KR20160060915A - Method for creating language model based on root and language processing apparatus thereof - Google Patents

Abstract

The present invention relates to a method for generation of a root-based language model and a language processing apparatus for the same. In particular, the method and the language processing apparatus are configured to extract each sentence, included in voice data, in a unigram form through language processing, analyze morphemes for each sentence in a unigram form, perform clustering based on roots of the morphemes, and match a cluster in an N-gram form, thereby overcoming a data shortage problem as in a class-based language model. Furthermore, the method and the language processing apparatus can be modified and utilized in such a way that using a linguistic relationship, an object which can be placed in front of a verb is mainly extracted in the case of the verb and the following relationship is extracted in the case of a noun having a postposition, thereby improving accuracy in language extraction. Furthermore, a new noun which is not present in a language model for voice recognition can be added in an N-gram form incorporating various expressions of a sentence, thereby providing higher voice recognition performance for a proper noun than that of a method of simply adding only the proper noun.

Description

TECHNICAL FIELD The present invention relates to a language-based language model generation method and a language processing apparatus for the same,

The present invention relates to a speech recognition technology, and more particularly, to a root-based language model generation method for automatically generating a statistical language model using a root of each word in speech recognition and natural language processing, and a language processing device .

Efforts to make the interface between machine and person more convenient and natural by using speech recognition technology have been progressed steadily at home and abroad, and as a result, it has been developed to a level that can process natural voiced speech beyond simple word recognition level . Speech recognition technology has evolved into a level that can be used in real life in various fields thanks to continuous technology development in the latter half of the 20th century. However, in order to be actively used in many application fields that we imagine, It is a fact that it is piled up.

In recent years, efforts to provide users with more convenient services by utilizing such natural language speech recognition processing technology have been taking place in domestic communication companies and financial institutions. In order to improve the difficulty of inputting through a limited keyboard in a smartphone, voice recognition support function becomes a necessary function in a wireless mobile device and the like.

In general, the speech recognition method is divided into several types according to the form of utterance, such as isolated word recognition, connected word recognition, continuous speech recognition, keyword spotting, etc. . Unlike lone word recognition, which recognizes individual words, continuous speech recognition is a method of finding sentences or consecutive word strings corresponding to a speech signal. As the number of words in a lexical dictionary increases, the number of word strings constituting a sentence increases significantly The more pronounced the number of words due to the variation of pronunciation between the word and the word, the more likely it is to be mistaken for words of similar pronunciation.

Especially, the language model in speech recognition is a model constructed by collecting the connectivity between words from the text corpus by statistical method so that the sentence that the user uttered is recognized as the correct sentence. Unigram, Bigram, and Trigram are often used in language models. The unigram uses the probability of a word and does not use the past word which is located immediately in front of it. In addition, Biagram and Triagram use probabilities that depend on the first one and two words respectively. The use of such a language model allows grammatically valid word sequences to be recognized, minimizing the search space of words and sentences, thereby improving recognition performance and shortening the search time.

However, if a statistical language model used for speech recognition and natural language processing is constructed, a data shortage phenomenon may occur. In the case of a deformation language and spoken language on the web, Thereby degrading the performance of the apparatus.

Korean Registered Patent No. 10-0511247 B1, registered on Aug. 23, 2005 (Name: Language Modeling Method in Speech Recognition System)

If a statistical language model used in conventional speech recognition and natural language processing is structured, data shortage may occur. In the case of a target language such as a variation language and spoken language on the web, In order to solve the problem of deteriorating the performance of speech recognition, an object of the present invention is to provide a speech recognition apparatus and a speech recognition method in which each sentence included in speech data is subjected to language processing and extracted as a unigram form, , A method for generating a root-based language model in which each root of each morpheme is clustered and matched in an N-gram form, and a language processing apparatus for the same.

According to another aspect of the present invention, there is provided a method for generating a root-based language model, the method including the steps of: collecting at least one piece of voice data; Extracting the separated sentences in a linguistic form and extracting them in a unigram form; analyzing a morpheme for each sentence in a unigram form extracted by the language processing apparatus; A step of constructing a cluster around each root, and a step of matching a cluster constituted by the language processing apparatus in an N-gram form.

Further, in the method of generating a root-based language model according to the present invention, the collecting step may include a step of collecting voice data including at least one of natural language processing, speech recognition, And converting the collected voice data into a text.

In the method for generating a root-based language model according to the present invention, the step of extracting in the form of a unicram is a computer-readable record in which a program for executing a processing language processing of a sentence in a corpus unit or a space- Media.

