System and method for language learning through ai-based literacy diagnosis and training

Abstract

A language learning system and method are provided. A language learning system according to one embodiment of the present invention may include: a diagnostic module that provides a diagnostic text corresponding to a user level to a user, provides a diagnostic problem corresponding to each of one or more language ability types based on the diagnostic text to the user, and receives a response from the user to the diagnostic problem; a language ability evaluation module that calculates a type-specific grade corresponding to each of the one or more language ability types based on the user's response to the diagnostic problem, and identifies a type of language ability weakness based on the type-specific grade; and a problem generation module that generates a learning text and a learning problem corresponding to the learning text based on the type of language ability weakness and provides them to the user.

Description

Technology Field

[0001] The present invention relates to a language learning system and method. More specifically, it relates to an AI-based language learning system and method that improves language ability by diagnosing a user's language ability and providing personalized learning problems using generative AI. Background Technology

[0003] Language ability is a complex capacity necessary for humans to acquire, understand, and express information through language, and includes various types such as vocabulary, factual understanding, inferential understanding, critical understanding, and creative understanding.

[0004] Literacy is one of the core elements of language ability, referring to the ability to go beyond simply reading and writing letters to accurately understand the content of information, grasp context, analyze it critically, and utilize it creatively. Particularly for elementary and middle school students, acquiring the level of language proficiency required by the curriculum for each grade level serves as an important foundation for academic achievement.

[0005] Conventional language learning systems have operated by having teachers provide students with passages and have them solve problems. This approach had limitations in assessing individual students' language proficiency levels and providing personalized learning.

[0006] Therefore, there is a growing need for a language learning system capable of diagnosing the level of each student's language ability by type and generating and providing personalized learning problems based on the diagnostic results. Prior art literature

[0008] Korean Patent Publication No. 10-2859061 (Registered on September 9, 2025) The problem to be solved

[0009] The technical problem of the present invention is to provide a language learning system and method that diagnoses a user's language ability by type, such as lexical comprehension, factual comprehension, inferential comprehension, critical comprehension, and creative comprehension, and automatically generates and provides customized learning problems based on the diagnosis results.

[0010] Another technical objective of the present invention is to provide a language learning system and method that identifies a user's weak areas in each learning domain regarding listening, speaking, reading, and writing, and provides customized learning by reflecting the identified weak areas.

[0011] Another technical objective of the present invention is to provide a language learning system and method capable of continuously improving a user's language ability by managing the user's incorrect answers, identifying types of weakness based on accumulated incorrect answer data, and reflecting them in the generation of subsequent questions.

[0012] The technical problems of the present invention are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0014] A language learning system according to embodiments of the present invention for solving the above technical problem may include: a diagnostic module that provides a diagnostic text corresponding to a user level to a user, provides a diagnostic problem corresponding to each of one or more language ability types based on the diagnostic text to the user, and receives a response from the user to the diagnostic problem; a language ability evaluation module that calculates a type-specific grade corresponding to each of the one or more language ability types based on the user's response to the diagnostic problem, and identifies a type of language ability weakness based on the type-specific grade; and a problem generation module that generates a learning text and a learning problem corresponding to the learning text based on the type of language ability weakness and provides them to the user.

[0015] In one embodiment, the system further includes a learning area evaluation module that performs an evaluation by learning area based on the user's learning activity data corresponding to each of one or more learning areas and identifies weak areas by learning area, and the problem generation module can generate the learning passage and the learning problem based on the weak areas by learning area along with the type of language ability weakness.

[0016] As one embodiment, the one or more types of language ability may include one of lexical comprehension, factual comprehension, inferential comprehension, critical comprehension, and creative comprehension.

[0017] In one embodiment, the diagnostic module assigns at least one of the one or more language ability types to each of the diagnostic problems, and the language ability evaluation module aggregates correct and incorrect answers for each of the language ability types assigned to each of the diagnostic problems to calculate a grade for each type, and can identify a language ability type corresponding to a grade for each type that is below a preset grade standard as a language ability weakness type.

[0018] In one embodiment, the problem generation module can dynamically generate the learning text and the learning problem corresponding to the type of language ability weakness using a generative language model.

[0019] In one embodiment, the user level is composed of a plurality of steps corresponding to a grade and a semester, and the diagnostic fingerprint may include a fingerprint linked to the curriculum of the grade and semester corresponding to the user level.

[0020] In one embodiment, the learning area evaluation module includes one or more evaluation units corresponding to each of the one or more learning areas, and the one or more learning areas may include one of a listening area, a speaking area, a reading area, and a writing area.

[0021] In one embodiment, the one or more evaluation units include a speech evaluation unit, and the speech evaluation unit preprocesses the user's voice data in the speech area and calculates the speech rate and the fundamental frequency of the voice based on the preprocessed voice data and compares them with each reference value calculated based on the statistical average value of the same attribute group.

[0022] In one embodiment, the speech evaluation unit calculates the frequency variability of the speech based on the preprocessed speech data, and the frequency variability of the speech is obtained by dividing the fundamental frequency of the speech per frame into preset intervals and calculating the difference value between adjacent intervals, and can classify the speech performance type of the user by comparing the speech rate and the frequency variability of the speech with threshold values ​​corresponding to each.

[0023] In one embodiment, the one or more evaluation units include a writing evaluation unit, and the writing evaluation unit generates a first evaluation result based on whether one or more preset keywords are included in the writing area response, generates a second evaluation result by calculating the semantic similarity between the writing area response and the correct answer, and can calculate a final evaluation score based on the first evaluation result and the second evaluation result.

[0024] In one embodiment, the system further includes a level management module for adjusting the user level, wherein the level management module can adjust the user level based on a diagnostic score calculated based on the user's response corresponding to the diagnostic problem.

[0025] In one embodiment, the level management module may adjust the user level by assigning weights to each of a plurality of sequentially calculated diagnostic scores, and assigning a higher weight to the most recently calculated diagnostic score.

[0026] As an embodiment, the system further includes an error management module that stores an error that occurred in the diagnostic problem and provides the stored error to the user, wherein the error management module re-presents a problem corresponding to the error to the user, and if the user's response to the re-presented problem does not match the correct answer, accumulates the number of error occurrences by language ability type of the error, and may identify a language ability type in which the accumulated number of error occurrences is greater than or equal to a preset number as an error vulnerability type and provide it to the problem generation module.

[0027] A language learning method performed by a computing device according to embodiments of the present invention to solve the above technical problem may include: providing a user with a diagnostic text corresponding to a user level and a diagnostic problem corresponding to each of one or more language ability types for said diagnostic text; calculating a type-specific grade corresponding to each of said one or more language ability types based on the user's response to said diagnostic problem and identifying a type of language ability weakness based on said type-specific grade; and generating a learning text and a learning problem corresponding to said learning text based on said type of language ability weakness and providing them to the user.

[0028] A computer program according to embodiments of the present invention for solving the above technical problem may be stored on a computer-readable recording medium to execute a language learning method. Effects of the invention

[0030] According to the embodiments of the present invention described above, by diagnosing a user's language ability by type and generating and providing personalized learning problems based on the diagnosis results, it is possible to effectively supplement each user's weak language ability type.

[0031] In addition, by identifying users' weak areas in listening, speaking, reading, and writing and reflecting them in customized learning, balanced improvement of language ability can be promoted.

[0032] In addition, by managing users' incorrect answers and reflecting accumulated incorrect answer data in the generation of subsequent questions, users can continuously improve their weak types that they repeatedly get wrong and gradually improve their language skills.

[0033] The effects of the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0035] Figure 1 shows the overall configuration of a language learning system according to one embodiment of the present invention. FIG. 2 is a flowchart exemplarily illustrating a method for identifying types of language ability weakness according to one embodiment of the present invention. FIG. 3 is a block diagram exemplarily illustrating the detailed configuration of a learning area evaluation module according to one embodiment of the present invention. FIG. 4 is a flowchart exemplarily illustrating the operation method of a speech evaluation unit according to one embodiment of the present invention. FIG. 5 is a flowchart exemplarily illustrating the operation method of a level management module according to one embodiment of the present invention. FIG. 6 is a flowchart exemplarily illustrating the operation method of an incorrect answer management module in one embodiment of the present invention. FIG. 7 is a flowchart exemplarily illustrating a language learning method according to one embodiment of the present invention. FIG. 8 is a block diagram illustrating the hardware configuration of a computing device used to implement various embodiments of the present invention. Specific details for implementing the invention

[0036] Hereinafter, embodiments of the present invention will be described in detail with reference to the attached drawings. The advantages and features of the present invention, and the methods for achieving them, will become clear by referring to the embodiments described below in detail together with the attached drawings. However, the technical concept of the present invention is not limited to the following embodiments but can be implemented in various different forms. The following embodiments are provided merely to complete the technical concept of the present invention and to fully inform those skilled in the art of the scope of the present invention, and the technical concept of the present invention is defined only by the scope of the claims.

[0037] It should be noted that when assigning reference numerals to the components of each drawing, the same components are assigned the same reference numeral whenever possible, even if they are shown in different drawings. Furthermore, in describing the present invention, if it is determined that a detailed description of related known components or functions could obscure the essence of the invention, such detailed description is omitted.

[0038] Unless otherwise defined, all terms used herein (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which the present invention pertains. Furthermore, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise. The terms used herein are for describing embodiments and are not intended to limit the present invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text.

[0039] Additionally, terms such as first, second, A, B, (a), (b), etc., may be used when describing the components of the present invention. These terms are intended merely to distinguish the components from other components, and the nature, order, or sequence of the components is not limited by such terms. Where it is stated that a component is "connected," "combined," or "joined" to another component, it should be understood that the component may be directly connected or joined to the other component, but that another component may also be "connected," "combined," or "joined" between each component.

[0040] In this invention, language ability may refer to the complex capabilities necessary for humans to acquire, understand, and express information through language. Language ability is not a specific single ability but is composed of a set of various sub-abilities; for example, it may include the ability to hear sounds and understand their meaning, the ability to express one's thoughts in speech or writing, and the ability to grasp the speaker's intent and respond appropriately. Language ability goes beyond the simple decoding of linguistic symbols to encompass the ability to understand the social and cultural context in which language is used and to utilize it appropriately, and may include literacy.

[0041] In this invention, literacy may refer to a linguistic ability focused on understanding and interpreting information. Literacy is not limited to the mere ability to decipher characters; it can encompass the ability to accurately grasp the content of given information, understand the communicator's intentions and perspectives, and actively construct meaning by connecting the information with one's own background knowledge and experiences. Furthermore, literacy includes high-order thinking skills that involve critically evaluating information and creatively reinterpreting it, and based on this, it can encompass the ability to linguistically understand and solve various problems encountered in real life. Such literacy can play an important role in academic achievement and overall daily social activities.

[0042] In the present invention, the type of language ability may refer to a type of detailed ability constituting the user's language ability, and may include at least one of lexical comprehension, factual comprehension, inferential comprehension, critical comprehension, creative comprehension, and combinations thereof.

[0043] In this invention, lexical comprehension may refer to the ability to accurately understand the meanings of words and idiomatic expressions, which are the basic units constituting stories or sentences, and to use them appropriately. Lexical comprehension includes vocabulary skills and can encompass the ability to grasp the dictionary meanings of various types of vocabulary—ranging from basic everyday words to Sino-Korean words and technical terms—used in presented information such as texts or audio, and to appropriately utilize the meanings of said vocabulary within a context. Furthermore, it can encompass the ability to understand the relationships between vocabulary, such as synonyms, antonyms, and polysemous words, and to accurately grasp the contextual meanings of words within the flow of sentences and paragraphs.

[0044] In the present invention, factual understanding may refer to the ability to accurately grasp facts or information explicitly revealed in a text. Factual understanding may encompass the ability to accurately understand the core content and detailed information explicitly revealed in presented information, such as text or audio. Specifically, it may encompass the ability to verify information such as facts, people, events, times, and places directly described in the presented information as they are, and to distinguish and understand the central content from the detailed content.

