KR102631382B1 - System for providing korean education service for korean native speaker - Google Patents

Abstract

A Korean language education service provision system is provided for users whose first language is Korean, a user terminal that inputs user information corresponding to demographic data including age, gender, region, and occupation, and an instructor terminal that conducts real-time lectures. , and a construction unit that builds a database by mapping cases where the standard pronunciation and actual pronunciation are different depending on demographic data including age, gender, region, and occupation, to demographic data, and user information input from the user terminal. As an input, an extraction unit extracts cases where the standard pronunciation is different from the actual pronunciation. After the user terminal utters a word or sentence containing a standard pronunciation different from the actual pronunciation, the uttered voice data and the reference data are used. It includes an education service providing server including a transmission unit that compares and transmits the difference data to the user terminal, and a lecture preparation unit that transmits the difference data to the instructor terminal before starting a real-time lecture on the instructor terminal.

Description

Korean education service provision system for users whose first language is Korean {SYSTEM FOR PROVIDING KOREAN EDUCATION SERVICE FOR KOREAN NATIVE SPEAKER}

The present invention relates to a Korean language education service provision system for users whose first language is Korean, and provides a platform for conducting speech education through an announcer who speaks standard pronunciation for users whose native language is Korean.

As the concept of communication ability emerged, Korean language education began to focus on actually using the language learned in education to acquire language knowledge. Recently, as the number of learners learning Korean has increased, speaking education research has also grown rapidly, and speaking teaching research centered on instructors who teach Korean and learner-centered speaking learning research are also increasing. In addition, in research on speaking education focused on academic purposes and Chinese-speaking learners, Korean speaking education for various purposes and language areas other than the Chinese-speaking area is required, such as Korean speech education for Koreans, that is, for general, employment, and tourism purposes. And, although speaking materials for intermediate and beginner levels are noticeably active, speaking materials and content for advanced level learners are relatively lacking.

At this time, a method was researched and developed to help correct pronunciation by comparing the similarity with the announcer's standard pronunciation or to improve the speech and pronunciation of the speech impaired. In this regard, the prior art, Korean Patent No. 10-1779361 (2017) Announced on September 18) and Korean Patent Publication No. 2021-0051278 (published on May 10, 2021), the pronunciation accuracy, speed, size, and intonation between the user's voice utterance received from the user terminal and the previously stored announcer's voice utterance are discussed. The similarity is analyzed using a pronunciation analysis algorithm, the user's pronunciation is different from the announcer's or the pronunciation that the user finds difficult is classified, and the configuration helps to practice pronunciation of the classified pronunciation, and is based on voice recognition and immersive virtual reality. Each configuration is disclosed to perform language training and to determine the presence of articulation disorder and the degree of articulation disorder based on articulation accuracy data preset according to consonant accuracy.

However, in the former case, since the pitch and frequency of the voice are compared when analyzing the similarity of pronunciation, even if the difference is mechanically derived, the average user has no idea how to practice pronunciation to correct this. There are many cases. In the latter case, the presence of articulation disorders is determined and a language training program is provided, but a structure for correcting differences or incorrect pronunciation is not disclosed. Speakers whose native language (L1) is Korean do not just use the standard pronunciation in everyday life, but sometimes use pronunciations that differ from the standard pronunciation. There are even cases where it sounds unnatural when pronounced with a standard pronunciation, and when pronounced with a realistic pronunciation, it feels natural and you can understand the speaker's intention. Accordingly, a platform for users whose native language (L1) is Korean, that allows them to pronounce the difference between the standard pronunciation and the actual pronunciation in a standard pronunciation like an announcer, while connecting them with an expert qualified to speak accurate Korean pronunciation. Research and development are required.

One embodiment of the present invention allows users whose native language is Korean to learn the difference between standard pronunciation and actual pronunciation when they want to receive speech education for purposes including employment, education, and speech training, and to enable the population of users whose native language is Korean. Based on statistical data, frequently used realistic pronunciations by age group, age group, and job group are constructed as big data to intensively learn standard pronunciation corresponding to realistic pronunciation, and the pronunciation of vowels and consonants is used to use accurate Korean pronunciation. It is possible to provide a Korean language education service provision system for users whose first language is Korean, which can provide a speech platform so that they can learn from an announcer who is a qualified expert.

However, the technical challenge that this embodiment aims to achieve is not limited to the technical challenges described above, and other technical challenges may exist.

As a technical means for achieving the above technical problem, one embodiment of the present invention inputs user information corresponding to demographic data including age, gender, region, and occupation, and outputs at least one instructor profile. A user terminal that selects an instructor, enters the class in real time and conducts the class, is registered as an instructor by verifying at least one qualification, and conducts a real-time lecture based on the selection of the user terminal. The instructor terminal, age, and gender , a construction unit that builds a database by mapping cases where the standard pronunciation and actual pronunciation are different according to demographic data including region and occupation, to demographic data, and standard pronunciation by inputting user information entered from the user terminal. An extraction unit that extracts cases where the actual pronunciation is different from the actual pronunciation, causes the user terminal to vocalize a word or sentence containing a standard pronunciation that is different from the actual pronunciation, and then compares the vocalized speech data with the reference data to generate the difference data. It includes an education service providing server including a transmission unit that transmits to the user terminal and a lecture preparation unit that transmits difference data to the instructor terminal before starting a real-time lecture on the instructor terminal.

The above-described means for solving the problem are merely illustrative and should not be construed as intended to limit the present invention. In addition to the exemplary embodiments described above, additional embodiments may be present in the drawings and detailed description of the invention.

According to one of the means for solving the problem of the present invention described above, when a user whose native language is Korean wants to receive speech education for purposes including employment, education, and speech training, he or she learns the difference between the standard pronunciation and the actual pronunciation, and learns the Korean language. Based on the demographic data of users whose native language is, frequently used realistic pronunciations by age group, age group, and occupation are constructed as big data to intensively learn standard pronunciation corresponding to realistic pronunciation, and vocalization of vowels and consonants. We can provide a speech platform so that people can learn from announcers who are qualified experts who can use accurate Korean pronunciation.

However, the effects that can be obtained from the present invention are not limited to the effects described above, and other effects may exist.

1 is a diagram illustrating a system for providing Korean language education services for users whose first language is Korean according to an embodiment of the present invention.
FIG. 2 is a block diagram illustrating an education service providing server included in the system of FIG. 1.
Figures 3 and 4 are diagrams for explaining an embodiment in which a Korean education service for users whose first language is Korean is implemented according to an embodiment of the present invention.
Figure 5 is an operation flowchart illustrating a method of providing a Korean language education service for a user whose first language is Korean according to an embodiment of the present invention.

Below, with reference to the attached drawings, embodiments of the present invention will be described in detail so that those skilled in the art can easily implement the present invention. However, the present invention may be implemented in many different forms and is not limited to the embodiments described herein. In order to clearly explain the present invention in the drawings, parts that are not related to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

Throughout the specification of the present invention, when a part is said to be “connected” to another part, this means not only “directly connected” but also “electrically connected” or “indirectly connected” with another element in between. Also includes cases where it is connected to.

Throughout the specification of the present invention, when a member is said to be located “on”, “in the upper part”, “at the top”, “below”, “at the bottom”, or “at the bottom” of another member, this means that a member This includes not only cases where a member is in contact with a member, but also cases where another member exists between two members.

Throughout the specification of the present invention, when a part is said to “include” a certain component, this means that it may further include other components rather than excluding other components unless specifically stated to the contrary.

Throughout the specification of the present invention, some of the operations or functions described as being performed by a terminal, apparatus, or device may instead be performed on a server connected to the terminal, apparatus, or device. Likewise, some of the operations or functions described as being performed by the server may also be performed in a terminal, apparatus, or device connected to the server.

Throughout the specification of the present invention, the term 'at least one' may be defined as a term including singular and plural, and even if the term 'at least one' does not exist, each component may exist in singular or plural, and singular Or, it is obvious that it can mean revenge. Additionally, whether each component is provided in singular or plural form may vary depending on the embodiment.

This specification is a successor patent to Korea Patent No. 10-2355960 (announced on February 8, 2022), which is the applicant's pre-registered patent. For structures and contents not described in this specification, refer to the applicant's pre-registered patent. It is decided to do so and will not be described in detail in this specification.