In the method of generating a root-based language model according to the present invention, the step of analyzing the morpheme may include separating at least one of a root, an irradiation, a macrophase, a frontal end, or a termination end And a step of storing the separated root by the part of speech by the language processing apparatus.

According to another aspect of the present invention, there is provided a method of generating a root-based language model, the method comprising the steps of: constructing a cluster by using a computer readable recording medium having recorded thereon a program for executing a clustering step on the basis of a central part, Lt; / RTI >

In the root-based language model generation method according to the present invention, the step of matching in the form of an N-gram may include a step of the language processing device confirming the cluster constituted by each root-based, (N-gram) type information by combining information obtained through morphological analysis with the inner root of a root-mean-square root.

In addition, in the root-based language model generation method according to the present invention, after the step of matching in the form of an N-gram, the language processing apparatus generates a language based on the result matched in an N- There is provided a computer-readable recording medium having recorded thereon a program for automatically generating a model.

The language processing apparatus according to an embodiment of the present invention separates speech data collected through a language collection module and a language collection module for collecting at least one speech data into sentences, A language processing module for analyzing the morpheme for each sentence in the extracted unigram form, constructing a cluster around each root of the analyzed morpheme, and matching the configured cluster in an N-gram form And a control unit.

According to the present invention, each sentence included in the speech data is language-processed and extracted in the form of a unicagram, the morphemes for each sentence in the form of a unicagram are analyzed, and clusters are centered around each root of the morpheme, grams, it is possible to solve the data shortage similar to the class-based language model.

In addition, it is possible to improve the accuracy of language extraction by using the verbal relation, extracting the predicate of the verb in the case of the verb, and transforming the relation in the case of the verb with the verb .

In addition, since a new noun not in the language model for speech recognition can be added in the form of an N-gram that reflects various expressions of the sentence, .

1 is a block diagram showing a configuration of a language processing apparatus according to an embodiment of the present invention.
2 is a flowchart illustrating a method of generating a root-based language model according to an embodiment of the present invention.
FIG. 3 is an exemplary diagram for explaining a root-based language model generation method according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the following description and the accompanying drawings, detailed description of well-known functions or constructions that may obscure the subject matter of the present invention will be omitted. It should be noted that the same constituent elements are denoted by the same reference numerals as possible throughout the drawings.

The terms and words used in the present specification and claims should not be construed to be limited to ordinary or dictionary meanings and the inventor is not limited to the concept of terminology for describing his or her invention in the best way. It should be interpreted as meaning and concept consistent with the technical idea of the present invention. Therefore, the embodiments described in the present specification and the configurations shown in the drawings are merely the most preferred embodiments of the present invention, and not all of the technical ideas of the present invention are described. Therefore, It is to be understood that equivalents and modifications are possible.

Hereinafter, a language processing apparatus according to an exemplary embodiment of the present invention is described as a representative example of a mobile communication terminal connected to a communication network to automatically generate a root-based language model and to apply a language model, The present invention can be applied to various terminals such as all information communication devices, multimedia terminals, wired terminals, fixed type terminals and IP (Internet Protocol) terminals. Also, the terminal may be a mobile phone, a portable multimedia player (PMP), a mobile Internet device (MID), a smart phone, a desktop, a tablet PC, a notebook, And an information communication device, which can be advantageously used in a mobile terminal having various mobile communication specifications.

Further, a processor mounted in the language processing apparatus according to the present invention can process program instructions for executing the method according to the present invention. In one implementation, the processor may be a single-threaded processor, and in other embodiments, the processor may be a multi-threaded processor. Further, the processor is capable of processing instructions stored on a memory or storage device.

A language processing apparatus for generating a root-based language model according to an embodiment of the present invention will be described.

1 is a block diagram showing a configuration of a language processing apparatus according to an embodiment of the present invention.

1, a language processing apparatus 100 according to the present invention includes a control unit 10, an input unit 20, a display unit 30, a storage unit 40, an audio processing unit 50, and a communication unit 60 do. Here, the control unit 11 includes a language acquisition module 11 and a language processing module 12, and the storage unit 40 includes voice data 41.

