KR20210060040A - Server and method for automatic assessment of oral language proficiency - Google Patents

Abstract

Disclosed are a foreign language speaking learning server and a method thereof. According to an embodiment of the present invention, the foreign language speaking learning server includes: a database storing learning situation data, model answer sentence data by learning situation, and learning data including a chunk of a model answer sentence and a learning quarter which is a learning situation matched with an answer of a user or the model answer sentence; a chunk output module creating and displaying a speaking hint chunk from model answer sentences by learning situation while waiting for utterance input from the user after the output of a learning situation selected by the user; a primary evaluation module receiving a voice signal from the user to classify the received voice signal into a chunk and compare the classified chunk with the chunk of the model answer sentence; a secondary evaluation module recognizing the received voice signal to evaluate each utterance evaluation factor including intonation, pronunciation and size; and a learning quarter creation module extracting a learning quarter in accordance with utterance evaluation results delivered from the primary and secondary evaluation modules. Therefore, the present invention is capable of more accurately evaluating the English speaking ability of a user.

Description

Foreign language speaking learning server and method {SERVER AND METHOD FOR AUTOMATIC ASSESSMENT OF ORAL LANGUAGE PROFICIENCY}

The present disclosure relates to a foreign language speaking learning server and method, specifically, a server that receives a user's voice, generates a selective learning branch based on the similarity of voice input to learn a foreign language, and automatically evaluates foreign language speaking ability according to evaluation factors And a method.

Unless otherwise indicated herein, the content described in this section is not prior art to the claims of this application, and inclusion in this section is not admitted to be prior art.

In recent years, foreign language learning, especially English or Chinese proficiency, has become more and more important. In particular, English functions as the world's official language (lingua franca) and English as an international language, which means that English has established itself as a means of global communication across national and cultural boundaries. . Despite the significant investment in English education in Korea, the effect of learning English is not great because the structure system of Korean and English is quite different and English is not the first language.

In particular, the English speaking area is the most vulnerable learning area for Koreans despite investing a lot of time and money in learning English. Among the areas of foreign language learning including speaking, listening, and writing, speaking is a learning area that is directly related to real life and should be considered the most important for smooth communication with foreigners, but it is also one of the most difficult learning areas for Koreans.

Various educational institutions have proposed various methods to improve speaking learning ability. For example, efforts are being made to improve speaking and listening learning ability by arranging native English instructors and expanding English classes, but this is only a way to differentiate the learning contents provided to learners, and accurately measure the elements of actual learners' foreign language utterance evaluation. It is far from a learning method in which you practice speaking and then guide you in areas that need to be improved. In addition, as described above, the series of learning processes that induce and practice English speaking to learners and measure the results can be conducted mostly by offline learning programs or tutors, so the user's temporal and spatial limitations for learning English speaking are limited. This is a very big situation.

In addition, since the English speaking learning program provided online is conducted by simply measuring the number of repetitions of speaking or reading the script, the learner writes and utters the answer appropriate to the situation, and the accuracy and pronunciation of the uttered English sentences. It is not possible to accurately measure various evaluation factors for fluently speaking foreign languages such as, intonation, etc.

1. Registered Patent Publication No. 10-1313268 (2013.09.24) 2. Registered Patent Publication No. 10-1943520 (2019.01.23)

It provides a selective learning branch based on the similarity between the user's foreign language voice input and the answer sentence, divides the voice input signal into chunks, and analyzes the voice of each chunk, Provides a foreign language speaking learning server and method that accurately analyzes speech evaluation items including intonation, fluency, and size.

In addition, it provides a foreign language speaking learning server and method in which a user can see the speaking hint chunks, write them directly, and then utter them by presenting an answer sentence for inducing speech in chunks.

In addition, in the embodiment, various learning branches corresponding to the learner's level are provided according to the result of evaluating the difficulty and fluency of the sentence uttered by the user, thereby providing individually differentiated learning contents according to the user's English speaking ability.

