KR20210073068A - Purification sentence generating apparatus for interactive question-anwering service and method thereof - Google Patents

Abstract

The present invention relates to an apparatus and method for generating a purified sentence for an interactive question-answering service, which comprises: a restoration processing module for receiving data for a text converted from voice, which are inputted from the outside through speech to text (STT) technology applied to the interactive question-answering service and restoring an error occurring due to features of the data for a text converted from voice into a correct sentence by using natural language processing technology; and a purification processing module for receiving data for sentences restored by the restoration processing module and changing the restored sentences based thereon into purified expression which is suitable for a situation and a context and can be easily understood by a user to generate a new purified sentence, thereby keeping quality of the interactive question-answering service consistent.

Description

PURIFICATION SENTENCE GENERATING APPARATUS FOR INTERACTIVE QUESTION-ANWERING SERVICE AND METHOD THEREOF

The present invention relates to an apparatus and method for generating a purified sentence for an interactive question and answer service, and more particularly, to a speech to text inputted from the outside through a speech to text (STT) technology applied to an interactive response service. Disclosed are an apparatus and method for generating a purified sentence for an interactive Q&A service that automatically generates data into a purified sentence suitable for the response service.

In general, a voice system utilized as a customer service uses a speech to text (STT) technology that converts recorded voice data into text. The voice-converted text data extracted in this way is being applied to various fields through analysis techniques such as natural language processing technology and text mining.

In most prior art, interactive data is limited to lexical processing technology to be applied to a chatbot system. In addition, there are many types of extracting keywords, subjects, emotions, etc. from text and providing them to counselors or making a database (DB).

However, when provided in this form, the counselor has to use the information to make an appropriate sentence, so it depends on the individual's ability, and this has a problem of causing a deviation in the quality of the Q&A service.

On the other hand, the existing chatbot system uses a knowledge base built on the basis of speech-to-text data or conversational data through a large amount of speech recognition (STT) technology to generate answers to the user's questions. Therefore, the quality of the chatbot is determined by the knowledge base.

However, building a knowledge base in the existing chatbot system is a costly operation, and modifications and additions also incur considerable costs. Therefore, most chatbot systems have a cumbersome problem of additionally performing post-processing to generate natural sentences.

Domestic Registered Patent No. 10-1858206 (2018.05.15. Announcement)

The present invention has been devised to solve the above-described problems, and an object of the present invention is to generate and provide a sentence so that an agent can directly use it when talking with a customer, so that the customer response service quality can be consistently maintained. An object of the present invention is to provide an apparatus and method for generating a purified sentence for a question and answer service.

Another object of the present invention is to restore lexical errors and omissions caused by the characteristics of speech-to-text data through speech recognition (STT) technology applied to an interactive question and answer service to the main sentence, and convert the restored data into sentences of purified expression. An object of the present invention is to provide an apparatus and method for generating a purified sentence for an interactive question and answer service, which can be automatically generated and directly provided to an agent to consistently maintain the quality of the interactive question and answer service.

In order to achieve the above object, a first aspect of the present invention uses a natural language processing technology by receiving voice-converted text data input from the outside through a speech to text (STT) technology applied to an interactive question-and-answer service. a restoration processing module for restoring an error occurring due to the characteristics of the inputted voice-converted text data to the front door; and reconstructed sentence data provided from the restoration processing module, and based on this, the restored sentence is converted into a purified expression in a form suitable for the situation and context and easy for the user to understand, thereby generating a new purified sentence. An object of the present invention is to provide a purified sentence generating apparatus for an interactive question and answer service including a processing module.

Here, the restoration processing module includes: a language model DB for storing and managing a preset Korean database corresponding to each English word; Based on natural language processing technology, at least one of preprocessing of space and misspelling errors, sentence separation, or cross-projection processing is based on changing the part written in English words in the input speech conversion text data to Korean with reference to the language model DB a preprocessor that does the work; abbreviation DB for storing and managing preset original strings corresponding to each abbreviation word into a database (DB); After receiving the pre-processed sentence data from the pre-processing unit, referring to the abbreviation DB based on this, and searching for the original string corresponding to the abbreviated word in the pre-processed sentence data, the abbreviated word is the searched original string Abbreviation restoration unit to restore to; a synonym DB for storing and managing one preset representative word corresponding to each synonym bundle into a database (DB); and when a plurality of synonyms exist in the sentence data restored to the original character string by referring to the synonym DB based on the received sentence data restored to the original string from the abbreviation restoration unit, one preset corresponding to the plurality of synonyms After searching for a representative word of , it is preferable to include a synonym processing unit for collectively replacing the plurality of synonyms with the searched representative word.

