KR102543926B1 - User Equipment with Artificial Inteligence for Forign Language Education and Method for Forign Language Education - Google Patents

Abstract

The present invention relates to a user device having an artificial intelligence function for foreign language education and providing a voice response according to a grammatical error/pronunciation error of a user's voice, and a method of providing foreign language education through online access of the user device.
A user device having an artificial intelligence function for foreign language education determines whether there is an error recognized in the process of recognizing the user's voice as a character, whether the error is a pronunciation error, or whether the error is a grammatical error, and determines whether the error is a grammatical error. In the case of a pronunciation error, a first voice response including an expression capable of replacing the erroneous part is provided through a speaker, and if the error is a grammatical error, a pre-prepared grammar corresponding to the erroneous part is provided. A second audio response containing the description is provided through the speaker.

Description

User Equipment with Artificial Intelligence for Foreign Language Education and Method for Foreign Language Education

The present invention relates to a user device having an artificial intelligence function for foreign language education and providing a voice response according to a grammatical error/pronunciation error of a user's voice, and a method of providing foreign language education through online access of the user device.

Recently, with the spread of artificial intelligence (AI) speakers, a service is being provided that controls household appliances through user voice recognition and, more advancedly, conducts conversations with AI characters.

Some users are also using it for learning foreign languages by inputting commands into these AI speakers in foreign languages and controlling them. However, the current AI speaker only recognizes the user's foreign language command and controls or reacts to the device accordingly, but cannot be used for higher level foreign language education.

In addition, a service that recognizes a user's voice and teaches pronunciation through online is also provided. However, these services only provide more accurate pronunciation corresponding to the user's voice, but do not provide a distinction as to whether there is a grammatical problem or a pronunciation problem in the user's foreign language voice.

In order to solve the above problems, in one embodiment of the present invention, it is efficiently classified whether there is an error in the user's foreign language voice, and if there is an error, whether it is a pronunciation problem or a grammatical problem, and appropriate education service according to this classification. want to provide

In addition, an embodiment of the present invention intends to provide an interface for providing an efficient education service to the user based on the above-described classification of the user's voice.

The object of the present invention is not limited to the above object, and various objects are presented through the following description.

In one aspect of the present invention for solving the above problems, in the user device having an artificial intelligence function for foreign language education, the voice input device for receiving the user's voice; a processor recognizing and processing the user's voice as text; and a speaker that provides a voice response to the user according to the processing of the processor, wherein the processor determines (a) whether there is an error recognized in the process of recognizing the user's voice as a text or whether the error is a pronunciation error. , or determining whether the error is a grammatical error, (b) if the error is a pronunciation error, providing a first voice response including an expression that can replace the erroneous part through the speaker; , (c) If the error is a grammatical error, the user device provides a second audio response including a pre-prepared grammatical explanation corresponding to the erroneous part through the speaker.

Here, the processor is used to recognize a first grammar model used to recognize the user's voice as a text and, after recognizing the user's voice as a text, a grammatical error present in the text-recognized language. A second grammar model can be used.

The second grammatical model may be configured to update each grammatical error and a grammatical description corresponding to each grammatical error in pairs based on learning based on data of the web server.

When there is no error, the processor may provide a third voice response corresponding to the content of the user's language recognized as text and the content of conversation through the speaker.

The processor extracts a feature from the user's voice, detects a candidate pronunciation sequence corresponding to the feature, selects a language that can be combined with a predetermined probability or more using a first grammar model from the candidate pronunciation sequence, and The user's voice can be recognized as text.

The processor may determine that the error is a pronunciation error when a language combinable using the first grammar model is not selected from among the candidate pronunciation sequences.

The processor selects a language that can be combined using the first grammar model from among the candidate pronunciation sequences, and when a difference between a character recognized with the highest probability and the selected language is greater than or equal to a predetermined standard, the error can be judged to be a pronunciation error.

In the operation (c), the processor receives a user's response to an expression capable of replacing the erroneous part through the voice input device, and converts the user's voice into text according to the user's response. can be recognized as

The processor may distinguish between a native language voice and a foreign language voice among the user voices, and perform operations (a) to (c) when the user voice is determined to be the foreign language voice.

The user's voice may include a command for changing the operation mode of the user device, and according to the command, the processor (1) processes the user's voice in units of one sentence and outputs the voice response to the speaker. or (2) a long text recording mode in which the user's voice is continuously recorded for a predetermined period of time and then the voice response is provided to the speaker.