The foreign language speaking learning server according to the embodiment stores learning data including learning situation data, model answer sentence data for each learning situation, chunk of the best answer sentence, and a learning branch that is a learning situation matched with the user's answer or the model answer sentence. Database; A chunk output module for generating and displaying a speaking hint chunk in the best answer sentence for each learning situation during a time period for waiting for the user's speech input after the learning situation selected by the user is output; A primary evaluation module for receiving a user's voice signal, classifying the input voice signal into chunks, and comparing the classified chunk with the chunk of the model answer sentence; A second evaluation module for recognizing the input speech signal and evaluating each speech evaluation element including intonation, pronunciation, and size; And a learning branch generation module for extracting a learning branch according to the speech evaluation result transmitted from the first evaluation module and the second evaluation module. Includes.

The foreign language speaking learning method according to another embodiment includes learning situation data, model answer sentence data for each learning situation, a chunk of model answer sentences, and a learning branch that is a learning situation matched to the user's answer or model answer sentence. A first step of storing the learning data to be performed; The second step of generating and displaying, at the foreign language speaking learning server, a speaking hint chunk from the best answer sentences for each learning situation during a time period for waiting for the user's utterance input after the learning situation selected by the user is outputted; The foreign language speaking learning server receives the user's voice signal, classifies the input voice signal into chunks, and compares the classified chunk with the chunk of the model answer sentence; A fourth step of the foreign language speaking learning server recognizing the input voice signal and evaluating each speech evaluation element including intonation, pronunciation, and size; And a fifth step of extracting a learning branch according to the speech evaluation result transmitted from the first evaluation module and the second evaluation module, in the foreign language speaking learning server. Includes.

The foreign language speaking learning server and method according to the embodiment presents the learning situation to the user and induces the utterance of English sentences, and then presents the exemplary answer sentences in chunks. It does not stop, and maximizes the effect of learning English speaking by allowing the user to speak after writing using the suggested speaking hint chunk.

The voice signal input by the user is evaluated for each chunk according to evaluation factors including pronunciation, intonation, fluency, and accuracy, so that the user's English speaking ability can be more accurately evaluated.

By providing a variety of learning branches according to the accuracy and difficulty of the user-selected chunks and sentences spoken by English, even a user who is very good at speaking English is provided with a learning branch corresponding to his or her English proficiency so that they can continuously practice speaking English. To be there. In addition, users who are not familiar with speaking English are provided with a learning branch appropriate to their level of learning so that they can gradually improve their speaking skills.

The effects of the present invention are not limited to the above effects, and should be understood to include all effects that can be inferred from the configuration of the invention described in the detailed description or claims of the present invention.

1 is a diagram showing a foreign language speaking learning system according to an embodiment
2 is a diagram showing a data processing block of the learning server 100 according to an embodiment
3 is a flowchart showing a data processing process implemented in a learning server of a foreign language speaking learning system according to an embodiment;
4 is a flowchart showing a data processing process performed by a learning server or a user terminal installed with a speaking learning application according to an embodiment;
5 is a diagram showing a data processing procedure for recognizing a user's voice signal implication according to an embodiment;
6 and 7 are diagrams showing an embodiment of an actual implementation interface of a foreign language speaking learning application

Advantages and features of the present invention, and a method of achieving them will become apparent with reference to the embodiments described later in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below, but may be implemented in various different forms, and only these embodiments make the disclosure of the present invention complete, and the general knowledge in the technical field to which the present invention pertains. It is provided to completely inform the scope of the invention to the possessor, and the invention is only defined by the scope of the claims. The same reference numerals refer to the same components throughout the specification.

In describing embodiments of the present invention, if it is determined that a detailed description of a known function or configuration may unnecessarily obscure the subject matter of the present invention, a detailed description thereof will be omitted. In addition, terms to be described later are terms defined in consideration of functions in an embodiment of the present invention, which may vary according to the intention or custom of users or operators. Therefore, the definition should be made based on the contents throughout the present specification.