Preferably, the purification processing module comprises: a Chinese character DB for storing and managing preset pure Korean words corresponding to each Chinese character word into a database (DB); After receiving the restored-processed sentence data from the restoration processing module and referring to the Chinese-character DB based on it, searching for a pure Korean word corresponding to a Chinese-character word consisting of a noun or a compound noun in the restored-processed sentence data, the corresponding Chinese character a noun refining unit for refining a word into the searched pure Korean word; a narrative dictionary DB that stores and manages a database (DB) of pre-set narrative refined words corresponding to each narrative high level including verbs; After receiving the sentence data of the noun refinement process from the noun refinement unit, and referring to the descriptive dictionary DB based on this, the descriptive refinement word corresponding to the descriptive exaltation including the verb is retrieved from the noun refinement processed sentence data, and then the corresponding verb is retrieved. It may include a narrative refinement unit that changes the included narrative exclamation word to the searched narrative refined word to finally generate a new purified sentence.

Preferably, the descriptive refining unit may extract morphemes and parts of speech of a sentence through morpheme analysis for a part composed of a noun and a ending or a verb and a ending.

Preferably, the purification processing module may further include a negative word DB for storing and managing preset purified words corresponding to each negative word having a negative meaning in a database (DB).

Preferably, the noun refinement unit searches for a purified word corresponding to the negative tendency word in the restored sentence data or the noun refinement processed sentence data with reference to the negative word DB, and then selects the word of the negative tendency. A purification process may be performed using the searched purified word.

Preferably, the descriptive refinement unit searches for a refined word corresponding to a negative tendency word in the noun refinement-processed sentence data or the narrative-type refinement-processed sentence data with reference to the negative word DB, and then the word of the negative tendency may be purified as the searched purified word.

A second aspect of the present invention is a method for generating a purified sentence using a restoration processing module and a purification processing module in an interactive Q&A service, (a) applied to the interactive Q&A service through the restoration processing module Using a natural language processing technology based on speech-to-text data input from the outside by speech recognition (Speech to Text, STT) technology, restoring an error occurring due to the characteristics of the inputted speech-to-text data to the front door; and (b) receiving the sentence data restored in step (a) through the purification processing module, and based on this, the restored sentence is converted into a purified expression in a form suitable for the situation and context and easy for the user to understand. An object of the present invention is to provide a method for generating a purified sentence for an interactive response service, which includes generating a new purified sentence by processing the change.

Here, in the step (a), (a-1) a separate language model DB based on natural language processing technology through a preprocessor provided in the restoration processing module is referred to, and the inputted speech converted text data is converted into English words. performing at least one of pre-processing of spaces and misspellings, sentence separation, and cross-projection processing on the basis of changing the marked part into Korean; (a-2) Referring to a separate abbreviation DB on the basis of the sentence data pre-processed in step (a-1) through the abbreviation restoration unit provided in the restoration processing module, the abbreviation from the pre-processed sentence data after searching for the original string corresponding to the word, restoring the corresponding abbreviated word to the searched original string; and (a-3) a sentence restored to the original character string by referring to a separate synonym DB based on the sentence data restored to the original character string in step (a-2) through the synonym processing unit provided in the restoration processing module When there are a plurality of synonyms in the data, it is preferable that the method includes searching for one preset representative word corresponding to the plurality of synonyms, and then collectively replacing the plurality of synonyms with the searched representative word.

Preferably, before the step (a-1), it may further include the step of storing and managing a preset Korean database corresponding to each English word through the restoration processing module and storing and managing it in the language model DB. have.

Preferably, before the step (a-2), the method may further include a step of converting a preset original string corresponding to each abbreviation word through the restoration processing module into a database (DB) and storing and managing the abbreviation DB in the abbreviation DB. have.

Preferably, before the step (a-3), a database (DB) of one preset representative word corresponding to each synonym bundle through the restoration processing module, further comprising the step of storing and managing the synonym DB can do.

Preferably, in step (b), (b-1) a separate Chinese character DB is referred to based on the sentence data restored in step (a) through a noun refinement unit provided in the purification processing module, and the restoration is performed. searching for a pure Korean word corresponding to a Chinese character word composed of a noun or a compound noun in the processed sentence data, and then refining the noun into the searched pure Korean word; and (b-2) referring to a separate narrative dictionary DB based on the sentence data processed by the noun purification in step (b-1) through the narrative purification unit provided in the purification processing module, and the noun purification processing sentence data After searching for a descriptive refined word corresponding to a descriptive adjective including a verb, changing the descriptive adjective including the corresponding verb to the searched descriptive refined word, finally generating a new purified sentence.

Preferably, before the step (b-1), it may further include the step of converting a preset pure Korean word corresponding to each Chinese character word through the purification processing module into a database (DB), and storing and managing it in the Chinese character DB. can

Preferably, before the step (b-2), the step of storing and managing the preset narrative refined words corresponding to each narrative adjective including the verb through the purification processing module into a database (DB) and storing and managing the narrative dictionary DB may include more.

Preferably, in step (b-2), the narrative refiner may extract morphemes and parts of speech of a sentence through morpheme analysis for a part composed of a noun and a ending or a verb and a ending.

Preferably, before or after step (b-1), by referring to a separate negative word DB through the noun-type refinement unit provided in the refinement processing module, negation in the restored-processed sentence data or the noun-refined-processed sentence data The method may further include a first step of purifying the word of the negative tendency into the searched purified word after searching for a refined word corresponding to the word of the propensity.