[0045] In the present invention, inferential understanding may refer to the ability to deduce content not directly revealed in the text or to infer causes and effects. Inferential understanding may encompass the ability to analyze content not explicitly stated in presented information, such as text or audio, based on context and background knowledge, and to logically derive conclusions. Specifically, it may encompass the ability to identify the communicator's intentions, hidden premises, and causal relationships not directly revealed in the presented information, and to synthesize the context and clues of the presented information to draw rational conclusions.

[0046] In this invention, critical understanding may refer to the ability to evaluate the arguments of a text or to compare and analyze information contained in a text from one's own perspective. Critical understanding may encompass the ability to evaluate the content of presented information, such as text or audio, as well as the reliability and logical validity of the arguments. Specifically, it may encompass the ability to analyze the communicator's perspective, the grounds for the arguments, and logical fallacies; to judge the factual accuracy of the presented information and the appropriateness of the arguments; and to compare different perspectives and form one's own viewpoint.

[0047] In this invention, creative comprehension may refer to the ability to understand a text expansively through emotional and creative thinking based on subjective responses to the text. Creative comprehension can encompass the ability to form emotional responses based on presented information, such as text or audio, to appreciate the emotions and atmosphere contained in the presented information, and to express them creatively or expand them into new ideas. Specifically, it can encompass the ability to reinterpret the content of the presented information from a new perspective by connecting it to one's own experiences and values, and to strengthen creative thinking skills based on inspiration obtained from the presented information.

[0048] Figure 1 shows the overall configuration of a language learning system according to one embodiment of the present invention.

[0049] Referring to FIG. 1, the language learning system (100) may include a diagnostic module (110), a language ability evaluation module (120), a problem generation module (130), a learning area evaluation module (140), a level management module (150), and / or an incorrect answer management module (160).

[0050] A language learning system (100) may refer to a learning platform that improves a user's language ability by diagnosing the user's language ability and generating and providing personalized learning passages and learning problems based on the diagnosis results.

[0051] In one embodiment, the language learning system (100) can diagnose the level of a user's language ability by type and collect learning activity data for listening, speaking, reading, and writing areas to identify weak areas for each learning area. Through this, the language learning system (100) can dynamically generate customized learning problems that reflect the identified language ability weak types and / or weak areas for each learning area using a generative language model.

[0052] In one embodiment, the language learning system (100) can systematically manage the user's incorrect answers and automatically adjust the user level. Through this, it can support the user in continuously improving their language ability in a learning environment optimized for their skill level.

[0053] In one embodiment, the language learning system (100) may be connected to communicate with a user through a user terminal (not shown). For example, the language learning system (100) may be connected to a user through a wired or wireless communication network, and the user may access the language learning system (100) through various types of computing devices such as a smartphone, tablet PC, laptop, or desktop computer. The language learning system (100) may provide the user with diagnostic fingerprints, diagnostic problems, learning fingerprints, and learning problems through the user terminal (not shown), and may receive the user's response through the user terminal (not shown).

[0054] The diagnostic module (110) provides the user with a diagnostic fingerprint corresponding to the user level, provides the user with a diagnostic problem corresponding to each of one or more language ability types based on the diagnostic fingerprint, and can receive the user's response to the diagnostic problem. The diagnostic module (110) can transmit the received user's response to the language ability evaluation module (120).

[0055] In one embodiment, a user level may be initially assigned to the user by an administrator or a teacher. Subsequently, the user level may be automatically adjusted by a level management module (150) according to the user's learning results. At this time, the user level may be composed of multiple stages corresponding to grades and semesters. For example, the user level may be composed of stages corresponding one-to-one to each grade and semester from the first semester of the first grade of elementary school to the second semester of the first grade of middle school, and may be designed so that the difficulty level increases step by step as the grade increases, such as Level 1 for the first semester of the first grade of elementary school, Level 2 for the second semester of the first grade of elementary school, and Level 3 for the first semester of the second grade of elementary school. Accordingly, the user may be provided with diagnostic passages and diagnostic questions that match their grade and semester level.

[0056] As one embodiment, one or more types of language ability may include lexical comprehension, factual comprehension, inferential comprehension, critical comprehension, and creative comprehension.

[0057] In one embodiment, the diagnostic text may include texts linked to the curriculum of the grade and semester corresponding to the user level. The diagnostic text is a text of difficulty corresponding to the user level, and the diagnostic module (110) may retrieve and provide diagnostic texts suitable for the user level from a separate database (not shown). For example, for a user corresponding to Level 3 (1st semester of 2nd grade elementary school), texts linked to the 1st semester of 2nd grade elementary school curriculum may be provided. Accordingly, the user can improve their language ability in a learning environment closely related to the current curriculum.

[0058] As an example, the passages linked to the curriculum may include passages from textbooks for the relevant grade and semester or passages from learning materials related to the curriculum. Additionally, it is possible to automatically generate new passages suitable for the curriculum level using a generative language model, or to provide passages written directly by administrators or teachers by registering them in a database. Accordingly, users can improve their language skills through various forms of passages closely related to the current curriculum.

[0059] In one embodiment, diagnosis may refer to the process of reading a diagnostic passage and solving diagnostic problems to evaluate the user's language proficiency level. The diagnosis may include a learning start diagnosis and / or a learning completion diagnosis. The learning start diagnosis is performed to assess the user's current language proficiency level before starting learning, and the learning completion diagnosis may be performed to verify changes in the user's proficiency after completing learning. The learning start diagnosis and the learning completion diagnosis are conducted in the same manner, and the degree of improvement in the user's language proficiency can be verified by comparing the results of the two diagnoses. Accordingly, the user can objectively verify changes in their language proficiency level before and after learning and directly experience the effects of learning.

[0060] In one embodiment, the diagnostic questions correspond to each of one or more types of language ability regarding the diagnostic passage and may be composed of at least one format among multiple-choice and true / false. For example, a total of 10 questions may be presented for the diagnostic passage, with 2 questions presented for each type of language ability. Multiple-choice questions may be composed in a manner where one of four options is selected, and true / false questions may be composed in a manner where the truth or falsity of the presented content is determined. Both multiple-choice and true / false questions can be applied to all types of language ability, including lexical comprehension, factual comprehension, inferential comprehension, critical comprehension, and creative comprehension. Through this, the user's language ability can be diagnosed in a balanced manner across five types.

[0061] In one embodiment, the diagnostic module (110) may provide diagnostic questions to the user by querying the language ability type information tagged for each diagnostic question in the database. In this case, tagging may refer to the act of an administrator or teacher marking the language ability type for each question when registering diagnostic questions in the database. For example, when an administrator or teacher registers diagnostic questions in the database, they may tag each question with one of the types of lexical comprehension, factual comprehension, inferential comprehension, critical comprehension, and / or creative comprehension, and the diagnostic module (110) may query the database for questions tagged with the corresponding type and provide them. For example, the language ability type may be tagged for each question, such as Question 1 being of the inferential comprehension type, Question 2 being of the factual comprehension type, and Question 3 being of the lexical comprehension type. Accordingly, the user can receive an objective diagnosis of their level by language ability type.

[0062] In one embodiment, the diagnostic module (110) may assign at least one type of language ability among one or more types to each diagnostic problem. In this case, assignment may mean the act of the diagnostic module (110) analyzing the content of the diagnostic problem and automatically assigning a language ability type suitable for the diagnostic problem. If a diagnostic problem is registered in a database without language ability type tagging, the diagnostic module (110) may analyze the content of the diagnostic problem and automatically assign a language ability type suitable for the diagnostic problem. For example, a diagnostic question can be automatically assigned a factual understanding type if it is in the form of verifying information specified in the text, such as "Which is the correct content directly mentioned by the author in the text?"; an inferential understanding type if it is in the form of inferring content not specified in the text, such as "Which is the most appropriate guess regarding content not appearing in this text?"; a lexical understanding type if it is in the form of asking about the meaning of vocabulary, such as "Which is the closest in meaning to the underlined word?"; a critical understanding type if it is in the form of evaluating the validity of an argument, such as "Which is the most appropriate evaluation of whether the author's argument is valid?"; and a creative understanding type if it is in the form of asking about appreciation and creative thinking, such as "Which is the most appropriate thought or new idea that came to mind after reading this text?". Through this, the language ability type of the diagnostic question can be automatically managed through the diagnostic module (110) even if the language ability type of the diagnostic question is not tagged in advance.

[0063] In one embodiment, the diagnostic questions may be configured to correspond to at least one learning area among the listening, speaking, reading, and writing areas. For example, the diagnostic questions corresponding to the reading area may be configured such that the user reads a diagnostic passage and answers the questions, and the diagnostic questions corresponding to the speaking area may be configured such that the user reads a diagnostic passage aloud and answers the questions. The diagnostic questions corresponding to the writing area may be configured such that the user writes the answers to given questions directly in text, and the diagnostic questions corresponding to the listening area may be configured such that the user listens to presented audio and answers the questions. Through this, the user's language ability can be comprehensively diagnosed across various learning areas.

[0064] The language ability evaluation module (120) can calculate a type-specific grade corresponding to each of one or more language ability types based on the user's response to the diagnostic problem, and can identify a type of language ability weakness based on the type-specific grade. The language ability evaluation module (120) can transmit the identified type of language ability weakness to the problem generation module (130).

[0065] As an example, the grade by type can be calculated by aggregating the user's correct and incorrect answers according to the language ability type assigned to each diagnostic question. In this case, the grade by type may refer to an indicator representing the level of the user's language ability type. For example, if both questions for each language ability type are answered correctly, it may be calculated as High; if only one question is answered correctly, as Medium; and if both questions are answered incorrectly, as Low. Specifically, if both questions of the Creative Comprehension type are answered correctly, Creative Comprehension may be calculated as High; if only one question of the Lexical Comprehension type is answered correctly, Lexical Comprehension may be calculated as Medium; and if both questions of the Inferential Comprehension type are answered incorrectly, Inferential Comprehension may be calculated as Low. In this case, Creative Comprehension may be classified as a strong type, Lexical Comprehension as a type requiring improvement, and Inferential Comprehension as a weak type requiring intensive study. Accordingly, the level of the user's language ability type can be distinguished and identified.

[0066] In one embodiment, a language ability weakness type may refer to a language ability type in which the user shows a relatively low level among one or more language ability types, and may mean a type that requires intensive learning.

[0067] In one embodiment, the language ability evaluation module (120) can identify a type of language ability corresponding to a grade that is lower than or equal to a preset standard grade among the grades by type as a type of language ability weakness. In this case, the preset standard grade may mean low or medium among the grades by type. For example, if the preset standard grade is 'medium', lexical comprehension and inferential comprehension that are grades by type of 'medium' or lower may be identified as types of language ability weakness. The identified types of language ability weakness may be reflected in the problem generation module (130) when generating learning problems. Through this, the user's weak language ability type can be precisely identified and reflected in customized learning.

[0068] In one embodiment, the diagnostic score may be calculated based on the number of correct answers the user gives to the diagnostic questions. For example, if 7 out of 10 questions are answered correctly, the diagnostic score may be calculated as 70 out of 100. The language ability evaluation module (120) may assign a compass grade to the user based on the diagnostic score. For example, a gold compass may be assigned if the diagnostic score is above a preset upper threshold, a silver compass if it is above a median threshold, and a bronze compass if it is below that. Through this, the user can intuitively check their overall level of language ability.

[0069] In one embodiment, the language ability evaluation module (120) can transmit the calculated diagnostic score to the level management module (150). The level management module (150) can adjust the user level based on the received diagnostic score. For example, if the diagnostic score is above a preset upper limit, the user level may be raised, and if it is below a lower limit, the user level may be lowered. Accordingly, the user's learning results are continuously reflected in the user level, thereby providing an optimized learning environment.

[0070] The problem generation module (130) can generate a learning passage and a learning problem corresponding to the learning passage and provide them to the user.