Prior to the description, the meaning of terms used in the present specification will be briefly described. It should be noted that the description of terms is to aid the understanding of the present specification and is not used in the sense of limiting the technical idea of the present invention unless explicitly described as limiting the present invention.

-Chunk

As an analysis unit combining a plurality of words constituting a sentence in a foreign language, in an embodiment, a sentence may be divided into a plurality of chunks according to intonation for pronouncing the sentence. In addition, chunks may be generated based on important words included in sentences. In addition, the speaking hint chunk according to the embodiment extracts a plurality of words from the exemplary answer sentence, and then randomly arranges the extracted words and provides the extracted words to the user, thereby recognizing the speaking hint chunk before the user utters, and using this to recognize the speaking hint chunk. Make it so that it can ignite. Speaking hint chunks provided in the embodiment may be provided in the embodiment to inspire English speaking learning motivation to learners whose English speaking ability is below a certain level, and to quickly improve the speaking learning effect.

-Learning quarter

It is another learning situation that can be followed according to the English sentences spoken by the user and the selected learning situation. In the embodiment, the learning branch may vary according to the sentence output in the learning situation, and the next level learning branch corresponding to the recognized user's speaking difficulty is generated by recognizing the speech spoken by the user in each learning situation. I can.

-Context

Context is an intention that considers surrounding situations, time, and environment beyond the simple superficial meaning of text, and is defined as context and context. In the embodiment, the context is the implication of the text analyzed to continue the conversation with the learner who speaks English.

1 is a diagram showing a foreign language speaking learning system according to an embodiment.

Referring to FIG. 1, a system for learning to speak a foreign language according to an embodiment may include a learning server 100 and a user terminal 200. The learning server 100 distributes the foreign language speaking learning application according to the embodiment to the user terminal 200 in a distributed manner, and updates the learning application installed in the user terminal 200. The learning server 100 may pre-build and store learning context data for driving a foreign language speaking learning application, model answer sentence data for each learning situation, chunk data of a model answer sentence, and learning branch data according to the chunk and spoken words. In addition, an artificial intelligence speech recognition program and reference data for speech evaluation including pronunciation, intonation, and size of a pronounced sentence may be pre-built.

The user terminal 200 receives and installs a foreign language speaking learning application from the learning server, and periodically updates the application so that the user can use the learning application through his terminal.

2 is a diagram showing a data processing block of the learning server 100 according to an embodiment.

2, the learning server 100 according to the embodiment includes a database 110, a chunk output module 120, a first evaluation module 130, a second evaluation module 140, and a learning branch generation module 150. ) Can be included.

The database 110 stores a series of data necessary for performing foreign language speaking learning and learning evaluation according to the embodiment. For example, the learning data 110 is stored in the database, and the learning data is necessary for determining the learning situation and speaking evaluation, such as the learning situation data, the best answer sentence data for each learning situation, the expected answer sentence, and the chunk of the model answer sentence and the expected answer sentence May contain a set of data. In the embodiment, the learning situation is information on various situations in everyday life that enables students to learn conversation. Specifically, information on various situations and conversations such as "choose a restaurant", "ask a way", "talk about hobby life", and "choose a movie" Can be. In the embodiment, the learning situation is pre-built and stored as video or audio. The parent method answer sentence data for each learning situation is the parent method answer sentence or expected answer sentence for each question presented to the user in order to induce the user's utterance in the learning situation. In an embodiment, a plurality of expected answer sentences may be configured according to the user's foreign language proficiency.

The chunk output module 120 outputs the learning situation selected by the user, extracts the sentence chunk of the model answer previously stored for each learning situation at the time the output is completed, and displays it as a speaking hint chunk. For example, when the user selects a learning situation called "order at a restaurant", the chunk output module 120 waits for the user's English utterance input after outputting a pre-built selected learning situation image. At this time, the chunk output module may extract a chunk of the best answer sentence matched to “order at a restaurant” selected by the user during the waiting time and display it as a speaking hint chunk.