Preferably, before the first step, the step of converting a preset purified word corresponding to each negative word having a negative meaning through the purification processing module into a database (DB), storing and managing the negative word DB may further include.

Preferably, before or after the step (b-2), by referring to a separate negative word DB through the narrative refinement unit provided in the refinement processing module, the noun refinement-processed sentence data or the narrative-type refinement-processed sentence data The method may further include a second step of purifying the word of the negative tendency into the searched purified word after searching for a refined word corresponding to the negative tendency word.

Preferably, before the second step, the step of converting a preset purified word corresponding to each negative word having a negative meaning through the purification processing module into a database (DB), storing and managing the negative word DB may further include.

A third aspect of the present invention provides a computer-readable recording medium in which a program capable of executing the above-described method for generating a purified sentence for an interactive question-and-answer service is recorded.

The method for generating a purified sentence for an interactive question and answer service according to the present invention may be implemented as a computer-readable code on a computer-readable recording medium. The computer-readable recording medium includes all kinds of recording devices in which data readable by a computer system is stored.

For example, computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, hard disk, floppy disk, removable storage device, and non-volatile memory (Flash Memory). , and optical data storage devices.

According to the apparatus and method for generating a purified sentence for an interactive question and answer service of the present invention as described above, the lexical expression generated by the characteristics of the speech conversion text data through the speech recognition (STT) technology applied to the interactive question and answer service is used. There is an advantage in that the quality of the interactive Q&A service can be consistently maintained by restoring errors and omissions to the main sentence and automatically generating the restored data as sentences of purified expression and directly providing it to the agent.

In addition, according to the present invention, the data can be intuitively utilized by restoring speech-converted text data through speech recognition (STT) technology for an interactive question-and-answer service and generating purified sentences. There is an advantage that can reduce the quality deviation that can be.

In addition, according to the present invention, it is possible to increase customer service satisfaction by reducing problems caused by the use of difficult terms and use of tones that do not fit the situation, and there is an advantage that the generated purified sentences can be utilized in the chatbot system.

In addition, according to the present invention, if the present invention is used when constructing a knowledge base for sentences in a purified form in the existing chatbot system, additional knowledge construction and post-processing costs can be reduced, and the quality of sentences generated by the chatbot can be improved. There are advantages that can be

1 is an overall block diagram illustrating an apparatus for generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention.
2 is an overall flowchart illustrating a method of generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention.
3 is a flowchart for explaining in detail the step of restoring and processing an error occurring due to the characteristics of the inputted speech text data to the front door in the method for generating a purified sentence for an interactive question and answer service according to an embodiment of the present invention. .
4 is a flowchart for specifically explaining a step of generating a new purified sentence by changing a purified expression to a purified expression in the method for generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention.

The above-described objects, features and advantages will be described below in detail with reference to the accompanying drawings, and accordingly, those of ordinary skill in the art to which the present invention pertains will be able to easily implement the technical idea of the present invention. In describing the present invention, if it is determined that a detailed description of a known technology related to the present invention may unnecessarily obscure the gist of the present invention, the detailed description will be omitted.

Terms including an ordinal number, such as first, second, etc., may be used to describe various elements, but the elements are not limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, a first component may be referred to as a second component, and similarly, a second component may also be referred to as a first component. The terms used in the present application are only used to describe specific embodiments, and are not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise.

The terms used in the present invention have been selected as currently widely used general terms as possible while considering the functions in the present invention, which may vary depending on the intention or precedent of a person skilled in the art, the emergence of new technology, and the like. In addition, in a specific case, there is a term arbitrarily selected by the applicant, and in this case, the meaning will be described in detail in the description of the corresponding invention. Therefore, the term used in the present invention should be defined based on the meaning of the term and the overall content of the present invention, rather than the name of a simple term.

When a part "includes" a certain element throughout the specification, this means that other elements may be further included, rather than excluding other elements, unless otherwise stated. In addition, terms such as "...unit" and "module" described in the specification mean a unit that processes at least one function or operation, which may be implemented as hardware or software, or a combination of hardware and software. .

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, the embodiments of the present invention illustrated below may be modified in various other forms, and the scope of the present invention is not limited to the embodiments described below. The embodiments of the present invention are provided to more completely explain the present invention to those of ordinary skill in the art.

Each block in the accompanying block diagram and combinations of steps in the flowchart may be executed by computer program instructions (execution engine), which computer program instructions may be executed by a processor of a general-purpose computer, special-purpose computer, or other programmable data processing equipment. It may be mounted so that the instructions, which are executed by the processor of a computer or other programmable data processing equipment, create means for performing the functions described in each block of the block diagram or in each step of the flowchart. These computer program instructions may also be stored in a computer usable or computer readable memory which may direct a computer or other programmable data processing equipment to implement a function in a particular manner, and thus It is also possible for the instructions stored in the block diagram to produce an article of manufacture containing instruction means for performing the functions described in each block of the block diagram or each step of the flowchart.