[0071] In one embodiment, the problem generation module (130) can generate a learning passage and a learning problem corresponding to the learning passage based on a type of language ability weakness and provide them to the user. For example, if the type of language ability weakness is lexical comprehension, the problem generation module (130) can generate a learning problem that primarily includes problems involving identifying the meaning of words used in the passage within the context or finding synonyms and antonyms. If the type of language ability weakness is factual comprehension, the module can generate a learning problem that primarily includes problems involving accurately verifying information such as facts, people, and events explicitly described in the passage. If the type of language ability weakness is inferential comprehension, the module can generate a learning problem that primarily includes problems involving inferring content not explicitly stated in the context of the passage or identifying causal relationships. If the type of language ability weakness is critical comprehension, the module can generate a learning problem that primarily includes problems involving evaluating the validity of claims and evidence in the passage or analyzing the author's perspective. If the type of language ability weakness is creative comprehension, the module can generate a learning problem that primarily includes problems involving appreciating the emotions and atmosphere contained in the passage, connecting them to one's own experiences, or expanding them into new ideas. Through this, personalized learning that intensively supplements the user's weak language ability types can be enabled.

[0072] In one embodiment, the problem generation module (130) can adjust the difficulty of the learning passage and / or the difficulty of the learning problem according to the user level.

[0073] In one embodiment, the difficulty level of the learning passage may be adjusted by reflecting at least one of vocabulary level, sentence length, passage length, and / or subject familiarity. In this case, the vocabulary level may refer to the difficulty standards of vocabulary presented in the curriculum of the grade and semester corresponding to the user level. For example, for Level 3 (1st semester of 2nd grade elementary school), a passage composed of basic vocabulary mainly consisting of two syllables and vocabulary presented in the curriculum of that grade may be generated, and for Level 7 (1st semester of 4th grade elementary school), a passage composed of vocabulary according to the curriculum standards of that grade, including Sino-Korean words and subject-specific vocabulary, may be generated. Sentence length may refer to the average number of words in each sentence constituting the passage. For example, for Level 3, a passage composed of a simple sentence structure of 5 to 7 words may be generated, and for Level 7, a passage containing a complex sentence structure of 10 to 15 words may be generated. Passage length may refer to the total number of characters constituting the passage. For example, at Level 3, passages of approximately 200 to 300 characters may be generated, and at Level 7, passages of approximately 500 to 700 characters may be generated. Topic familiarity can be determined based on topics covered in the curriculum of the grade and semester corresponding to the user level; topics covered in the curriculum of that grade may be classified as having higher familiarity, while topics not covered may be classified as having lower familiarity. For example, at Level 3, passages on familiar topics such as family, school, and animals covered in the curriculum of that grade may be generated, while at Level 7, passages on more abstract topics such as the environment, society, and history covered in the curriculum of that grade may be generated.

[0074] In one embodiment, the difficulty of a learning problem may be adjusted by reflecting at least one of the number of inference steps, the similarity of answer choices, and the indirectness of deriving the correct answer. In this case, the number of inference steps may refer to the number of pieces of information or logical connections that must be verified in the text to derive the correct answer. For example, at Level 3, a Level 1 inference problem that can be directly verified in a single sentence of the text may be presented; at Level 5 (first semester of the third grade of elementary school), a Level 2 inference problem that requires deriving the correct answer by connecting two sentences of the text may be presented; and at Level 7, a Level 3 or higher inference problem that requires deriving a conclusion by sequentially connecting information scattered across multiple sentences of the text may be presented. The similarity of answer choices may refer to the semantic and morphological similarity between the correct answer and the incorrect answer choices in a four-option multiple-choice question, and may be measured based on the number of incorrect answer choices that contain the same topic word or have a similar sentence structure. For example, Level 3 may consist of incorrect answer choices that are semantically clearly distinguishable from the correct answer; Level 5 may include one or two incorrect answer choices that contain expressions somewhat similar to the correct answer; and Level 7 may be configured to require precise judgment by including multiple incorrect answer choices that contain the same subject words as the correct answer or have similar sentence structures. The indirectness of deriving the correct answer may refer to an indicator that distinguishes, based on whether the correct answer is directly stated in the text, the degree to which it can be directly verified in the text as direct, the degree to which the correct answer must be derived by interpreting parts of the text as semi-direct, and the degree to which it must be interpreted by synthesizing the context and background knowledge of the text as indirect.For example, at Level 3, direct questions may be presented in which the correct answer can be derived by verifying the content directly described in the passage; at Level 5, semi-direct questions may be presented in which the correct answer must be derived by interpreting parts of the passage; and at Level 7, questions may be presented in which the correct answer must be derived by indirectly interpreting content not explicitly stated in the passage through the meaning between the lines and the context. Accordingly, learning passages and learning questions of a difficulty level optimized for the user's level can be automatically provided.

[0075] In one embodiment, the problem generation module (130) can dynamically generate learning passages and learning problems corresponding to types of language ability weakness using a generative language model. For example, if the user's type of language ability weakness is inferential comprehension and critical comprehension, the generative language model can generate learning passages and learning problems that can focus on supplementing the corresponding type of weakness. Specifically, the generative language model can generate a learning passage in which the flow of cause and effect is clearly revealed, and for the type of inferential comprehension learning problem, it can generate a problem that infers content not explicitly stated in the context of the passage, such as "Which is the most appropriate prediction regarding what will happen in the future if environmental pollution worsens in the passage?", and for the type of critical comprehension learning problem, it can generate a problem that evaluates the validity of an argument, such as "Which is the most appropriate evaluation regarding whether the solution to environmental pollution proposed by the author is valid?". Accordingly, learning passages and learning problems optimized for the user's type of weakness can be generated and provided in real time.

[0076] In one embodiment, the generative language model of the problem generation module (130) can generate learning passages and learning questions by reflecting difficulty adjustment factors according to the user level. For the learning passages, vocabulary level, sentence length, passage length, and / or familiarity with the topic may be reflected, and for the learning questions, the number of inference steps, similarity of answer choices, and / or indirectness of deriving the correct answer may be reflected. For example, in the case of Level 5 (first semester of the third grade of elementary school), the generative language model can generate a learning passage composed of vocabulary at the level of the curriculum for that grade, a sentence structure of 10 to 15 words, and a passage length of approximately 400 to 500 characters, and the learning questions may consist of a two-step inference question in which the correct answer must be derived by connecting two sentences in the passage, and a question containing an incorrect answer choice that includes expressions partially similar to the correct answer. Accordingly, learning passages and learning questions of difficulty optimized for the user's level can be automatically generated and provided.

[0077] In one embodiment, the problem generation module (130) can generate learning passages and learning problems by inputting a prompt into a generative language model. In this case, the prompt may refer to an input text that structures and includes condition information necessary for the generative language model to generate learning passages and learning problems. The prompt may include a user level, a type of language ability weakness, a weakness area by learning domain, and / or a difficulty adjustment factor as parameters.

[0078] For example, the prompt may include vocabulary level, sentence length, passage length, and topic familiarity corresponding to the user level as passage generation conditions, and question type, number of inference steps, choice similarity, and indirectness of deriving the correct answer corresponding to types of language ability weakness as question generation conditions.

[0079] Specifically, when the user level is Level 5 (1st semester of 3rd grade elementary school) and the type of language ability weakness is inferential comprehension, the prompt can be configured as follows: "[Passage Generation Conditions] Vocabulary Level: 1st semester of 3rd grade elementary school curriculum level, Sentence Length: 10 to 15 words, Passage Length: 400 to 500 characters, Topic: Curriculum-linked topic, "[Question Generation Conditions] Type of Language Ability Weakness: Inferential Comprehension, Number of Inference Stages: 2 stages, Similarity of Options: Medium, Indirectness of Deriving the Answer: Semi-direct, Question Format: 4-option multiple choice, Number of Questions: 3 inferential comprehension questions out of 5 questions".

[0080] As an example, the prompt may reflect the information in the fingerprint generation conditions if additional vulnerable areas by learning area are identified.

[0081] For example, if the speaking area is identified as a weak area by learning area, "text type: conversational, mainly short sentences" may be added to the text generation conditions of the prompt, and if the writing area is identified as a weak area by learning area, "question format: includes subjective format" may be added to the question generation conditions. Accordingly, the question generation module (130) can dynamically generate user-customized learning texts and learning questions by inputting a prompt that comprehensively reflects the user's language ability weakness type, weak area by learning area, and user level into a generative language model.

[0082] In one embodiment, the problem generation module (130) can perform verification on the output result of the generative language model. Specifically, the problem generation module (130) can verify whether the learning text and learning problem output by the generative language model satisfy the conditions included in the prompt.

[0083] For example, the problem generation module (130) can check whether the number of characters in the generated learning passage is included within the passage length range specified in the prompt, whether the number of generated learning questions and the distribution ratio of questions by language ability type match the conditions specified in the prompt, and / or whether the options of the four-option multiple-choice questions consist of four.

[0084] If there are items that do not meet the conditions as a result of verification, the problem generation module (130) can regenerate the learning text and / or learning problem by re-entering a correction prompt containing correction instructions for the items that do not meet the conditions into the generative language model.

[0085] For example, if the number of characters in a generated training passage exceeds a specified range, a correction prompt such as "reduce the passage length to a range of 400 to 500 characters" can be re-entered into the generative language model. Accordingly, the quality of the output of the generative language model can be managed to satisfy preset conditions, and the reliability and consistency of the training passages and training questions provided to the user can be ensured.

[0086] In one embodiment, when the problem generation module (130) generates learning problems using a generative language model, it may prioritize selecting problems corresponding to a type of language ability weakness already calculated for the user, and additionally select problems corresponding to other types of language ability that are not weak in language ability, thereby configuring the learning problems in a mixed form of multiple types of language ability. For example, if the user's type of language ability weakness is lexical comprehension, based on 5 problems per set, 3 problems of the lexical comprehension type and the remaining 2 problems may be selected from factual comprehension, inferential comprehension, critical comprehension, and creative comprehension that are not weak types. Accordingly, the user can maintain a balanced overall level of language ability by learning various types of language ability evenly while intensively supplementing the weak type.

[0087] In one embodiment, when an error vulnerability type is transmitted from the error management module (160), the problem generation module (130) may combine the language ability vulnerability type and the error vulnerability type to form a vulnerability type pool, and recalculate the problem distribution ratio based on the pool. Specifically, types corresponding to both the language ability vulnerability type and the error vulnerability type may be classified as priority types to allocate the largest number of problems, types corresponding to only one of them may be classified as general vulnerability types to allocate problems sequentially, and the remaining problems may be composed of other types. For example, if the language ability vulnerability type is critical comprehension and inferential comprehension and the error vulnerability type is inferential comprehension and lexical comprehension, inferential comprehension may be classified as a priority type to allocate 2 problems, critical comprehension and lexical comprehension may be classified as general vulnerability types to allocate 1 problem each, and the remaining 1 problem may be composed of other types. In this way, when there are types of weak errors, a distribution ratio reflecting the weak errors is applied prior to the basic problem distribution principle, thereby enabling personalized learning that allows users to more intensively improve the types of errors they repeatedly get wrong.

[0088] In one embodiment, the problem generation module (130) can construct learning problems by identifying a preceding type based on a sequential hierarchical structure among language ability types. Specifically, the problem generation module (130) can identify a preceding type corresponding to the stage immediately preceding a type of language ability weakness and construct learning problems by prioritizing and additionally selecting problems of that preceding type. At this time, the sequential hierarchical structure can be defined in the order of lexical understanding (Level 1), factual understanding (Level 2), inferential understanding (Level 3), critical understanding (Level 4), and creative understanding (Level 5), and can reflect the characteristic of language ability development that in order to acquire a language ability type of a higher level, the preceding type of the stage immediately preceding must be possessed first.

