Intelligent non-face-to-face foreign language education system and method for non-native language learner

Abstract

An embodiment of the present invention may comprise: a server; and a user terminal which accesses the server and receives educational content to learn a non-native language, wherein the server includes: a service provision unit for providing a web service and an intermediary service for video-based education between a leaner and an instructor; a question development unit for constructing a model for learner-specific reinforcement learning to select questions and provide same to the user terminal on the basis of learner information and question difficulty information; a reading evaluation index calculation unit for calculating a reading evaluation index of a non-native language learner; and a content matching unit for providing content corresponding to an evaluation result of the reading evaluation index calculation unit to the user terminal.

Description

Intelligent non-face-to-face foreign language education system and method for non-native language learners

[0001] The present invention relates to an intelligent non-face-to-face foreign language education system and method for non-native language learners, which performs a reading index evaluation and a reading level evaluation for non-native language learners, thereby enabling foreign language reading education.

[0002] Edutech is a compound word of education and technology, and refers to improving existing services or providing new services by connecting ICT technology to education.

[0003] Amid the recent pandemic, which has highlighted the importance of the non-face-to-face (Untact) industry, unprecedented online classes and attendance at elementary, middle, and high schools are increasing interest in the non-face-to-face edutech industry. The growing adoption of smart devices and the growing number of Generation Z members familiar with them are also accelerating the spread of edutech.

[0004] Furthermore, a survey conducted by an educational institution in April 2023 revealed that seven out of ten adult learners prefer digital learning methods. This trend was also observed among foreigners living in Korea or abroad.

[0005] Accordingly, as the importance of online education grows, problems such as the difficulty of self-directed learning, uniform education provided regardless of learner level, and insufficient diagnosis of learning level are becoming more prominent.

[0006] The present invention has been devised to solve the above-mentioned conventional problems, and the purpose of the present invention is to provide an intelligent non-face-to-face foreign language education system and method for non-native language learners, which can provide customized content by evaluating the reading level and reading evaluation index of non-native language learners so as to suggest an educational environment and guidelines suitable for the learning environment of learners learning non-native languages.

[0007] An embodiment of the present invention includes a server and a user terminal that accesses the server to receive educational content in order to learn a non-native language, and the server may include a service provider that provides a web service and an intermediary service for video education between learners and instructors, a question development unit that builds a model for learner-specific reinforcement learning and selects questions based on learner information and question difficulty information and provides the selected questions to the user terminal, a reading evaluation index calculation unit that calculates a reading evaluation index of a non-native language learner, and a content matching unit that provides content to the user terminal according to the evaluation results of the reading evaluation index calculation unit.

[0008] In the above embodiment, the item development unit may include a question bank linked to its own server or a cloud server as a database for storing selected items, an item selection module for selecting items based on item difficulty information calculated based on the results of a learner's basic test and providing the selected items to the user terminal, and an item creation module for providing the selected items to the user terminal based on the evaluation items selected by the item selection module based on the learner's evaluation level.

[0009] Additionally, the question creation module can create questions on the user terminal based on learner information and question difficulty information, and can recalculate the difficulty of subsequent questions presented to the learner by predicting the correct answer rate for the next question during the problem-solving process.

[0010] Additionally, the question creation module can create questions with difficulty levels that reflect the learner's learning information, and then adjust the difficulty of the next question based on whether the user's terminal inputs a correct or incorrect answer.

[0011] In addition, the question creation module calculates the expected value of the current question based on whether the current question is correct or incorrect and the question difficulty value, and applies the calculated expected value to the reinforcement learning model of the Markov Decision Process (MDP) to select and create the next question with the same difficulty level or a question of the same category but different level.

[0012] In the above embodiment, the reading evaluation index calculation unit may include a reading evaluation index measurement module that collects the learner's learning history by distinguishing between a native language environment and a non-native language environment, subdivides the collected learning history by age and education level, and calculates a reading evaluation index; and a reading evaluation index judgment module that implements natural language processing in the native language or non-native language through an answer sheet submitted through the solution of a question, and determines the learner's non-native language reading level based on the learner's native language learning achievement and non-native language grade by applying important word weights through a word matrix distribution technique between documents.

[0013] The reading evaluation index measurement module of the above example can calculate the reading evaluation index by applying the formula of the Korean-style English book readability index measurement formula = (Vocabulary difficulty) X Sentence difficulty (SCI) = (sum of the 'difficulty X TOKEN ratios of each group' resulting from the execution of the modified Range program) X (SCI obtained using the syntactic formula of Botel and Granowsky).

[0014] Another embodiment of the present invention utilizes any one of the above embodiments, comprising: a step S230 of confirming a learner's learning level; and a step S300 of recommending content according to the learner's learning level; Including, step S230 comprises: step S231) where the server collects and stores learner information from at least one of a user terminal and an institution terminal; step S232) where the server sets a learner level through the collected age and grade of the learner; step S233) where the server sets the difficulty of the question according to the learner level; step S234) where the server selects and recommends questions having a learning level and a set difficulty to the user terminal; step S235) where the server presents the selected test questions to the user terminal; step S236) where the server scores the answer sheet of the test question input from the user terminal and recalculates the question difficulty according to the result to adjust the question difficulty range; step S237) where the server presents the next recommended question having the adjusted question difficulty; step S238) where the server recalculates the question difficulty according to whether the presented question is correct or incorrect and presents the next question, repeating the process, but checking whether the set order of questions has been reached; S239) The server may include a step of storing the results in a database when the set sequence of items has been presented to the learner terminal.

[0015] An embodiment of the present invention includes a server and a user terminal that accesses the server to learn a non-native language and receives educational content, wherein the server sequentially provides a plurality of set evaluation items to the user terminal, and includes a level evaluation unit that evaluates a foreign language reading level of a non-native language learner based on the results thereof, and a content matching unit that provides content to the user terminal according to the evaluation results of the level evaluation unit, wherein the plurality of evaluation items are characterized by being composed of a first group consisting of a plurality of evaluation items having a large range of difficulty from the initial start item to the evaluation items that are subsequently presented, and a second group consisting of evaluation items of the same level that are sequentially presented and the level of difficulty varies depending on whether the answer is correct or incorrect.

[0016] The above embodiment further includes a question development unit that stores and selects foreign language learning questions provided to a user terminal, and the question development unit includes a question bank that is linked to its own server or a cloud server as a database that stores the selected questions, a question selection module that selects questions based on question difficulty information calculated according to the results of a basic test conducted by a learner and provides the selected questions to the user terminal, a question issuing module that provides the selected questions suggested based on the evaluation questions selected by the question selection module according to the learner's evaluation level to the user terminal, and a question evaluation module that maintains, deletes, and changes the level of the questions stored in the question bank based on at least one of the number of questions, the evaluation level of a learner who inputted a correct answer, and the evaluation level of a learner who inputted an incorrect answer.