And, since the computer program instructions may be mounted on a computer or other programmable data processing equipment, a series of operating steps are performed on the computer or other programmable data processing equipment to create a computer-executed process to create a computer or other program It is also possible that instructions for performing the possible data processing equipment provide steps for carrying out the functions described in each block of the block diagram and each step of the flowchart.

In addition, each block or step may represent a module, segment, or portion of code comprising one or more executable instructions for executing specified logical functions, and in some alternative embodiments the blocks or steps referred to in the block or steps. It should be noted that it is also possible for functions to occur out of sequence. For example, it is possible that two blocks or steps shown one after another may be performed substantially simultaneously, and also the blocks or steps may be performed in the reverse order of the corresponding functions, if necessary.

1 is an overall block diagram illustrating an apparatus for generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention.

Referring to FIG. 1 , an apparatus for generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention largely includes a restoration processing module 100 and a purification processing module 200 . Meanwhile, since the components shown in FIG. 1 are not essential, the purified sentence generating apparatus for an interactive question and answer service according to an embodiment of the present invention may have more components or fewer components. .

Hereinafter, components of an apparatus for generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention will be described in detail as follows.

The restoration processing module 100 receives the voice-converted text data input from the outside through Speech to Text (STT) technology applied to the interactive question-and-answer service, and converts the voice input from the outside using natural language processing technology. It performs the function of restoring and processing errors that occur due to the characteristics of text data to the main text.

As shown in FIG. 1, the restoration processing module 100 is largely a language model DB 110, a preprocessor 120, an abbreviation DB 130, an abbreviation restoration unit 140, a synonym DB 150, and The synonym processing unit 160 and the like are included.

Here, the language model DB 110 performs a function of storing and managing preset Korean corresponding to each English word into a database (DB).

The preprocessor 120 is based on changing the part written in English words into Korean in the voice-converted text data input from the outside with reference to the language model DB 110 based on natural language processing technology, and spacing and typo errors, sentence separation , and/or performs a function of performing at least one pre-processing of liver projection processing.

The abbreviation DB 130 performs a function of storing and managing a database (DB) of a preset original character string corresponding to each abbreviation word.

The abbreviation restoration unit 140 receives the pre-processed sentence data from the pre-processing unit 120 and based on this, refers to the abbreviation DB 130 and searches for the original string corresponding to the abbreviated word in the pre-processed sentence data. After that, a function of restoring the corresponding abbreviated word to the searched original string is performed.

The synonym DB 150 performs a function of storing and managing one preset representative word corresponding to each synonym bundle into a database (DB).

The synonym processing unit 160 receives the sentence data restored to the original character string from the abbreviation restoration unit 140 and refers to the synonym DB 150 based on the received sentence data when a plurality of synonyms exist in the sentence data restored to the original character string. After searching for one preset representative word corresponding to a plurality of synonyms, a function of collectively replacing the plurality of synonyms with the searched representative word is performed.

And, the purification processing module 200 receives the restored processed sentence data from the restoration processing module 100, and based on this, the restored processed sentence is a purified expression in a form suitable for the situation and context and easy for the user to understand. It performs the function of generating a new purified sentence by processing the change.

That is, the refinement processing module 200 serves to change the restored-processed sentence into a refined expression suitable for the situation and context and easy to understand by the user, rather than finding and correcting the error.

As shown in FIG. 1 , the refinement processing module 200 includes a Chinese character DB 210 , a noun refinement unit 220 , a narrative dictionary DB 230 , and a narrative refinement unit 240 . In addition, the purification processing module 200 applied to an embodiment of the present invention may further include a negative word DB 250 and the like. On the other hand, since the components shown in FIG. 1 are not essential, the purifying processing module 200 applied according to an embodiment of the present invention may have more or fewer components than that.

Here, the Chinese character DB 210 performs a function of storing and managing preset pure Korean words corresponding to each Chinese character word into a database DB.

The noun refining unit 220 receives the restored sentence data from the restoration processing module 100 and refers to the Chinese character DB 210 based on the received sentence data from the restoration processing module 100, and in the restored sentence data, a Chinese word word composed of a noun and/or a compound noun. After searching for a pure Korean word corresponding to , a function of refining the corresponding Chinese word to the searched pure Korean word is performed.

In addition, the noun refinement unit 220 refers to the negative word DB 250 to search for a purified word corresponding to a negative tendency word in the restored sentence data and/or the noun refinement processed sentence data, and then It is possible to perform a function of refining the word of the negative tendency into the searched purified word.

The narrative dictionary DB 230 performs a function of storing and managing a database (DB) of pre-set narrative refined words corresponding to each narrative exclamation including verbs.

The descriptive refinement unit 240 receives the noun refinement-processed sentence data from the noun refinement unit 220, and refers to the descriptive dictionary DB 230 based on this, and corresponds to the descriptive exaltation including the verb in the noun-refined-processed sentence data. After searching for a refined descriptive word to be used, a function of finally generating a new purified sentence is performed by changing a descriptive high word including the corresponding verb to the searched descriptive refined word.