The user device may be wirelessly connected to one or more in-house devices using an Internet of Things (IoT) method, and may control the in-house devices through at least one of a native language voice and a foreign language voice among the user voices.

Meanwhile, in another aspect of the present invention for solving the above problems, in a method for providing foreign language education through online access of a user device, by recognizing the user's foreign language sentence voice pronunciation input through the microphone of the user device Conversion into characters, determining an error in the sentence converted into characters, and if the erroneous part is a grammatical error, a pre-prepared grammatical explanation corresponding to the erroneous part is highlighted on the erroneous part is displayed on the display device of the user device together with, and when the erroneous part is a pronunciation error, an expression that can replace the erroneous part is displayed on the display device of the user device along with highlighting of the erroneous part. A method for providing foreign language education is provided, which includes displaying on a display device, recommending, and presenting a recommended pronunciation of an exemplary sentence including the recommended expression through a speaker of the user device.

Recommending an expression capable of replacing the erroneous part recommends a plurality of expressions and displays them on a display device of the user device, and determines an expression selected through an input device of the user device as the recommended expression; It may include storing a mapping relationship between an erroneous portion of the user's voice pronunciation of a foreign language sentence and the determined recommended expression in a database of a server providing foreign language education, and learning the server based on artificial intelligence.

Corresponding data between the erroneous part and the prepared grammatical explanation is increased in the database of the server providing the foreign language education by artificial intelligence, and when it is determined that the erroneous part is a grammatical error, the It may be the case that there is a pre-prepared grammar explanation corresponding to the part with the error.

According to the embodiments of the present invention as described above, it is possible to efficiently classify whether there is an error in the user's foreign language voice, and if there is an error, whether it is a pronunciation problem or a grammatical problem, and provide appropriate education services according to this classification. can

In addition, according to an embodiment of the present invention, an interface for providing an efficient education service to a user may be provided based on the above-described classification of the user's voice.

Effects according to the present invention are not limited to the above-mentioned effects, and various effects are presented through the following description.

1 is a diagram for explaining the concept of a user device according to an embodiment of the present invention.
2 is a diagram for explaining an operating method of a user device according to an embodiment of the present invention.
3 is a diagram for explaining in detail a process of recognizing a user's voice according to an embodiment of the present invention.
4 is a diagram for explaining a concept of controlling an in-house device through a user device according to an embodiment of the present invention.
5 is a diagram for explaining a method of providing foreign language education according to an embodiment of the present invention.

Hereinafter, the description of the present invention will be described in detail with reference to the drawings.

1 is a diagram for explaining the concept of a user device according to an embodiment of the present invention.

Referring to FIG. 1 , a user device having an artificial intelligence function for foreign language education according to the present embodiment may have the form of an AI speaker 10. In addition, the AI speaker 10 may be connected wirelessly or wired to a user device having various software processing functions, such as a smart phone 20a or a laptop computer 20b.

In the following description, the user device may be built into the AI speaker 10 itself, and unlike the voice input device and speaker function for providing a voice interface with the user, the AI speaker 10 and the processor that processes this information may be provided in another user device 20a or 20b connected to the AI speaker 10.

Specifically, the user device according to the present embodiment includes a voice input device that receives a user's voice; a processor recognizing and processing the user's voice as text; and a speaker providing a voice response to the user according to the processing of the processor. Here, the user's voice may be a native language voice signal or a foreign language voice signal, which will be described later.

The voice input device is preferably provided in the AI speaker 10 shown in FIG. 1, but may be provided in other user devices (20a, 20b, etc.).

2 is a diagram for explaining an operating method of a user device according to an embodiment of the present invention.

The processor according to the present embodiment may first determine whether there is an error recognized in the process of recognizing the user's voice as text (S310). If there is an error, the process may additionally determine whether the error is a pronunciation error or a grammatical error (S320).

If the error is a pronunciation error, the processor may provide a first voice response including an expression capable of replacing the erroneous part through the speaker. For example, the user intended the foreign language sentence “Today, I'm gonna finish this patent specification drafting.” , and when it is determined that the word 'specification' is a pronunciation problem, the user device according to the present embodiment sends “Do You meant 'patent specification?'