[0017] In addition, the level evaluation unit includes a CAT (Computerized Adaptive Testing) logic that increases the difficulty of the next evaluation question when the user terminal's correct answer to the evaluation question is confirmed, and decreases the difficulty of the next question when the correct answer is incorrect, and an artificial intelligence model that executes the CAT logic, but calculates the correct answer rate for other questions of the same level based on whether the previously given evaluation question was correct or not, and predicts and adjusts the level of difficulty of the next question.

[0018] And, the artificial intelligence model can be the SAKT (Self-Attentive Knowledge Tracing) model, which is a model that applies the attention mechanism of the Transformer to knowledge tracing.

[0019] In addition, the level evaluation unit is characterized by providing the user terminal with evaluation items belonging to the second group in succession at the same level, and moving the level of the evaluation items according to whether there is a continuation of correct and incorrect answers.

[0020] And, another embodiment of the present invention includes a step of registering learner information including a native language, learning information such as native language level, and at least one of age and affiliated school from at least one of a user terminal and an institution terminal in a server, b) a step of determining a starting level through the learner information in the server, transmitting a plurality of evaluation questions according to the starting level to the user terminal, and calculating the level of the learner by adjusting the level of the evaluation questions based on whether the evaluation questions are answered correctly or incorrectly, and c) a step of providing, in the server, customized content according to the learner's evaluation result to the user terminal, wherein in step b), the plurality of evaluation questions are composed of a first group consisting of a plurality of evaluation questions having a large range of difficulty from an initial starting question to subsequent evaluation questions, and a second group consisting of evaluation questions of the same level that are answered consecutively and have a level of difficulty that varies depending on whether the answer is answered correctly or incorrectly.

[0021] In addition, step b) includes: step b-1) setting the starting level of the initially transmitted evaluation item; step b-2) if the starting level is greater than the minimum level, a step of issuing an evaluation item corresponding to the corresponding level to the user terminal, and sequentially issuing evaluation items of the first group by adjusting the difficulty level according to the result; step b-3) confirming the correct or incorrect answer through the user terminal's reply to the final evaluation item of the first group, and determining the level calculated by the final evaluation item as the starting level of the second group; step b-4) providing a plurality of evaluation items corresponding to the starting level in the second group to the user terminal in batches or sequentially, and observing a pattern according to the correct and incorrect answers of the user terminal, and changing the level of the next evaluation items if the answer is consecutively incorrect or correct; step b-5) if a pattern is observed and the level difficulty level is changed, a step of sequentially or collectively presenting a plurality of evaluation items of a set number having the changed level difficulty level to the user terminal, and observing a pattern according to the correct and incorrect answers of the user terminal, and changing the level of the next evaluation items; b-6) A step is included in calculating a level based on whether the user terminal gave correct or incorrect answers to multiple evaluation items presented sequentially or in batches in the second group, and calculating an average value of the calculated levels to determine the final level.

[0022] And, step b-6) can lower the final level by a set level if the correct and incorrect answers to the evaluation items are mixed alternately.

[0023] In addition, step b-6) is characterized by lowering the level by a set level when a pattern is observed in which a correct answer follows an incorrect answer and then an incorrect answer appears again.

[0024] In addition, the method further includes a step of evaluating the question based on at least one of the number of times the question was asked, the evaluation level of a learner who entered a correct answer, and the evaluation level of a learner who entered an incorrect answer, stored in a question bank linked to an in-house server or a cloud server as a database for storing selected questions, and deciding whether to maintain, delete, or change the level.

[0025] Therefore, the present invention provides detailed individual learning consulting by dividing the learner's level in the local environment into multiple levels, and by providing not only individual progress learning but also step-by-step learning, it can provide motivation so that the learner can reach the goal without losing interest by predicting the performance and suggesting the necessary parts for each individual.

[0026] In addition, the present invention can provide a reading index and evaluation system, enabling evaluation and guidance of learning abilities for all age groups capable of learning, and can further enhance learning efficiency for non-native language learners by providing customized content based on an evaluation of the learner's foreign language reading level.

[0027] Figure 1 is a block diagram illustrating an embodiment of the present invention.

[0028] Figure 2 is a block diagram illustrating the server of Figure 1.

[0029] Figure 3 is a block diagram illustrating the item development department.

[0030] Figure 4 is a block diagram illustrating a reading evaluation index calculation unit.

[0031] [Correction pursuant to Rule 91, dated 02.10.2024]

[0032] Figure 6 is a diagram illustrating the item selection process.

[0033] Figure 7 is a drawing showing an example of a level chart in an embodiment of the present invention.

[0034] Figure 8 is a graph showing an example of level calculation using delta values.

[0035] Figure 9 is a flowchart illustrating an intelligent non-face-to-face reading level assessment method for non-native language learners according to the present invention.

[0036] Figure 10 is a flowchart illustrating step S100.

[0037] Figure 11 is a flowchart illustrating another embodiment of level evaluation.

[0038] Figure 12 is a flowchart illustrating step S300.

[0039] While the present invention is susceptible to numerous modifications and embodiments, specific embodiments are illustrated in the drawings and described in detail. This is not intended to limit the present invention to specific embodiments, but rather to encompass any modification, equivalent, or alternative within the spirit and technical scope of the present invention for connecting and / or securing structures extending in different directions.

[0040] The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise.

[0041] In this specification, terms such as “include” or “have” are intended to specify the presence of a feature, number, step, operation, component, part or combination thereof described in the specification, but should be understood not to exclude in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts or combinations thereof.

[0042] Hereinafter, a preferred embodiment of an intelligent non-face-to-face foreign language education system and method for non-native language learners according to the present invention will be described in detail with reference to the attached drawings.

[0043] FIG. 1 is a block diagram illustrating an embodiment of the present invention, and FIG. 2 is a block diagram illustrating the server of FIG. 1.

[0044] Referring to FIGS. 1 and 2, the present invention may include a plurality of user terminals (200), an organization terminal (300), a server (100), and a database (400).

[0045] The user terminal (200) accesses the server (100) and receives content to learn a non-native language.

[0046] The institution terminal (300) is an educational institution terminal (300) that conducts non-native language learning for multiple user terminals (200) connected through a server (100). For this purpose, the server (100) provides a video education system, and the institution terminal (300) can conduct native language and / or non-native language learning through the video education system provided by the server (100).