The input unit 20 receives various information such as numbers and character information and transmits various signals to the control unit 10 in response to the setting of various functions and the function control of the language processing apparatus 100. The input unit 20 may include at least one of a keypad and a touchpad that generates an input signal according to a user's touch or operation. At this time, the input unit 20 may be configured in the form of a single touch panel (or a touch screen) together with the display unit 30 to simultaneously perform the input and display functions. The input unit 20 may be any type of input device that can be developed in addition to an input device such as a keyboard, a keypad, a mouse, a joystick, and the like. In particular, the input unit 20 according to the present invention senses a series of input signals generated in the root-based language model generation process and transmits the input signals to the control unit 10.

The display unit 30 displays information on a series of operation states, operation results, and the like that occur during the performance of the functions of the language processing apparatus 100. In addition, the display unit 30 can display menus of the language processing apparatus 100, user data input by the user, and the like. The display unit 30 may be a liquid crystal display (LCD), a thin film transistor LCD (TFT-LCD), a light emitting diode (LED), an organic light emitting diode LEDs, active matrix organic light emitting diodes (AMOLED), active matrix OLEDs, retina displays, flexible displays, and three-dimensional displays. In this case, when the display unit 30 is configured as a touch screen, the display unit 30 may perform some or all of the functions of the input unit 20. [ In particular, the display unit 30 according to the present invention outputs all the screen information generated in the root-based language model generation process.

The storage unit 40 is a device for storing data, and includes a main storage device and an auxiliary storage device, and stores an application program required for the functional operation of the language processing device 100. [ The storage unit 40 may include a program area and a data area. Here, when the language processing apparatus 100 activates each function corresponding to a request from the user, the language processing apparatus 100 executes the corresponding application programs under the control of the control unit 10 to provide each function. In particular, the storage unit 40 according to the present invention includes an operating system for booting the language processing apparatus 100, a program for collecting voice data, a program for separating voice data on a sentence basis, A program for analyzing a morpheme for each sentence in a uniagram form, a program for forming a cluster around each root of a morpheme, and a program for matching a cluster in an N-gram form. Also, the storage unit 40 stores voice data input from the outside and voice data received from another apparatus.

The audio processing unit 50 performs a function of transmitting an audio signal input from a speaker SPK or a microphone MIC for reproducing and outputting an audio signal to the control unit 10. The audio processing unit 50 converts an analog audio signal input through a microphone into a digital format, and transmits the audio signal to the controller 10. The audio processing unit 50 may convert an audio signal of a digital format output from the control unit 10 into an analog signal and output it through a speaker. In particular, the audio processing unit 50 according to the present invention outputs an effect sound or an execution sound generated in the root-based language model generation process.

The communication unit 60 performs a function for transmitting and receiving data through a communication network (not shown) with another device. Here, the communication unit 60 includes RF transmitting means for up-converting and amplifying the frequency of the transmitted signal, RF receiving means for low-noise amplifying the received signal and down-converting the frequency. The communication unit 60 may include at least one of a wireless communication module (not shown) and a wired communication module (not shown). The wired communication module is for transmitting / receiving data by wire. In particular, the communication unit 60 according to the present invention collects voice data in cooperation with an external device.

Here, the communication network performs a series of data transmission / reception operations for data transmission and information exchange between the language processing apparatus 100 and other devices. In particular, various types of communication networks may be used for the communication network. For example, the communication network may be a wireless LAN (WLAN), a Wi-Fi, a Wibro, a WiMAX, a high speed downlink packet access Speed Downlink Packet Access) or a wired communication method such as Ethernet, xDSL (ADSL, VDSL), HFC (Hybrid Fiber Coax), FTTC (Fiber to the Curb), FTTH (Fiber To The Home) Can be used. Meanwhile, the communication network is not limited to the above-described communication methods, and may include all other known or later-developed communication methods in addition to the communication methods described above.

The control unit 10 may be an operating system (OS) and a process unit for driving each configuration. In particular, the control unit 10 according to the present invention collects voice data.

The control unit 10 separates the collected voice data on a sentence-by-sentence basis, and language-processes each separated sentence to extract it as a unigram form. Then, the control unit 10 analyzes morphemes for each sentence in the extracted unigram form.

The control unit 10 forms a cluster around each root of the analyzed morpheme. Then, the control unit 10 matches the configured clusters in an N-gram form. For example, an N-gram type may include a Bigram type, a Trigram type, and the like. Thereafter, the control unit 10 automatically generates a language model based on the result matched in the form of an N-gram.

In order to perform the functions of the language processing apparatus 100 more effectively, the control unit 10 includes a plurality of modules, which include a language acquisition module 11 and a language processing module 12 .

The language collection module 11 collects and manages voice data input from the outside and voice data received from another device. Here, the language collection module 11 collects speech data including natural language processing, speech recognition, transformed language, and spoken language, and converts the collected speech data into text.