In one embodiment, a response such as 'Yes, you are right' is received through the voice input device as a user's response to the speaker's voice, that is, an expression that can replace the erroneous part, and the user's According to the response, the user's voice may be recognized as text. Then, the user device according to the present embodiment may provide a third voice response to be described later through a speaker.

Meanwhile, when the error is a grammatical error, a second audio response including a pre-prepared grammatical explanation corresponding to the erroneous part may be provided through the speaker. For example, when a user inputs a foreign language sentence “Today, I gonna finish this patent specification drafting”, it is recognized that there is a grammatical error in the user's voice based on a grammatical error model described later, and accordingly, a second voice response can be provided. In this case, since the user device can sufficiently know the intention of the user's intended sentence, it can provide a second voice response including a third voice response for the case without an error, which will be described later.

In the above example, the processor of the user device sends “Good luck Aaron, go for it! But, you'd better put 'be verb' in front of 'gonna'”.

On the other hand, if there is no error in the user's voice in step S310, the user device according to the present embodiment may provide a third voice response corresponding to the content of the user's language recognized as text and the content of conversation through the speaker. there is. With the recent development of AI technology, the user device according to the present embodiment can conduct a natural language conversation with the user and, as will be described later, can control home appliances wirelessly connected to the Internet of Things (IoT).

3 is a diagram for explaining in detail a process of recognizing a user's voice according to an embodiment of the present invention.

Referring to FIG. 3 , a device 100 performing voice recognition may include a feature extractor 110, a candidate pronunciation string detector 120, and a language selector 140. The feature extractor 110 extracts feature information on the input voice signal. The candidate pronunciation sequence detector 120 detects at least one candidate pronunciation sequence from the extracted feature information. The language selection unit 140 selects the final language recognized as speech based on information on the probability of appearance of each candidate pronunciation sequence. In addition, the language selection unit 140 may select a final voice-recognized language based on the appearance probability information of the detected word by detecting a word corresponding to each candidate pronunciation sequence using the pronunciation dictionary 150 . The appearance probability information of a word means information on a probability that a corresponding word appears in a voice-recognized language when voice recognition is performed. Hereinafter, each component of the device 100 will be described in detail.

When a voice signal is received, the feature extractor 110 may detect only a voice part uttered by an actual speaker and extract information representing characteristics of the voice signal. The information indicating the characteristics of the voice signal may include, for example, information capable of indicating the shape of the mouth or the position of the tongue according to the waveform to which the voice signal belongs. A pronunciation sequence corresponding to the voice signal may be detected based on the feature information extracted by the feature extractor 110 .

The candidate pronunciation sequence detector 120 may detect at least one candidate pronunciation sequence (pronunciation variants) that may be matched with the speech signal by using the characteristic information of the extracted speech signal and the acoustic model 130 . A plurality of candidate pronunciation sequences may be detected according to the voice signal. For example, since pronunciations of 'th' and 'd' are similar, a plurality of candidate pronunciation sequences including similar pronunciations of 'th' and 'd' may be detected for the same voice signal. Candidate pronunciation sequences may be detected in word units, but are not limited thereto, and candidate pronunciation sequences may be detected in various units such as phonemes, phonemes, and syllables.

The acoustic model 130 may include information for detecting a candidate pronunciation sequence from feature information of a speech signal. Also, the acoustic model 130 may be generated through a statistical method from a large amount of voice data. For example, the acoustic model 130 may be generated from unspecified plurality of speech data or speech data collected from a specific speaker. Accordingly, there may be an acoustic model 130 that can be individually applied in voice recognition according to a speaker.

The language selection unit 140 may use the pronunciation dictionary 150 and the grammar model 160 to obtain information on the probability of occurrence of each candidate pronunciation sequence detected by the candidate pronunciation sequence detection unit 120 . Then, the language selection unit 140 selects the final language recognized as speech based on the information about the probability of occurrence of each candidate pronunciation sequence.

Also, the language selector 140 may use the pronunciation dictionary 150 to obtain a word corresponding to each candidate pronunciation sequence. The language selector 140 may obtain an appearance probability value of a word corresponding to each candidate pronunciation sequence using the grammar model 160 . The language selector 140 may finally select a candidate pronunciation sequence having the largest occurrence probability value for a word. A word corresponding to the finally selected pronunciation sequence may be output as a voice-recognized word.

The pronunciation dictionary 150 may include information necessary to obtain a word corresponding to the candidate pronunciation sequence detected by the candidate pronunciation sequence detector 120 . The pronunciation dictionary 150 may be constructed from a sequence of pronunciations obtained according to the phonological change of each word.