[0047] The server (100) can provide a web service that enables video education to support video education of an institution terminal (300) and a user terminal (200), information sharing through the web, and data search.

[0048] In addition, the server (100) conducts a non-native language reading index evaluation test for non-native language learners based on artificial intelligence, analyzes the results, outputs a diagnostic result report, and provides artificial intelligence-based personalized consulting that specifically suggests future reading methods and learning directions by considering the individual learner's native language environment, the learner's reading index, and the learner's knowledge level.

[0049] The database (400) may be built as an integrated system with the server (100) or as a separate, independent system. The database (400) may include a learner DB (410) in which learner information is stored, a content DB (420) in which native and non-native language educational content is stored, and an evaluation DB (430) in which learners' non-native language reading indices and learning evaluation results are stored.

[0050] The learner DB (410) can store personal information of the learner and academic schedule information including progress and completion of educational content currently being taken and previously taken.

[0051] The content DB (420) can store educational content for non-native language learners, such as textbook information provided by the institution terminal (300), various educational materials uploaded by other educational institutions or instructors, and textbooks.

[0052] The evaluation DB (430) can store information on the non-native language reading evaluation index and level of each non-native language learner, evaluation information including the learning level based on the evaluation results of each learner, and information on selected items with the difficulty level adjusted based on the evaluation results of each learner.

[0053] The server (100) may include a service provision unit (110), a question development unit (120), a reading evaluation index calculation unit (130), a content matching unit (140), and a level evaluation unit (150).

[0054] The service provision department (110) provides web services and video education services.

[0055] The video education service can provide a whiteboard and document view screen sharing for video education between learners and institutional terminals (300) (including instructors), a learning dashboard that can confirm attendance through event accumulation such as speaking time, webcam time, and messages, and a video education center management function for conducting individual 1:1 or 1:N sessions.

[0056] The item development unit (120) selects and provides items based on the individual learner's assessment results. Here, the item development unit (120) builds a learning model for individual learner reinforcement learning using an artificial intelligence engine, thereby selecting and providing items based on learner information and item difficulty information. This is illustrated in Figure 3.

[0057] Figure 3 is a block diagram illustrating the item development department.

[0058] Referring to FIG. 3, the item development unit (120) may include a question bank (121) storing questions for non-native language learners, a question selection module (122) that selects questions for each learner, a question creation module (123) that outputs questions selected according to the learner level and evaluation results to a user terminal (200), and a question evaluation module (124) that evaluates the questions.

[0059] The question bank (121) is a database (400) that stores questions (problems) selected by the question selection module (122). The question bank (121) is constructed as a database that stores actual structured and unstructured raw data generated through an artificial intelligence engine (e.g., machine learning) by linking a database (400) structure designed based on DB modeling to an internal server (100) or a cloud server (100). An example of such a database is presented in FIG. 5.

[0060] Additionally, the question bank (121) according to the present invention can automatically present test questions, including the number of questions produced, the content type of the test question, and the test question type, by dividing them into learning stages and detailed levels. An example of such a question bank (121) is shown in Table 1 below.

[0061]

[0062] The item selection module (122) outputs item difficulty information based on the results of a basic test and the learner information stored in the learner DB (410), and selects items based on the information and provides the results to the user terminal (200). An example of this is introduced in Table 2 below.

[0063]

[0064]

[0065] Referring to Table 2, the item selection module (122) can present a standard value classified into multiple stages based on the test item ratio (%) of the non-native language (e.g., pronunciation, reading, vocabulary, grammar, word set) and the question type and grammar detail table by ratio.

[0066] [Revised 02.10.2024 pursuant to Rule 91] Here, the item selection module (122) can digitize criteria for various question types based on the assessment questions presented for learner testing. Furthermore, combinations of selected items can be constructed in a tree-like form.

[0067] The question presentation module (123) can provide selected questions to the user terminal (200) and / or the institution terminal (300) based on the evaluation questions selected by the question selection module (122) according to the learner's evaluation level.

[0068] For example, when a user terminal (200) logs in and inputs learner data (e.g., native language, native language level, age, nationality, level of non-native language selected for learning), the question creation module (123) retrieves question difficulty information from a question bank and transmits selected evaluation questions to the user terminal (200).

[0069] At this time, the question creation module (123) may, for example, present a question with a difficulty level that reflects the learning data as the first question presented, and then, based on whether the user terminal (200) inputs a correct or incorrect answer, predict the correct answer rate for the next question and adjust the difficulty level of the next question. The process is described with reference to FIG. 6 and Table 3 below.

[0070] First, referring to FIG. 6, the question creation module (123) calculates the expected value of the current question (e.g., reward=0.6) based on whether the first question is correct or incorrect and the question difficulty value, and applies the calculated expected value to a reinforcement learning model converted into a Markov Decision Process (MDP) to select and create the next question or a question of a different level in the same category.

[0071] In addition, the question creation module (123) can adjust whether the first question presented is correct or incorrect and the difficulty level of the next question as described above, and the range of difficulty can be set according to the set level as shown in Table 3 below.

[0072] Table 3 contains values ​​for adjusting the item difficulty level according to level.

[0073] Learning stage question difficulty level note A 0~0.2 difficult problem B 0.21~0.40 slightly difficult problem C 0.41~0.60 average D 0.61~0.80 easy problem E 0.80~1 very easy problem

[0074] For example, item difficulty can be adjusted within the range of -0.5 to 0.5 depending on the learner's learning stage based on Table 3 above.

[0075] In addition, when a user terminal (200) logs in and inputs learner data (e.g., native language, native language level, age, nationality, level of non-native language selected for learning), the question creation module (123) retrieves question difficulty information from the question bank (121) and transmits selected evaluation questions to the user terminal (200). This selection of evaluation questions is implemented by a learner-tailored question selection algorithm based on feedback.

[0076] The learner-tailored item selection algorithm is implemented by a transfer learning (Transformer) model that infers relationships, such as which words a root is related to within a sentence structure.

[0077] At this time, the question presentation module (123) can, for example, present a question with a difficulty level that reflects the learning data as the first question presented, and then adjust the difficulty level of the next question by predicting the correct answer rate of the next question based on whether the user terminal (200) inputs a correct or incorrect answer.

[0078] The question evaluation module (124) evaluates questions stored in the question bank (121) and determines whether to keep or delete them. Here, the question evaluation module may be equipped with an artificial intelligence model equipped with a big data algorithm.