[0089] For example, if the user's language ability weakness type is critical comprehension (Layer 4), the problem generation module (130) can configure learning problems by prioritizing and additionally selecting problems of the inferential comprehension (Layer 3) type, which is the preceding type. Meanwhile, if an error weakness type is transmitted from the error management module (160), the learning problems can be configured by identifying the preceding type for each type within the weakness type pool, and by prioritizing the preceding type of the type classified as the highest priority type in other type problems. Accordingly, customized learning is possible to improve the user's language ability more systematically and fundamentally by reinforcing the preceding type that is the root cause of the weakness type, rather than simply repeating learning only the weakness type.

[0090] In one embodiment, the problem generation module (130) can generate learning passages and learning questions based on the learning domain-specific weak areas and / or user levels received from the learning domain evaluation module (140) along with the type of language ability weakness. For example, if the type of language ability weakness is inferential comprehension and the speaking domain is weak, the problem generation module (130) can generate a learning passage consisting of conversational and short sentences composed of vocabulary levels and sentence lengths appropriate for the user level, and generate learning questions that primarily include inferential comprehension type questions that infer content not explicitly stated in the context of the passage. As another example, if the type of language ability weakness is critical comprehension and the writing domain is weak, the problem generation module (130) can generate a learning passage in which claims and evidence are clearly described at a difficulty level appropriate for the user level, and generate learning questions that primarily include subjective learning questions of the critical comprehension type that evaluate the validity of the writer's claims. As another example, if the type of language ability weakness is factual comprehension and the listening area is weak, the problem generation module (130) can generate a learning text containing repetitive expressions and clear pronunciation at a difficulty level suitable for the user level, and generate learning problems that focus on factual comprehension type problems that verify factual information explicitly described in the presented audio. Accordingly, by providing learning texts and learning problems that comprehensively reflect the type of language ability weakness, the weak area by learning area, and the user level, integrated customized learning is possible that complements both the user's weaknesses by language ability type and weaknesses by learning area.

[0091] In one embodiment, the type of learning problem is determined based on the type of weakness in language ability, and the problem may be composed of at least one format among multiple-choice and true / false. The learning problem may consist of a total of 5 sets, and 5 questions may be presented in each set. In the case of multiple-choice questions, two attempts to solve the problem may be provided, and if an incorrect answer is selected on the first attempt, a hint may be provided and a chance to retry may be given. In the case of true / false questions, only one response may be given, and whether the answer is correct may be displayed immediately upon selection.

[0092] In one embodiment, gold, silver, and bronze badges may be awarded based on the user's response to a learning problem, that is, the result of solving the learning problem. If the user fails to obtain a gold badge, they may attempt the corresponding learning again, and the score may be calculated based on the number of questions answered correctly at once. In one embodiment, learning problems may be provided corresponding to exploration areas marked on an exploration map, and the user can intuitively check their current learning progress through the exploration map. Accordingly, the user can participate proactively in learning while feeling a sense of accomplishment in a gamified learning environment.

[0093] The learning area evaluation module (140) can independently collect user learning activity data corresponding to each of one or more learning areas in each learning area to perform an evaluation by learning area and identify weak areas by learning area. The learning area evaluation module (140) can transmit the identified weak areas by learning area to the problem generation module (130) and / or level management module (150). Accordingly, the learning area evaluation module (140) can identify the user's weak areas by learning area across the entire listening, speaking, reading, and writing areas and promote balanced improvement of language ability. The detailed configuration of the learning area evaluation module (140) will be described later with reference to FIG. 3.

[0094] In one embodiment, the learning area evaluation module (140) may include one or more evaluation units corresponding to each of one or more learning areas. The one or more evaluation units may collect user learning activity data in each learning area and perform evaluations for each learning area. For example, it may include a speaking evaluation unit (141) corresponding to the speaking area, a reading evaluation unit (142) corresponding to the reading area, a writing evaluation unit (143) corresponding to the writing area, and a listening evaluation unit (144) corresponding to the listening area.

[0095] In one embodiment, one or more learning areas may include at least one of a listening area, a speaking area, a reading area, and a writing area. Each learning area may be conducted based on a learning text, and the user's language ability may be comprehensively evaluated in various aspects for each learning area. For example, listening learning may be conducted by listening to presented audio and understanding the content, speaking learning by reading a learning text aloud and understanding the content, reading learning by reading a learning text and measuring reading speed and comprehension, and writing learning by reading a learning text and writing answers directly to given questions. Accordingly, the user can achieve balanced improvement in language ability across all areas of listening, speaking, reading, and writing.

[0096] In one embodiment, learning activity data may refer to data independently collected during the user's learning process in each learning area. Learning activity data in the speaking area may include voice data collected through a microphone while the user reads a passage aloud, and may include the user's speech rate and the fundamental frequency of the voice. Learning activity data in the reading area may include the dwell time the user spends on the diagnostic passage or learning passage page, and the reading speed may be calculated based on the time the user spends on the passage page. Learning activity data in the writing area may include writing area responses entered by the user directly writing text in response to a given question, and the writing area responses may be received in a subjective form where the user freely constructs and inputs sentences. Learning activity data in the listening area may include the result of the user solving problems after listening to the presented audio, that is, the correct or incorrect answer result of the problem-solving.

[0097] As an example, weak areas for each learning area can be identified based on evaluation results for each learning area. For example, in the case of the speaking area, weak areas for each learning area can be identified by classifying speaking performance types based on the user's speech speed and frequency variability of the voice, and in the case of the reading area, weak areas for each learning area can be identified by classifying reading performance types based on the user's reading speed and the correct answer rate for solving problems. In the case of the writing area, if the final evaluation score for the writing area response is below a preset threshold value, it can be identified as a weak area for each learning area, and in the case of the listening area, if the correct / incorrect answer result for solving problems is below a preset threshold value, it can be identified as a weak area for each learning area.

[0098] In one embodiment, the learning area evaluation module (140) may transmit the identified weak areas for each learning area to the problem generation module (130) and / or the level management module (150). The problem generation module (130) may generate learning passages and learning questions by reflecting the received weak areas for each learning area along with the type of language ability weakness, and the level management module (150) may adjust the user level by reflecting the received information on the weak areas for each learning area in the calculation of the auxiliary level adjustment score. Accordingly, the user's weak areas for each learning area are simultaneously reflected in the generation of customized learning questions and the adjustment of the user level, enabling more precise and balanced improvement of language ability.

[0099] The specific operation method of the learning area evaluation module (140) will be described later with reference to FIGS. 3 and FIGS. 4.

[0100] The level management module (150) can adjust the user level based on the diagnostic score calculated based on the user response corresponding to the diagnostic question. The level management module (150) can transmit the adjusted user level to the diagnostic module (110), the language ability evaluation module (120), the learning area evaluation module (140), and / or the question generation module (130). Accordingly, diagnostic passages, diagnostic questions, learning passages, and / or learning questions suitable for the user's learning level can be continuously provided.

[0101] In one embodiment, the level management module (150) can adjust the user level according to the level adjustment score. In this case, the level adjustment score is a score calculated based on the user's diagnostic score and may refer to a score that serves as a criterion for determining whether to adjust the user level upward or downward. The level management module (150) can adjust the user level upward if the level adjustment score is above a preset upper limit, adjust the user level downward if it is below a lower limit, and maintain the user level in other cases. For example, if the preset upper limit is 80 points and the lower limit is 20 points, the user level may be adjusted upward by one step if the level adjustment score is 80 points or higher, the user level may be adjusted downward by one step if the level adjustment score is 20 points or lower, and the user level may be maintained if the level adjustment score is greater than 20 points but less than 80 points. Level adjustment may be performed after the diagnosis is completed. Accordingly, the user can be provided with a learning environment of progressively higher levels as their skills improve.

[0102] In one embodiment, the level management module (150) can adjust the user level by assigning weights to each of the multiple diagnostic scores calculated sequentially. Specifically, the level management module (150) can adjust the user level by assigning a high weight to the most recently calculated diagnostic score. For example, if a user obtains 60 points in the first diagnostic test, 70 points in the second diagnostic test, and 80 points in the third diagnostic test, the level management module (150) can calculate the level adjustment score by assigning the highest weight to the most recent third diagnostic score and assigning lower weights in the order of the second and first tests. Accordingly, the user's recent learning results are more reflected in the level adjustment, making it possible to make an accurate level adjustment that matches the user's current skill level.

[0103] The specific operation method of the level management module (150) will be described later with reference to FIG. 5.

[0104] The incorrect answer management module (160) can store incorrect answers that occurred in the diagnostic questions and provide the stored incorrect answers to the user. Specifically, the incorrect answer management module (160) can store incorrect answers that occurred in the diagnostic questions to generate an incorrect answer list and provide the stored incorrect answer list to the user. At this time, the incorrect answer list may include the questions the user answered incorrectly, the language ability type of the questions, and the time of occurrence of the incorrect answers, and can be checked at any time upon the user's request.

[0105] In one embodiment, the incorrect answer management module (160) may re-present a problem to the user that corresponds to an incorrect answer stored in the incorrect answer list, and identify a type of weakness in the incorrect answer based on the user's response to the re-presented problem and transmit it to the problem generation module (130). Accordingly, the user can continuously improve the type of weakness that is repeatedly incorrect and gradually improve their language ability.

[0106] In one embodiment, an incorrect answer is a question that the user answered incorrectly in a diagnostic question, and can be automatically stored in an incorrect answer management module (160). The stored incorrect answers are provided to the user upon the user's request, and if the user solves the re-presented question again and answers correctly, the incorrect answer can be automatically deleted from the incorrect answer list. On the other hand, if the user fails to answer the re-presented question correctly, the incorrect answer remains in the incorrect answer list and can be repeatedly re-presented until the user answers correctly.

[0107] In one embodiment, the incorrect answer management module (160) can store and manage incorrect answers that occurred in diagnostic problems and incorrect answers that occurred in learning problems.

[0108] In one embodiment, when a user's response to a re-issued question does not match the correct answer, the number of errors is accumulated for each language ability type of the error, and a language ability type with an accumulated number of errors exceeding a preset number can be identified as an error vulnerability type. For example, if the preset number is 3, if an error is accumulated 3 or more times in a question of the inferential comprehension type, inferential comprehension can be identified as an error vulnerability type. At this time, since the number of errors is accumulated independently for each language ability type, for example, if an error of the inferential comprehension type is accumulated 3 times, only inferential comprehension is identified as an error vulnerability type, and if an error of the critical comprehension type is accumulated 2 times, critical comprehension may not be identified as an error vulnerability type. The identified error vulnerability types can be reflected based on a vulnerability type pool composed by summing the language ability vulnerability types and error vulnerability types when the question generation module (130) generates a learning question. Accordingly, the vulnerability types that the user repeatedly makes mistakes on can be systematically managed and intensively supplemented.

[0109] The specific operation method of the incorrect answer management module (160) will be described later in FIG. 6.

[0110] The user is connected to communicate with the language learning system (100) through a user terminal (not shown) and can receive diagnostic fingerprints, diagnostic problems, learning fingerprints, and / or learning problems and input responses to them.

[0111] In one embodiment, users may include learners, teachers, and / or administrators, and the functions and screens provided may vary depending on the user's role. For example, learners may be provided with learning functions such as reading passages, solving problems, and reviewing incorrect answers, while teachers and administrators may be provided with functions to check and manage the learners' diagnostic results and learning status.

[0112] In one embodiment, teachers and / or administrators can view individual learner diagnostic reports, which provide achievement levels by language ability type in the form of a radar chart and may include vocabulary requiring review, reading patterns, exploration levels, etc. Additionally, teachers and administrators can view statistical information such as group diagnostic scores, study frequency, and scores by item. Accordingly, teachers and administrators can systematically identify the learners' language ability levels and provide effective learning guidance.