Pronunciation of a word is inconsistent because it can be changed according to the words before and after the word or position in a sentence, characteristics of the speaker, and the like. In addition, the appearance probability value means a probability that the current word appears or a probability that the current word appears simultaneously with a specific word. The device 100 may perform speech recognition in consideration of context by using an appearance probability value for a word or pronunciation sequence included in the grammar model 160 .

The device 100 may perform speech recognition by obtaining words in a candidate pronunciation sequence using the pronunciation dictionary 150 and obtaining appearance probability information for words corresponding to the candidate pronunciation sequence using the grammar model 160. . The device 100 is not limited thereto, and instead of obtaining a word corresponding to a pronunciation string through the pronunciation dictionary 150, the device 100 may obtain occurrence probability information for a candidate pronunciation string from the grammar model 160.

For example, in the case of Korean, when the candidate pronunciation sequence detection unit 120 detects a candidate pronunciation sequence 'hakkkyo', the language selector 140 selects the detected candidate pronunciation sequence 'hakkkyo' and As a corresponding word, the word 'school' can be obtained using the pronunciation dictionary 150 . As another example, in the case of English, when the candidate pronunciation sequence detection unit 120 detects a candidate pronunciation sequence 'skul', the language selector 140 selects a word corresponding to the detected candidate pronunciation sequence 'skul', and pronounces the word. The word 'school' can be obtained using the dictionary 150 . The language selector 140 may finally select a word corresponding to the voice signal based on the appearance probability information for the word 'school' or 'school' and output the selected word.

The grammar model 160 may include occurrence probability information for words. Appearance probability information may exist for each word. The device 100 may obtain information about appearance probabilities of words included in each candidate pronunciation sequence from the grammar model 160 .

For example, when the word A appears before the current word B appears, the grammar model 160 may include P(B|A), which is information about a probability that the current word B appears. In other words, the appearance probability information of word B, P(B|A), may be conditional on the appearance of word A before the appearance of word B. As another example, the grammar model 160 may include words A and C before the appearance of word B, that is, P(B|A C) conditional on the appearance of a plurality of words. In other words, P(B|A C) can be conditional on the occurrence of both words A and C before the occurrence of word B. As another example, the grammar model 160 may include occurrence probability information P(B) for word B, which is not a conditional probability. P(B) means the probability that word B appears during speech recognition.

The device 100 may finally determine the voice-recognized word based on the appearance probability information of a word corresponding to each candidate pronunciation sequence in the language selector 140 using the grammar model 160 . That is, the device 100 may finally determine a word having the highest occurrence probability information as a speech recognition word. The language selector 140 may output the voice-recognized words as text.

Meanwhile, the language selector 140 according to an embodiment of the present invention may use the grammatical error model 170 to determine whether there are grammatical errors in the user language recognized using the grammatical model 160 .

That is, the processor according to the present embodiment includes not only the first grammar model 160 used to recognize the user's voice as a text, but also the user's voice as a text, and then the text that exists in the recognized language. It can be seen as additionally using the second grammar model 170 used to recognize grammatical errors.

The second grammar model 170 may be configured to update each grammatical error and a grammatical explanation corresponding to each grammatical error in pairs based on learning based on data of the web server. For example, it may be configured to update grammatical errors (eg, tense, singular/plural expressions, syntax, etc.) frequently generated by users accumulated through online education services, and descriptions of these grammatical errors in pairs. .

Of course, there may not be a one-to-one relationship between a grammatical error and an explanation of the grammatical error. For example, in the example of the second voice response of FIG. 2, 'grammar description of the phrase 'be going to' may be paired, and 'grammar description of the phrase 'be going to' may be connected in various ways as well as the example of FIG. 2. It can be linked to grammatical errors in sentences.

Summarizing the contents of FIG. 3, the processor according to the present embodiment extracts a feature from the user's voice (110), detects a candidate pronunciation sequence corresponding to the feature (120), An operation of recognizing the user's voice as a text may be performed by selecting a language that can be combined with a predetermined probability or higher using the 1 grammar model 160 (140).

Here, if a language that can be combined using the first grammar model 160 is not selected from among the candidate pronunciation sequences, it may be determined that the error is a pronunciation error.