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

1 is a diagram illustrating a system for providing Korean language education services for users whose first language is Korean according to an embodiment of the present invention. Referring to FIG. 1, the Korean education service providing system 1 for a user whose first language is Korean includes at least one user terminal 100, an education service providing server 300, and at least one instructor terminal 400. may include. However, since the Korean education service providing system 1 for users whose first language shown in FIG. 1 is Korean is only an embodiment of the present invention, the present invention is not limitedly interpreted through FIG. 1.

At this time, each component of FIG. 1 is generally connected through a network (Network, 200). For example, as shown in FIG. 1, at least one user terminal 100 may be connected to the education service providing server 300 through the network 200. In addition, the education service providing server 300 may be connected to at least one user terminal 100 and at least one instructor terminal 400 through the network 200. Additionally, at least one instructor terminal 400 may be connected to the education service providing server 300 through the network 200.

Here, the network refers to a connection structure that allows information exchange between each node, such as a plurality of terminals and servers. Examples of such networks include a local area network (LAN) and a wide area network (WAN). Wide Area Network, Internet (WWW: World Wide Web), wired and wireless data communication network, telephone network, wired and wireless television communication network, etc. Examples of wireless data communication networks include 3G, 4G, 5G, 3rd Generation Partnership Project (3GPP), 5th Generation Partnership Project (5GPP), Long Term Evolution (LTE), World Interoperability for Microwave Access (WIMAX), and Wi-Fi. , Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), RF (Radio Frequency), Bluetooth network, NFC ( It includes, but is not limited to, Near-Field Communication (Near-Field Communication) network, satellite broadcasting network, analog broadcasting network, and DMB (Digital Multimedia Broadcasting) network.

In the following, the term at least one is defined as a term including singular and plural, and even if the term at least one does not exist, each component may exist in singular or plural, and may mean singular or plural. This should be self-explanatory. In addition, whether each component is provided in singular or plural form may be changed depending on the embodiment.

At least one user terminal 100 may be a terminal of a user who wants to learn Korean using a web page, app page, program, or application related to a Korean education service for users whose first language is Korean. At this time, the user may be a Korean whose native language is Korean and who wants to correct or learn pronunciation. At this time, the user terminal 100 may be a terminal that connects to the education service providing server 300, prints the instructor's profile, selects the instructor, and participates in the class. Here, the user terminal 100 may be connected one-to-one with the instructor, but depending on the embodiment, it may be a terminal that can participate in various class types such as one-to-many, many-to-many, and many-to-one.

Here, at least one user terminal 100 may be implemented as a computer capable of accessing a remote server or terminal through a network. Here, the computer may include, for example, a laptop equipped with a navigation system and a web browser, a desktop, a laptop, etc. At this time, at least one user terminal 100 may be implemented as a terminal capable of accessing a remote server or terminal through a network. At least one user terminal 100 is, for example, a wireless communication device that guarantees portability and mobility, and includes navigation, personal communication system (PCS), global system for mobile communications (GSM), personal digital cellular (PDC), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) ) It may include all types of handheld-based wireless communication devices such as terminals, smartphones, smartpads, and tablet PCs.

The education service providing server 300 may be a server that provides a Korean education service web page, app page, program, or application for users whose first language is Korean. And, when registering a profile from the instructor terminal 400 in order to verify the qualifications of at least one instructor terminal 400, the education service providing server 300 determines whether the person has actually received announcing training, is a broadcaster, or is an announcer. It may be a server that verifies and registers, etc. Additionally, the education service providing server 300 may be a server that allows the user terminal 100 to participate in a real-time class when the verified instructor's profile is uploaded and the user terminal 100 selects the instructor. In addition, the education service providing server 300 collects the ratings and reviews of each instructor from the user terminal 100 and accumulates them into a log, divides the speech, which is the voice utterance of the user terminal 100, into levels, tests it, and conducts a speech test. It may be a server that issues certificates issued by the association.

Here, the education service providing server 300 may be implemented as a computer that can access a remote server or terminal through a network. Here, the computer may include, for example, a laptop equipped with a navigation system and a web browser, a desktop, a laptop, etc.

At least one instructor terminal 400 may be an instructor terminal that uses a web page, app page, program, or application related to a Korean language education service for users whose first language is Korean. At this time, the instructor terminal 400 may be a terminal that transmits a profile to the education service provision server 300 to prove one's career or qualifications and allows viewing of each institution's qualifications in the blockchain, which will be described later. there is. If the instructor terminal 400 is registered after completing qualification verification, it may be a terminal connected to the user terminal 100 and a real-time lecture. Additionally, the instructor terminal 400 may be a terminal that receives reviews and ratings of each user terminal 100 as feedback.

Here, at least one instructor terminal 400 may be implemented as a computer capable of accessing a remote server or terminal through a network. Here, the computer may include, for example, a laptop equipped with a navigation system and a web browser, a desktop, a laptop, etc. At this time, at least one instructor terminal 400 may be implemented as a terminal capable of accessing a remote server or terminal through a network. At least one instructor terminal 400 is, for example, a wireless communication device that ensures portability and mobility, and includes navigation, Personal Communication System (PCS), Global System for Mobile communications (GSM), Personal Digital Cellular (PDC), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) ) It may include all types of handheld-based wireless communication devices such as terminals, smartphones, smartpads, and tablet PCs.

Figure 2 is a block diagram for explaining the education service providing server included in the system of Figure 1, and Figures 3 and 4 show a Korean education service for a user whose first language is Korean according to an embodiment of the present invention. This is a drawing to explain an implemented embodiment.

Referring to FIG. 2, the education service providing server 300 includes a construction unit 310, an extraction unit 320, a transmission unit 330, a lecture preparation unit 340, a registration unit 350, and a storage unit 360. , it may include a pronunciation visualization unit 370, a vowel analysis unit 380, a speech unit 390, an evaluation unit 391, a big data visualization unit 393, and a recognition performance improvement unit 395.

The education service providing server 300 according to an embodiment of the present invention or another server (not shown) operating in conjunction with at least one user terminal 100 and at least one instructor terminal 400 whose first language is Korean. When transmitting a Korean language education service application, program, app page, web page, etc. for a user, at least one user terminal 100 and at least one instructor terminal 400 are provided for a user whose first language is Korean. You can install or open Korean education service applications, programs, app pages, web pages, etc. Additionally, a service program may be run on at least one user terminal 100 and at least one instructor terminal 400 using a script executed in a web browser. Here, a web browser is a program that allows the use of web (WWW: World Wide Web) services and refers to a program that receives and displays hypertext written in HTML (Hyper Text Mark-up Language), for example, Netscape. , Explorer, Chrome, etc. Additionally, an application refers to an application on a terminal and includes, for example, an app running on a mobile terminal (smartphone).

Referring to FIG. 2, the construction unit 310 builds a database by mapping cases where the standard pronunciation and the actual pronunciation are different according to demographic data including age, gender, region, and occupation, to the demographic data. can do. Cases where the standard pronunciation is different from the actual pronunciation may include cases where the standard pronunciation is not pronounced in reality. Speakers whose native language is Korean do not just use the standard pronunciation in their daily lives, but sometimes use pronunciations that are different from the standard pronunciation. There are even cases where it sounds unnatural when pronounced with a standard pronunciation, and when pronounced with a realistic pronunciation, it feels natural and you can understand the speaker's intention.

[Table 1]

In the above utterance, if you pronounce [barayo], [jolji], [heulgi], and [gimbap] according to the standard pronunciation, most native Korean speakers will feel awkward and may even not understand the meaning. Rather, it is naturally accepted to pronounce it as non-standard pronunciations such as [Baraeyo], [Jjolji], and [Heugi], or as [Gimbbap] among the plural standard pronunciations. In this way, there is a realistic pronunciation that is frequently realized in everyday spoken language and is naturally accepted by native Korean speakers.

<Real pronunciation>

First, looking at the real pronunciation from the perspective of Korean phonology, although the real pronunciation is different from the standard pronunciation, it can be understood as the pronunciation that many people utter in actual everyday spoken language. In addition, it is a pronunciation that Korean speakers actually use a lot, and it is a pronunciation that needs to be the subject of pronunciation education for Korean speakers to communicate. If real pronunciation is defined as a pronunciation that is different from the standard pronunciation and is realized frequently in spoken language, the dialectal pronunciation of each region can also be included in the real pronunciation.