The language collection module 11 separates actual speech data and noise from the collected speech data. That is, since the noise data collected from the outside is mixed with surrounding noise, a noise removal process is performed for reliable speech recognition. For example, the language acquisition module 11 may apply a discrimination technique based on a forward search, a psychoacoustic-based estimation technique, or an improved spectral subtraction technique to remove noise in the collected voice data.

The language processing module 12 separates the speech data collected through the language acquisition module 11 on a sentence-by-sentence basis, and language-processes each separated sentence to extract it as a uniagram form. At this time, the language processing module 12 processes each sentence in a processing unit in a corpus unit or a space unit. Here, the corpus contains texts of a single language or several languages and can be used for statistical analysis and hypothesis verification, or for checking the occurrence of language rules within a specific language area and for justifying the rules. For example, a corpus may contain information about the parts of speech (verbs, nouns, adjectives, etc.) of each word.

The language processing module 12 analyzes morphemes for each sentence in the extracted unigram form. Here, the language processing module 12 separates at least one of the root, the irradiation, the affix, the prefix, and the termination term in each sentence that has been subjected to the language processing, and stores the separated root by the parts of speech. That is, the language processing module 12 analyzes one or more morphemes according to the language unit words that lose their meaning if they are further analyzed as the smallest unit having meaning in a language.

In addition, morphemes can be classified into independent morphemes, dependent morphemes, vocabulary morphemes (substantial morphemes), and grammatical morphemes (morphemes). For example, an independent morpheme is a morpheme that can be used alone, usually a morpheme that constitutes a noun and some adverbs, a dependency morpheme is a morpheme that can not be used alone, and a part of speech composed of independent morpheme and dependency morpheme is morpheme, . In addition, vocabulary morpheme is a morpheme that clarifies the grammatical relation between morphemes and other morphemes and constituent parts. In the case of the atmosphere, it has an independent meaning, but sometimes it can not stand alone, and includes nouns, verbs, adjectives, and adverbs . In addition, grammatical morphemes are grammatical morphemes that combine morphemes to form meaningful language expressions, including stem, affix, and investigation.

The language processing module 12 forms a cluster around each root of the analyzed morpheme. At this time, the language processing module 12 clusters on the basis of a central part indicating a substantial meaning. In other words, the language processing module 12 extracts a root, which is a meaningful part, and the root is a case where the word is divided according to the center of the meaning, that is, the part having the central meaning in the word and is used with the affix having the secondary meaning . In addition, a root is a necessary part when making a word except for a few words, a word consisting of only one root is singly, and a case where two or more roots are combined to form a word is called a compound word. When combined, the word is called a derivative.

The language processing module 12 matches the clusters in an N-gram form. Here, the language processing module 12 identifies clusters constituting each root base, and forms an N-gram type by combining the information obtained through morphological analysis with the root in the cluster. Then, the language processing module 12 automatically generates a language model based on the result matched in the form of an N-gram.

On the other hand, the memory mounted on the language processing apparatus 100 stores information in the apparatus. In one implementation, the memory is a computer-readable medium. In one implementation, the memory may be a volatile memory unit, and in other embodiments, the memory may be a non-volatile memory unit. In one implementation, the storage device is a computer-readable medium. In various different implementations, the storage device may include, for example, a hard disk device, an optical disk device, or any other mass storage device.

Although the present specification and drawings describe exemplary device configurations, the functional operations and subject matter implementations described herein may be embodied in other types of digital electronic circuitry, or alternatively, of the structures disclosed herein and their structural equivalents May be embodied in computer software, firmware, or hardware, including, or in combination with, one or more of the foregoing. Implementations of the subject matter described herein may be embodied in one or more computer program products, i. E. One for computer program instructions encoded on a program storage medium of the type for < RTI ID = 0.0 & And can be implemented as a module as described above. The computer-readable medium can be a machine-readable storage device, a machine-readable storage substrate, a memory device, a composition of matter that affects the machine readable propagation type signal, or a combination of one or more of the foregoing.

FIG. 2 is a flowchart illustrating a root-based language model generation method according to an embodiment of the present invention. FIG. 3 is a diagram illustrating a root-based language model generation method according to an embodiment of the present invention.

Referring to FIGS. 2 and 3, in the root-based language model generation method according to the present invention, the language processing apparatus 100 collects voice data input from the outside and voice data received from another apparatus in step S11 Management. Here, the language collection module 11 collects speech data including natural language processing, speech recognition, transformed language, and spoken language, and converts the collected speech data into text.

