to indicate that it is a verb because the part of 'Eajiha' is a verb.

In addition, if a rule related to 'Ughha' is searched in the index of the semantic correction rule dictionary 42, and there is a rule related to 'Ughha', there is a possibility of semantic writing. The stylistic analyzer 40 extracts a rule having an index of 'Yuhha' with the identification number to create a rule list. In the rule list, the 'alley road' following the 'shrugged alley', which is the subject of inspection, matches the rule of replacing 'shrugged alley' with 'shrugged alley', and thus generates and stores a substitute word 'shrugged alley'.

4 is a flowchart illustrating a method of checking spelling of Korean according to a preferred embodiment of the present invention. The configuration of the present invention with reference to Figure 4 is a text input step (S10), word extraction step (S20), general dictionary analysis step (S30), compound noun analysis step (S40), multiword analysis step (S50), semantic style Analysis step (S60), morpheme determination step (S70), alternative word determination step (S80) and display step (S90) and buffer word clause determination step (S100).

In the text input step S10, a plurality of texts are input by a signal of a keyboard such as a space bar, a tap, or an enter.

The word extracting step (S20) extracts the word by separating the input word using parentheses, spaces, special characters, and numbers as delimiters, and stores several target words in a buffer.

The general dictionary analysis step (S30), the morphological analysis of the target word, compared to the above-described general dictionary 12, if there is a substitute word correcting the error in the target word is stored in the buffer, and with the meaning in the target word Determine the possibility of stylistic errors.

In the morphological analysis, the target phrases are classified for each part of speech, and a specific number is stored in the target phrase, and a group number is assigned for each specific group of parts of speech so that it can be used in a compound noun analysis step S40 to be described later. The possibility of an error in the semantic stylistic in the target word is searched in the semantic correction rule dictionary 42 and if there is an index associated with it, the identification number is stored in the target word, and if there is an identification number, the semantic sentence analysis step (S60) will be described later. In the case where there is no identification number, it is possible to reduce the spell check time by omitting the step of the semantic character analysis step (S60).

The compound noun analysis step (S40), if there is a rule corresponding to the compound noun constraint rule dictionary 22 goes through the step of searching for the compound noun constraint rule (S42), if the compound noun according to the constraint rules The spaces to be spaced between nouns are stored to generate and output alternate words with spaces between nouns in a display step S80 to be described later. Herein, the compound noun constraint 22 is defined in the following specific rules, and various other rules of compound nouns are possible.

1) (body, exercise, illness) + 'no factory' (ex) basketball factory-> basketball factory

2) (Animals, Plants) + 'No Garden' (ex) Tiger Garden-> Tiger Garden

3) Company + (no vehicles, machines, parts of the body, sports, illness, food, clothes)

ex) Company plane-> company plane

The multiple word analysis step (S50) compares the multiple word misuse word and the target word based on the constructed multiple word dictionary (32), and whether the target word is the beginning word of the multiple word misuse word of the multiple word dictionary (32). The search step (S52) is determined to determine. Here, if the start phrase and the target phrase of the majority word misuse word coincide, a pointer to the majority word dictionary 32 is recorded in the target word.

Accordingly, the multi-word phrase is recognized as one test subject through the multi-word dictionary 32 during the spell check. In this way, if a substitute word correcting an error of the semantic sentence is generated, the process proceeds to the display step (S90). .

In addition, the multi-word analysis step (S50) receives information from the multi-word dictionary 32 as many times as the user inputs the number of words to check the target word, the multi-word dictionary consisting of the number of words multi-word dictionary 32 In the comparative search, when the start phrase and the target phrase of the multiple word dictionary 32 coincide with each other, the multiple words are integrally reflected as one test object in the morpheme analysis.

The semantic sentence analysis step (S60), if there is an identification number likely to be a semantic sentence error in the general dictionary analysis step (S30), all the rules that are searched in the index from the semantic correction rule dictionary index to the target phrase analysis Gather, step S62 of creating a rule list. Here, by referring to the front and rear words of the target word in the rule list, through the step (S64) of determining whether a replacement word is generated by selecting the applicable correction rule, the display moves to the display step (S80). If there is no substitute word, go to the buffer word clause determination step (S100).