<Types of realistic pronunciation>

The types of actual pronunciation can be categorized into universal phonological phenomena, changes in noun final consonants, utilization, and individual words and morphemes, as shown in Table 2 below. Table 3 summarizes the actual pronunciation by dividing it into replacement, omission, addition, and abbreviation. Looking at the degree of occurrence, the proportion of real pronunciations corresponding to replacements is high, and most of the replacements correspond to vowel replacements. Among vowel exchanges, the degree of realization of o → u, uh → eu, ah → uh, and e →, corresponding to vowel rise, is high, and a significant portion of actual pronunciation is related to vowel rise. In particular, pronunciations in which the o of particles or endings are realized as u account for a high proportion of all real pronunciations, and the proportion of o → right substitutions among real pronunciations is overwhelming.

[Table 2]

[Table 3]

The types of realistic pronunciation that are the subject of education can be summarized as in Table 4.

[Table 4]

Types such as semi-vowel dropping, articulatory position assimilation, vowel reduction, and ㅎ elimination are realized for economy of pronunciation in everyday fast speech, and include initial consonants, vowel replacement due to vowel elevation, ㄹ addition, umlaut, and end-of-word consonants. Replacement, etc. is a pronunciation corresponding to the new form in the process of language change where new and old forms coexist. This pronunciation is related to phonetic changes in Korean. In addition, there are cases where the standard pronunciation and the actual pronunciation are different depending on the user's information, such as dialect, accent, intonation, etc. for each region. Accordingly, cases where the actual pronunciation and standard pronunciation differ depending on the user's region, gender, occupation, age, etc. are classified and stored, and a database is created so that users can learn in advance about cases where the standard pronunciation is different from the actual pronunciation before entering the class. Build it. The extraction unit 320 can extract cases where the standard pronunciation and the actual pronunciation are different by using user information input from the user terminal 100. The user terminal 100 may input user information corresponding to demographic data. Depending on the region where each user was born, the environment in which they lived, and the group they currently belong to, the actual pronunciation may differ from the standard pronunciation. At this time, after extracting this data, first learn which pronunciation is the standard pronunciation and which pronunciation you misunderstood as the standard pronunciation before meeting the instructor. Of course, it is possible to provide the entire standard pronunciation equally to all users, but if there is a pronunciation that is frequently used incorrectly, it is important to focus on correcting it, so practice first with the pronunciation that is frequently and frequently incorrectly made in the group.

The transmission unit 330 causes the user terminal 100 to vocalize a word or sentence containing a standard pronunciation that is different from the actual pronunciation, and then compares the vocalized voice data with the reference data to obtain difference data. Can be transmitted to the user terminal 100.

The lecture preparation unit 340 may transmit difference data to the instructor terminal 400 before starting a real-time lecture on the instructor terminal 400. The user terminal 100 can output at least one instructor profile, select one instructor, enter the class in real time, and then proceed with the class. The instructor terminal 400 is registered as an instructor by verifying at least one qualification condition and can conduct a real-time lecture by selecting the user terminal 100.

In other words, the lecture preparation unit 340 controls the operation of the user terminal 100 so that at least one instructor profile is output from the user terminal 100, and then the user displays at least one instructor profile on the user terminal 100. If it is detected that the first instructor profile has been selected as one of the instructor profiles, the user then enters the first class (class, class content) of the professor (first professor) corresponding to the first instructor profile. It can be controlled so that the first class can be listened to in real time.

At this time, when the user selects the first instructor profile, the lecture preparation unit 340 controls the user terminal 100 to proceed with the preset class fee payment process, and the user progresses through the preset class fee payment process. Only when the terminal 100 detects that the preset class fee for admission to the first class has been paid, it can be controlled to allow the user to enter the first class.

Here, the preset class fee payment process includes i) a first process of determining whether there is at least one user-owned payment card pre-registered in the user terminal 100, ii) a plurality of payments in the first process. If it is determined that the payment card is pre-registered in the user terminal 100, a method of obtaining card benefit information for each of the plurality of pre-registered payment cards is obtained from the website of the card company corresponding to each of the plurality of payment cards. Process 2, iii) Based on the card benefit information of each of the plurality of payment cards obtained in the second process and the attribute information of the class providing site where the first class (class content) is provided, the plurality of payment cards A third process of deriving the expected discount amount expected upon payment of the preset class fee corresponding to the first class, and iv) at least one of the expected discount amounts for each payment card derived in the third process. If the expected discount amount is different, the payment card with the highest expected discount amount among the plurality of payment cards is selected (primary selection) as the payment card, and the selected payment card (primary selection) is selected as the payment card. It may include a fourth process of actually making payment for the preset class fee using the payment card.

Here, the preset class fee corresponding to the first class may mean information about the class fee (amount) that the user must pay to listen to the first class.

Additionally, the class provision site refers to a site where professors' classes (class content) are provided, and may refer to a site provided by the education service provision server 300.

Additionally, in the third process, the attribute information of the class providing site may include, for example, if a card company affiliated with the class providing site exists, discount rate information for each affiliated card company. In addition, the expected discount amount considered in the third process refers to information about the amount expected to be discounted when paying with a payment card. For example, if 1,000 won is discounted when paying with the first payment card, It can mean 1,000 won.

In addition, the preset class fee payment process is v) For example, when performing the fourth process, at least one expected discount amount among the expected discount amounts for a plurality of payment cards is different, but the expected discount amount among the plurality of payment cards is different. Since there are at least two payment cards with the highest amount, the one payment card with the highest expected discount amount among the plurality of payment cards is selected as the payment card (i.e., selected first) ), if it is impossible to do so, then, based on the current progress of the preset class fee payment process, from the website of the card company corresponding to each of at least two payment cards with the highest expected discount amount. , Obtain information on the insufficient card performance amount at the current time for each of the at least two or more payment cards, and then obtain information on the insufficient card performance amount for each of the at least two or more payment cards. The payment card with the lowest card performance amount information is selected as the payment target card (secondary selection), and the actual payment for the preset class fee is made with the payment target card (secondary selection). It may include a fifth process.

In addition, the preset class fee payment process is similar to the fifth process. vi) For example, when performing the fourth process, the expected discount amount for each of the plurality of payment cards is the same, so that among the plurality of payment cards, If it is impossible to select the payment card with the highest expected discount amount as the payment card (i.e., select first), the payment process will be based on the current time when the preset class fee payment process is in progress. , obtain information on the insufficient card performance amount at the current time for each of the plurality of payment cards from the website of the card company corresponding to each of the plurality of payment cards, and then use the plurality of payment cards Among each of the information on the insufficient card performance amount, the payment card with the lowest card performance amount information is selected (secondary selection) as the payment target card, and the selected payment card (secondarily selected) is the payment card. It may include a sixth process of actually making payment for the preset class fee.

For example, the above-mentioned preset class fee payment process may be a process that is automatically performed only when the user enters a consent check in the consent box for automatically performing the preset class fee payment process on the user terminal 100.

The lecture preparation unit 340 ensures that the preset class fee for the first class is paid through the above-described preset class fee payment process, so that among the plurality of payment cards held by the user, the primary expected discount amount is the highest. Payment is to be made by card first, but if it is impossible to identify (select) the card with the highest expected discount amount, payment can be made first to the card with the lowest card performance amount information. This allows card payments to be made automatically so that users can enjoy more benefits when holding multiple cards, thereby ensuring reasonable consumption.

The register 350 can establish an instructor pool by setting at least one qualification condition and registering at least one instructor terminal 400 by verifying at least one qualification condition. At least one qualification may be one of those who have received announcing training, a broadcaster and announcer, a voice actor, or a singer. In addition, instructors registered in the constructed instructor pool may include instructors belonging to a specific company (entertainment, company, etc.) and individual freelance instructors who are not affiliated with a specific company and work individually.

For example, when the register 350 detects that a request to provide personal class information has been made from the user terminal 100 after the instructor pool is established, it extracts the information of individual freelance instructors from among the plurality of instructors registered in the instructor pool and creates a list. It can be generated and provided to the user terminal 100.

At this time, when providing the list, the register 350 provides the remaining information (e.g., educational background, experience, educational activity area of the instructor, and Only the instructor's needs information) is provided primarily, and then, based on the remaining information excluding personal information, the user selects the first individual freelance instructor among the plurality of individual freelance instructors in the list and provides 1:1 training. When a request for personal training is made, a notification message related to the 1:1 personal training request is sent to the first instructor terminal, which is the instructor terminal of the first personal freelance instructor, and then approval is received as a response to the notification message from the first instructor terminal. If it is detected that an acceptance response has been made among the responses and rejection responses, the personal information of the first individual freelance instructor corresponding to the first instructor terminal that responded yes may be provided to the user terminal 100 so that the user can confirm it. . In addition, when the register 350 detects that the consent response has been made, the user's personal information (for example, name, age, and mobile phone number) is provided to the first instructor terminal so that the first instructor of the first instructor terminal can confirm it. You can do it.