[0113] As an example, a teacher can use a generative language model to create diagnostic passages and diagnostic questions, and / or learning passages and learning questions. For example, if a teacher inputs the title "Changes in Spring" and a level of 3 (first semester of second grade elementary school), the generative language model can generate a new diagnostic passage regarding "Changes in Spring" consisting of vocabulary focused on basic words that match the vocabulary level required for the first semester of second grade elementary school curriculum, simple sentence structures of 5 to 7 words, and a passage length of approximately 200 to 300 characters. At this time, for the generated diagnostic passage, the generative language model can automatically generate diagnostic questions corresponding to each type of lexical comprehension, factual comprehension, inferential comprehension, critical comprehension, and creative comprehension. For example, a diagnostic question of the lexical comprehension type such as "Which is closest in meaning to the underlined word?" may be generated, and a diagnostic question of the inferential comprehension type such as "Which is the most appropriate guess regarding what changes will occur in trees when spring comes?" may be generated. The teacher can review and modify the generated diagnostic passages and diagnostic questions and register them in a database. Similarly, when a teacher inputs the title of a learning passage and the user level, the generative language model can automatically generate learning passages and questions appropriate for that level, and the teacher can review and modify the generated results to register them in the database. In this way, since passages and questions that do not currently exist can be newly generated to match the curriculum level through the generative language model, teachers can efficiently organize customized diagnostic and learning materials tailored to class characteristics while reducing the burden of directly writing diagnostic and learning passages.

[0114] In one embodiment, a learner can check their current learning progress through an exploration map based on diagnostic results. The exploration map may be configured such that exploration areas are opened sequentially according to the learner's diagnostic results and learning progress, and each exploration area may consist of learning problems corresponding to a specific type of weakness in language ability. For example, if a user completes the learning problems in a specific exploration area and obtains a gold badge, the next exploration area may be opened, and if a silver or bronze badge is obtained, the user may challenge that exploration area again or proceed to the next exploration area.

[0115] In one embodiment, the exploration map may display the badges, compass ranks, and learning progress acquired by the user to date, allowing the user to intuitively grasp their learning goals and progress status through the exploration map. Accordingly, the gamified learning environment can promote self-directed learning participation and continuously maintain learning motivation.

[0116] FIG. 2 is a flowchart exemplarily illustrating a method for identifying types of language ability weakness according to one embodiment of the present invention.

[0117] The operation method described in the embodiment of FIG. 2 can be performed by the language ability evaluation module (120) of FIG. 1. Therefore, if the performing entity is omitted in the following steps, it is assumed that the performing entity is the language ability evaluation module (120). In the description of this embodiment, content identical to that described in FIG. 1 above may be omitted to avoid duplication of description.

[0118] In step S110, the correct and incorrect answers of the user are tallied according to the language ability type assigned to each diagnostic question. For example, if the diagnostic questions consist of a total of 10 questions with 2 questions for each language ability type, questions 1 and 2 may be classified as lexical comprehension type, questions 3 and 4 as factual comprehension type, questions 5 and 6 as inferential comprehension type, questions 7 and 8 as critical comprehension type, and questions 9 and 10 as creative comprehension type. The language ability evaluation module (120) can tally the correct and incorrect answers based on the user's response to the questions classified by each type. For example, if the user answers question 1 as correct and question 2 as incorrect, the correct answer for the lexical comprehension type is tallied as 1 and the incorrect answer as 1, and the correct and incorrect answers for each language ability type can be classified and tallied in this manner.

[0119] In step S120, grades by type are calculated based on the tally of correct and incorrect answers by type.

[0120] As an example, the grade by type may be calculated based on the number of correct answers relative to the number of questions asked for each type of language ability. For example, if two questions are asked for each type of language ability, the grade may be calculated as High if both questions are answered correctly, Medium if only one question is answered correctly, and Low if both questions are answered incorrectly. Based on the above example, since the number of correct answers for the Lexical Comprehension type is 1, Lexical Comprehension may be calculated as Medium; since the number of correct answers for the Factual Comprehension type is 2, Factual Comprehension may be calculated as High; since the number of correct answers for the Inferential Comprehension type is 0, Inferential Comprehension may be calculated as Low; since the number of correct answers for the Critical Comprehension type is 1, Critical Comprehension may be calculated as Medium; and since the number of correct answers for the Creative Comprehension type is 2, Creative Comprehension may be calculated as High.

[0121] In step S130, types whose grades by type are below a preset standard grade are identified as types with weak language ability. At this time, the preset standard grade may be set to low or medium among high, medium, and low. For example, if the standard grade is set to medium, lexical comprehension (medium), inferential comprehension (low), and critical comprehension (medium) may be identified as types with weak language ability, while factual comprehension (high) and creative comprehension (high) may not be identified as types with weak language ability. The identified types with weak language ability are transmitted to the problem generation module (130) and can be reflected in the generation of learning problems.

[0122] In step S140, a diagnostic score is calculated based on the user's response to the diagnostic problem.

[0123] In one embodiment, the diagnostic score may be calculated based on the number of correct answers the user gives to the diagnostic questions. For example, if the user answers 6 out of 10 questions correctly, the diagnostic score may be calculated as 60 points. The language ability evaluation module (120) may assign a compass grade to the user based on the calculated diagnostic score. For example, a gold compass may be assigned if the diagnostic score is 80 points or higher, a silver compass if it is 50 points or higher but less than 80 points, a bronze compass if it is less than 50 points, and a silver compass if the diagnostic score is 60 points.

[0124] In step S150, the identified language ability weakness type is transmitted to the problem generation module (130), and the calculated diagnostic score is transmitted to the level management module (150).

[0125] Accordingly, the problem generation module (130) can generate learning problems that focus on the type of language ability weakness received, and the level management module (150) can adjust the user level based on the diagnostic score received. In this way, by transmitting the results of identifying the type of language ability weakness and the diagnostic score to the respective modules, customized learning that intensively supplements the user's weakness type and precise adjustment of the user level can be achieved simultaneously.

[0126] FIG. 3 is a block diagram exemplarily illustrating the detailed configuration of a learning area evaluation module according to one embodiment of the present invention.

[0127] Referring to FIG. 3, the learning area evaluation module (140) may include a speaking evaluation unit (141), a reading evaluation unit (142), a writing evaluation unit (143), and a listening evaluation unit (144).

[0128] The speech evaluation unit (141) receives user voice data in the speech area, preprocesses the received voice data, calculates the speech rate and the fundamental frequency of the voice, and compares them with the corresponding reference values. The detailed operation of the speech evaluation unit (141) will be described later with reference to FIG. 4.

[0129] The reading evaluation unit (142) can calculate the reading speed based on the time taken by the user to read the learning text in the reading area.

[0130] As an example, the reading speed may be calculated based on the dwell time a user spends on the fingerprint page. For example, if it takes 60 seconds for a user to read a 300-character fingerprint, the reading speed may be calculated as 5 characters per second. In this case, the reading speed may be determined as high or low based on the average reading speed of users at the same level as a reference value, and as high if it is above the reference value, and low if it is below the reference value.

[0131] In one embodiment, the reading evaluation unit (142) can classify the user's reading performance type based on reading speed and the accuracy rate of solving reading area learning problems. At this time, the reading speed threshold value may be set based on the average reading speed of users of the same level, and the accuracy rate threshold value may be a preset value, for example, set to 60%. The reading performance type may be classified into four types as follows, depending on the combination of reading speed and accuracy rate. If the reading speed is greater than or equal to the threshold value and the accuracy rate is also greater than or equal to the threshold value, it may be classified as Type 1, which means that the user reads the passage quickly while accurately understanding the content. If the reading speed is greater than or equal to the threshold value but the accuracy rate is less than the threshold value, it may be classified as Type 2, which means that the user read the passage quickly but did not sufficiently understand the content, i.e., a state where comprehension is reduced due to speed reading. If the reading speed is less than the threshold value but the accuracy rate is greater than or equal to the threshold value, it may be classified as Type 3, which means that the user reads the passage slowly but accurately understands the content. If the reading speed is below the threshold value and the accuracy rate is also below the threshold value, it may be classified as Type 4, which means that the user did not fully understand the content even though they read the passage slowly.

[0132] In one embodiment, the reading evaluation unit (142) can identify whether there is weakness in the reading area and the cause of the weakness based on the reading performance type. If the reading performance type is classified as Type 2 or Type 4, the reading area can be identified as a weak area for each learning area. In this case, if classified as Type 2, feedback such as "You read quickly but your understanding of the content is insufficient. Try reading slowly to grasp the content" may be provided to the user, and the problem generation module (130) can generate a learning passage consisting of short sentences that encourage slow reading. If classified as Type 4, feedback such as "You read slowly but your understanding of the content is insufficient. You need to practice grasping the key content of the passage" may be provided to the user, and the problem generation module (130) can generate a learning passage in which the key content is clearly described and a learning problem that focuses on factual understanding type questions. Accordingly, the user's reading speed and comprehension can be comprehensively evaluated to provide customized feedback and learning tailored to the cause of the weakness.

[0133] The writing evaluation unit (143) can evaluate the writing area response received from the user in the writing area. At this time, the writing area response is a subjective answer entered by the user by writing it directly in text in response to a given question in the writing area, and may correspond to the learning activity data collected by the writing evaluation unit (143) to perform the evaluation of the writing area.

[0134] In one embodiment, the writing evaluation unit (143) can generate a first evaluation result based on whether one or more pre-set keywords are included in the writing area response. At this time, the pre-set keywords can be directly set by an administrator or teacher based on the core concept of the correct answer, or they can be automatically extracted from the correct answer using a generative language model. For example, if "photosynthesis," "light," and "carbon dioxide" are set as the correct answer keywords for the question "Explain the process of plants making nutrients on their own," and the user's writing area response includes "photosynthesis" and "light" but does not include "carbon dioxide," the first evaluation result can be calculated as including two out of three keywords.

[0135] In one embodiment, the writing evaluation unit (143) can generate a secondary evaluation result by calculating the semantic similarity between the writing area response and the correct answer. Semantic similarity can be calculated by quantifying the semantic similarity between the writing area response and the correct answer using a semantic similarity calculation model. At this time, the semantic similarity calculation model may be identical to or separate from the generative language model. For example, regarding the above problem, if the user responds "Plants receive sunlight and make nutrients themselves" and the correct answer is "Plants use light to produce nutrients through photosynthesis," the semantic similarity calculation model can analyze the semantic similarity of the two sentences and calculate a high similarity score. On the other hand, if the user responds "Plants absorb water through roots and grow," a low similarity score may be calculated because it does not include the core content of photosynthesis and nutrient production.

[0136] In one embodiment, the writing evaluation unit (143) can calculate a final evaluation score based on the first evaluation result and the second evaluation result. Specifically, the writing evaluation unit (143) can calculate a final evaluation score by assigning a pre-set weight to the first evaluation result and the second evaluation result, respectively. The weight can be adjusted depending on which element is considered more important: whether the core keyword is included or the overall meaning is conveyed. For example, if a higher score is to be given to an answer that conveys meaning even without using the keyword accurately, the final evaluation score can be calculated by assigning a weight of 40% to the first evaluation result and 60% to the second evaluation result.

[0137] As an example, if the final writing evaluation score is below a preset threshold, the writing section may be identified as a weak area by learning area. For instance, if the preset threshold is 60 points, and the user's final writing evaluation score is less than 60 points, the writing section may be identified as a weak area by learning area. In this case, if the first evaluation score is low, feedback such as "You did not sufficiently include the core keywords of the correct answer. Try writing your answer focusing on the core concepts" may be provided to the user, and if the second evaluation score is low, feedback such as "The overall meaning of the answer is insufficiently conveyed. Try writing your answer in complete sentences based on the content of the passage" may be provided to the user. Accordingly, not only the inclusion of keywords but also the semantic completeness of the answer are comprehensively evaluated, and customized feedback tailored to the cause of weakness can be provided to precisely improve the user's writing ability.

[0138] The listening evaluation unit (144) can evaluate the listening area by collecting the results of the user solving problems after listening to the voice presented in the listening area, that is, the correct or incorrect answers to the problem-solving results. Based on the evaluation results, the listening evaluation unit (144) can identify whether there is a weakness in the listening area and transmit the identified weakness to the learning area evaluation module (140).