In this process, partial sentence analysis techniques are applied according to patterns, which are used to determine semantic relations in a broader range. The partial sentence analysis technique is based on the results of morphological analysis of each word, and checks whether or not it can be used as a predicate, an object, an adverb, a tubular adjective, and the like, and expands accordingly to further generate a list of applicable rules in an internal routine.

For example, if a part of speech such as an adjective or an adverb such as 'tangled blood', 'tangled blood', or 'tangled blood' is changed, a rule that is corrected to 'tangky-> tangling' is applied. Will generate

The morpheme determination step (S70), if there is no morpheme of the target word, go through a correction routine step (S72) of analyzing whether the stylistic is appropriate through a test such as spacing or phoneme replacement. The phoneme substitution is to replace the consonants or vowels when a word that does not fit the phrase is found to generate a substitute word by searching for a suitable word in meaning. At this time, after replacing various consonants and vowels and applying them to the target word, if there is no substitute word with no semantic error, it is indicated as an error word, and then goes to the buffer word determination step (S100) to check the next word stored in the buffer. If there is a substitute word, it is stored in the buffer and the display moves to the display step S90 to output the substitute word and the related help.

The alternative word determination step (S80) is displayed when there is a substitute word stored in various steps such as the alternative word or the compound noun analysis step (S40), the analysis correction routine step (S72), etc. according to the general dictionary search step (S30) described above. In step S90, if there is no substitute word when the target word is correct without error, then the next word to be checked is moved to the buffer word determination step (S100).

In the displaying step (S90), a substitute word and a help related thereto are generated and displayed on the screen, and if the strength of the stylistic error of the substitute word is weak, the user may select the substitute word and the help word so that the user can select it. If the strength of the error is strong, it is strongly suggested to replace the substitute word and the help to inform the user about the strength of the stylistic error.

Hereinafter, the operation flow of the present invention through the table will be described in detail.

Stage 1 The thread I used yesterday was entangled. Tier 2 yesterday Used Thread Smash Entangled Tier 3 yesterday Using N Thread end Smash Intertwined Was All 4 steps a little Intertwined Was All 5 steps Tangled All

Table 1 shows the Korean spelling check method according to a preferred embodiment of the present invention in five steps.

In the text input step (S10) of the first step, as shown in Table 1, the sentence "the thread used yesterday was entangled."

In the word extraction step (S20) of the second step, it can be separated and extracted as 'yesterday', 'used', 'thread', 'splitting', 'entangled'.

In step 3, the general dictionary analysis step (S30) generates a substitute word 'a little' and stores it in a buffer, and then displays the substitute word and related help in the display step (S90). In the buffer word determination step (S100), if a word remains in the buffer, the processor returns to the word extraction step (S20) and checks the next word, “tangled,” as the target word.

Next, in the general dictionary analysis step (S30) of step 4, the rule of `` tangled '' in the morphological analysis, and `` tangled '' in the semantic grammar correction rule dictionary (42) to check whether there is a stylistic error possibility If there is a rule, the identification number is stored.

Finally, in step 5, a list of rules is created in the semantic correction rule dictionary 42, which indexes 'entangled' in the semantic sentence analysis step (S60), and then the appropriate rule is selected by searching the words before and after the 'entangled'. do.

As a specific example of the rule list creation, the following three rules can be searched in the semantic style correction rule dictionary 42 as a rule applicable to the word “entangled”, and can be created as a list in Table 2.

Rule 1 Entangle -> Tangle Rule 2 Threads / lines / thinks are entangled. -> The thread / line thought is tangled. Rule three Tofu / blood / oil are entangled. -> Tofu / blood / oil tangle

Table 2 shows a list of rules indexed 'entangled'.

Referring to Table 2, when a total of three rules were applied to the target word, the 'thread' of 'thread / line / think is entangled-> thread / line / think is entangled' in rule 2 In conformity with, the applicable constraints can be applied to the target clause.