In addition, the narrative refinement unit 240 may perform a function of extracting morphemes and parts of speech of a sentence through morpheme analysis for a part composed of a noun and a ending and/or a verb and a ending.

In addition, the narrative refinement unit 240 refers to the negative word DB 250 to search for the purified word corresponding to the negative tendency in the sentence data subjected to the noun purification process and/or the sentence data processed by the narrative purification process, It is possible to perform a function of purifying the word of the negative tendency into the searched purified word.

Additionally, the negative word DB 250 performs a function of storing and managing preset purified words corresponding to each negative word having a negative meaning in a database DB.

Hereinafter, a method for generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention will be described in detail.

2 is an overall flowchart illustrating a method of generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention, and FIG. 3 is a purified sentence for an interactive Q&A service according to an embodiment of the present invention. It is a flowchart for specifically explaining the step of restoring and processing an error occurring due to the characteristics of the speech converted text data input in the generating method to the front door, and FIG. 4 is a purification for an interactive question and answer service according to an embodiment of the present invention. It is a flowchart for specifically explaining the step of generating a new purified sentence by changing the sentence generation method to a purified expression.

1 to 4 , in the method for generating a purified sentence for an interactive question and answer service according to an embodiment of the present invention, first, voice recognition applied to the interactive question and answer service through the restoration processing module 100 ( Using natural language processing technology based on speech-to-text data input from the outside by Speech to Text (STT) technology, errors occurring due to the characteristics of the inputted speech-to-text data are restored to the front door and processed (S100).

As shown in FIG. 3 , the step S100 refers to a separate language model DB 110 based on natural language processing technology through the pre-processing unit 120 provided in the restoration processing module 100, and the inputted Performing at least one preprocessing task of spacing and misspelling errors, sentence separation, and/or inter-projection processing on the basis of changing the part written in English words in the speech conversion text data to Korean (S110), and a restoration processing module ( Referring to a separate abbreviation DB 130 based on the sentence data pre-processed in step S110 through the abbreviation restoring unit 140 provided in 100), corresponding to the abbreviation word in the pre-processed sentence data After searching for the original character string, restoring the corresponding abbreviated word to the searched original character string (S120) and the sentence restored to the original character string in step S120 through the synonym processing unit 160 provided in the restoration processing module 100 When a plurality of synonyms exist in the sentence data restored to the original string by referring to a separate synonym DB 150 based on the data, after searching for one preset representative word corresponding to the plurality of synonyms, the plurality It may include a step (S130) of collectively replacing the synonyms of the searched representative words.

Here, in the pre-processing operation of step S110, the operation of changing the part written in English words into Korean may restore "mobile web" to "mobile web", for example.

In addition, when the synonym processing operation of step S130 is used in various forms such as "App", "app", "application", etc., one representative word may be set and replaced with a consistent word.

On the other hand, although not shown in the drawing, before the step S110, the preset Korean corresponding to each English word is converted into a database (DB) through the restoration processing module 100, and stored and managed in the language model DB 110 may further include.

In addition, before the step S120, it may further include the step of converting a preset original character string corresponding to each abbreviation word through the restoration processing module 110 into a database DB and storing and managing the abbreviation DB 130. .

In addition, before the step S130, a database (DB) of one preset representative word corresponding to each synonym bundle through the restoration processing module 110 may further include the step of storing and managing the synonym DB 150 may be

On the other hand, the above-described steps S110 to S130 are preferably implemented sequentially, but are not limited thereto, and may be implemented by changing the order according to the needs of the applied system.

Thereafter, based on the sentence data restored in step S100 through the purification processing module 200, the restored-processed sentence is changed to a purified expression in a form suitable for the situation and context and easy for the user to understand, and a new A purified sentence is generated (S200).

As shown in FIG. 4, in step S200, a separate Chinese character DB 210 is created based on the sentence data restored in step S100 through the noun refinement unit 220 provided in the purification processing module 200. With reference, after searching for a pure Korean word corresponding to a Chinese character word made up of a noun and/or a compound noun in the restored sentence data, the noun purification process of the Chinese character word into the searched pure Korean word (S210) and , by referring to a separate narrative dictionary DB 230 based on the sentence data processed by the noun purification in step S210 through the narrative purification unit 240 provided in the purification processing module 200, the noun purification processing sentence data After searching for the descriptive refined word corresponding to the descriptive adjective including the verb, changing the descriptive adjective including the corresponding verb to the searched descriptive refined word to finally generate a new purified sentence (S220) and the like.

Here, in the case of the noun refinement of step S210, it refers to a case in which a noun or a compound noun is combined with a noun, and unnecessary Chinese characters or words with negative tendencies can be purified.

In addition, the Chinese character DB 210 is composed of, for example, words such as "unrequested", "request", "reissue", etc. or words consisting of affixes and nouns and purified expressions. In addition, a frequently used Chinese character dictionary is used to turn difficult terms into refined expressions.