[0079] Accordingly, the item evaluation module (124) re-evaluates the difficulty of each item through at least one of the correct answer rate and incorrect answer rate of each learner by level and the number of times the item was given to each learner by level by the level evaluation unit (150), and can maintain, change, or delete the item according to the evaluation result.

[0080] For example, the item evaluation module (124) can determine that if the correct answer rate of learners of the same level is above or below a set range (e.g., 50-80%), the difficulty of the item can be changed upward or downward, or if the probability that the item was presented to learners of the same level is above or below the set range, the item can be maintained, and if it is below the set range, the item can be deleted.

[0081] Figure 4 is a block diagram illustrating a reading evaluation index calculation unit.

[0082] Referring to FIG. 4, the reading evaluation index calculation unit (130) may include a reading evaluation index measurement module (131) and a reading evaluation index judgment module (132).

[0083] The reading evaluation index measurement module (131) collects the learner's learning history through artificial intelligence by distinguishing between native language environments and non-native language environments, and calculates the reading evaluation index by subdividing the collected learning history by age and academic background.

[0084] Here, the reading evaluation index measurement module (131) can derive readability by measuring "word length, word frequency, word familiarity, sentence length, sentence complexity" for English learners, which is necessary for learners in the environment of the country where the learner resides, rather than the ESL standard evaluation values, which are the measurement methods of the existing Lexile index and AR index used in a specific native language environment.

[0085] For example, the Lexile measure measures 'Sentence Length + Word Frequency', and the AR measure measures 'Average Sentence Length + Average Word Length + Vocabulary Grade Level + The Number of words in book'.

[0086] The reading index and evaluation index used in the Korean language environment are derived by measuring the same “word length, word frequency, word familiarity, sentence length, and sentence complexity” as in English.

[0087] That is, the reading evaluation index measurement module (131) applies the evaluation index of a non-native language environment to the native language environment to calculate the reading evaluation index in the country of residence (home country).

[0088] An example of such a reading evaluation index can be calculated as shown in Table 4 below.

[0089] Classification evaluation index measurement contents (example) Target technology evaluation index example Measurement formula (example) Korean-style English book readability index measurement formula = (Vocabulary difficulty) X Sentence difficulty (SCI) = (Sum of each 'difficulty X TOKEN ratio' of the results of executing the modified Range program) X (SCI obtained using the syntactic formula of Botel and Granowsky) Readability vocabulary elements Word length, word frequency, word familiarity Sentence elements Sentence length, sentence complexity

[0090] The table above shows an example of measuring the readability index of a Korean-style English book. The reading evaluation index measurement module (131) calculates the reading evaluation index using the above measurement formula. The reading evaluation index judgment module (132) recalculates the difficulty of the item and stores the result after determining whether the answer is correct or incorrect after solving the recommended items based on learner information and item difficulty information.

[0091] Here, the reading evaluation index determination module (132) implements natural language processing in the native or non-native language through the answer sheet submitted through the question explanation, and can derive the influence of similarity between the native and non-native languages ​​based on frequency. This similarity derivation can utilize, for example, the cosine similarity algorithm or Jaccard similarity.

[0092] In particular, the reading evaluation index determination module (132) applies important word weights through a word matrix distribution technique between documents, and presents the learner's achievement in the native language learning process and the non-native language grade-based reading level based on the learner's test results, so that content can be recommended based on the learner's reading evaluation index results.

[0093] This reading evaluation index provides detailed guidance on extensive reading and intensive reading based on age, academic achievement, or grade level, as there are differences in educational environments across countries even when reading levels are the same.

[0094] That is, the reading evaluation index determination module (132) determines the reading evaluation index of the user terminal (200) and provides extensive reading and intensive reading guides according to the learner's native language environment.

[0095] The content matching unit (140) calculates the learner's level by accumulating the learner's learning information stored in the database (400) and calculates the learner's learning efficiency by content to recommend customized content for the learner.

[0096] The level evaluation unit (150) combines the logic of the CAT method (Computerized Adaptive Testing) and an artificial intelligence model that executes the CAT logic (Computerized Adaptive Testing) to evaluate the foreign language learning level (e.g., reading level) of a non-native language learner and produce the results.

[0097] The logic of the CAT (Computerized Adaptive Testing) method is that if the learner gets the question right, the difficulty of the next question is increased, and if the learner gets the question wrong, the difficulty of the next question is lowered.

[0098] For example, if the assessment problems presented to the learner are numbered 1 to 100, and each problem is ranked from easiest (1) to most difficult (100), then if the learner solves problem 50 and passes, he or she will be presented with problem 67. If he or she also passes problem 67, he or she will be presented with problem 78. If he or she fails problem 78, he or she will be presented with problem 72.

[0099] And, if the learner passes question 72, he / she solves question 77, and if he / she passes question 77, the final level of this student is determined to be level 77.

[0100] In other words, CAT Logic (Computerized Adaptive Testing) gradually reduces the difficulty level for each problem, so that when the problem is solved, the final difficulty level corresponds to the student's final level.

[0101] However, although CAT Logic (Computerized Adaptive Testing) allows for accurate individual evaluation, if there is an error in the difficulty of the test questions themselves, it can affect the final level determination, so an accurate level measurement is necessary, and for this, the difficulty of the questions must be accurate.

[0102] Therefore, in order to increase the accuracy of the difficulty level, the level evaluation unit (150) of the present invention is equipped with an artificial intelligence model that can adjust the difficulty level when presenting the next problem by calculating the expected correct answer rate for the remaining problems in the problem bank (121) when a student has solved some of the previous problems.

[0103] Here, it is desirable that the artificial intelligence model be the SAKT (Self-Attentive Knowledge Tracing) model, which is a model that applies the attention mechanism of the Transformer to knowledge tracing.

[0104] That is, in the present invention, the level evaluation unit (150) performs individual evaluation through CAT logic (Computerized Adaptive Testing), and calculates the learner's reading evaluation through difficulty adjustment by the SAKT (Self-Attentive Knowledge Tracing) model.

[0105] More specifically, the level evaluation unit (150) sequentially presents evaluation items to the learner according to a level chart established through collaboration between the aforementioned CAT logic (Computerized Adaptive Testing) and the artificial intelligence model, and predicts the correct answer rate for the next question based on whether the current question is answered correctly or incorrectly, adjusts the difficulty of the question presented to the learner, and presents and evaluates the next question to calculate the foreign language level of the non-native learner.

[0106] At this time, the level evaluation unit (150) sets a level chart that divides the learner's level and the level of difficulty into upper, middle, and lower, and calculates a delta value according to the difference in level depending on whether the question presented at each level is correct or incorrect through the level chart.