In the actual sentence, 'a little' is an adverb, which is included in the extended range of partial sentence analysis technology that can apply the index-related part even if the part of the index is changed or the adverb is entered. Can be. Since we found an applicable rule, we stored a substitute in the buffer that the thread we used yesterday is a bit tangled.

In the display step (S90) to display a substitute word stored in the buffer, the word used yesterday a little tangled, and the corresponding help on the screen, and in the buffer word determination step (S100) to check whether there are remaining words in the buffer If there are no remaining words, you're done with a spell check.

According to the present invention having the configuration described above, a word having a form of a compound noun in an improper sense has an effect of suggesting a substitute word for spacing between nouns.

In addition, the present invention has the effect of recognizing a conventional part-of-speech consisting of a large number of words at the time of spell check.

In addition, the present invention has the effect of presenting a substitute word suitable for a word before and after a target word having a semantically erroneous word in the flow of context.

Claims (5)

In the Korean spell checker that provides error-corrected alternatives, After dividing the input word into morphemes, the target word is stored in a buffer, and then compared with the general dictionary (12) consisting of a part-of-speech dictionary, a glossary dictionary, a ending search dictionary and a terminology dictionary, and if there is an error, A general dictionary analyzer 10 for storing the identification number if the target phrase has a stylistic error; A compound noun analyzer (20) for storing positions to be spaced between words if the target phrase does not match the rule, compared with the compound noun constraint rule dictionary (22) consisting of rules related to compound nouns; A multiple word analyzer (30) for recognizing the majority word as one morpheme if the target word matches the starting phrase of the multiple word dictionary (32); And If the identification number is stored in the target phrase, the correction rule is searched in the semantic style correction rule dictionary 42 to generate a rule list, and one of the rule list is based on the front and rear words of the target phrase. Korean spelling checker comprising a; semantic parser 40 for generating and storing alternative words by selecting a rule. In the Korean spelling checker providing a corrected alternative, A text input step (S10) in which text for refining an error is input; A word extracting step (S20) of storing a target word obtained by separating the text into words in a buffer; After analyzing the target phrase in morpheme, compared with the general dictionary consisting of a part-of-speech dictionary, a glossary dictionary, a parent research dictionary, and a terminology dictionary, if there is an error, the corrected substitute word is stored, and if there is a possibility of stylistic error General dictionary analysis step of storing the identification number (S30); If the target phrase does not match the compound noun constraint rule dictionary (22), compound noun analysis step of storing a space to be spaced between words (S40); A multiple word analysis step (S50) of recognizing the majority word as one morpheme if the target word is matched with the start phrase of the multiple word dictionary 32; A rule list is generated by searching a rule indexing the target word in the semantic correction rule dictionary 42, and a substitute word is selected by selecting one of the rule lists based on the front and rear words of the target word. Generating and storing semantic analysis step (S60); If there is no morpheme of the target phrase, the morpheme determination step (S70) is passed to the correction routine step (S72) for applying spacing and phoneme substitution, and if the morpheme is present, the substitute word determination step (S80); If there is no substitute word for the target word, go to the buffer word determination step (S100), and if there is a substitute word, go to the display step (S90); If there is a substitute word for the target word, a display step of outputting the screen with help (S90); And Korean word spelling comprising a buffer word determination step of converting the target word to the next word if there is a next word that comes after the target word stored in a buffer, and ending the spell check if there is no stored next word; method of inspection. According to claim 2, The general dictionary analysis step (S30), Korean spelling check method characterized in that the abbreviated target phrases are analyzed by morphemes decoded by the program code. According to claim 2, The multiple word analysis step (S50), Korean spelling check method characterized in that for searching the majority word consisting of the number of words input by the user in the multiple word dictionary (32). The method of any one of claims 2 to 4, wherein the semantic style analysis step (S60), The method of spelling Korean, characterized in that the target word phrase is converted into a variety of parts of speech and the stylistic correction rule dictionary (42) in which the stylistic rules are stored.

Images