[0139] In one embodiment, in the listening area, the user can selectively activate a fingerprint viewing function while listening to the audio. When the fingerprint viewing function is activated, the user can understand the content by referring to the fingerprint displayed on the screen along with the audio; when the fingerprint viewing function is deactivated, the user must understand the content by listening only to the audio. Through this, the difficulty level of listening learning can be flexibly adjusted according to the user's learning level and needs.

[0140] In one embodiment, if the correct answer rate for solving problems in the listening area is below a preset threshold value, the listening area may be identified as a weak area for each learning area. For example, if the preset threshold value for the correct answer rate is 60%, and the user's correct answer rate for solving problems in the listening area is less than 60%, the listening area may be identified as a weak area for each learning area. At this time, feedback such as "You have difficulty listening to the audio and understanding the content. Please activate the text view function and study by referring to the text along with the audio" may be provided to the user, and the problem generation module (130) may improve listening comprehension by generating a learning text that includes repetitive expressions and clear pronunciation. Accordingly, the user's listening comprehension can be objectively evaluated, and customized feedback and learning tailored to the cause of weakness can be provided to effectively compensate for weaknesses in the listening area.

[0141] FIG. 4 is a flowchart exemplarily illustrating the operation method of a speech evaluation unit according to one embodiment of the present invention.

[0142] The operation method described in the embodiment of FIG. 4 can be performed by the speech evaluation unit (141) of FIG. 3. Therefore, if the performing entity is omitted in the following steps, it is assumed that the performing entity is the speech evaluation unit (141). In the description of this embodiment, content identical to that described in FIG. 1 to FIG. 3 above may be omitted to avoid duplication of description.

[0143] In step S210, user voice data is received in the speaking area. For example, voice data may be received through the microphone while the user reads aloud a learning text displayed on the screen. At this time, before starting speaking learning, the microphone usage environment can be checked and a test can be performed to verify whether the voice is recognized correctly.

[0144] In step S220, the received voice data is preprocessed.

[0145] In one embodiment, the preprocessing of voice data may include a process of removing noise from the voice data. For example, noise removal may be performed using methods such as Spectral Subtraction, which extracts a noise profile from a section of voice data where there is no speech for a certain period of time and subtracts the extracted noise profile from the entire voice data, or a Wiener Filter, which estimates and removes noise using the statistical characteristics of the signal. Through this, background noise such as ambient noise, wind sound, and mechanical sound can be removed.

[0146] As an example, the preprocessing of voice data may include a process of filtering voice frequency bands that a person can utter from the voice data. For example, voice frequency band filtering may be performed by applying a band-pass filter that passes only signals in the range of approximately 80 Hz to 8000 Hz, which is the frequency band that a person can utter. Low-frequency signals below 80 Hz and high-frequency signals above 8000 Hz may be removed as they are determined to be outside the range of human speech, and only signals in the range of 80 Hz to 8000 Hz may be extracted and subsequently used to calculate the speech rate and the fundamental frequency of the voice. Accordingly, the quality of the voice data can be improved, thereby increasing the accuracy of calculating the speech rate and the fundamental frequency of the voice.

[0147] In step S230, the speech rate and the fundamental frequency of the voice are calculated based on the preprocessed voice data.

[0148] In one embodiment, the speaking speed can be calculated as the number of syllables per second (SPS) based on the time taken for a user to read a learning text aloud. In this case, when the learning text is registered in the database, the total number of syllables of the text may be stored together, or the system may analyze the text of the text to automatically calculate and store the number of syllables. For example, if it takes a user 20 seconds to read a learning text aloud that has a total of 100 syllables, the speaking speed can be calculated as 5 syllables per second (5 SPS) by dividing 100 syllables by 20 seconds.

[0149] As an example, the fundamental frequency of a voice refers to the number of periodic vibrations of the vocal cords when a person speaks, and can be expressed in Hz (Hertz) as an indicator of the pitch (tone) of the voice. A higher fundamental frequency indicates a higher voice tone, while a lower fundamental frequency indicates a lower voice tone. The fundamental frequency of a voice can be calculated by dividing preprocessed voice data into frames at regular time intervals, extracting the fundamental frequency from each frame, and averaging them. For example, preprocessed voice data can be divided into frames of 10ms intervals, and the fundamental frequency can be extracted from each frame using a method such as an autocorrelation function, and then the average value of all frames can be calculated to determine the fundamental frequency of the voice. Accordingly, the appropriateness of the user's voice tone can be evaluated by comparing the fundamental frequency of the voice with the average fundamental frequency of users of the same gender and level.

[0150] In step S240, the speech rate and the fundamental frequency of the voice are compared with their respective reference values ​​calculated based on the statistical average value of the same attribute group. Here, the same attribute group may refer to a set of users having the same gender and / or the same user level as the user.

[0151] In one embodiment, speech rate serves as an indicator of a user's reading fluency, and the user's speaking level can be objectively evaluated by comparing the average speech rate of a group of the same attribute as a reference value. If the user's speech rate is below the reference value, it may be judged as slow speech, and if it is above the reference value, it may be judged as fast speech. At this time, the normal range of the fundamental frequency of the voice may vary depending on gender and age. For example, since the fundamental frequency of men generally ranges from approximately 85 Hz to 180 Hz and that of women from approximately 165 Hz to 255 Hz, the user's voice tone can be objectively evaluated within the group by setting the average fundamental frequency of users of the same gender and / or the same level as a reference value and comparing it. If the user's fundamental frequency exceeds the reference value, the voice tone may be judged as high, and if it is below the reference value, the voice tone may be judged as low. Accordingly, the user's speech rate and voice tone can be evaluated simply and intuitively without complex statistical calculations.

[0152] In one embodiment, the speech rate and the fundamental frequency of the voice may be evaluated by comparing them to a reference range calculated based on the standard deviation of the same attribute group. If the speech rate is one standard deviation or more lower than the reference value, the speech may be judged to be slow, and if it is one standard deviation or more higher, the speech may be judged to be fast; similarly, if the fundamental frequency of the voice is one standard deviation or more higher than the reference value, the voice tone may be judged to be high, and if it is one standard deviation or more lower, the voice tone may be judged to be low. In this case, one standard deviation may refer to a statistical measure indicating how spread out the data values ​​of the same attribute group are from the mean value. For example, if the average speech rate of female users at Level 5 is 6 SPS and the standard deviation is 1 SPS, the speech may be judged to be slow if the speech rate is less than 5 SPS (mean - 1 standard deviation), and fast if it is greater than 7 SPS (mean + 1 standard deviation). Accordingly, an objective and fair evaluation of speaking can be made by comparing the user's speech rate and / or voice tone with the same attribute group.

[0153] In step S250, the frequency variability of the voice is calculated based on the preprocessed voice data.

[0154] As an example, frequency variability of speech may be an indicator representing how much the fundamental frequency of speech changes over time during the utterance process. Regarding frequency variability, it can be judged that the higher the frequency variability, the more natural the user's speech exhibits intonation and various pitch variations, while the lower the frequency variability, the more monotonously the user speaks without changes in intonation. For example, when a user reads a learning text naturally, significant changes in the fundamental frequency may occur, such as the intonation dropping at the end of a sentence or rising during a question; however, when reading monotonously, the fundamental frequency may remain constant with almost no change.

[0155] As an example, the frequency variability of speech can be obtained by dividing the fundamental frequency of speech per frame into pre-set intervals and calculating the difference value between adjacent intervals. Specifically, the fundamental frequency per frame can be divided into pre-set intervals to calculate the average fundamental frequency per interval, and the absolute value of the fundamental frequency difference between adjacent intervals can be calculated and averaged to determine the frequency variability. In this case, the pre-set interval may refer to a interval in which speech data is divided into fixed time units, and the size of the interval may be set considering the precision of speech analysis and computational efficiency. For example, the fundamental frequency per frame of pre-processed speech data can be divided into intervals of 0.1 seconds to calculate the average fundamental frequency per interval, and the absolute value of the fundamental frequency difference between two adjacent intervals can be calculated and averaged to determine the frequency variability. Specifically, if the average fundamental frequency of the interval from 0 to 0.1 seconds is 200 Hz and the average fundamental frequency of the interval from 0.1 to 0.2 seconds is 220 Hz, the absolute value of the difference between the intervals is calculated as 20 Hz. In this way, the absolute value of the difference between all adjacent intervals can be calculated and averaged to determine the frequency variability. By calculating frequency variability based on the absolute value of the difference between adjacent intervals in this manner, it is possible to measure the amount of change in fundamental frequency over time, which is difficult to capture when simply calculating the average frequency of the entire interval, and to measure intonation changes during the speech process more precisely.

[0156] In step S260, the user's speech performance type is classified by comparing the speech rate and the frequency variability of the voice with corresponding threshold values. At this time, the threshold values ​​serve as reference values ​​for classifying the speech performance type and may include a speech rate threshold and a frequency variability threshold.

[0157] In one embodiment, the utterance rate threshold may be set based on the average utterance rate of the same attribute group, and the frequency variability threshold may be set based on the average frequency variability of the same attribute group. In this case, the utterance rate threshold may be set to be the same as the utterance rate reference value used to determine whether the utterance rate is fast or slow in step S240. For example, if the average utterance rate of the same attribute group is 6 SPS, the utterance rate threshold may be set to 6 SPS, and if the average frequency variability of the same attribute group is 15 Hz, the frequency variability threshold may be set to 15 Hz.

[0158] In one embodiment, speech performance types may be classified according to a combination of speech rate and frequency variability. For example, if the speech rate is above the speech rate threshold and the frequency variability is below the frequency variability threshold, it may be classified as a first type, which may mean that the user reads quickly but the voice is monotonous. If the speech rate is below the speech rate threshold and the frequency variability is below the frequency variability threshold, it may be classified as a second type, which may mean that the user reads slowly and the voice is also monotonous. If the speech rate is above the speech rate threshold and the frequency variability is above the frequency variability threshold, it may be classified as a third type, which may mean that the user reads quickly and the pitch of the voice is varied. If the speech rate is below the speech rate threshold and the frequency variability is above the frequency variability threshold, it may be classified as a fourth type, which may mean that the user reads slowly but the pitch of the voice is varied. When the speaking performance type is Type 1 or Type 2, that is, when the frequency variability is below the frequency variability threshold, the speech is judged to be monotonous and lacks natural intonation, so the speaking area may be identified as a weak area by learning area.

[0159] In step S270, the vulnerability of the speaking area is identified based on the type of speaking performance. If the type of speaking performance is classified as Type 1 or Type 2, the speaking area may be identified as a vulnerable area by learning area and transmitted to the learning area evaluation module (140). On the other hand, if the type of speaking performance is classified as Type 3 or Type 4, the change in intonation of the voice is judged to be sufficient, so the speaking area may not be identified as a vulnerable area by learning area.

[0160] In one embodiment, if the speaking area is identified as a weak area by learning area, the learning area evaluation module (140) can transmit the relevant information to the problem generation module (130). The problem generation module (130) can generate learning passages and learning problems by reflecting the weakness of the speaking area. For example, if the speaking performance type is Type 1 (fast speaking speed + monotonous voice), the problem generation module (130) can generate learning passages consisting of short sentences that encourage slow reading, and generate learning problems centered on conversational passages that allow focus on intonation and emotional expression. If the speaking performance type is Type 2 (slow speaking speed + monotonous voice), the problem generation module (130) can generate learning passages with a reference reading speed indicated to improve both speaking speed and intonation, and generate learning problems consisting of rhythmic sentences. Accordingly, customized learning can be provided to identify the causes of weakness according to the user's speaking performance type and effectively compensate for them.

[0161] FIG. 5 is a flowchart exemplarily illustrating the operation method of a level management module according to one embodiment of the present invention.

[0162] The operation method described in the embodiment of FIG. 5 can be performed by the level management module (150) of FIG. 1. Therefore, if the performing entity is omitted in the following steps, it is assumed that the performing entity is the level management module (150). In the description of this embodiment, content identical to that described in FIG. 1 to FIG. 4 above may be omitted to avoid duplication of description.