[0107] And the difficulty of the evaluation items is adjusted so that the gap between delta values ​​gradually decreases to reach an appropriate level.

[0108] An explanation of such level charts and delta values ​​is provided with reference to FIGS. 7 and 8.

[0109] Fig. 7 is a drawing showing an example of a level chart in an embodiment of the present invention, and Fig. 8 is a graph showing an example of level calculation using a delta value.

[0110] Referring to Figures 7 and 8, the level chart is divided into multiple levels of learners in the vertical direction (e.g., K, 1, 2, 3, 4, ...12), and the difficulty of each level is classified as high (0.9), medium (0.6), and low (0.3). Here, levels 1, 2, 3, ...12 are classified based on the American grades. That is, level 1 means the difficulty level of a first-grade student in the United States, and grade 12 corresponds to the English level of a 12th-grade student in the United States. Level K corresponds to the English level used by children before entering school.

[0111] Therefore, the difficulty for each level is determined by adding the level value to the high, medium, and low difficulty values.

[0112] For example, the low difficulty of level K is 0.3, the low difficulty of level 1 is 1.3, which is 1 level added to 0.3 of the low difficulty, and the low difficulty of level 2 is 2.3, which is 0.3 and the level value added to 2.

[0113] Also, the medium difficulty of level 1 is 0.6 plus the level value 1, which is 1.6, and the medium difficulty of level 2 is 0.6 plus the level value 2, which is 2.6.

[0114] The delta value can be considered a transition value between levels based on whether the answer is correct or incorrect. For example, the delta value can be set to any of the following values: 4, 3, 2, 1, or 0.5, and is set for each question.

[0115] For example, the delta value for each item can be set by level as a structure of 3 (K level), 2 (1st level), 1 (3rd level), 1 (4th level), 0.5 (5th level), 0.5 (6th level), 0.5 (7th level), .... 0.5 (12th level) for correct answers, and 4 (K level), 3 (1st level), 2 (2nd level), 2 (3rd level), 1 (4th level), 1 (5th level), 0.5 (6th level), 0.5 (7th level), .... 0.5 (12th level) for incorrect answers.

[0116] Alternatively, if the learner is identified as a beginner through learner information, the delta value for the evaluation items of the first group is set to 0.5.

[0117] That is, in the present invention, the greater the difference between the difficulty level of a question and the learner's level, the greater the difficulty of the next question. This change in difficulty is calculated using the delta value. As shown in the graph of Figure 8, this delta value gradually decreases in magnitude, and the level assessment is conducted by considering the difficulty level of the evaluation questions and the order of questions so that the learner reaches an appropriate level.

[0118] Here, the level evaluation unit (150) can calculate the learner's initial starting level based on information (e.g., study period, overseas training, age, region, grade) input through the user terminal (200) using an artificial intelligence model. In addition, the artificial intelligence model of the level evaluation unit (150) adjusts the difficulty of the next question based on a predicted value based on whether the evaluation question was answered correctly or incorrectly using a big data algorithm.

[0119] And CAT Logic (Computerized Adaptive Testing) selects and presents evaluation items based on difficulty adjustment by an artificial intelligence model, and calculates the learner's evaluation level based on the results.

[0120] In addition, the learner level evaluation items conducted in the level evaluation unit (150) may be composed of a first group (e.g., 40% of all items) that estimates the level by increasing the range of the delta value, and a second group (e.g., 60% of all items) that estimates the final level based on the achievement by repeatedly asking items of the same level.

[0121] For example, Group 1 includes the first 8 questions when there are 20 questions in total.

[0122] For example, Group 2 includes the last 12 questions when there are 20 questions in total. Here, Group 1's starting questions can start at a medium difficulty level of the minimum level, or at a medium difficulty level of N (N>K, where K is the minimum level) higher than the minimum level.

[0123] Additionally, the level of the starting question of Group 2 is the level of the last question of Group 1.

[0124] Additionally, the final evaluation level is determined by the average of the last item and the multiple items preceding it in Group 2.

[0125] The present invention includes an intelligent non-face-to-face foreign language education system for non-native language learners, and below, an intelligent non-face-to-face foreign language education method for non-native language learners is described.

[0126] Figure 9 is a flowchart illustrating an intelligent non-face-to-face education method for non-native language learners according to the present invention.

[0127] Referring to FIG. 9, the present invention includes a step S100 of registering learner information input through a user terminal (200), a step S200 of evaluating the learner's level, a step S300 of providing customized content according to the learner's level, and a step S400 of evaluating questions stored in a question bank (121).

[0128] Step S100 is the step where the server (100) collects and stores learner information from a user terminal (200) and / or an institution terminal (300). The user terminal (200) logs into the server (100) via the web or an app and inputs learning information such as the user's native language and native language level, as well as information such as age and school affiliation. Therefore, the server (100) stores the input information in the learner database (410).

[0129] Step S200 is the step where the server (100) evaluates the learner's level. The service provider (110) sets the learner's level based on the entered learner information. The learner's level can be automatically set based on the entered information, such as age and grade. The level evaluation unit (150) verifies the set learner's level, sets the evaluation start level, evaluates the learner's level using the question bank (121), and calculates the results. A more detailed description will be provided below with reference to FIG. 10.

[0130] Figure 10 is a flowchart illustrating step S200.

[0131] Referring to FIG. 10, the level evaluation unit (150) includes a step S211 in which the level evaluation unit (150) receives learner information input from a user terminal (200) and sets a starting level difficulty, a step S212 in which the starting level difficulty calculation result is confirmed, a step S213 in which the level difficulty is adjusted by sequentially presenting evaluation questions of the first group, a step S214 in which the evaluation questions of the second group are presented, a step S215 in which the level is moved by repeatedly testing the same difficulty questions in the evaluation questions of the second group, a step S216 in which the pattern of correct and incorrect answers is observed, a step S217 in which the pattern is reset, and a step S218 in which the final evaluation level is calculated by calculating an average value.

[0132] Step S211 is the step where the level evaluation unit (150) calculates the starting level of the evaluation items through learner information.

[0133] Step S212 is a step in which the level evaluation unit (150) checks whether the calculated starting level is a value greater than the minimum evaluation level, level K.

[0134] Step S213 is a step in which, if the starting level is higher than the K level, the level evaluation unit (150) issues the first evaluation question corresponding to the level to the user terminal (200), and adjusts the difficulty level based on the result to sequentially issue the evaluation questions of the first group.