Meanwhile, in an embodiment, the model answer and the speaking hint chunk extracted according to the user's English proficiency may be different. Specifically, in the learning context of “Order at a restaurant” selected by the user, “May I take your order?” When the sentence "" is outputted, information on the speaking hint chunk extracted from the sentence of the best answer of various levels matched with the output sentence is provided. Specifically, the sentences for the best answer with low difficulty may be “I'll have a shrimp burger and coke”, “shrimp burger and coke, please”, etc., and at this time, speaking by extracting a plurality of words constituting the above sentence You can print hint chunks. Another example is I'll have shrimp burger with avocado and Diet Coke. Please remove cucumber and use this coupon. I'll show my cell phone” extracts and outputs the speech hint chunk from multiple sentences. In an embodiment, the difficulty of a sentence for a model answer may be determined by the number of words, vocabulary, and information included in the sentence, and may be stored in advance.

The primary evaluation module 130 receives the user's voice signal, classifies the input voice signal into chunks, and compares the classified chunk with the chunk matched with the best answer sentence. For example, the primary evaluation module 130 calculates a match rate by comparing each of the classified chunks with words included in the chunks of the best answer sentence, and when the calculated match rate is greater than or equal to a set value, Let's create a learning branch according to the chunk.

In an embodiment, the primary evaluation module 130 includes the number of words included in the chunk of the voice signal input by the user exceeding the number of words included in the chunk of the best answer sentence, and each of the words included in the chunk of the voice signal input by the user The difficulty level and the context of the learning situation are analyzed, and if the analyzed context is above a certain level, a learning branch is generated based on the voice signal chunk input by the user.

The secondary evaluation module 140 evaluates the utterance evaluation element. In the embodiment, the speech evaluation factor is a factor for evaluating the user's speech fluency. In the embodiment, the intonation, pronunciation, size, and speed of the input voice signal may be included. The secondary evaluation module 140 provides the evaluation result of each of the speech evaluation elements compared with the model answer sentence data previously stored in the database, and when the average score of the speech evaluation element is less than a preset minimum value, the next learning branch is not generated. , It is possible to repeatedly learn a specific sentence by outputting the previous learning situation again. In addition, in an embodiment, the secondary evaluation module 140 may analyze all voice data in which the user uttered a specific sentence, visually convert how each evaluation element is changing for each episode, and provide it to the user. Through this, the user can intuitively grasp which part of the speech evaluation element is vulnerable when speaking his or her English, and can objectively grasp whether his or her ability is increasing according to the amount of speaking practice.

The learning branch generation module 140 learns according to the user's voice signal evaluation result transmitted from the first evaluation module 130 and the second evaluation module 140, the learning situation exposed to the user, and the last sentence included in the learning situation. Extract the branch. For example, in the learning branch generation module 140, the last sentence output in the learning situation “choose a restaurant” is “Would you like to go to a seafood restaurant? In the case of, a plurality of situations that can appear as answers to the output sentence may be extracted as a learning branch. In an embodiment, learning branch 1_ “Ok, I like Sea food”, learning branch 2_ “No, I want to eat steak”, and the like may be extracted.

In addition, in the embodiment, the learning branch generation module 140 extracts the speaking hint chunk from the sentence matched to the learning situation and displays it to the user, analyzes the voice uttered by the user, and creates a plurality of next learning branches according to the analysis result. Create and enter the next learning branch selected by the user.

In addition, in the embodiment, when the user does not utter the output chunk and utters another sentence, the learning branch generation module 140 recognizes this and generates a new learning branch. Continuing the above-described example, the user “Would you like to go to a seafood restaurant?” included in the learning context information. In response to the last sentence, "sorry, I don't want to eat anything because I'm not feeling well right now." In the case where the user directly utters and utters the sentence of, the learning branch generation module 140 recognizes the implications of the sentence uttered by the user. Thereafter, the learning branch generation module 140 may extract the next learning situation. In an embodiment, the learning branch generation module 140 may analyze sentences included in the learning context data to extract a sentence uttered by the user and a plurality of learning situations representing contexts of a certain level or higher as a next learning branch. Specifically, in the embodiment, learning situations such as "Ask the sick place" and "Find a hospital" may be presented as the next learning quarter, or the sentence uttered by the user and the learning situation with the highest context may be automatically output as the next learning quarter. .