In other words, the register 350 allows the user to check only the remaining information of the individual freelance instructors in the list when the user wants to receive 1:1 personal education (Korean language training) from a personal freelance instructor, and allows the user to check the remaining information of the individual freelance instructors. Personal information can be shared between the instructor and the user only when a 1:1 personal education request is made and the instructor responds with approval. In providing Korean education services, this register 350 reduces unnecessary exposure of personal information between instructors and users and ensures that personal information is exchanged only when an approval response is made, thereby raising the risk of misuse of personal information such as prank calls. can be lowered, and measures such as reporting can be taken quickly when damage occurs, and when a user wants 1:1 personal education, the instructor is selected based on ability (experience, academic background, etc.) rather than selecting the instructor based on person or age. can be selected.

In addition, the register 350, for example, when constructing an instructor pool, when a request for instructor registration is made from a first instructor terminal that is one of at least one instructor terminal, at least one qualification is registered from the first instructor terminal. Evidential data for verification of conditions is uploaded, and through analysis of the uploaded evidential data, it is determined whether the first instructor, who is the instructor of the first instructor terminal, satisfies at least one qualification condition (i.e., whether he or she has received announcing training). , whether the voice actor is correct, etc.) can be determined. Afterwards, if the register 350 determines that the first instructor satisfies at least one qualification condition, it can build an instructor pool by registering the instructor information of the first instructor in an instructor DB (not shown).

Here, the supporting materials may include, for example, an announcing training certificate or, in the case of a singer, album information, etc. In addition, instructor information may include the instructor's name, age, cell phone number, educational background, career, instructor's educational activity area, and instructor's needs information.

At this time, once the instructor information of the first instructor is registered in the instructor DB (not shown), the register 350 registers the schools (elementary school, middle school, high school) from which the first instructor graduated based on the academic background information in the instructor information. , and university) request and receive life record information about the first instructor targeting corresponding school terminals, and also collect information on posts uploaded by the first instructor by accessing the first instructor's SNS account. , By analyzing the plurality of life record information about the first instructor and the collected posting information received in relation to the first instructor, it can be determined whether the first instructor has a history of theft, history of school violence, and a tendency toward violence. there is.

Here, the existence of a history of theft and a history of school violence can be determined through analysis (character analysis, semantic analysis, etc.) of data recorded in life record information and posting information, and as an example, within the life record information of the first instructor. If phrases (text, data) such as 'tends to steal things from friends', 'frequent fights with friends', etc. are recorded, the register 350 determines that the first instructor has a history of theft and school violence. You can judge.

In addition, whether a tendency toward violence exists can be determined, for example, by whether a preset number (for example, 10) or more of words related to violence exist in life record information and posting information. That is, the register 350 may determine that the first instructor has a tendency toward violence if there are more than a preset number of words related to violence in the received life record information and collected posting information. Here, words related to pukryeok may include words corresponding to assault, violence, quarrel, fight, throwing, and breaking.

At this time, the register 350 may use, for example, a previously learned artificial intelligence model when determining whether the first instructor has a history of theft, history of school violence, and a tendency to violence.

In addition, the register 350, for example, comprehensively analyzes a plurality of life record information received in relation to the first instructor and post information collected in relation to the first instructor, thereby providing information on the first instructor's theft history, school violence history, and If it is determined that at least one of the violent tendencies exists, the instructor information of the first instructor registered in the instructor DB can be automatically deregistered (i.e., deleted). According to this, the register 350 allows only instructors who have no history of theft, school violence, and a tendency to violence to be registered in the instructor DB or instructor pool and provide lectures to users through the education service provision server 300. there is.

In addition, the register 350, for example, comprehensively analyzes a plurality of life record information received in relation to the first instructor and post information collected in relation to the first instructor, thereby providing the first instructor as personality type information. It can be derived by guessing the instructor's MBTI (Myers Briggs Type Indicator) information. The register 350 can derive the MBTI information of each instructor for each of the plurality of instructors registered in the instructor DB (or instructor pool), and then links the derived MBTI information of each instructor with the instructor information to determine the instructor's status. It can be registered in DB.

At this time, as previously known, the types of MBTI information include, for example, 16 types: ISTJ, ISFJ, INFJ, INTJ, ISTP, ISFP, INFP, INTP, ESTP, ESFP, ENFP, ENTP, ESTJ, ESFJ. , ENFJ, and ENTJ, and detailed descriptions of the characteristics of each type will be omitted.

At this time, when the register 350 detects that an MBTI-based instructor recommendation request has been made from the user terminal 100 while MBTI information of a user already registered in the user terminal 100 exists, the instructor DB (or instructor pool) Based on the MBTI information of each of the plurality of instructors already registered in ), among the plurality of instructors already registered, instructors with MBTI information that matches the user's MBTI information are extracted as recommended instructors, and the information on the recommended instructors is then extracted. Can be provided on the screen of the user terminal 100.

Alternatively, when the user has previously entered preferred MBTI information, the register 350 extracts as recommended instructors instructors whose MBTI information matches the preferred MBTI information previously entered by the user from among a plurality of pre-registered instructors. Information on recommended instructors may be provided on the screen of the user terminal 100.

This register 350 provides information on recommended instructors, allowing the user to select one instructor among instructors with a personality similar to or identical to the user's (in particular, instructors with the same MBTI as the user's MBTI). You can enter the instructor's class and receive education, or you can select one of the instructors with the MBTI information you prefer and enter that instructor's class to receive education.

The storage unit 360 uses voice speech data collected from at least one instructor terminal 400 to calculate pitch, first formant (F1), pitch ratio, pitch bandwidth, and average within the pitch bandwidth. A standard pronunciation database can be built by collecting acoustic timbre analysis data including pitch position and speech rate. At this time, pitch is the height of the sound and is expressed as frequency.

<F1 and pitch ratio>

Formants are one of the important elements that determine the fundamental characteristics of an individual's voice. In the case of pitch, it can be sufficiently changed by adjusting the vocal cords and it is possible to produce a natural voice within an appropriate level. However, since the first formant (F1) is fundamentally related to the length of the vocal tract, there are limits to artificial adjustment. Once F1 is determined, the second formant (F2) is calculated proportionally depending on which vowel is produced, so it follows the frequency characteristics of F1. For example, when expressing the ratio relationship between F1 and pitch, the average pitch of general women used to compare the ratio of F1/F0 can be used. F1 basically reflects body size, so it decreases in that order: children, women, and men. A low F1 in the normal listening range affects the weight and stability of the voice. On the other hand, if F1 is high, it will affect the relatively light and lively feeling. This reflects the overall vocal characteristics of men, women, and children. Additionally, depending on F1, the frequencies of F2 and F3 are all affected, which affects the overall frequency resonance shape. On the other hand, the higher the ratio between pitch and F1, the more the voice feels subdued. This is a different factor from the absolute height of the pitch.

<Pitch Bandwidth>

Since a person's vocal structure and characteristics, including the vocal cords, are innate factors, the pitch that is suitable for one is determined to some extent. On the other hand, the pitch bandwidth in a speaking situation is not related to the innate vocal structure, and can be interpreted as a measure that can be used to determine how fluidly the speaker changes from the basic pitch when speaking. Pitch bandwidth reflects how dynamic the pitch changes are. In tense situations or depressed emotions, the activity of vocal cords decreases, and the most representative example is the vocalization of a robot.

Typical robot vocalizations have a constant pitch, so the pitch bandwidth is very narrow, and this is a typical example of vocalizations that express blunt emotions. On the contrary, in order to elicit affection from the other person or increase the ability to convey the targeted content, the activity of the vocal cords must inevitably be increased. Artificially excessive pitch bandwidth may be unnatural to listen to, but narrow pitch bandwidth inevitably reduces concentration and causes boredom. Ultimately, if the listener is within the level of natural hearing, a wide pitch bandwidth is advantageous.