[0135] For example, if the starting level is level 1 of medium difficulty, the level evaluation unit (150) issues an evaluation question corresponding to level 1 of medium difficulty to the user terminal (200), checks whether the answer is correct or incorrect, checks the delta value of the corresponding section, adjusts the difficulty of the next evaluation question, and issues the next question.

[0136] That is, the level evaluation unit (150) sequentially issues evaluation items of the first group that have a large delta value among all evaluation items.

[0137] Here, the level evaluation unit (150) changes the level difficulty only when the delta value is 0.5 and the result is consecutively correct or incorrect.

[0138] Step S214 is a step of confirming the correct or incorrect answer through the response of the user terminal (200) to the final evaluation item of the first group, and determining the level calculated by the final evaluation item as the level of the starting item of the second group.

[0139] Step S215 sequentially presents the first evaluation item of Group 2, and after a set number of evaluation items have been submitted, patterns are observed through the results. The level of the evaluation items in Group 2 is calculated based on the final evaluation item in Group 1.

[0140] Accordingly, the level evaluation unit (150) sequentially transmits a set number of evaluation items with the same level as the evaluation items of the second group to the user terminal (200) and observes the pattern according to the correct and incorrect answer results.

[0141] For example, the level evaluation unit (150) moves up one level if three consecutive questions are correct or two consecutive questions are incorrect. That is, the level evaluation unit (150) raises one level if three consecutive questions are correct, and lowers one level if two consecutive questions are incorrect.

[0142] Step S216 is a step for resetting when a pattern is observed and then moving to a new level by observing a new pattern. That is, the level evaluation unit (150) moves to a new level and resets the observed pattern information when a certain pattern is observed after repeated testing using evaluation items of the same level as described above. In addition, the level evaluation unit (150) moves to a new level when the above-described pattern is observed after repeated testing using multiple evaluation items with changed levels.

[0143] Step S217 is a step in which the level evaluation unit (150) determines the average value as the learner's reading level through the results of the last evaluation items in the second group.

[0144] For example, when the last problem of the second group is 4 evaluation items of level 8, the level is calculated based on whether each problem is correct or incorrect, and the average of the levels calculated for each of the 4 problems is calculated. At this time, if 1 problem of level 8 is correct, it is calculated as 9, and if it is incorrect, it is calculated as 7, and if there are 2 correct answers and 2 incorrect answers, it is 18+14=32, and this is divided by 4 to calculate the average value as 8. Therefore, the learner's reading level is determined as level 8.

[0145] Here, the level evaluation section (150) can lower the final level by 1 if the last four questions are in a mixed format of correct and incorrect answers, such as OXOX (correct answer O, incorrect answer X).

[0146] In addition, the level evaluation unit (150) can lower the final level by 1 if the last 4 questions are a pattern of two consecutive correct answers followed by an incorrect answer, such as XOOX.

[0147] The learner's reading level evaluation of the level evaluation unit (150) is registered in the server (100) and provided to the user terminal (200).

[0148] In addition, the level evaluation unit (150) determines the difficulty of the starting question of the first group as the medium difficulty of the minimum level K based on the learner's information of step S211, and conducts the test by applying the same delta value of 0.5 to all evaluation questions of the first group.

[0149] That is, the present invention further includes a step S221 of presenting the evaluation items of the first group to the user terminal (200) at the minimum level K to check correct and incorrect answers, and a step S222 of calculating the difficulty level by applying the same delta value to the evaluation items of the first group to move the level.

[0150] Step S221 is a step in which, if the level evaluation unit (150) determines that the learner is a beginner, it determines the minimum level as K level, issues an evaluation question of K level to the user terminal (200), and checks whether the answer is correct or incorrect.

[0151] Here, if the learner is a beginner, the level evaluation unit (150) sets the delta value for the evaluation items in Group 1 to 0.5. Furthermore, the evaluation items in Group 1 are determined to have a medium difficulty level of Level K, and evaluation items according to the corresponding difficulty level are sequentially presented to the user terminal (200) and checked for correct or incorrect answers.

[0152] This is the step where the S222 level evaluation unit (150) moves the level depending on whether the user terminal (200) has given a correct or incorrect answer. At this time, the level evaluation unit (150) moves the level difficulty (e.g., from medium difficulty of level K to high difficulty of level K) if there are consecutive correct answers, and conversely, lowers the level difficulty if there are consecutive incorrect answers.

[0153] Here, the level evaluation section (150) does not go down any further even if the evaluation question of the lower difficulty level of the K level is incorrect, as the lower difficulty level of the K level is the minimum level.

[0154] Alternatively, the level evaluation unit (150) raises the difficulty level to 1 level lower only when a problem of high difficulty level K is answered correctly 3 times in a row.

[0155] The evaluation results by the first group are applied as the level of difficulty of the starting items of the second group, and steps S213 to S217 described above are performed.

[0156] Additionally, step S200 includes step S230 as another embodiment.

[0157] Step S230 is described with reference to Fig. 11. Fig. 11 is a flowchart illustrating step S230.

[0158] Referring to FIG. 11, step S230 includes step S231 for storing learner information, step S232 for setting a learner level, step S233 for setting a range of difficulty changes, step S234 for recommending questions, step S235 for presenting test questions, step S236 for recalculating question difficulty and storing information, step S237 for generating the next question, step S238 for checking the number of test questions, and step S239 for storing the results.

[0159] Step S231 is the step where the server (100) collects and stores learner information from a user terminal (200) and / or an institution terminal (300). The user terminal (200) logs into the server (100) via the web or an app and inputs learning information such as the user's native language and native language level, as well as information such as age and school affiliation. Therefore, the server (100) stores the input information in the learner database (410).

[0160] Step S232 is the step of setting the learner level on the server (100). The service provider (110) sets the learner level based on the entered learner information. The learner level can be automatically set based on the entered information, such as age and grade.

[0161] Step S233 is the step of changing and resetting the difficulty level based on the learner's level. The item development unit (120) changes and sets the difficulty level based on the learner's level. At this time, the item development unit (120) can set the range of difficulty changes based on the learner's level confirmed through subsequent testing.

[0162] Step S234 is a step in which the server (100) selects and recommends questions to the user terminal (200) according to the learning level. The question development unit (120) can recommend questions classified and stored in the question bank (121) according to the learning level.

[0163] Step S235 is a step in which the server (100) presents test questions selected by the question development unit (120) to the user terminal (200). Here, the recommended test questions of the server (100) may be the questions recommended in step S234.

[0164] Step S236 recalculates and stores the item difficulty. The item development department (120) grades the answer sheets for the recommended test questions and recalculates the item difficulty based on the results. Furthermore, the item development department (120) adjusts the learner's item difficulty range based on the test results.