In addition, in the embodiment, when the speaking hint chunk according to the output learning situation is not displayed, the learning branch generation module 150 recognizes the inputted user's speech signal after outputting the learning situation, so that words and chunks included in the speech signal are not displayed. And a learning branch corresponding to the voice signal input by the user is generated according to the recognition result by grasping the sentence accuracy, recognizing the implication of the user's voice signal.

In addition, in the embodiment, the chunk output module 120 randomly combines n words constituting the chunk of the model answer sentence matched to the learning situation selected by the user, and n! Can create utterance hint chunks. In the embodiment, the speaking hint chunk is a chunk in which some words constituting a keyword of a good answer sentence or some words constituting a good answer sentence are combined in order to induce a user to appropriately answer an English sentence output last in a learning situation. . Specifically, the sentence included in the next learning quarter is “What do you call this place?” In the case of, the chunk can be generated in a different order and the number of words extracted from the words What, do, you, call, this, and place, which are words constituting a sentence. For example, speaking hint chunks are [called, this place, What do], [this place, called, What do], [called, What do, this place], [What do, called, this place], [this place, It can be created as What do, called], [What do, this place, called], etc.

Hereinafter, a method of learning to speak a foreign language will be sequentially described. Since the operation (function) of the foreign language speaking learning method according to the embodiment is essentially the same as that of the foreign language speaking learning server and the system, the overlapping description with FIGS. 1 and 2 will be omitted.

3 is a flowchart illustrating a data processing process implemented in a learning server of a foreign language speaking learning system according to an embodiment.

In step S310, learning situation data is stored in the database of the learning server. The learning situation data is a video that presents a conversation situation to induce a user to speak a foreign language, and specific conversation situations such as'introducing yourself' and'finding a way' may be stored for each level of difficulty. In an embodiment, the learning context data may be stored as video or audio format data.

In step S320, the best answer sentences for each learning situation are stored in the database of the learning server for each difficulty level, and learning branches matching the parent method answer sentences are also stored for each difficulty level. In step S330, the parent method answer sentence is divided into chunks, and in step S340, the words included in the divided chunks are generated according to the model answer sentence. In an embodiment, the speaking hint chunk may be generated by extracting two or more words from the chunks separated from the best answer sentences and randomly arranging the order of the extracted words. Specifically, if n words are included in the chunk, 2 to n-1 words may be extracted in step S340. In this case, the number of cases of word extraction for generating the speaking hint chunk may be calculated by Equation 1.

(n 은 모범 답안 문장을 구성하는 단어 수)Equation 1: Number of word extraction cases for generating speaking hint chunks =

(n is the number of words that make up the best answer sentence)

In the embodiment, since a speaking hint chunk is generated by selecting 2 or more and n-1 or less words from a model answer sentence consisting of n words, the speaking hint chunk that can be generated from a model answer sentence including n words The number can be calculated through Equation 1 using a combination. In Equation 1 according to the embodiment, the number of cases in which the speaking hint chunk is formed by selecting one word or all n words from the best answer sentence is excluded, but the speaking hint chunk generation is performed according to the user's foreign language learning ability. The number equation for extracting the word for may be set differently.

In the embodiment described in Equation 1, the final hint chunk is generated by combining the arrangement order of the words after two or more words are extracted, so the number of arrangement cases for each word extraction is calculated by Equation 2 and is calculated as a plurality of extracted chunks. Hint chunk sentences may be generated.