Accordingly, when the overall pitch height and frequency characteristics and the pitch bandwidth and bandwidth ratio of each instructor are represented by the average spectrum around the pitch, it can be a reference value for comparison with the user's vocalization. Human frequency hearing is close to a logarithmic scale, as can be seen in musical scales that rise in multiples per octave. Therefore, the bandwidth ratio according to the pitch position rather than the linear pitch bandwidth is meaningful. Here, the pitch bandwidth is the range from the center frequency with the highest energy in the pitch range of the entire average spectrum of the speaker's voice until the energy drops by 3dB, and the bandwidth ratio is the result of dividing this by the average pitch.

<Average pitch position within pitch bandwidth>

People have a tendency to go lower in their voice when finishing a sentence. In other words, the position of the average pitch within the pitch bandwidth is determined depending on the time invested in finishing the sentence or the slope of the curve lowering the pitch. For example, based on the position occupied by the average pitch within the pitch bandwidth, the ratio of low and high pitch bands within the pitch bandwidth considering auditory characteristics can be considered in relation to intonation. The average pitch mainly indicates only the pitch used, but the ratio of the low-pitched band and the high-pitched tone band can be divided through the position of the average pitch within the pitch bandwidth. If the ratio of the low-pitched range of the corresponding number is high, it reflects the speaker's tendency to finish calmly at the end of the sentence or at the middle comma. Conversely, if the low-pitched range ratio is low, it reflects a tendency to finish sentences relatively quickly.

The F0 falling slope of these sentence endings also affects the impression. This number does not represent the entire sentence-ending vocalization characteristics, and it is not possible to distinguish between good and bad instructor's vocalization characteristics through this. F _LP in Equation 1 below is the position ratio of the low-pitched tone band occupied by the average pitch within the pitch bandwidth, and the positional ratio of the high-pitched tone band is F _HP . Here, F _L and F _H represent the minimum and maximum frequencies of the pitch bandwidth, respectively, and F _C is the average pitch.

[Equation 1]

<Ignition speed>

Speech speed is a measure of how many words a speaker utters per second and indicates the speed at which a speaker speaks. Since the appropriate speed varies depending on which sentence is spoken and under what circumstances, it is impossible to determine whether the speed has been properly controlled based on the average speaking speed alone, but it is possible to determine a certain degree of speed tendency through the average. At this time, the standard for measuring speed may be syllables per second (SPS).

The pronunciation visualization unit 370 expresses voice data, which is a voice utterance input from the user terminal 100, as time series data defined on the time axis, and uses a probabilistic Self-Organizing Map (SOM-VAE) that reflects the structure that changes on the time axis. With Variational Auto Encoder) algorithm, the difference between the standard pronunciation of the standard pronunciation database and the error pronunciation of the user terminal 100 can be visualized. Research on voice visualization includes, for example, a visualization system that graphically expresses the shape of the lips and the position of the tongue for each voice and displays them in a 3D animation, or uses a time delay neural network to obtain phoneme characteristics and then displays them in color. There is a way to express it. In addition, the standard pronunciation of words and sentences and the characteristics expressed in spectrum and color of the user's pronunciation are simultaneously displayed to enable comparison.

Since these visualization methods only extract and visualize the characteristics of individual voices, they provide information about how far the user's pronunciation is from the standard pronunciation and how to speak to get closer to the standard pronunciation. There is a downside to not being able to do it. Accordingly, one embodiment of the present invention decides to use a model that displays on the screen how far the user's pronunciation is from the standard pronunciation by introducing the concept of relative distance. Users can check how far their pronunciation is from the standard pronunciation, and at the same time, change the shape of their tongue, mouth, or mouth to learn the standard pronunciation before the instructor's class.

It configures the standard pronunciation and error pronunciation space visualization module, and its main function is to extract features from speech, perform clustering, and then map them on the screen. The ultimate goal is to extract multi-mode features such as duration, intensity, temporal pattern of pitch, temporal change pattern of formant frequency, voice envelope pattern, and amount of expression from input voice. , construct a scale that can express the distance between standard pronunciation and error pronunciation as the difference between these features, and enable the difference in pronunciation according to this scale to be expressed visually as a relative distance on a two-dimensional computer screen. The biggest feature is that it creates a pronunciation space that expresses the difference in pronunciation of a word or phrase as a distance on the computer screen, allowing users to adjust their vocal organs and articulation methods to correct their own pronunciation and approach standard pronunciation. There is excellence that can be achieved. At this time, standard pronunciation refers to the correct pronunciation of an expert, for example, an announcer.

<Voice feature extraction>

Because the causes of pronunciation errors are complex, multimodal features must be addressed, making it difficult to simply compare the differences between standard pronunciation and error pronunciation. For example, the important qualities that distinguish between flat, aspirated, and tense sounds are the pitch and amount of air of the trailing vowel. These acoustic qualities can be expressed as [da], [ta], and [ta] pronunciations, for example, pitch. is not displayed as a single number (frequency), but appears as a pattern in time. Therefore, in the standard pronunciation, the word ganda should be tense and pronounced as [gatta], but if it is pronounced as [gatda] due to a user's pronunciation error, it will indicate the difference between the pitch patterns for [tta] and [da], that is, the degree of error. A measure or method that can be used is needed. Moreover, designing a multimodal scale that uses both pitch in the form of patterns and quantity in the form of simple numbers is not a simple task. Therefore, it is possible to use a method that receives time series data as input and processes it without applying a clustering algorithm based on point data.

The acoustic characteristics of Korean consonants can be characterized by vocal fold vibration time (Voice Onset Time, VOT), fundamental frequency of the trailing vowel, strength of airflow during burst, closure duration (CD), etc. VOT plays a decisive role in distinguishing the plain, tense, and aspirated sounds of Korean plosives. In the articulation method of plosives, at the word-initial position, the vocal cord vibration time (VOT) and the length of the trailing vowel play an important role in distinguishing plosives. In position, closure duration (CD) and the length of the preceding vowel play an important role. Looking at the distribution of VOT values at the word-initial position, that is, in the case of initial consonants, tense consonants [ㅃ, ㄲ, ㄸ] have a distribution of less than 20ms, aspirated consonants [ㅍ, ㅋ, ㅌ] have a distribution of 60~80ms, and plain consonants [ㅃ, ㄲ, ㄸ] have a distribution of less than 20ms. [ㅂ, ㄱ, ㄷ] is observed to have a distribution of 40 to 60 ms. In the case of word-middle, the distribution is reduced to less than 20ms for tense sounds, the distribution is in the 50~60ms range for aspirated sounds, and the flat sounds are voiced. Additionally, the pitch and closure duration of trailing vowels have properties that distinguish tense, aspirated, and flat sounds.

Considering the above-mentioned data, VOT, pitch, and CD can be set as features to distinguish consonants. The characteristic of the vowel is that three formants F1, F2, and F3 can be obtained and used from voice data, which is detailed in Korean Patent No. 10-2355960 (announced on February 8, 2022), which is the applicant's pre-registered patent. Since it is described, contents not described in this specification will refer to the above-mentioned registered patent. To obtain formants, a method based on LPC (Linear Predictive Coding) analysis is widely used as follows.

① First, the voice data is divided into frames of 25ms in length. After multiplying the Hamming Window as a preprocessing operation, a high-pass all pole filter is applied. When the number of voice samples in each frame is expressed as N and the nth sample data of the ith frame is expressed as si(n), the LPC model is a linear combination of the previous p voice sample values as follows: It is assumed that an approximate expression can be made.

[Equation 2]

The coefficient vector (vec)a=(a1,...,ap) is a value that minimizes the MSE (Mean Square Error) between the voice signal si(n) and the prediction signal (hat)si(n), and is expressed in Equation 2 below: It is obtained by finding the solution of a linear equation such as .

[Equation 3]

② Convert the filter coefficient vector (vec)a to LP Cepstrum Coefficient (vec)c=(c1,...,cn) as follows.

[Equation 4]

③. If you sequentially find the peak points from low frequencies in the LP spectrum, you can get F1, F2, and F3.

<Clustering of time series data>

Voice data is expressed as time series data defined on the time axis. Therefore, a clustering algorithm that can reflect the changing structure on the time axis is required. The recently announced stochastic SOM-VAE (Self-Organizing Map with Variational Auto Encoder) algorithm provides performance that guarantees smooth continuity when mapping high-dimensional time series data to low-dimensional space. By applying this technique, information loss can be prevented because time series data can be used directly instead of expressing formants that change on the time axis as point data called average values.