Looking at an example of processing the nouns into pure Korean words using the Chinese character DB 210, the sentence "Use of the service as previously announced is available the next day." can be used the next day." That is, "notice" is refined to "information", and the sentence is refined to "next day" so that the user can easily understand it rather than "next day".

On the other hand, in the narrative form purification process of step S220, the narrative form including the verb plays a role of suitably purifying the customer service. For example, if you use an excessively awkward adjective, it will rather be a burden to the customer or inconvenient to communicate, so change it to a pure expression. In addition, a purified expression is generated by changing it to a word suitable for the context according to the situation.

That is, since the descriptive refinement of step S220 consists of a noun and a suffix or a verb and a suffix, the corresponding part extracts the morpheme and part-of-speech of the sentence through morphology analysis and uses the descriptive dictionary DB 230 to replace that part with a pure word.

Looking at an example of purifying awkward high-level words and unnecessary long expressions using this descriptive dictionary DB 230, the sentence of “Please refer to it and input it accurately.” The sentence is refined to “Please refer to it and enter it correctly.” can do it That is, since "refer to you" is an awkward exclamation, it is refined to "refer to", and "please refer" is unnecessarily long, which may make the user feel uncomfortable. Therefore, it is succinctly refined to "I hope" to create a sentence.

On the other hand, the above-described steps S210 and S220 are preferably implemented sequentially, but are not limited thereto, and may be implemented by changing the order according to the needs of the applied system.

Additionally, although not shown in the drawing, before the step S210, a preset pure Korean word corresponding to each Chinese character word is converted into a database (DB) through the purification processing module 200, and stored and managed in the Chinese character DB 210. It may include further steps.

In addition, before and/or after step S210, a separate negative word DB 250 is referred to through the noun-type refinement unit 220 provided in the purification processing module 200, and the restored processed sentence data and/or the The method may further include a first step of searching for a purified word corresponding to a negative word from the noun purification-processed sentence data, and then purifying the word of the negative tendency into the searched purified word.

That is, the negative word purification through the noun refinement unit 220 serves to reduce the use of words having negative meanings due to the characteristics of the response service. For example, a sentence of "The text message notification service of the desired rate of return cannot be used in the next month" may be refined to "The text notification service of the desired rate of return will not be used in the next month". That is, "notification" gives a coercive feeling, so it is refined to "notification", and a sentence is generated by refining it to "exclusion" rather than a word giving a negative feeling of "impossible".

In addition, before the first step, through the purification processing module 200, a preset purified word corresponding to each negative word having a negative meaning is converted into a database (DB) and stored in the negative word DB (250) and It may further include a step of managing.

In addition, before the step S220, through the purification processing module 200, the preset descriptive refined words corresponding to each of the descriptive high words including the verb are converted into a database (DB), and the step of storing and managing them in the descriptive dictionary DB 230 is further performed. may include

In addition, before and/or after the step S220, by referring to a separate negative word DB 250 through the narrative refinement unit 240 provided in the refinement processing module 200, the sentence data and/or The method may further include a second step of searching for a refined word corresponding to the negative tendency word from the descriptive refinement-processed sentence data, and then purifying the word of the negative tendency into the searched purified word.

In addition, before the second step, through the purification processing module 200, a preset purified word corresponding to each negative word having a negative meaning is converted into a database (DB) and stored in the negative word DB (250) and It may further include a step of managing.

An example of a process in which a sentence is finally generated through restoration and purification from an original text input from the outside using the purified sentence generating method for an interactive question and answer service according to an embodiment of the present invention described above will be described below. same.

First, the original text input from the outside through the voice recognition (STT) technology applied to the interactive Q&A service (e.g., “The maximum amount of jeonse loan you want to apply for is up to 200 million won, and the related contents are notified in bulk before the start of the next day notification time. In addition, please refer to the application for related inquiries.") restored through the restoration processing module 100 (e.g., "The maximum amount of jeonse loan you want to apply for is 200 million won, and related details start the next day's notification time. Batch notice is given before. Also, please refer to the app for related inquiries."), after finding and correcting the error, it is changed to a purified expression in a form suitable for the situation and context and easy for the user to understand through the purification processing module 200 process to create a new purification sentence (eg, "The maximum amount of jeonse loan you have applied for is up to 200 million won, and related information will be provided before the start of the next day's guide time. Also, please refer to the app for related inquiries.").

That is, the original text contains a number of errors such as spacing errors and the use of one word in multiple forms. "App" is restored as one representative word "app" by synonym processing. In the sentence restored through the space processing and the synonym processing unit 160 , the Chinese characters and excessive high-level words are purified so that the customer can easily understand them by the purification processing module 200 . For example, the amount, number, and amount are refined to "maximum" rather than "best", and to "information at once" for easier understanding than "batch notification".

The purified and finally generated sentences are provided directly to the counselor, so that a consistent and intuitive response service can be provided to the customer unlike the existing ones where the agent's individual dependence is high.

Meanwhile, the method for generating a purified sentence for an interactive Q&A service according to an embodiment of the present invention may also be implemented as a computer-readable code on a computer-readable recording medium. The computer-readable recording medium includes all types of recording devices in which data readable by a computer system is stored.