Equation 2: Number of array cases for each word extraction =

(n is the number of words that make up the best answer sentence)

Equation 2 is an equation using permutation that calculates the number of cases in which r out of n different numbers are selected and arranged in consideration of the order. In an embodiment, the number of speaking hint chunks that can be calculated in a parent method answer sentence including n words is the number of cases in which 2 to n-1 words are selected and then arranged as described in Equation 2 It can be calculated by the summation formula. In the embodiment, the number of cases in which one word is selected from a sentence of the best answer or all of the n words are selected to list the selected word, but a hint chunk is composed of one or n words according to the user's foreign language learning ability. When extracting as words, an equation for calculating the number of arrangement cases for each word extraction may also be different.

In step S350, speaking hint chunks matched for each difficulty level according to the parent method answer sentence and the model answer sentence are registered.

4 is a flowchart illustrating a data processing process performed by a user terminal installed with a learning server or a speaking learning application according to an embodiment.

Referring to FIG. 4, when the application is executed in step S400, a user terminal that installs a speaking learning application outputs a learning situation according to a user selection and outputs a speaking hint chunk corresponding to the learning situation. In step S410, the user's voice information input to the smart terminal is recognized. In step S420, chunks are classified from the recognized voice information, and each of the classified chunks is analyzed. In this case, the number of words included in each of the classified chunks, the level of each word, and sentence contents may be analyzed.

In step S430, the similarity between the analyzed chunk in the user's voice and the speaking hint chunk is calculated. In step S440, the similarity of the chunk analyzed from the user's voice and the speaking hint chunk is compared to determine whether the similarity is greater than or equal to a preset value. In step S440, the degree of similarity may be calculated by determining the word matching rate and the meaning matching rate included in each chunk. If the calculated similarity is less than the preset value, step S450 is entered to recognize the user's voice signal. If the calculated similarity is greater than or equal to the preset value, the process proceeds to step S460, and the next learning branch to be entered is extracted according to the similarity. In the embodiment, when a plurality of speaking hint chunks are presented in step S460, a hint chunk that is most similar to a voice spoken by a user among the plurality of hint chunks is identified, and a different learning situation is entered by entering a learning branch matched to the identified hint chunk. You can continue learning by printing it out.

In step S470, evaluation scores of speech evaluation elements of the user's voice signal are calculated. In the embodiment, evaluation scores for each speech evaluation element such as pronunciation, speed, intonation, and size are calculated and converted into visual data so that the user can intuitively quickly know which evaluation elements are vulnerable.

In step S480, the average score of the user's utterance evaluation factor is compared with a preset minimum score. If the average of the utterance evaluation factor scores is less than the minimum score, step S490 goes to step S490, outputs the sentence of the model answer matched with the output learning situation, and induces repeated speech of the model answer. If the average of the score for the utterance evaluation element exceeds the minimum score, the process goes to step S500, outputs the evaluation score for each utterance evaluation element, and outputs the next learning branch extracted or generated according to the user's answer.

5 is a diagram illustrating a data processing procedure for recognizing an implication of a user's voice signal according to an embodiment.

Referring to FIG. 5, in step S452, a user terminal analyzes a voice signal to generate chunks by dividing an input sentence into a plurality of words. In step S454, the chunk of the voice uttered by the user is analyzed, and a context between the chunk of the voice uttered by the user and the output learning situation is calculated. In step S456, the number of words per chunk and the word difficulty uttered by the user are evaluated, and in step S458, the next learning branch is extracted according to the evaluation result. For example, in step S458, a plurality of learning situations matched according to the evaluation result may be extracted as the next learning branch. In an embodiment, two learning situations having the highest context values may be extracted as a next learning branch according to a chunk uttered by a user and a context value calculated for each learning situation.

6 and 7 are diagrams showing an embodiment of an actual implementation interface of a foreign language speaking learning application.