In contrast, the processor selects a combinable language from among the candidate pronunciation sequences using the first grammar model 160, and the difference between a character recognized with the highest probability and the selected language is determined by using the first grammar model 160. If it is greater than or equal to the standard, it may be determined that the error is a pronunciation error.

In the above description, it is assumed that the voice spoken by the user is in a foreign language, but it is not necessary to be limited thereto.

For example, the processor according to an embodiment of the present invention may distinguish between a native language voice and a foreign language voice among the user voices. When the processor determines that the user's voice is the foreign language voice, the processor may be configured to perform the above-described operation with respect to FIGS. 2 and 3 .

Meanwhile, in one embodiment of the present invention, the user's voice may include a command for changing the operation mode of the user device. According to the command, the processor performs (1) a question-and-answer mode in which the user's voice is processed in sentence-by-sentence units and the voice response is provided to the speaker, or (2) the user's voice is continuously recorded for a predetermined time and then the voice is It can optionally operate in a palm print mode providing a response to the speaker.

(1) Q&A mode

Unlike the foregoing, the user may explain the corresponding mode for each foreign language on the user device, such as a mobile phone or a laptop computer, in advance. In general, it is difficult to speak long sentences at the beginner level of the foreign language, and a one-sentence question is provided through the speaker of the user device according to the present embodiment according to the situation, and the user's one-sentence answer is received and the above-mentioned foreign language We can provide educational services.

In the question-and-answer mode, it is desirable to increase the probability of a user's pronunciation error among grammatical errors and pronunciation errors, but it is not necessary to be limited thereto.

(2) Long Record Mode

Unlike the above, it is the same that the user can explain the corresponding mode for each foreign language on the user device such as a mobile phone or a laptop computer in advance. In general, at the advanced level of the foreign language, for example, it may be effective for education to receive about 1 minute of user speech as if writing a diary and to provide a general conversation (eg, the above-described third voice response).

While having a natural language conversation with the user in a foreign language, grammatical improvement in the user's speech may be provided at a predetermined frequency. Of course, the grammar point frequency may be adjusted through a conversation with a user device (eg, an AI speaker) according to the present embodiment. For example, you can adjust the frequency by saying “Bob (speaker character name), reduce the frequency of error correction”.

Meanwhile, in one embodiment of the present invention, the following education mode may be provided in addition to the above-described modes.

(3) expression conversion

Various expressions can be informed by notifying a more formal expression or a more casual expression for a user voice expression having the same meaning.

For example, when the user says 'I'm going to finish this patent specification drafting', the user device according to the present embodiment converts it into a more casual expression according to a preset mode or a separate inquiry from the user. It can tell you 'I'm gonna finish this patent specification drafting', or vice versa by changing your speech to something more formal.

As a similar function, it can suggest exemplary American English pronunciation and British English pronunciation for the user's English expression.

(4) Lecture mode

A lecture may be provided to a user through a user device according to an embodiment of the present invention represented by an AI speaker. During a lecture, there may be interactions that require a user's voice response.

If visual display is required during audio lectures, as will be described later, visual data can be displayed through a TV, monitor, etc. connected through a wireless communication method such as an IoT method. If it is difficult for the user to access the monitor/TV at the present time regarding the display of such visual data, it may be commanded to skip it and proceed.

In addition, in this lecture mode or separately, the user device according to an embodiment of the present invention may provide a user's education scheduling service, and may change and suggest educational contents provided according to the achievement level of each educational course.

(5) advance mode

A user may frequently inquire about the meaning of a word or phrase in a foreign language, and in response to this, the user device according to the present embodiment can notify the user of the meaning/accurate pronunciation of the word or phrase by being connected to the web.

4 is a diagram for explaining a concept of controlling an in-house device through a user device according to an embodiment of the present invention.

As shown in FIG. 4, the user device 10 may be wirelessly connected to one or more in-home devices 201, 202, 203, 204, 205, and 206 in an IoT manner, and The device may include a mobile phone 201 shown in FIG. 4, another AI speaker 202, a washing machine 203, a robot vacuum cleaner 204, an air conditioner 205, a refrigerator 206, and the like, and need to be limited thereto. There is no

The user device 10 according to the present embodiment may control the in-house devices 201 , 202 , 203 , 204 , 205 , and 206 through at least one of a native language voice and a foreign language voice among the user voices. In addition, in the process of controlling the in-house devices 201, 202, 203, 204, 205, and 206 through the user's foreign language voice, the foreign language education described above with reference to FIGS. 2 and 3 may be provided.