Briefly looking at the flow of the SOM-VAE model, input voice data It is expressed approximately by mapping to the nearest node Zq. In the SOM map space, learning is performed by applying the Markov transition model, through which the transition from the current state Ztq to the next state Zt+1q can be predicted with probability. In this way, continuous time series data can be expressed in discrete expressions, which can then be mapped into the original time series data space through a decoder. The feature map constructed by the SOM-VAE algorithm can be used to represent the relative distance of the voice on the computer screen.

<Distribution of tense pronunciation>

Gyeongeumization is a phenomenon in which a plain sound is pronounced as a hard sound, and the word Hapbeop is pronounced as [hapj]. If the error pronunciation was entered as [Habeop], [Haj], [Hotbeop], etc., four pronunciation groups including these and the standard pronunciation [Haj] were formed, and the results of constructing the SOM-VAE feature map were calculated. It can be displayed. Standard pronunciation and error pronunciation are each distributed with nodes. The distribution of nodes forms a cluster, and when this feature map is mapped onto a computer screen, the relative distance of the user's error pronunciation from the standard pronunciation can be displayed.

<Distribution of nasal pronunciation>

The nasalization phenomenon is a phenomenon in which consonants belonging to a non-nasal category are pronounced as nasal sounds, and is the same phenomenon as the word last year being pronounced as [Jangnyeon]. If the error pronunciation related to this is entered as [janyeon], [natural], and [jannyeon], four pronunciation groups including the standard pronunciation can be formed, a SOM-VAE feature map can be constructed, and the results can be displayed.

<Distribution of voiced pronunciation>

In the Korean consonant system, there is only one consonant [ㄹ]. The voicing phenomenon is a phenomenon that occurs when the nasal sound [ㄴ] is changed to the voice sound [ㄹ], and is like the word convenience being pronounced as [pulli]. If the error pronunciation related to word convenience is entered as [Peoeri], [Peooni], and [Pyeonni], four pronunciation groups including the standard pronunciation can be formed and a SOM-VAE feature map can be constructed. In addition, there are organic phonetic pronunciations, ni-eun (ㄴ)-added pronunciations, and palatal phonetic pronunciations, and each pronunciation data can be mapped onto the screen according to formant characteristics.

When a voice utterance is recognized from the user terminal 100, the vowel analysis unit 380 collects a voice sample, analyzes the vowels A, I, and U in the voice sample using the Praat program, and then generates a standard pronunciation voice utterance. After calculating the frequency difference for each sound zone, a vowel triangle is created between the user's voice utterance and the standard pronunciation voice utterance, and the instructor terminal (400) is used to guide mouth opening and lip movement for vowels where differences occur. ) can be transmitted. In Korean, vowels serve as a carrier to produce consonants. When analyzing the difference between the vowels produced by the announcer in a natural sentence and analyzing the patterns produced by acoustically analyzing the user's pronunciation, the user's pronunciation It can be used as basic information on maps.

<Voice sample collection>

To collect the user's voice, individual voice samples are recorded as wav files using Roland (R-09HR), and the sampling rate can be set to 44kHz. At this time, it is possible to construct as meaningful words as possible for users' natural reading of the text, and only the part where there is no phonetic transition of the vowel following the open rupture of the plosive can be analyzed. To prevent students from being nervous or stopping to read each syllable when reading Korean sentences, you can read the practice sentences first and then read the target sentence.

<Voice sample analysis>

Using the Praat program, you can analyze the vowels /ah, i, u/ in collected voice samples. In order to analyze the sound range of the vowel included in the target word, three sound ranges were analyzed by analyzing the flat range from the beginning of the vowel range after the burst of the plosive until the transition to another phoneme occurs. You can. At this time, an independent samples t test can be performed to examine the frequency difference between the sound patterns (F1, F2, F3) of the representative vowels of the user and the instructor, which is the control group.

<Vowel triangle>

The size of the vowel angle is related to speech intelligibility, and the narrower the area of the vowel angle, the lower the speech intelligibility. Likewise, vowel triangle is the same concept, but in context vowels in the same environment, Korean learners sometimes have a small vowel triangle area, which reduces the clarity of pronunciation. This reduction is due to tension for articulation production and reduction of lip movement, resulting in reduction of the vocal tract. Accordingly, if the size of the vowel triangle of Korean learners is smaller than that of the general public and the vowels /ah, uh, u/ are placed further back, it is calculated by expanding the mouth opening and movement of the lip space to improve pronunciation clarity. Language pronunciation education or guidance is needed. If these characteristics are identified, the instructor can set the direction of instruction by having the user wear an aperture device such as a cochlear implant or continuously providing systematic pronunciation training.

The speech unit 390 may conduct a speech test by dividing the levels of users of the user terminal 100 and then issue a speech certificate issued by a preset organization or association.

The evaluation unit 391 selects an instructor from the user terminal 100 and conducts a real-time lecture with the instructor terminal 400, then requests the instructor to input ratings and reviews, and accumulates the ratings and reviews for each instructor. You can save it.

The big data converting unit 393 can construct big data by labeling voice data for voice utterances input from the user terminal 100 with demographic data and mapping it with the voice utterance text. In the case of English, not only speech texts such as the TIMIT speech corpus but also natural speech labels such as the Buckeye speech corpus are provided, making speaker information as well as specific speech data public. Since the pronunciation form corresponding to each word is in the form of a text-based pronunciation symbol, it can be easily converted to a pronunciation dictionary, and pronunciation variations can be extracted by comparing it with a public pronunciation dictionary such as the CMU pronunciation dictionary. For the Korean speech corpus, there is the Seoul Speech Corpus provided by the National Institute of the Korean Language in the case of reading, and in the case of spontaneous speech, there is the recently released Seoul Corpus, but it contains the voices actually uttered by people in each group, group, or occupation. Difficult to obtain on a large scale. Even if it is obtained, it is necessary to go through a process of labeling each user's demographic data. In one embodiment of the present invention, each user repeatedly practices pronunciation for a long period of time, and each user's region, gender, occupation, etc. are collected. This makes it possible to obtain voice data for each group without labeling demographic data.

When generating STT (Speech to Text), the recognition performance improvement unit 395 learns by inputting voice data and text, so that even if voice data corresponding to voice utterance that varies depending on demographic data is input, the text is It can be printed. STT (Speech To Text) is a field of voice recognition and refers to the process of converting human voice language into text data through computer interpretation. Because input is made through human voice rather than through a keyboard or other input device, it is the basis for various technologies such as HCI (Human Computer Interaction), telematics, artificial intelligence assistant, and ChatBot. STT converts the speaker's voice into text through a mechanical algorithm, and recently, the voice recognition engine for STT is improving accuracy by using sound and language models through deep learning algorithms.

In addition to the traditional voice recognition algorithm, HMM (Hidden Markov Model), it is showing higher accuracy than in the past by applying DNN (Deep Neural Network) and RNN (Recurrent Neural Network) techniques, which are widely used as deep learning-based algorithms. The process of performing STT preprocesses the input voice and outputs text results by comparing it with a model trained on a lot of voice data. Therefore, the output results may appear differently depending on various technologies, such as how diverse the voice data was trained with and how specialized the model was created. Therefore, in order to accurately evaluate and test performance, it is necessary to evaluate how accurately the input audio has been converted to text. At this time, the necessary technology is text similarity evaluation.

Text similarity measurement is an important research field in the field of Natural Language Processing (NLP), and has many areas of application with the recent development of artificial intelligence (AI) technology. String similarity measurement scores the semantic similarity between two strings and is a measure of how similar the two data are. In the field of data science, measuring similarity between data is the basis for classification and clustering of data and the basis for applying improved algorithms. When comparing similarity to data, simple 1:1 matching may be simple, but text converted through voice has many variables. When converting English to text, it is relatively advantageous to measure similarity because English is made up of alphabets and spacing is clear, but in Korean, the initial consonant, middle consonant, and final consonant make up one letter, and accurate spacing is difficult. . In addition, although the meaning is sufficiently conveyed due to the rules of initial pronunciation and linking rules, there are many cases where the correct Hangul spelling and pronunciation are different. Therefore, the similarity evaluation for Hangul must be approached in a different way from that for English to achieve a more accurate evaluation.