[0163] In step S310, a diagnostic score is received from the language ability evaluation module (120). At this time, the diagnostic score may refer to a score calculated by converting the number of correct answers from the user to the diagnostic questions into a maximum of 100 points. For example, if the user answers 6 out of 10 questions correctly, the diagnostic score may be calculated as 60 points and transmitted to the level management module (150).

[0164] In step S320, weights are assigned to each of the multiple diagnostic scores calculated sequentially, with a higher weight assigned to the most recently calculated diagnostic score.

[0165] In one embodiment, the level management module (150) may give a high weight to the most recently calculated diagnostic score. Since the user's language ability may change as learning progresses, the recent diagnostic score may be reflected with a higher weight than the initial diagnostic score to more accurately reflect the user's current skill level. For example, if the user obtains 60 points in the first diagnostic test, 70 points in the second diagnostic test, and 80 points in the third diagnostic test and the score is improving, if the weights are set to 0.2 for the first test, 0.3 for the second test, and 0.5 for the third test, the weighted diagnostic score can be calculated as 60×0.2 + 70×0.3 + 80×0.5 = 12 + 21 + 40 = 73 points. On the other hand, if a user obtains 80 points in the first diagnosis, 70 points in the second diagnosis, and 60 points in the third diagnosis, and the score is decreasing, applying the same weighting can result in a weighted diagnosis score of 80×0.2 + 70×0.3 + 60×0.5 = 16 + 21 + 30 = 67 points. In this way, by assigning a higher weighting to the recent diagnosis score, upward level adjustments are reflected more quickly for users whose scores are improving, and downward level adjustments are appropriately reflected for users whose scores are decreasing, thereby enabling accurate level adjustments that match the user's current skill level.

[0166] In step S330, a level adjustment score is calculated based on the weighted diagnostic score.

[0167] In one embodiment, the level adjustment score may be calculated by summing the weighted diagnostic scores calculated in step S320. In this case, the level adjustment score may be calculated by reflecting more diagnostic score history as the number of diagnoses accumulates, and if the number of diagnoses is 1, the corresponding diagnostic score may be determined as the level adjustment score. Accordingly, the level adjustment score may be calculated by reflecting the user's entire diagnostic history while giving greater weight to the recent skill level.

[0168] In another embodiment, the level adjustment score may be calculated based on the basic level adjustment score and the auxiliary level adjustment score. The basic level adjustment score may be calculated by summing the weighted diagnostic scores calculated in step S320. The auxiliary level adjustment score may be calculated based on the weak area information for each learning area received from the learning area evaluation module (140). Specifically, the level management module (150) receives information on weaknesses for each learning area (speaking, reading, writing, and listening) from the learning area evaluation module (140), and may calculate the auxiliary level adjustment score based on the number of learning areas identified as weak areas for each learning area. For example, the auxiliary level adjustment score may be calculated by applying a pre-set deduction score (e.g., -3 points) per weak area for each learning area. Specifically, if the speaking and reading areas are identified as weak areas by learning area, the supplementary level adjustment score can be calculated as 2 × (-3 points) = -6 points; and if all four learning areas are identified as weak areas by learning area, the supplementary level adjustment score can be calculated as 4 × (-3 points) = -12 points. Subsequently, the level adjustment score can be calculated by summing the basic level adjustment score and the supplementary level adjustment score. For example, if the basic level adjustment score is 73 points and the supplementary level adjustment score is -6 points, the level adjustment score can be calculated as 73 + (-6) = 67 points. Accordingly, by additionally reflecting weak areas by learning area in addition to the diagnostic score in the level adjustment, it is possible to perform a level adjustment that more precisely reflects the user's overall language ability level.

[0169] In step S340, the user level is adjusted and saved based on the level adjustment score.

[0170] In one embodiment, a preset upper limit may represent a reference score indicating that the user has sufficiently acquired the level of language proficiency required at the current level, and a preset lower limit may represent a reference score indicating that the user has not met the level of language proficiency required at the current level.

[0171] In one embodiment, the level management module (150) may raise the user level if the level adjustment score is above a preset upper limit, lower the user level if it is below a lower limit, and maintain the user level otherwise. For example, if the level adjustment score is 80 points or higher, the user level may be raised by one step, and if it is 20 points or lower, the user level may be lowered by one step. Specifically, if the level adjustment score is 73 points, the user level may be maintained because it falls between the upper limit (80 points) and the lower limit (20 points). Level adjustment may be performed after the diagnosis is completed, and the adjusted user level may be stored in a database (not shown) and reflected in subsequent diagnosis and learning. Accordingly, the user may be provided with a learning environment of progressively higher levels as their skills improve.

[0172] FIG. 6 is a flowchart exemplarily illustrating the operation method of an incorrect answer management module according to one embodiment of the present invention.

[0173] The operation method described in the embodiment of FIG. 6 can be performed by the incorrect answer management module (160) of FIG. 1. Therefore, if the performing entity is omitted in the following steps, it is assumed that the performing entity is the incorrect answer management module (160). In the description of this embodiment, content identical to that described in FIG. 1 to FIG. 5 above may be omitted to avoid duplication of description.

[0174] In step S410, incorrect answers generated from the diagnostic problem are received and stored.

[0175] In one embodiment, an incorrect answer may refer to a question for which the user gave an incorrect response to a diagnostic question, and may be received from the language ability evaluation module (120) and automatically stored in the incorrect answer list. At this time, the incorrect answer list may include the question the user answered incorrectly, the language ability type of the question, and / or the time of the incorrect answer. For example, if the user answers question number 5 of the inferential comprehension type incorrectly, the question, the language ability type (inferential comprehension), and the time of the incorrect answer may be stored in the incorrect answer list.

[0176] In one embodiment, the incorrect answer management module (160) can store and manage incorrect answers that occurred in diagnostic problems and incorrect answers that occurred in learning problems. For example, as the user proceeds with learning after completing the diagnosis, incorrect answers from both diagnostic problems and learning problems may be accumulated and stored in an incorrect answer list, and the user can check the incorrect answers from diagnostic problems and learning problems separately from the incorrect answer list or review them by integrating them. Accordingly, by managing the incorrect answers that occurred in diagnostic problems and learning problems in an integrated manner, the user's weak type can be identified more precisely.

[0177] In step S420, a question corresponding to the stored incorrect answer is re-issued upon user request, and a user response to the re-issued question is received.

[0178] In one embodiment, the user may request the incorrect answer management module (160) to re-issue the question at a time when they wish to check the list of incorrect answers and review it. For example, if the user requests a review of question number 5 of the inferential understanding type from the list of incorrect answers, the incorrect answer management module (160) may re-issue the question to the user and receive the user's response.

[0179] In step S430, check whether the user's response to the re-asked question matches the correct answer.

[0180] If the user's response to the re-asked question matches the correct answer, proceed to step S440.

[0181] If the user's response to the re-asked question does not match the correct answer, proceed to step S450.

[0182] In step S440, delete the incorrect answer from the incorrect answer list.

[0183] In one embodiment, if a user's response to a re-issued question matches the correct answer, the corresponding incorrect answer can be automatically deleted from the incorrect answer list. Accordingly, incorrect answers resolved by the user are removed from the incorrect answer list, and only incorrect answers that have not yet been resolved remain in the list, allowing for continuous review.

[0184] In the S450 stage, the number of incorrect answers is accumulated for each language ability type corresponding to the incorrect answer.

[0185] In one embodiment, if a user's response to a re-presented question does not match the correct answer, i.e., is an incorrect answer, the number of incorrect answers may be accumulated for each type of language ability. In this case, the number of incorrect answers may be accumulated independently for each type of language ability. For example, if a user selects a first incorrect answer to a re-presented question of the inferential comprehension type, the number of incorrect answers for the inferential comprehension type may be accumulated as 1, and if the user selects an incorrect answer again to the same type of question thereafter, it may be accumulated as 2. Separately, if an incorrect answer is selected for a question of the critical comprehension type, the number of incorrect answers for the critical comprehension type may be accumulated independently as 1.

[0186] In step S460, check if the cumulative number of incorrect answers for each language ability type is greater than or equal to a preset number.

[0187] If the cumulative number of incorrect answers exceeds a preset number, proceed to step S470.

[0188] If the cumulative number of incorrect answers is less than a preset number, the current step is terminated without identifying the type of vulnerability to incorrect answers.

[0189] In one embodiment, the cumulative number of incorrect answers may be accumulated by increasing by 1 for each language ability type whenever a user selects an incorrect answer to a question re-presented through the incorrect answer management module (160), and if the incorrect answer is correct, it may be deleted from the incorrect answer list and excluded from the calculation of the cumulative number of incorrect answers for that type.

[0190] In one embodiment, the preset number may refer to a threshold number for identifying a specific language ability type as an error-prone type when the user repeatedly selects an incorrect answer in that type. For example, if the preset number is 3 times, the process may proceed to step S470 if the cumulative number of incorrect answers in the inferential comprehension type is 3 or more.

[0191] In step S470, language ability types with a cumulative number of incorrect answers exceeding a preset number are identified as types vulnerable to incorrect answers.

[0192] In one embodiment, the language ability type identified as the error-vulnerable type is transmitted to the problem generation module (130) and can be reflected based on a pool of vulnerable types formed by summing the language ability vulnerable type and the error-vulnerable type when generating learning problems after the next diagnosis. For example, the problem generation module (130) may combine the language ability vulnerable type and the error-vulnerable type into a union, but treat duplicate types as one and classify them as the highest priority type to be reflected in the generation of learning problems. Specifically, if the language ability vulnerable type is critical comprehension and inferential comprehension, and the error-vulnerable type is inferential comprehension and lexical comprehension, the combined result may be three types: critical comprehension, inferential comprehension, and lexical comprehension. At this time, since inferential comprehension is included in both the language ability vulnerable type and the error-vulnerable type, it is classified as the highest priority type, and critical comprehension and lexical comprehension may be classified as general vulnerable types. The problem generation module (130) can generate learning problems by reflecting this, with 5 problems per set, consisting of 2 problems of the inferential comprehension type, which is the highest priority type, 1 problem each of the critical comprehension and lexical comprehension types, which are general weak types, and the remaining 1 problem of other types. Accordingly, beyond simply reflecting the diagnosis results, it is possible to additionally identify the types that the user repeatedly gets wrong during the error review process and reflect them in the learning problems generated after the next diagnosis, thereby enabling customized learning that continuously tracks the user's actual weak types to identify them more precisely and intensively supplement them.

[0193] FIG. 7 is a flowchart exemplarily illustrating a language learning method according to one embodiment of the present invention.

[0194] The operation method described in the embodiment of FIG. 7 can be performed by the language learning system (100) of FIG. 1. Therefore, if the performing entity is omitted in the following steps, it is assumed that the performing entity is the language learning system (100). In the description of this embodiment, content identical to that described in FIG. 1 to FIG. 6 above may be omitted to avoid duplication of description.

[0195] In step S510, diagnostic fingerprints and diagnostic problems corresponding to the user level are provided to the user.

[0196] In one embodiment, the diagnostic module (110) may look up a diagnostic fingerprint and a diagnostic problem corresponding to the user level received from the level management module (150) in a database (not shown) and provide them to the user.

[0197] In one embodiment, the diagnostic text may include text linked to the curriculum of the grade and semester corresponding to the user level, and the diagnostic question may be composed of at least one format among multiple choice and true / false as a question corresponding to each of one or more types of language ability.

[0198] In step S520, a user response corresponding to the diagnostic problem is received.

[0199] In one embodiment, the diagnostic module (110) can receive a response to a diagnostic problem from a user and transmit it to the language ability evaluation module (120).

[0200] As one embodiment, the user response corresponding to the diagnostic problem may be received by selecting one of four options in the case of a multiple-choice question, or by selecting O or X in the case of an O / X question.

[0201] In step S530, a grade by type is calculated based on user responses corresponding to the diagnostic problem, and types of language ability weakness are identified.