As shown in (a) of FIG. 6, in the embodiment, when a foreign language speaking learning application is executed in the user terminal, various learning situations are presented according to the user's English level. When the user selects a specific learning situation, as shown in Fig. 6B, the image data constructed as the learning situation is reproduced, and the user's speaking is induced. 6C and 6D are diagrams showing output of a learning branch according to an embodiment. In an embodiment, a learning branch as shown in FIGS. 6C and 6D may be generated and output according to a result of analyzing a speech signal input by a user and a result of analyzing a context with an output learning situation. The user checks the output learning branch, selects one of a plurality of learning branches, and utters an appropriate answer for the learning branch. As shown in (c) and (d) of FIG. 6, it is possible to induce a user's utterance without outputting a hint chunk according to a user selection or program setting.

7E and 7F are diagrams showing speech hint chunks according to an embodiment. As shown in (e) and (f) of FIG. 7, the speaking hint chunk extracts the model answer sentence according to the learning situation and the user's voice recognition result, and then generates and outputs the speaking hint chunk from the extracted model answer sentence. can do. When the user displays the speaking hint chunk, the user checks the hint chunk and utters a sentence appropriate to the learning situation. After the utterance, as shown in FIG. 7(g), the user terminal recognizes the voice input sentence and analyzes and evaluates the speech signal for each utterance evaluation element. In an embodiment, as shown in (h) of FIG. 7, each evaluation element such as intonation, speed, and size is evaluated, and converted into visual data so that the user can accurately and intuitively recognize his or her speech element evaluation information. Print it out.

The foreign language speaking learning server and method according to the embodiment presents the learning situation to the user and induces the utterance of English sentences, and then presents the exemplary answer sentences in chunks. It does not stop, and improves the English speaking learning effect by allowing the user to speak after writing using the suggested speaking hint chunk.

The voice signal input by the user is evaluated for each chunk according to evaluation factors including pronunciation, intonation, fluency, and accuracy, so that the user's English speaking ability can be more accurately evaluated. In addition, by providing a variety of learning branches according to the accuracy and difficulty of the chunks selected by the user and the sentences spoken in English, even users who are very good at speaking English are provided with a learning branch corresponding to their English proficiency and continuously practice speaking English. To be able to do it. In addition, users who are not familiar with speaking English are provided with a learning branch appropriate to their level of learning so that they can gradually improve their speaking skills.

The disclosed contents are only examples, and various changes may be made by those of ordinary skill in the art without departing from the gist of the claims claimed in the claims, so the scope of protection of the disclosed contents is limited to the above-described specific It is not limited to the examples.

Claims (14)