5 is a diagram for explaining a method of providing foreign language education according to an embodiment of the present invention.

In the description up to FIG. 4, it is assumed that the user device is an AI speaker user device, but the user device according to another embodiment of the present invention allows the user to use a laptop computer, a desktop computer, or other type of display device having a display unit 400 such as a monitor. It may also be provided through a user device.

For example, in the method of providing foreign language education through online access of a user device according to the present embodiment, a user's foreign language sentence voice pronunciation input through a microphone of the user device is recognized, converted into text, and converted into text. Errors in converted sentences can be judged.

If the erroneous part is a grammatical error, a prepared grammatical explanation corresponding to the erroneous part is shown in FIG. can be displayed as shown in That is, in the example shown in FIG. 5, since there is a grammatical error in which the 'be verb' is missing between 'I' and 'gonna', the 'I gonna' part 410 is highlighted as an error part, and thus In response, “Good luck Aaron, go for it! But, you'd better put 'be verb' in front of 'gonna'.

In addition, in the above example, “View grammar explanation 420 for the [be going to] phrase” is additionally displayed, and when the corresponding view explanation 420 is clicked, the corresponding specific grammar explanation or grammar lecture can be taken. can do.

On the other hand, if the part with the error is a pronunciation error, an expression that can replace the part with the error is displayed on the display device 400 of the user device together with the highlight of the part with the error to recommend it. , a recommended pronunciation of an exemplary sentence including the recommended expression may be presented through a speaker of the user device. Through this, it is possible to provide education on exemplary pronunciation to the user through recommended pronunciation. Such exemplary pronunciation can be provided as a more fluent pronunciation to the user by considering the liaison/silent relationship between words.

On the other hand, recommending an expression that can replace the erroneous part recommends a plurality of expressions and displays them on the display device 400 of the user device, and recommends an expression selected through the input device of the user device. expression, storing the mapping relationship between the erroneous part of the user's foreign language sentence voice pronunciation and the determined recommended expression in the database of the server providing foreign language education, and learning the server based on artificial intelligence. may also include As such an interaction with the user, the user device according to the present embodiment can evolve itself to suit a specific user.

On the other hand, the corresponding data of the erroneous part and the prepared grammatical explanation is increased in the database of the server providing the foreign language education by artificial intelligence, and when it is determined that the erroneous part is a grammatical error, , it may be a case where there is a pre-prepared grammar explanation corresponding to the part with the error.

Various embodiments of the present invention as described above are provided for a specific understanding of the present invention, and the present invention is not limited to the specific examples of the above description, and user devices having artificial intelligence functions for foreign language education provided throughout the specification, and It should be judged based on the foreign language teaching method.

As described above, the user device and foreign language education method equipped with an artificial intelligence function for foreign language education according to the present invention can be widely used as a new education method in the non-face-to-face era.

Claims (8)

In a user device having an artificial intelligence function for foreign language education,
a voice input device that receives a user's voice;
a processor for recognizing and processing the user's voice as text; and
A speaker for providing a voice response to a user according to the processing of the processor;
the processor,
(a) determining whether the user's voice error is a pronunciation error or a grammatical error;
(b) when the error is a pronunciation error, providing a first audio response including an expression capable of replacing the erroneous part through the speaker;
(c) if the error is a grammatical error, providing a second audio response including a pre-prepared grammatical explanation corresponding to the erroneous part through the speaker;
The user's voice includes a command for changing the operation mode of the user device;
According to the command, the processor performs (1) a question-and-answer mode in which the user's voice is processed in sentence-by-sentence units and the voice response is provided to the speaker, or (2) the user's voice is continuously recorded for a predetermined time and then the voice is and selectively operating in a palm print mode providing a response to the speaker. According to claim 1,
the processor,
A first grammar model used to recognize the user's voice as text; and
After recognizing the user's voice as text, the user device uses a second grammar model used to recognize grammatical errors present in the text-recognized language. According to claim 2,
wherein the second grammar model is configured to update each grammatical error and a grammatical description corresponding to each grammatical error in pairs on a learning basis based on data of a web server. According to claim 1,
the processor,
If there is no error, the user device provides a third voice response corresponding to the content of the user's language recognized as text and the content of conversation through the speaker. delete delete delete delete

Images