The language processing apparatus 100 separates the speech data collected in step S13 on a sentence-by-sentence basis, and performs language processing on each separated sentence to extract it as a uniagram form. At this time, the language processing apparatus 100 processes each sentence in a speech processing unit in units of a corpuscule or a space. Here, the corpus contains texts of a single language or several languages and can be used for statistical analysis and hypothesis verification, or for checking the occurrence of language rules within a specific language area and for justifying the rules. For example, a corpus may contain information about the parts of speech (verbs, nouns, adjectives, etc.) of each word.

The language processing apparatus 100 analyzes the morpheme for each sentence in the unigram form extracted in step S15. Here, the language processing apparatus 100 separates at least one of a root, an irradiation, a close-up, a line ending, or a termination end in each language-processed sentence, and stores the separated root by the parts of speech. That is, the language processing apparatus 100 analyzes one morpheme or several morphemes according to a language unit word for which meaning is lost if it is further analyzed as the smallest unit having a meaning in one language.

In addition, morphemes can be classified into independent morphemes, dependent morphemes, vocabulary morphemes (substantial morphemes), and grammatical morphemes (morphemes). For example, an independent morpheme is a morpheme that can be used alone and is usually a morpheme that constitutes a noun and some adverbs, a dependency morpheme is a morpheme that can not be used alone, and a part of speech composed of independent morpheme and dependency morpheme is morpheme, . In addition, vocabulary morpheme is a morpheme that clarifies the grammatical relation between morphemes and other morphemes and constituent parts. In the case of the atmosphere, it has an independent meaning, but sometimes it can not stand alone, and includes nouns, verbs, adjectives, and adverbs . In addition, grammatical morphemes are grammatical morphemes that combine morphemes to form meaningful language expressions, including stem, affix, and investigation.

The language processing apparatus 100 forms a cluster around each root of the morpheme analyzed in step S17. At this time, the language processing apparatus 100 clusters on the basis of a central part indicating a substantial meaning. That is, the language processing apparatus 100 extracts a root, which is a meaningful part, and the root is a case in which a word is divided according to the center of meaning, that is, a part having a central meaning in a word, . In addition, a root is a necessary part when making a word except for a few words, a word consisting of only one root is singly, and a case where two or more roots are combined to form a word is called a compound word. When combined, the word is called a derivative.

The language processing apparatus 100 matches the cluster in the form of an N-gram in step S19. Here, the language processing apparatus 100 identifies clusters constituting each root base, and forms an N-gram form by combining the information obtained through morphological analysis with the root in the cluster. Then, the language processing apparatus 100 automatically generates a language model based on the result matched in the form of an N-gram.

That is, as shown in FIG. 3, the language processing apparatus 100 extracts each sentence of the voice data in the form of a unigram. For example, the language processing apparatus 100 has processed each sentence. I did, but I did, I applied, I applied, I applied, I ... 'Can be extracted in the form of a unigram.

Then, the language processing apparatus 100 analyzes morphemes for each sentence in the extracted unigram form. For example, the language processing apparatus 100 separates the "had" into a morpheme of "VV + odd +" and separates the morpheme into "VV + odd" (VV) + I + I ', and' Do (VV) + I + Da 'are separated by the morpheme. Further, the language processing apparatus 100 separates the 'application request' into the morphemes of 'application (NN) + Ha + ward', and separates the morpheme of 'application (NN) + Ha + And separates the 'application' into the 'application (NN) + morphology'.

The language processing apparatus 100 performs root-based trustering based on the morpheme analysis result. For example, the language processing apparatus 100 clusters to '(VV)', 'did', 'did', and 'clusters' to 'application (NN) and clusters to'.

Thereafter, the language processing apparatus 100 matches clusters in an N-gram form. For example, the language processing apparatus 100 may match in a bi-gram form of an N-gram. In other words, the language processing apparatus 100 matches the bi-gram matching of "I had applied" and "I applied, and I applied, but bi-gram matching", "I had hedged" and "I canceled, I had a connection and I had a connection, I had a connection, I had a connection, I had a connection, I had a connection, I had a connection, I had a connection, I had a connection,

Computer-readable media suitable for storing computer program instructions and data include, for example, magnetic media such as hard disks, floppy disks, and magnetic tape, compact disk read only memory (CD-ROM) A magneto-optical medium such as a floppy disk and an optical recording medium such as a digital video disk, a magneto-optical medium such as a floppy disk, and a read only memory (ROM) Access Memory), a flash memory, an erasable programmable ROM (EPROM), and a semiconductor memory such as an Electrically Erasable Programmable ROM (EEPROM). The processor and memory may be supplemented by, or incorporated in, special purpose logic circuits. Examples of program instructions may include machine language code such as those generated by a compiler, as well as high-level language code that may be executed by a computer using an interpreter or the like. Such a hardware device may be configured to operate as one or more software modules to perform the operations of the present invention, and vice versa.