[0165] Step S237 is the step of presenting the next recommended problem. The question development unit (120) presents the next problem to the user terminal (200) with the recalculated question difficulty.

[0166] Step S238 is the step where the item development unit (120) checks whether the set order of the test items has been reached after presenting the questions. The item development unit (120) recalculates the question difficulty based on whether the presented questions are correct or incorrect, and repeats the process of presenting questions with the recalculated difficulty, while checking whether the set order of the questions has been reached.

[0167] Step S239 is the step where the item development department (120) saves the results in the learning DB when the items have been presented in the set order.

[0168] Step S300 is where the server recommends customized content based on the learner's level assessment results. The content matching unit (140) calculates the learner's learning level and learning efficiency for each content, and then provides customized content.

[0169] Here, learning efficiency by content is calculated based on the ratio of correct and incorrect answers to assessment questions following the learner's previous learning of the content. The content matching unit (140) then sends a message to the user terminal (200) recommending learning content with a predicted value for high learning efficiency in the learner's desired non-native language.

[0170] This is explained with reference to Fig. 12.

[0171] Figure 12 is a flowchart illustrating step S300.

[0172] Referring to FIG. 12, step S300 includes step S310 in which the content matching unit (140) collects the learner's learning and learning result information, step S320 in which learner information is analyzed, step S330 in which the learner's learning level and learning level are calculated, step S340 in which the learner's learning efficiency for each content is calculated, and step S350 in which learner-tailored content is recommended.

[0173] Step S310 is a step in which the content matching unit (140) collects the learner's learning and learning result information stored in the database (400). Here, the learning and learning result information may be information produced in step S230.

[0174] Step S320 analyzes learner information. The content matching unit (140) analyzes learner information such as the learner's age, grade, and native language.

[0175] Step S330 is the step where the content matching unit (140) calculates the learner's learning level and learning level. Here, the content matching unit (140) can calculate the learning level based on the reading index evaluation results of the reading evaluation index calculation unit (130).

[0176] Step S340 is where the content matching unit (140) calculates the learner's learning efficiency for each content. Here, the content matching unit (140) can calculate the learning efficiency for each content based on the correct and incorrect answer ratios for assessment questions following the learner's previous content learning. Alternatively, the content matching unit (140) can predict and calculate the learner's learning efficiency for each content based on the learner's reading evaluation index using an artificial intelligence learning model.

[0177] Step S350 is where the content matching unit (140) recommends learning content with a predicted high learning efficiency for the non-native language the learner wishes to learn. The content matching unit (140) can generate and provide learning content that maximizes efficiency by using the learner's learning level, learning information, and reading evaluation index, all calculated using artificial intelligence techniques. The content matching unit (140) then provides the generated content to the user terminal (200).

[0178] Step S400 is the step of evaluating the problems stored in the problem bank (121) and deleting, maintaining, or changing the level. The question development department considers the number of times the problems stored in the problem bank (121) were presented, the evaluation level of learners who entered correct answers, and the evaluation level of learners who entered incorrect answers, and decides whether to delete, maintain, or change the level of the problems stored in the problem bank (121).

[0179] The purpose of evaluating the problem bank (121) is to select problems according to the learner's level and to increase the learning effect for the learner.

[0180]

[0181] Although the present invention has been described above with limited embodiments and drawings, the present invention is not limited thereto, and various modifications and variations are possible by those skilled in the art to which the present invention pertains.

[0182]

[0183] Those skilled in the art will appreciate that the present invention can be implemented in modified forms without departing from the essential characteristics of the above-described description. Therefore, the disclosed methods should be considered illustrative rather than restrictive. The scope of the present invention is set forth in the claims, not the foregoing description, and all differences within the scope equivalent thereto should be construed as being encompassed by the present invention.

Claims

1. Server; and Includes a user terminal that accesses a server to learn a non-native language and receives educational content; The server is Service provision department that provides web services and intermediary services for video education between learners and instructors; A question development department that builds a model for learner-specific reinforcement learning, selects questions based on learner information and question difficulty information, and provides them to user terminals; A reading assessment index calculation unit that calculates the reading assessment index of non-native language learners; and An intelligent non-face-to-face foreign language education system for non-native language learners, including a content matching unit that provides content to a user terminal based on the evaluation results of a reading evaluation index calculation unit.

2. In claim 1, the item development department A question bank that is linked to your own server or cloud server as a database that stores selected questions; The item selection module is a module that selects items based on the item difficulty information calculated based on the results of the learner's basic test and provides the items to the user terminal; and An intelligent non-face-to-face foreign language education system for non-native language learners, including a question generation module that provides selected questions to a user terminal (200) based on evaluation questions selected from a question selection module according to the learner's evaluation level.

3. In claim 2, the question creation module An intelligent, non-face-to-face foreign language education system for non-native language learners, characterized by presenting questions to the user terminal based on learner information and question difficulty information, and recalculating the difficulty of subsequent questions presented to the learner by predicting the correct answer rate for the next question during the problem-solving process.

4. In claim 2, the question creation module An intelligent, non-face-to-face foreign language education system for non-native language learners, characterized by presenting questions with a difficulty level reflecting the learner's learning information, and then adjusting the difficulty of the next question based on whether the user's terminal inputs a correct or incorrect answer.