The process for evaluating the accuracy of the STT engine is generally as follows. Prepare an original text file that exactly matches the first test audio voice file. The similarity is compared between the text result file produced through the second STT engine and the text file of the original audio. Although the evaluation process is very simple, there may be some errors in evaluating the recognition rate. First, there may be errors when creating the original text file. For example, when a word has both a standard pronunciation and a real pronunciation, the meaning may be differentiated depending on the pronunciation. This can be said to be a case where there is a certain difference in meaning between the actual pronunciation and the standard pronunciation. In the case of a prefix, a meaning different from the original meaning may be created as a result of the prefix. [Jjolda] is a tense pronunciation of the [ㅈ] in [Jjolda] at the beginning, and when it is used to mean ‘to be psychologically scared,’ it is generally pronounced as [jjolda]. In a situation where [dozing off] is used in this sense, if the speaker pronounces it in a realistic pronunciation, such as ‘Hey, don’t doze [get scared].’ or ‘Are you dozing off [get scared]?’, STT is dictated as [Don’t get scared]. This is treated as an error because it has not been converted to text properly. At this time, if [Don't doze off] is pronounced as [Don't be scared], that is, [Standard pronunciation] - [Real pronunciation] are mapped and stored, and linked with [Text] - [Standard pronunciation], even if the realistic pronunciation is entered. After converting to standard pronunciation, it can be converted to standard language, that is, text, without error.

At this time, when mapping [standard pronunciation] and [realistic pronunciation], the database is enriched by mapping the voice data, that is, the voice data input when each user reads the [text] corresponding to the standard pronunciation, to [realistic pronunciation]. . Each user undergoes a test before receiving training. Before training, the actual pronunciation used by the general public is the same, so within this period, that is, before the instructor determines or judges that the user is below a certain level, Only the input voice utterance can be extracted and used as real pronunciation. Since [text]-[standard pronunciation] is already mapped and stored in the textbook or content, only the actual pronunciation uttered by the user needs to be entered and connected to it. When using this, users can practice, and developers developing STT engines can be provided with a rich voice corpus. Developers can use it as is without preprocessing because it is already labeled without hiring a labeler, making it highly useful and economical. am.

In addition, although not shown in the drawing, the education service providing server 300 operates in the education materials database (not shown) when it detects that the user has clicked on the textbook menu on the user terminal 100 in order to enable the user to purchase Korean education materials. It may include a textbook management unit (not shown) that provides a list of educational textbooks, which is a list of a plurality of Korean educational textbooks already registered in the city, on the screen of the user terminal 100.

At this time, when providing a list of educational teaching materials, the teaching materials management unit (not shown) displays a wish list in which information on at least one wish list teaching material selected by the user is displayed in one area of the screen of the user terminal 100 where the list of educational teaching materials is displayed. An exposure icon can also be provided. At this time, the size of the wish list exposure icon may be set to have a size corresponding to 1/10 of the total size of the screen of the user terminal 100, for example. That is, the wish list exposure icon may be provided in a size smaller than the overall size of the screen of the user terminal 100.

The textbook management unit (not shown) may allow the user to store at least one Korean education textbook that the user wants to purchase among a plurality of Korean education textbooks included in the education textbook list as a wish list textbook in the wish list exposure icon.

At this time, the wishlist exposure icon may include a plurality of exposure areas (i.e., 5 exposure areas) where information on the wishlist textbook is exposed as much as a preset number (for example, 5). Accordingly, for example, the user can store up to five wish list teaching materials in the wish list exposure icon.

At this time, the textbook management unit (not shown) prepares the wishlist exposure icon to have a size corresponding to 1/10 of the total screen size of the user terminal 100, and configures the wishlist exposure icon to include a plurality of exposure areas. Accordingly, the user can check the wish list textbooks saved in the wish list exposure icon and the list of educational materials displayed on the screen of the user terminal together on one screen through the wish list exposure icon and purchase at least one educational material of their choice. It can be made possible.

Additionally, the wishlist exposure icon may include a lowest price search button provided in the day area.

For example, the textbook management department (not shown) may wish to purchase four Korean education textbooks, including the 1st to 4th Korean education textbooks among the plurality of Korean education textbooks in the education textbook list, and select the wish list textbooks (i.e. , 4 wish list textbooks), if it is detected that the lowest price search button has been clicked while the wish list exposure icon is saved, in response to the user's click on the lowest price search button, a plurality of online textbook sales sites ( For example, for Interpark, Aladdin, Coupang, Kyobo Bookstore, Auction, Youngpoong Bookstore, G Market, etc., 4 wishlist textbooks (i.e. 4 corresponding to 4 Korean education textbooks) already stored in the wishlist exposure icon. For each of the four wishlist textbooks, a lowest price search is conducted for the same educational textbooks with the same textbook name and author name (or publisher) as each of the four wishlist textbooks, and then based on the results of the lowest price search, the four wishlist textbooks are It can be controlled so that the lowest price search result information for each of the four wishlist textbooks is exposed in the bottom area of each of the four exposure areas within the wishlist exposure icon where information on the wishlist textbooks is exposed.

For example, in the first exposure area of the wishlist exposure icon, 'the first wishlist textbook (first Korean education textbook) with the name of the first textbook and the name of the first author' is stored, and in the second exposure area, 'the second textbook' is stored. The 2nd wishlist textbook (2nd Korean education textbook) with the name of the textbook and the 2nd author is stored, and in the third exposure area, the 3rd wishlist textbook (3rd Korean education textbook) with the name of the 3rd textbook and the name of the 3rd author is stored. 'Textbook)' is stored, and in the fourth exposure area, '4th wish list textbook (4th Korean education textbook) with the name of the 4th textbook and the name of the 4th author' is stored, and in this situation, the user clicks the lowest price search button. Let's say you clicked.

Afterwards, the textbook management department (not shown) may conduct a search for the lowest price for each of the four wish list textbooks targeting a plurality of textbook sales sites. At this time, the results of the lowest price search, i) with respect to the first wish list textbook, , it is being sold for 8,000 won on the first textbook sales site (e.g., Youngpoong Bookstore), 9,500 won on the second textbook sales site (e.g., G Market), and 10,000 won on the third textbook sales site (e.g., Aladdin). Confirmed, ii) Regarding the second wish list textbook, 9,000 won on the first textbook sales site (for example, Youngpoong Bookstore), 9,550 won on the fourth textbook sales site (for example, Interpark), and 9,550 won on the fifth textbook sales site (for example, Interpark) It has been confirmed that it is being sold for 9,700 won on Home & Shopping) and 10,500 won on the 6th textbook sales site (for example, Maruseojae). iii) Regarding the 3rd wish list textbook, it is sold for 10,500 won on the 6th textbook sales site (for example, Youngpoong). It is confirmed that it is being sold for 7,600 won on the second textbook sales site (e.g., G Market) and 8,050 won on the second textbook sales site (e.g., G Market). iv) Regarding the fourth wish list textbook, it is sold for 7,600 won on the second textbook sales site (e.g., G Market). Let's say it is being sold for 15,000 won on the 5th textbook sales site (e.g., Home & Shopping), 9,900 won on the 6th textbook sales site (e.g., Maruseo), and 10,900 won.

In this case, the textbook management unit (not shown) responds to the user's click input on the lowest price search button, and as a result of the lowest price search, the first wish list textbook is displayed in the lower area of the first exposure area in the wish list exposure icon. '8,000 won (Youngpoong Bookstore)' as the lowest price search result information, '9,000 won (Youngpoong Bookstore)' as the lowest price search result information for the second wish list textbook in the lower part of the second exposure area, and '9000 won (Youngpoong Bookstore)' as the lowest price search result information of the second wish list textbook in the lower part of the second exposure area. In the lower daily area, the lowest price search result information for the 3rd wish list textbook is '7,600 Won (Youngpoong Bookstore)', and in the lower daily area of the 4th exposure area, the lowest price search result information for the 4th wish list textbook is '9,900 won (Home& The user terminal 100 can be controlled so that information called 'shopping)' is exposed.

This textbook management unit (not shown) provides a wish list exposure icon including a lowest price search button, so that when a user wishes to purchase at least one Korean education textbook among a plurality of Korean education textbooks, the user wishes to purchase the wish list textbook. The lowest price search for a number of stored Korean education textbooks (i.e., 4 wish list textbooks, for example) can be easily done all at once (simultaneously) by simply pressing the lowest price search button, allowing users to It is possible to provide convenience in purchasing at least one wish list textbook.

