KR102648139B1 - Server and method for providing AI chatbot tutor for complete learning-based personalized learning support - Google Patents

Abstract

One aspect of the present invention provides an AI chatbot tutor providing server for personalized learning support based on complete learning. The AI chatbot tutor providing server for complete learning-based personalized learning support includes at least one processor and a memory that stores instructions that instruct the at least one processor to perform at least one step. It can be included. At least one step includes collecting learner information, performing a diagnostic assessment to evaluate the learner's current learning level to create a learning session appropriate for the learner's learning level, and determining the learner's learning goals based on the results of the diagnostic assessment. Steps to determine recommended learning topics, steps to proceed with learning according to the determined learning goals and learning plan, and tracking the learner's learning progress in real time, and if the learner does not continue to progress in a specific learning content, Steps to determine progress barriers and provide additional support to progress the learning; If the learner is determined to have fully understood the topic, a notification can be created to encourage the learner to move on to the next learning topic; and ending the learning session. It may include a step of analyzing the learner's learning performance based on the step of determining the learner's understanding and updating the learner's profile.

Description

Server and method for providing AI chatbot tutor for complete learning-based personalized learning support {Server and method for providing AI chatbot tutor for complete learning-based personalized learning support}

The present invention relates to a server and method for providing an AI chatbot tutor for supporting personalized learning based on complete learning. More specifically, it relates to a server and method for providing personalized learning based on complete learning to provide customized learning for learners by providing a chatbot tutor using artificial intelligence. This relates to a server and method for providing AI chatbot tutors.

The digital education environment has developed in a variety of ways, with an increasing emphasis on personalized learning experiences. In addition, due to the active research and introduction of non-face-to-face education systems due to the recent pandemic (COVID-19), digital-based non-face-to-face education can no longer be viewed as unfamiliar, and the market is gradually expanding its scope.

In addition, with recent advances in computing technology, a variety of chatbot technologies using natural language processing algorithms based on deep learning or machine learning have emerged, and various attempts are being made to utilize them in the learning field.

However, individual learners have various variables such as learning tendencies, learning methods, learning abilities, and areas of interest, and even when using high-performance natural language processing modules, it is still difficult to provide a completely personalized learning program.

Recently, some interactive learning methods have been introduced with the introduction of chatbots in the digital education environment, but these also cannot provide accurate, contextual, and relevant answers to complex educational questions, and are only suitable for comprehensive and iterative learning. The current situation is that only the same answer is provided.

In other words, questions in learning should be asked from the perspective of mutual exchange of knowledge compared to conversation in everyday life, but existing studies are experiencing difficulties because this perspective is not reflected.

Accordingly, there is a need for research on providing AI chatbot tutors to support personalized learning based on complete learning that can reflect the learner's personal learning tendencies and performance and continuously track the learner to progress learning.

Domestic registered patent No. 10-2232343

The purpose of the present invention to solve the above problems is to provide a server and method for providing an AI chatbot tutor for supporting personalized learning based on complete learning.

One aspect of the present invention for achieving the above object is to provide an AI chatbot tutor providing server for personalized learning support based on complete learning.

The AI chatbot tutor providing server for complete learning-based personalized learning support includes at least one processor and a memory that stores instructions that instruct the at least one processor to perform at least one step. It can be included.

At least one step includes collecting learner information, performing a diagnostic assessment to evaluate the learner's current learning level to create a learning session appropriate for the learner's learning level, and determining the learner's learning goals based on the results of the diagnostic assessment. Steps to determine recommended learning topics, steps to proceed with learning according to the determined learning goals and learning plan, and tracking the learner's learning progress in real time, and if the learner does not continue to progress in a specific learning content, Steps to determine progress barriers and provide additional support to progress the learning; If the learner is determined to have fully understood the topic, a notification can be created to encourage the learner to move on to the next learning topic; and ending the learning session. It may include a step of analyzing the learner's learning performance based on the step of determining the learner's understanding and updating the learner's profile.