For example, computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, hard disk, floppy disk, removable storage device, and non-volatile memory (Flash Memory). , and optical data storage devices.

In addition, the computer-readable recording medium may be distributed in a computer system connected through a computer communication network, and stored and executed as readable code in a distributed manner.

Although the above-described preferred embodiment of the apparatus and method for generating a purified sentence for an interactive question and answer service according to the present invention has been described, the present invention is not limited thereto, and the claims, detailed description of the invention, and accompanying drawings It is possible to carry out various modifications within the scope of the invention, and this also belongs to the present invention.

100: restoration processing module,
110: language model DB,
120: preprocessing unit,
130: Abbreviation DB,
140: abbreviation restoration unit,
150: Synonym DB,
160: synonym processing unit;
200: purification processing module,
210: Chinese character DB,
220: noun refinement unit,
230: narrative dictionary DB,
240: narrative refinement unit,
250: negative word DB

Claims (20)

Through Speech to Text (STT) technology applied to the interactive Q&A service, it receives the voice-converted text data input from the outside, and uses natural language processing technology to correct errors caused by the characteristics of the input-to-speech text data. Restoration processing module for restoration processing to the front door; and
Refining processing of receiving the restored sentence data from the restoration processing module and changing the restored processed sentence into a purified expression in a form suitable for the situation and context and easy to understand by the user based on the received sentence data to generate a new purified sentence A purified sentence generator for an interactive question and answer service including a module.
According to claim 1,
The restoration processing module,
a language model DB for storing and managing preset Korean corresponding to each English word into a database (DB);
Based on natural language processing technology, at least one of preprocessing of space and misspelling errors, sentence separation, or cross-projection processing is based on changing the part written in English words in the input speech conversion text data to Korean with reference to the language model DB a preprocessor that does the work;
abbreviation DB for storing and managing preset original strings corresponding to each abbreviation word into a database (DB);
After receiving the pre-processed sentence data from the pre-processing unit, referring to the abbreviation DB based on this, and searching for the original string corresponding to the abbreviated word in the pre-processed sentence data, the abbreviated word is the searched original string Abbreviation restoration unit to restore to;
a synonym DB for storing and managing one preset representative word corresponding to each synonym bundle into a database (DB); and
When a plurality of synonyms exist in the sentence data restored to the original character string by referring to the synonym DB based on the received sentence data restored to the original character string from the abbreviation restoring unit, one preset corresponding to the plurality of synonyms and a synonym processing unit for collectively replacing the plurality of synonyms with the searched representative words after searching for a representative word.
According to claim 1,
The purification processing module,
Chinese character DB for storing and managing preset pure Korean words corresponding to each Chinese character word into a database (DB);
After receiving the restored-processed sentence data from the restoration processing module and referring to the Chinese-character DB based on it, searching for a pure Korean word corresponding to a Chinese-character word consisting of a noun or a compound noun in the restored-processed sentence data, the corresponding Chinese character a noun refining unit for refining a word into the searched pure Korean word;
a narrative dictionary DB that stores and manages a database (DB) of pre-set narrative refined words corresponding to each narrative high level including verbs;
After receiving the sentence data of the noun refinement process from the noun refinement unit, and referring to the descriptive dictionary DB based on this, the descriptive refinement word corresponding to the descriptive exaltation including the verb is retrieved from the noun refinement processed sentence data, and then the corresponding verb is retrieved. and a narrative refiner for finally generating a new purified sentence by changing the included narrative exclamation word to the searched narrative refined word.
4. The method of claim 3,
The descriptive refinement unit extracts morphemes and parts of speech of a sentence through morphological analysis of a part composed of a noun and ending or a verb and ending.
4. The method of claim 3,
The purification processing module,
A negative word DB is further included for storing and managing preset purified words corresponding to each negative word having a negative meaning in a database (DB),
The noun refinement unit searches for a purified word corresponding to the negative tendency word in the restored sentence data or the noun refinement processed sentence data with reference to the negative word DB, and then refines the searched word of the negative tendency. An apparatus for generating a purified sentence for an interactive question and answer service, characterized in that the purification process is performed as a type word.
4. The method of claim 3,
The purification processing module,
A negative word DB is further included for storing and managing preset purified words corresponding to each negative word having a negative meaning in a database (DB),
The descriptive refinement unit searches for a refined word corresponding to a negative tendency word in the noun refinement-processed sentence data or the narrative-type refinement-processed sentence data with reference to the negative word DB, and then searches for the word of the negative tendency in the searched A purified sentence generating apparatus for an interactive Q&A service, characterized in that the purified words are purified.
A method of generating a purified sentence using a restoration processing module and a purification processing module in an interactive question and answer service, the method comprising:
(a) the speech input using natural language processing technology based on speech-to-text data input from the outside by speech-to-text (STT) technology applied to the interactive question-and-answer service through the restoration processing module Restoring an error occurring as a characteristic of the converted text data to the main text; and