Hereinafter, the operation process according to the configuration of the education service providing server of FIG. 2 described above will be described in detail using FIGS. 3 and 4 as an example. However, it will be apparent that the embodiment is only one of various embodiments of the present invention and is not limited thereto.

Referring to FIG. 3, (a) the education service providing server 300 receives age, gender, occupation, grade, group affiliation, region, etc. from the user terminal 100, and selects standards frequently used or incorrect in the group to which the user belongs. Extract real pronunciation that is different from pronunciation. (b) Then, the education service providing server 300 collects the voice data collected from the user terminal 100 as realistic pronunciation, performs data labeling as in (c), and later uses data to improve the accuracy of the STT engine. use. As shown in (d), the education service providing server 300 allows the user terminal 100 to practice correcting the actual pronunciation that is different from the standard pronunciation, and then uses the verified instructor as shown in (a) of FIG. 4 in (b) and Let's apply together so we can listen. As shown in (c), the real pronunciation collected from the user terminal 100 is used to improve STT, but if the voice utterance uttered by the user has a high similarity to the standard pronunciation, it is not included as the real pronunciation. As shown in (d), STT can be dictated in standard language by any group of users across the country, and furthermore, it can be recognized and utilized without error in various chat agents or voice-based modules such as AI chatbots or AI assistants.

Matters that are not explained in FIGS. 2 to 4 regarding the method of providing Korean education services for users whose first language is Korean are explained in FIG. 1 regarding the method of providing Korean education services for users whose first language is Korean. Since the content is the same as the content described or can be easily inferred from the content described, the description below will be omitted.

FIG. 5 is a diagram illustrating a process in which data is transmitted and received between components included in the Korean education service providing system for users whose first language in FIG. 1 is Korean according to an embodiment of the present invention. Hereinafter, an example of the process of transmitting and receiving data between each component will be described with reference to FIG. 5, but the present application is not limited to this embodiment, and the process shown in FIG. 5 according to the various embodiments described above It is obvious to those skilled in the art that the process of transmitting and receiving data can be changed.

Referring to Figure 5, the education service provision server builds a database by mapping cases where the standard pronunciation and actual pronunciation are different according to demographic data including age, gender, region, and occupation, to demographic data. (S5100).

And, the education service providing server extracts cases where the standard pronunciation and the actual pronunciation are different from the user information input from the user terminal (S5200), and selects a word or sentence containing a standard pronunciation different from the actual pronunciation from the user terminal to the user. After the terminal is made to produce a voice speech, the spoken voice data is compared with the reference data and the difference data is transmitted to the user terminal (S5300).

In addition, the education service providing server transmits difference data to the instructor terminal before starting a real-time lecture on the instructor terminal (S5400).

The sequence between the above-described steps (S5100 to S5400) is only an example and is not limited thereto. That is, the order between the above-described steps (S5100 to S5400) may change, and some of the steps may be executed simultaneously or deleted.

Matters that are not explained regarding the method of providing Korean education services for users whose first language in FIG. 5 are Korean are previously described in FIGS. 1 to 4 regarding the method of providing Korean education services for users whose first language is Korean. Since the content is the same as the content described or can be easily inferred from the content described, the description below will be omitted.

A method of providing a Korean language education service for a user whose first language is Korean according to an embodiment described with reference to FIG. 5 includes a recording medium containing instructions executable by a computer, such as an application or program module executed by a computer. It can also be implemented in form. Computer-readable media can be any available media that can be accessed by a computer and includes both volatile and non-volatile media, removable and non-removable media. Additionally, computer-readable media may include all computer storage media. Computer storage media includes both volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data.

The method of providing a Korean language education service for a user whose first language is Korean according to an embodiment of the present invention described above includes an application installed by default on a terminal (this may include programs included in a platform or operating system, etc., installed by default on the terminal). It may be executed by an application (i.e., a program) that the user installs directly on the master terminal through an application providing server such as an application store server, an application, or a web server related to the service. In this sense, the method of providing a Korean language education service for a user whose first language is Korean according to an embodiment of the present invention described above is implemented as an application (i.e., a program) installed by default on the terminal or directly installed by the user, and is implemented through the terminal. It may be recorded on a computer-readable recording medium such as .

The description of the present invention described above is for illustrative purposes, and those skilled in the art will understand that the present invention can be easily modified into other specific forms without changing the technical idea or essential features of the present invention. will be. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. For example, each component described as unitary may be implemented in a distributed manner, and similarly, components described as distributed may also be implemented in a combined form.

The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present invention. do.

Claims (4)

a user terminal for inputting user information corresponding to demographic data;
Instructor terminal that conducts real-time lectures; and
Including education service provision servers,
The education service provision server is,
A construction unit that constructs a database by mapping cases where standard pronunciation and actual pronunciation are different according to demographic data including age, gender, region, and occupation, to demographic data;
An extraction unit that extracts cases where the standard pronunciation and the actual pronunciation are different from the user information input from the user terminal;
A transmission unit that causes the user terminal to utter a word or sentence containing a standard pronunciation different from the actual pronunciation by voice, then compares the uttered voice data with reference data and transmits the difference data to the user terminal;
a lecture preparation unit that transmits the difference data to the instructor terminal before starting a real-time lecture on the instructor terminal;
Using voice speech data collected from at least one instructor terminal, collect acoustic timbre analysis data including pitch, first formant and pitch ratio, pitch bandwidth, average pitch position within the pitch bandwidth, and speech rate. a storage unit that builds a standard pronunciation database;
Voice data, which is a voice utterance input from the user terminal, is expressed as time series data defined on the time axis, and standard pronunciation is made using the probabilistic SOM-VAE (Self-Organizing Map with Variational Auto Encoder) algorithm that reflects the changing structure on the time axis. A pronunciation visualization unit that visualizes the difference between the standard pronunciation of the database and the error pronunciation of the user terminal;
When a voice utterance is recognized from the user terminal, a voice sample is collected, the vowels a, i, and u are analyzed in the voice sample using the Praat program, and the frequency difference between the standard pronunciation voice utterance and each sound type is calculated, and then the user A vowel analysis unit that generates a vowel triangle between the voice utterance and the standard pronunciation voice utterance and transmits it to the instructor's terminal to guide mouth opening and lip movement for vowels where there is a difference;
A speech department that conducts a speech test by dividing the level of the user of the user terminal and then issues a speech certificate issued by a preset organization or association;
An evaluation unit that selects an instructor on the user terminal and conducts a real-time lecture with the instructor terminal, then requests the instructor to enter ratings and reviews, and accumulates and stores the ratings and reviews for each instructor;
A big data unit that builds big data by labeling voice data for voice utterances input from the user terminal with demographic data and mapping it with the voice utterance text;
When generating STT (Speech to Text), a recognition performance improvement unit that learns by inputting voice data and text so that text is output even if voice data corresponding to voice utterances that vary depending on demographic data are input; and
When it is detected that a textbook menu has been clicked on the user terminal, it includes a textbook management unit that provides an educational textbook list, which is a list of a plurality of Korean educational textbooks already registered in the educational textbook DB, on the screen of the user terminal,
The textbook management department,
When providing the list of educational materials, a wishlist exposure icon that exposes information on at least one wishlist textbook selected by the user is provided in an area of the screen of the user terminal where the list of educational materials is displayed,
The wish list exposure icon is provided to have a size corresponding to 1/10 of the total size of the screen of the user terminal, and a wish list of a pre-set number is provided so that the user can store a pre-set number of wish list teaching materials. By providing a plurality of exposed areas where the information of the teaching materials is exposed, the wish list teaching materials stored by the user in the wish list exposure icon and the list of educational teaching materials displayed on the screen of the user terminal are displayed through the wish list exposure icon. A Korean education service provision system for users whose first language is Korean, which allows them to purchase desired educational materials while checking them on the screen. According to paragraph 1,
The lecture preparation department,
When the user terminal detects that the first instructor profile has been selected as one of at least one instructor profile, the user enters the first class, which is the class of the first professor corresponding to the first instructor profile, 1 A Korean education service provision system for users whose first language is Korean, which controls the ability to listen to classes in real time. According to paragraph 1,
The education service provision server is,
A register that establishes an instructor pool by setting at least one qualification condition and registering at least one instructor terminal by verifying the at least one qualification condition;
It further includes,
A system for providing Korean language education services for users whose first language is Korean, wherein the at least one qualification is to be one of those who have received announcing training, broadcasters and announcers, voice actors, and singers. delete

Images