In addition, according to the progress of the diagnostic evaluation and learning session, it may include the step of creating a user interface (UI) that provides a report analyzing the learner's learning performance, learning strengths and weaknesses.

At this time, the diagnostic evaluation to determine the learner's learning level infers the learner's learning level based on the received learner information and extracts problems from the diagnostic evaluation problems stored in the learning data database (DB), and selects learning problems based on artificial intelligence. Using the model, diagnostic evaluation problems appropriate for the learner's level are selected from among the diagnostic evaluation problems stored in the learning data database (DB) and provided to the learner through the user terminal. Diagnostic evaluation is performed through an adaptive quiz method or classification. It may be characterized by performing a diagnostic evaluation using a quiz-based method.

Meanwhile, the collected learner information is classified into at least one category according to attributes, with age, economic background, cultural background, and language ability classified into social background categories, and grade, subject of interest, academic performance, learning goals, major, and preference. Classify subjects, target occupations, and research interests into academic goal categories, and study habits by classifying learning style, group activity preference, online learning affinity, time management efficiency, learning speed, learning efficiency, study habits, main study time, and main study methods. It further includes a step of classifying into categories, calculates cultural economic indicators from the average of economic background and cultural background classified into social background categories, and calculates language age integration index from the average of language ability and age conversion scores. The score is determined from the difference between the appropriate age of the subject the learner wants to learn and the learner's age, and the step of calculating the social background score from the weighted sum of the calculated cultural economic index and language age integration index, grade conversion score, academic performance, The learning goal score is calculated from the average of the learning goals and research interests, but the grade conversion score is determined by the difference between the appropriate grade of the learning topic the learner wants to learn and the learner's grade, online learning friendliness, time management efficiency, and learning speed. , calculating a learning habit score from the weighted sum of learning efficiency and main learning time, calculating a learner information score from the weighted sum of the calculated social background score, learning goal score, and learning habit score, the calculated learner information score It may include the step of creating a learner group by clustering learners included within a predetermined error range and providing the same diagnostic evaluation problem to the learners clustered in the same learner group.

In addition, collecting learning path data generated from the learner's learning progress, classifying the collected learning path data into at least one category according to attributes, including the number of interactions per learning session, average duration per interaction, and total learning session. Time spent, average time spent on learning sessions, number of accessed learning materials, and time spent for each learning material are classified as learning participation indicators, and evaluation scores, percentage of correct answers, percentage of incorrect answers, response time, and grade trends are classified as learning ability indicators. The step of classifying the number of feedback and feedback time into learning feedback index items, the product of the number of interactions per learning session and the average duration per interaction, the product of the number of accessed learning materials and the time spent for each learning material, and the learning session. A step of calculating the learning participation score by dividing the total time required and the average time spent in the learning session; A step of calculating the learning ability score by dividing the sum of the correct answer rate and performance trend by the sum of the incorrect answer rate and response time; Number of feedback and feedback A step of calculating a learning feedback score from the average of time, a step of calculating a learning path recommendation score based on the weighted sum of the calculated learning participation score, learning ability score, and learning feedback score, and the calculated learning path recommendation score having a predetermined error It may include the step of creating a learner path group by clustering learners within the scope.

When using the AI chatbot tutor providing server and method for supporting personalized learning based on complete learning according to the present invention as described above, a personalized educational experience tailored to the individual needs and learning style of the learner can be provided.

In addition, it is possible to suggest an efficient learning direction through continuous monitoring and analysis of the learner's learning performance, and it is possible to improve learner satisfaction and learning effectiveness.

The effects that can be obtained from the present invention are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below. will be.