(b) receiving the sentence data restored in step (a) through the purification processing module, and changing the restored sentence based on this into a purified expression in a form suitable for the situation and context and easy for the user to understand A method of generating a purified sentence for an interactive Q&A service, comprising the step of generating a new purified sentence by processing.
8. The method of claim 7,
The step (a) is,
(a-1) It is basic to change the part written in English words in the input speech converted text data to Korean by referring to a separate language model DB based on natural language processing technology through the preprocessor provided in the restoration processing module. performing at least one pre-processing operation of spacing and misspelling errors, sentence separation, and inter-projection processing;
(a-2) Referring to a separate abbreviation DB on the basis of the sentence data pre-processed in step (a-1) through the abbreviation restoration unit provided in the restoration processing module, the abbreviation from the pre-processed sentence data after searching for the original string corresponding to the word, restoring the corresponding abbreviated word to the searched original string; and
(a-3) The sentence data restored to the original character string by referring to a separate synonym DB based on the sentence data restored to the original character string in step (a-2) through the synonym processing unit provided in the restoration processing module When a plurality of synonyms exist in , after searching for one preset representative word corresponding to the plurality of synonyms, the step of collectively replacing the plurality of synonyms with the searched representative word is an interactive question and answer method comprising the steps of: How to create a refining sentence for a service.
9. The method of claim 8,
Prior to the step (a-1), the method further comprising the step of storing and managing a preset Korean language corresponding to each English word through the restoration processing module into a database (DB) and storing and managing the language model DB A method of generating purified sentences for an interactive Q&A service.
9. The method of claim 8,
Prior to the step (a-2), the method further comprising the step of converting a preset original character string corresponding to each abbreviation word through the restoration processing module into a database (DB) and storing and managing the abbreviation DB in the abbreviation DB A method of generating purified sentences for an interactive Q&A service.
9. The method of claim 8,
Prior to the step (a-3), the method further comprises the step of converting one preset representative word corresponding to each synonym bundle through the restoration processing module into a database (DB) and storing and managing the synonym DB in the database (DB). A method of generating purified sentences for an interactive Q&A service.
8. The method of claim 7,
The step (b) is,
(b-1) Referring to a separate Chinese character DB on the basis of the sentence data restored in step (a) through the noun-type refiner provided in the purification processing module, from the restored sentence data to a noun or a compound noun after searching for a pure Korean word corresponding to the constructed Chinese character word, the step of refining the noun into the searched pure Korean word; and
(b-2) In the sentence data subjected to the noun purification by referring to a separate narrative dictionary DB based on the sentence data processed in step (b-1) through the narrative purification unit provided in the purification processing module Conversational question-and-answer, characterized in that it includes the step of searching for a descriptive refined word corresponding to a descriptive adjective including a verb, and then changing the descriptive adjective including the corresponding verb to the searched descriptive refined word to finally generate a new purified sentence How to create a refining sentence for a service.
13. The method of claim 12,
Prior to the step (b-1), the method further comprising the step of storing and managing a preset pure Korean word corresponding to each Chinese character word through the purification processing module into a database (DB) and storing and managing the Chinese character DB A method of generating purified sentences for an interactive Q&A service.
13. The method of claim 12,
Prior to the step (b-2), the conversion processing module further comprises the step of storing and managing a pre-set narrative refined word corresponding to each narrative adjective including a verb into a database (DB), and storing and managing the narrative dictionary DB. A method of generating a purified sentence for an interactive question and answer service, characterized in that.
13. The method of claim 12,
In the step (b-2), the descriptive refinement unit extracts the morpheme and part-of-speech of the sentence through morpheme analysis for a part composed of a noun and a ending or a verb and a ending. How to create a sentence.
13. The method of claim 12,
Before or after step (b-1),
After referring to a separate negation DB through the noun refining unit provided in the refining processing module, a refining type word corresponding to a negative tendency word is searched for in the restored sentence data or the noun refining processed sentence data, The method of generating a purified sentence for an interactive question and answer service, characterized in that it further comprises a first step of purifying the word of the negative tendency into the searched purified word.
17. The method of claim 16,
Prior to the first step, the method further includes the step of converting a preset purified word corresponding to each negative word having a negative meaning through the purification processing module into a database, and storing and managing the negative word DB in the negative word DB. A method of generating a purified sentence for an interactive question and answer service, characterized in that
13. The method of claim 12,
Before or after step (b-2),
After referring to a separate negation DB through the narrative purification unit provided in the purification processing module, the purified word corresponding to the negative tendency word is searched for in the noun purification-processed sentence data or the narrative-type purification processed sentence data, The method for generating a purified sentence for an interactive question and answer service, characterized in that it further comprises a second step of purifying the word of the negative tendency into the searched purified word.
18. The method of claim 17,
Prior to the second step, the method further includes a step of converting a preset purified word corresponding to each negative word having a negative meaning through the purification processing module into a database, storing and managing the negative word DB in the negative word DB A method of generating a purified sentence for an interactive question and answer service, characterized in that
A computer-readable recording medium in which a program capable of executing the method of any one of claims 7 to 19 by a computer is recorded.

Images