While the specification contains a number of specific implementation details, it should be understood that they are not to be construed as limitations on the scope of any invention or claim, but rather on the description of features that may be specific to a particular embodiment of a particular invention Should be understood. Certain features described herein in the context of separate embodiments may be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment may also be implemented in multiple embodiments, either individually or in any suitable subcombination. Further, although the features may operate in a particular combination and may be initially described as so claimed, one or more features from the claimed combination may in some cases be excluded from the combination, Or a variant of a subcombination.

Likewise, although the operations are depicted in the drawings in a particular order, it should be understood that such operations must be performed in that particular order or sequential order shown to achieve the desired result, or that all illustrated operations should be performed. In certain cases, multitasking and parallel processing may be advantageous. Also, the separation of the various system components of the above-described embodiments should not be understood as requiring such separation in all embodiments, and the described program components and systems will generally be integrated together into a single software product or packaged into multiple software products It should be understood.

It should be noted that the embodiments of the present invention disclosed in the present specification and drawings are only illustrative of specific examples for the purpose of understanding and are not intended to limit the scope of the present invention. It will be apparent to those skilled in the art that other modifications based on the technical idea of the present invention are possible in addition to the embodiments disclosed herein.

In the present invention, each sentence included in speech data is subjected to language processing to extract it as a uniagram form, analyze the morpheme for each sentence in the form of a uniagram, cluster it around each root of the morpheme, . Accordingly, the present invention can solve the data shortage problem similar to the class-based language model. In addition, it is possible to improve the accuracy of language extraction by using the verbal relation, extracting the predicate of the verb in the case of the verb, and transforming the relation in the case of the verb with the verb . In addition, since a new noun not in the language model for speech recognition can be added in the form of an N-gram that reflects various expressions of the sentence, . This is not only a possibility of commercialization or sales, but also a possibility of being industrially applicable since it is practically possible to carry out clearly.

100: language processing apparatus 10: control unit
11: Language collection module 12: Language processing module
20: input unit 30: display unit
40: storage unit 41: audio data
50: audio processor 60:

Claims (8)

The language processing apparatus collecting at least one voice data;
Separating the collected speech data on a sentence-by-sentence basis, language processing each of the separated sentences, and extracting the speech data in a unigram form;
Analyzing a morpheme for each sentence in the extracted unigram form;
Constructing a cluster around each root of the analyzed morpheme; And
And the language processing apparatus matches the configured cluster in an N-gram form. The method of claim 1, wherein said collecting comprises:
Collecting speech data including at least one of natural language processing, speech recognition, transformed language, and spoken language; And
And the language processing apparatus converts the collected voice data into text. The method of claim 1, wherein extracting in the form of a unicram
And the language processing apparatus processes the sentence in a processing unit of a corpuscary unit or a spacing unit. The method of claim 1, wherein analyzing the morpheme comprises:
Separating at least one of a root, an irradiation, a lexical root, a prefix end, or a termination end in each language-processed sentence by the language processing apparatus; And
And the language processing apparatus classifies the separated root according to part-of-speech and stores the divided root. The method of claim 1, wherein configuring the cluster comprises:
And the clustering step is performed on the basis of the central part in which the language processing apparatus has a substantial meaning. 2. The method of claim 1, wherein matching in the N-gram form comprises:
Identifying the cluster consisting of each root based on the language processing device; And
And the language processing apparatus forms an N-gram form by combining the information obtained through morphological analysis with the root in the cluster. 2. The method of claim 1, wherein after the matching in the N-gram form,
And the language processing apparatus automatically generates a language model based on a result matched in the form of the N-gram. A language collection module for collecting at least one voice data; And
Separating the speech data collected through the language collection module on a sentence basis, language processing each of the separated sentences, extracting the speech data in a unigram form, analyzing morphemes for each sentence in the extracted unigram form, A language processing module for constructing a cluster around each root of the analyzed morpheme and matching the configured cluster in an N-gram form;
Wherein the language processing apparatus comprises:

Images