5. In claim 2, the question creation module An intelligent, non-face-to-face foreign language education system for non-native language learners, characterized by calculating the expected value of the current question based on whether the current question is correct or incorrect and the question difficulty value, and applying the calculated expected value to a reinforcement learning model of a Markov Decision Process (MDP) to select and present the next question with the same difficulty level or a question of a different level in the same category; 6. In claim 1, the reading evaluation index calculation unit A reading evaluation index measurement module that collects the learner's learning history by distinguishing between native and non-native language environments, and calculates a reading evaluation index by dividing the collected learning history by age and academic background; and An intelligent non-face-to-face foreign language education system for non-native language learners, characterized by including a reading evaluation index determination module that implements natural language processing in the native or non-native language through the answer sheet submitted through the solution of the question, and determines the non-native language reading level of the learner based on the learner's native language learning achievement and non-native language grade by applying important word weights through the word matrix distribution technique between documents; 7. In claim 6, the reading evaluation index measurement module Korean-style English book readability index measurement formula = (Vocabulary difficulty) X Sentence difficulty (SCI) = (sum of each 'difficulty X TOKEN ratio of each group' as a result of executing the modified Range program) X (SCI obtained using the syntactic formula of Botel and Granowsky), An intelligent, non-face-to-face foreign language education system for non-native language learners, characterized by calculating a reading evaluation index by applying the formula; 8. An intelligent non-face-to-face foreign language education method for intelligent non-face-to-face non-native language learners using any one of claims 1 to 7, It includes a step S230 for checking the learner's learning level and a step S300 for recommending content according to the learner's learning level; Step S230 S231) A step in which the server collects and stores learner information from at least one of a user terminal and an institution terminal; S232) A step of setting the learner level through the age and grade of the learner collected by the server; S233) A step in which the server sets the difficulty level of the question according to the learner's level; S234) A step in which the server selects and recommends items with a learning level and set difficulty to the user terminal; S235) A step in which the server presents selected test items to the user terminal; S236) A step in which the server scores the answer sheet for the test item input from the user terminal, recalculates the item difficulty level based on the result, and adjusts the item difficulty range; S237) A step in which the server presents the next recommended problem having an adjusted item difficulty; S238) A step of repeating the process of recalculating the question difficulty level based on whether the server has given the correct or incorrect answer to the given question and presenting the next question, while checking whether the set order of questions has been reached; and S239) A step of storing the results in a database when the server has presented the items in the set order to the learner's terminal; An intelligent non-face-to-face foreign language education method for intelligent non-face-to-face non-native language learners, including the step.

9. Server; and A user terminal that accesses a server to learn a non-native language and receives educational content; The server is A level evaluation unit that sequentially provides evaluation items set as multiple items to a user terminal and evaluates the foreign language reading level of a non-native language learner based on the results; Includes a content matching unit that provides content to the user terminal according to the evaluation results of the level evaluation unit; The level evaluation department CAT (Computerized Adaptive Testing) logic that increases the difficulty of the next evaluation question when the user terminal's correct answer to the evaluation question is confirmed, and lowers the difficulty of the next question when the answer is incorrect. An intelligent, non-face-to-face foreign language education system for non-native language learners, characterized by including an artificial intelligence model that executes CAT logic, calculates the correct answer rate for other questions of the same level based on whether the previously given evaluation question was correct, and predicts and adjusts the level of difficulty of the next question.

10. In claim 9, the plurality of evaluation items are An intelligent, non-face-to-face foreign language education system for non-native language learners, characterized by comprising a first group consisting of multiple assessment questions with a large range of difficulty from the initial question to the subsequent assessment questions, and a second group consisting of assessment questions of the same level that are presented consecutively and whose level of difficulty varies depending on whether the answer is correct or incorrect.

11. In claim 9, further comprising a question development unit for storing and selecting foreign language learning questions provided to a user terminal; The item development department A question bank that is linked to your own server or cloud server as a database that stores selected questions; The item selection module is a module that selects items based on the item difficulty information calculated based on the results of the learner's basic test and provides the items to the user terminal; A question selection module that provides selected questions to the user terminal based on the evaluation questions selected from the question selection module according to the learner's evaluation level; and An intelligent non-face-to-face foreign language education system for non-native language learners, comprising: a question evaluation module that maintains, deletes, and changes the level of a question based on at least one of the number of questions stored in a question bank, the evaluation level of a learner who entered a correct answer, and the evaluation level of a learner who entered an incorrect answer; 12. In claim 9, the artificial intelligence model An intelligent, non-face-to-face foreign language education system for non-native language learners, characterized by the SAKT (Self-Attentive Knowledge Tracing) model, which is a model that applies the attention mechanism of Transformer to knowledge tracing.

13. In claim 10, the level evaluation unit An intelligent, non-face-to-face foreign language education system for non-native language learners, characterized by providing evaluation items belonging to the second group to a user terminal in succession at the same level and moving the level of the evaluation items according to whether there are consecutive correct and incorrect answers. 14.a) A step of registering learner information including at least one of the native language, native language level, and age and affiliated school from at least one of the user terminal and the institution terminal on the server; b) A step of determining a starting level through learner information on a server, transmitting multiple evaluation items according to the starting level to the user terminal, and calculating the learner's level by adjusting the level of the evaluation items through whether the evaluation items are correct or incorrect; and c) a step of providing customized content based on the learner's evaluation results to the user terminal from the server; including; In step b) multiple evaluation items are An intelligent, non-face-to-face foreign language education method for non-native language learners, characterized by comprising a first group consisting of multiple evaluation questions with a large range of difficulty between the initial question and the subsequent evaluation questions, and a second group consisting of evaluation questions of the same level that are sequentially asked and have a level of difficulty that varies depending on whether the answer is correct or incorrect.

15. In claim 14, step b) b-1) Step for setting the starting level of the evaluation items to be first transmitted; b-2) If the starting level is greater than the minimum level, a step of sequentially issuing evaluation items corresponding to the level to the user terminal, and adjusting the difficulty level based on the result, and sequentially issuing evaluation items of the first group; b-3) A step of confirming the correct or incorrect answer through the user terminal's response to the final evaluation item of the first group and determining the level calculated by the final evaluation item as the starting level of the second group; b-4) A step of providing multiple evaluation items corresponding to the starting level in the second group to the user terminal in batches or sequentially, observing the pattern according to the correct and incorrect answers of the user terminal, and changing the level of the next evaluation items if the answer is correct or incorrect in succession; b-5) When a pattern is observed and the level difficulty is changed, a step of sequentially or collectively presenting a set number of evaluation items with the changed level difficulty to the user terminal, and observing the pattern according to the correct and incorrect answers of the user terminal to change the level of the next evaluation items; and b-6) A step of calculating a level according to whether the user terminal has given correct or incorrect answers to multiple evaluation items presented sequentially or in batches in the second group, and calculating an average of the calculated levels to determine the final level; An intelligent non-face-to-face foreign language education method for non-native language learners, including the step.

16. In claim 15, step b-6) is An intelligent, non-face-to-face foreign language education method for non-native language learners, characterized in that if the correct and incorrect answers to the evaluation items are mixed in an alternating manner, the final level is lowered by a set level.

17. In claim 15, step b-6) is An intelligent non-face-to-face foreign language education method for non-native language learners, characterized by: when a pattern of consecutive correct answers followed by incorrect answers and then incorrect answers again is observed, the level is lowered by a set level; 18. An intelligent non-face-to-face foreign language education method for non-native language learners, further comprising: a step of evaluating the question based on at least one of the number of times the question was presented, the evaluation level of a learner who entered a correct answer, and the evaluation level of a learner who entered an incorrect answer, stored in a question bank linked to an in-house server or a cloud server as a database for storing selected questions, and determining whether to maintain, delete, or change the level;

Images