In the foreign language speaking learning server,
A database for storing learning data including learning context data, model answer sentence data for each learning situation, chunk of the model answer sentence, and learning branch which is a learning situation matched with the user's answer or the model answer sentence;
A chunk output module for generating and displaying a speaking hint chunk in the best answer sentence for each learning situation during a time period for waiting for the user's speech input after the learning situation selected by the user is output;
A primary evaluation module for receiving a user's voice signal, classifying the input voice signal into chunks, and comparing the classified chunk with the chunk of the model answer sentence;
A second evaluation module for recognizing the input speech signal and evaluating each speech evaluation element including intonation, pronunciation, and size; And
A learning branch generation module for extracting a learning branch according to a speech evaluation result transmitted from the first evaluation module and the second evaluation module; A foreign language speaking learning server containing a.
According to claim 1, The first evaluation module; silver
The match rate is calculated by comparing the chunks classified in the user's voice input with the words included in the chunks of the best answer sentence, and if the calculated match rate is greater than or equal to a set value, learning branches according to the chunks included in the voice signal input by the user A foreign language speaking learning server, characterized in that to generate a.
The method of claim 1, wherein the speaking hint chunk is
A foreign language speaking learning server, characterized in that it is generated to induce a user's utterance by extracting two or more of a plurality of words constituting the best answer sentences matched for each learning situation and arranging them at random.
According to claim 1, The first evaluation module; silver
When the number of words included in the chunk of the voice signal input by the user exceeds the number of words included in the chunk of the sample answer sentence, the difficulty level of each word included in the chunk of the voice signal input by the user and the context of the learning situation ( context) and, when the analyzed context is above a certain level, extracting a learning branch based on a voice signal chunk input by a user.
According to claim 1, The second evaluation module; silver
Provides the evaluation result of each of the utterance evaluation elements compared with the model answer data previously stored in the database, and if the average score of the utterance evaluation element is less than a preset minimum value, the previous learning situation is repeated without extracting the next learning branch. Foreign language speaking learning server, characterized in that to output.
The method of claim 1, further comprising: the learning branch generation module; silver
When the hint chunk matched to the output learning situation is not displayed, the user's voice signal input after the learning event is output is recognized, the accuracy of words, chunks and sentences included in the voice signal is identified, and the user's voice A foreign language speaking learning server, characterized in that for generating a learning branch corresponding to the voice signal input by the user according to the recognition result by recognizing the signal implications.
According to claim 1, The chunk output module; silver
After randomly selecting from 2 to n-1 of n words constituting the chunk of the model answer sentence matched to the learning situation selected by the user, the selected words are combined and n! A foreign language speaking learning server, characterized in that generating three utterance hint chunks.
In the method of learning to speak a foreign language,
The foreign language speaking learning server includes: a first step of storing learning data including learning context data, model answer sentence data for each learning situation, chunk of the model answer sentence, and learning branch, which is a learning situation matched to the user's answer or the model answer sentence;
A second step of generating and displaying, at the foreign language speaking learning server, a speaking hint chunk from the best answer sentences for each learning situation during a time period for waiting for the user's utterance input after the learning situation selected by the user is output;
The foreign language speaking learning server receives the user's voice signal, classifies the input voice signal into chunks, and compares the classified chunk with the chunk of the model answer sentence;
A fourth step of the foreign language speaking learning server recognizing the input voice signal and evaluating each speech evaluation element including intonation, pronunciation, and size; And
A foreign language speaking learning method comprising; a fifth step of extracting, by the foreign language speaking learning server, a learning branch according to the speech evaluation result transmitted from the first evaluation module and the second evaluation module.
According to claim 8, The third step; Is
The match rate is calculated by comparing the chunks classified in the user's voice input with the words included in the chunks of the best answer sentence, and if the calculated match rate is greater than or equal to a set value, learning branches according to the chunks included in the voice signal input by the user A foreign language speaking learning method, characterized in that to generate.
The method of claim 8, wherein the speaking hint chunk is
A foreign language speaking learning method, characterized in that it is generated to induce a user's utterance by extracting at least two of a plurality of words constituting a sentence for a model answer matched by learning situation and arranging them at random.
According to claim 8, The third step; Is
When the number of words included in the chunk of the voice signal input by the user exceeds the number of words included in the chunk of the sample answer sentence, the difficulty level of each word included in the chunk of the voice signal input by the user and the context of the learning situation ( analyzing context); And
Extracting a learning branch based on the voice signal chunk input by the user when the analyzed context is greater than or equal to a predetermined level; A foreign language speaking learning method comprising a.
According to claim 8, The fourth step; Is
Providing an evaluation result of each of the speech evaluation elements compared with the model answer data previously stored in the database; And
If the average score of the speech evaluation factor is less than or equal to a preset minimum value, outputting a previous learning situation once again without extracting a next learning branch; A foreign language speaking learning method comprising a.
According to claim 8, The fifth step; Is
When the hint chunk matched to the output learning situation is not displayed, the user's voice signal input after the learning event is output is recognized, the accuracy of words, chunks and sentences included in the voice signal is identified, and the user's voice A foreign language speaking learning method, characterized in that for generating a learning branch corresponding to the voice signal input by the user according to the recognition result by recognizing the signal implications.
According to claim 8, The second step; Is
Randomly selecting from 2 to n-1 of n words constituting the chunk of the model answer sentence matched to the learning situation selected by the user; And
Selected words are combined and n! Generating four utterance hint chunks; foreign language speaking learning method comprising a.

Images