[0202] In one embodiment, the language ability evaluation module (120) can calculate a grade by type based on user responses and identify types of language ability weakness and transmit them to the problem generation module (130).

[0203] In one embodiment, the language ability evaluation module (120) can calculate a diagnostic score and transmit it to the level management module (150). Subsequently, the level management module (150) can adjust the user level based on the received diagnostic score.

[0204] In the S540 step, learning passages and learning questions are generated and provided to the user based on the type of language ability weakness.

[0205] In one embodiment, the problem generation module (130) can generate learning passages and learning questions based on identified types of language ability weakness and provide them to the user. In this case, the learning passages and learning questions can be dynamically generated using a generative language model, and the difficulty level can be adjusted according to the user level.

[0206] As an example, the learning problems may consist of a total of 5 sets, and 5 questions may be presented in each set. In the case of multiple-choice questions, two attempts to solve the problem may be given, and if an incorrect answer is selected on the first attempt, a hint may be provided and a chance to try again may be given.

[0207] In step S550, user response and learning activity data corresponding to the learning problem are received.

[0208] In one embodiment, a user response corresponding to a learning problem may refer to an answer that the user selects or writes and submits regarding the learning problem. For example, in the case of a multiple-choice question, an answer in which one of four options is selected, or in the case of an O / X question, an answer in which O or X is selected may correspond to the user response. The problem generation module (130) may transmit the received user response to the language ability evaluation module (120).

[0209] In one embodiment, learning activity data may refer to data collected independently for each learning area during the process of a user solving learning problems. For example, in the speaking area, a speaking evaluation unit (141) may collect the user's voice data through a microphone; in the reading area, a reading evaluation unit (142) may measure the time the user spends on the learning text page; in the writing area, a writing evaluation unit (143) may receive a writing area response that the user has written and entered directly as text; and in the listening area, a listening evaluation unit (144) may receive the problem-solving result in which the user listens to the presented voice and solves the problem.

[0210] In one embodiment, the learning area evaluation module (140) can perform an evaluation by learning area based on collected learning activity data and identify weak areas by learning area, and the identified weak areas by learning area can be transmitted to the problem generation module (130) and reflected in the learning passages and learning problems generated thereafter. Accordingly, changes in the user's language ability level and weak areas by learning area are continuously reflected, thereby providing an optimized customized learning environment.

[0211] According to the embodiments of the present invention described so far, by diagnosing a user's language ability by type and generating and providing personalized learning problems based on the diagnosis results, it is possible to effectively supplement each user's weak language ability type.

[0212] In addition, by identifying users' weak areas in listening, speaking, reading, and writing and reflecting them in customized learning, balanced improvement of language ability can be promoted.

[0213] In addition, by managing users' incorrect answers and reflecting accumulated incorrect answer data in the generation of subsequent questions, users can continuously improve their weak types that they repeatedly get wrong and gradually improve their language skills.

[0214] Hereinafter, with reference to FIG. 8, an exemplary computing device (500) in which the methods described in various embodiments of the present invention are implemented will be described. For example, the computing device (500) of FIG. 8 may be the language learning system (100) of FIG. 1.

[0215] FIG. 8 is a block diagram illustrating the hardware configuration of a computing device used to implement various embodiments of the present invention.

[0216] As illustrated in FIG. 8, a computing device (500) may include one or more processors (510), a bus (550), a communication interface (570), a memory (530) for loading a computer program (591) executed by the processor (510), and a storage (590) for storing the computer program (591). However, FIG. 8 only illustrates components related to embodiments of the present invention. Therefore, a person skilled in the art to which the present invention pertains will understand that other general-purpose components may be included in addition to the components illustrated in FIG. 8.

[0217] The processor (510) controls the overall operation of each component of the computing device (500). The processor (510) may be configured to include at least one of a CPU (Central Processing Unit), MPU (Micro Processor Unit), MCU (Micro Controller Unit), GPU (Graphic Processing Unit), or any form of processor well known in the art of the present invention. Additionally, the processor (510) may perform operations for at least one application or program for executing a method / operation according to various embodiments of the present invention. The computing device (500) may have one or more processors.

[0218] The memory (530) stores various data, commands and / or information. The memory (530) may load one or more programs (591) from storage (590) to execute methods / operations according to various embodiments of the present invention. Examples of the memory (530) may be RAM, but are not limited thereto.

[0219] The bus (550) provides communication functions between components of the computing device (500). The bus (550) can be implemented as various types of buses, such as an address bus, a data bus, and a control bus.

[0220] The communication interface (570) supports wired and wireless internet communication of the computing device (500). The communication interface (570) may also support various communication methods other than internet communication. To this end, the communication interface (570) may be configured to include a communication module well known in the technical field of the present invention.

[0221] Storage (590) may store one or more computer programs (591) non-temporarily. Storage (590) may be configured to include non-volatile memory such as ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), flash memory, a hard disk, a removable disk, or any form of computer-readable recording medium well known in the art to which the present invention belongs.

[0222] A computer program (591) may include one or more instructions in which methods / operations according to various embodiments of the present invention are implemented. For example, the computer program (591) may include instructions for performing operations such as providing a diagnostic fingerprint corresponding to a user level and a diagnostic problem corresponding to each of one or more language ability types for said diagnostic fingerprint to a user, calculating a type-specific grade corresponding to each of said one or more language ability types based on the user's response to said diagnostic problem, and identifying a type of language ability weakness based on said type-specific grade, and generating a learning fingerprint and a learning problem corresponding to said learning fingerprint based on said type of language ability weakness and providing them to the user.

[0223] When a computer program (591) is loaded into memory (530), the processor (510) can perform methods / operations according to various embodiments of the present invention by executing one or more of the instructions.

[0224] The technical concept of the present invention described so far may be implemented as computer-readable code on a computer-readable medium. The computer-readable recording medium may be, for example, a removable recording medium (CD, DVD, Blu-ray disc, USB storage device, removable hard disk) or a fixed recording medium (ROM, RAM, computer-equipped hard disk). The computer program recorded on the computer-readable recording medium may be transmitted to another computing device via a network such as the Internet and installed on the other computing device, thereby allowing it to be used on the other computing device.

[0225] Although embodiments of the present invention have been described above with reference to the attached drawings, those skilled in the art will understand that the present invention may be implemented in other specific forms without altering the technical concept or essential features thereof. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. The scope of protection of the present invention shall be interpreted by the claims below, and all technical concepts within the equivalent scope shall be interpreted as being included within the scope of rights of the technical concept defined by the present invention. Explanation of the symbols

[0227] 100: Language Learning System 110: Diagnostic Module 120: Language Proficiency Assessment Module 130: Problem Generation Module 140: Learning Domain Evaluation Module 150: Level Management Module 160: Incorrect Answer Management Module

Claims

Claim 1 A diagnostic module that provides a diagnostic passage corresponding to a user level to the user, provides a diagnostic problem corresponding to each of one or more language ability types based on the diagnostic passage to the user, and receives the user's response to the diagnostic problem; a language ability evaluation module that calculates a type-specific grade corresponding to each of the one or more language ability types based on the user's response to the diagnostic problem, and identifies a type of language ability weakness based on the type-specific grade; and a problem generation module that generates a learning passage and a learning problem corresponding to the learning passage based on the type of language ability weakness and provides them to the user.It includes a learning area evaluation module that performs an evaluation by learning area based on the user's learning activity data corresponding to each of one or more learning areas and identifies weak areas by learning area, wherein the learning area evaluation module includes a speaking evaluation unit, wherein the speaking evaluation unit preprocesses voice data received from the user in the speaking area, calculates speech rate and frequency variability of the voice based on the preprocessed voice data, classifies the user's speaking performance type by comparing the speech rate and frequency variability of the voice with corresponding threshold values, and identifies whether there is a weakness in the speaking area based on the speaking performance type and transmits it to the problem generation module, wherein the frequency variability of the voice is obtained by dividing the fundamental frequency of the voice per frame into preset intervals and calculating the difference value between adjacent intervals, wherein the problem generation module generates the learning text and the learning problem by reflecting the weakness in the speaking area, and the one or more language ability types have a sequential hierarchical structure defined in the order of lexical understanding, factual understanding, inferential understanding, critical understanding, and creative understanding, and the problem generation module, based on the sequential hierarchical structure, immediately preceding the language ability weakness type A language learning system that identifies a prerequisite type corresponding to a stage and configures the learning problem by additionally selecting a problem of the said prerequisite type.; Claim 2 In claim 1, the problem generation module generates the learning passage and the learning problem based on the learning domain-specific weak areas along with the language ability weakness type, in a language learning system. Claim 3 delete Claim 4 A language learning system according to claim 1, wherein the diagnostic module assigns at least one of the one or more language ability types to each of the diagnostic problems, and the language ability evaluation module aggregates correct and incorrect answers for each of the language ability types assigned to each of the diagnostic problems to calculate the grade for each type, and identifies the language ability type corresponding to the grade for each type that is below a preset grade standard as the language ability weak type. Claim 5 A language learning system according to claim 1, wherein the user level is composed of a plurality of stages corresponding to a grade and a semester, and the diagnostic text includes a text linked to the curriculum of the grade and semester corresponding to the user level. Claim 6 A language learning system according to claim 1, wherein the learning area evaluation module comprises one or more evaluation units corresponding to each of the one or more learning areas, and the one or more learning areas comprise one of a listening area, a speaking area, a reading area, and a writing area. Claim 7 In claim 1, the speech evaluation unit calculates the speech rate and the fundamental frequency of the speech based on the preprocessed voice data and compares them with a reference value calculated based on the statistical average value of the same attribute group, in a language learning system. Claim 8 delete Claim 9 A language learning system according to claim 6, wherein one or more evaluation units include a writing evaluation unit, and the writing evaluation unit generates a first evaluation result based on whether one or more preset keywords are included in the writing area response, generates a second evaluation result by calculating semantic similarity between the writing area response and the correct answer, and calculates a final evaluation score based on the first evaluation result and the second evaluation result. Claim 10 A language learning system according to claim 1, further comprising an error management module that stores an error that occurred in the diagnostic problem and provides the stored error to the user, wherein the error management module re-presents a problem corresponding to the error to the user, and if the user's response to the re-presented problem does not match the correct answer, accumulates the number of error occurrences by language ability type of the error, and identifies a language ability type in which the accumulated number of error occurrences is greater than or equal to a preset number as an error vulnerability type and provides it to the problem generation module. Claim 11 A language learning method performed by a computing device comprises: providing a user with a diagnostic fingerprint corresponding to a user level and a diagnostic problem corresponding to each of one or more language ability types for said diagnostic fingerprint; calculating a type-specific grade corresponding to each of said one or more language ability types based on the user's response to said diagnostic problem, and identifying a type of language ability vulnerability based on said type-specific grade; preprocessing voice data received from said user in a speaking area, calculating speech rate and frequency variability of voice based on the preprocessed voice data, classifying the user's speaking performance type by comparing said speech rate and said frequency variability with corresponding threshold values, and identifying whether there is vulnerability in the speaking area based on said speaking performance type. A language learning method comprising the step of generating a learning passage and a learning problem corresponding to the learning passage based on the above-mentioned type of weakness in language ability and providing it to the user, wherein the frequency variability of the voice is obtained by dividing the fundamental frequency of the voice per frame into preset intervals and calculating the difference value between adjacent intervals, and wherein the one or more types of language ability have a sequential hierarchical structure defined in the order of lexical understanding, factual understanding, inferential understanding, critical understanding, and creative understanding, and wherein the step of generating and providing the learning problem to the user generates the learning passage and the learning problem by reflecting the weakness in the speaking area, and, based on the above-mentioned sequential hierarchical structure, identifies a preceding type corresponding to the stage immediately preceding the type of weakness in language ability, and additionally selects a problem of the preceding type to construct the learning problem. Claim 12 A computer program stored on a computer-readable recording medium to execute the language learning method described in paragraph 11.

Images