The above-described and other aspects, features and benefits of certain preferred embodiments of the present invention will become more apparent from the following description taken in conjunction with the accompanying drawings.
Figure 1 is an exemplary diagram showing the operating environment of an AI chatbot tutor providing server for personalized learning support based on complete learning according to an embodiment of the present invention.
Figure 2 is an example diagram showing an example of an AI chatbot tutor providing server for complete learning-based personalized learning support.
Figure 3 is an exemplary algorithm configuration diagram for providing accurate answers in a learning session through RAG (Retrieval-Augmented Generation).
Figure 4 is a flowchart showing an example operation of an AI chatbot tutor providing server for personalized learning support based on complete learning according to an embodiment of the present invention.
Figure 5 is a hardware configuration diagram of a context-specific curriculum implementation server for providing experiential education according to Figure 1.
It should be noted that throughout the drawings, like reference numerals are used to illustrate identical or similar elements, features and structures.

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

In describing the embodiments, description of technical content that is well known in the technical field to which the present invention belongs and that is not directly related to the present invention will be omitted. This is to convey the gist of the present invention more clearly without obscuring it by omitting unnecessary explanation.

For the same reason, some components are exaggerated, omitted, or schematically shown in the accompanying drawings. Additionally, the size of each component does not entirely reflect its actual size. In each drawing, identical or corresponding components are assigned the same reference numbers.

The advantages and features of the present invention and methods for achieving them will become clear by referring to the embodiments described in detail below along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms. The present embodiments are merely provided to ensure that the disclosure of the present invention is complete and to provide common knowledge in the technical field to which the present invention pertains. It is provided to fully inform those who have the scope of the invention, and the present invention is only defined by the scope of the claims. Like reference numerals refer to like elements throughout the specification.

At this time, it will be understood that each block of the processing flow diagrams and combinations of the flow diagram diagrams can be performed by computer program instructions. These computer program instructions can be mounted on a processor of a general-purpose computer, special-purpose computer, or other programmable data processing equipment, so that the instructions performed through the processor of the computer or other programmable data processing equipment are described in the flow chart block(s). It creates the means to perform functions. These computer program instructions may also be stored in computer-usable or computer-readable memory that can be directed to a computer or other programmable data processing equipment to implement a function in a particular manner, so that the computer-usable or computer-readable memory It is also possible to produce manufactured items containing instruction means that perform the functions described in the flowchart block(s). Computer program instructions can also be mounted on a computer or other programmable data processing equipment, so that a series of operational steps are performed on the computer or other programmable data processing equipment to create a process that is executed by the computer, thereby generating a process that is executed by the computer or other programmable data processing equipment. Instructions that perform processing equipment may also provide steps for executing the functions described in the flow diagram block(s).

Additionally, each block may represent a module, segment, or portion of code that includes one or more executable instructions for executing specified logical function(s). Additionally, it should be noted that in some alternative execution examples it is possible for the functions mentioned in the blocks to occur out of order. For example, it is possible for two blocks shown in succession to be performed substantially at the same time, or it is possible for the blocks to be performed in reverse order depending on the corresponding function.

At this time, the term '~unit' used in this embodiment refers to software or hardware components such as FPGA (field-programmable gate array) or ASIC (Application Specific Integrated Circuit), and '~unit' refers to what role perform them. However, '~part' is not limited to software or hardware. The '~ part' may be configured to reside in an addressable storage medium and may be configured to reproduce on one or more processors. Therefore, as an example, '~ part' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, and procedures. , subroutines, segments of program code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables. The functions provided within the components and 'parts' may be combined into a smaller number of components and 'parts' or may be further separated into additional components and 'parts'. Additionally, components and 'parts' may be implemented to regenerate one or more CPUs within a device or a secure multimedia card.

In describing the embodiments of the present invention in detail, the main focus will be on the example of a specific system, but the main point claimed in this specification is that the scope disclosed in this specification is applicable to other communication systems and services with similar technical background. It can be applied within a range that does not deviate significantly, and this can be done at the discretion of a person with skilled technical knowledge in the relevant technical field.

Figure 1 is an exemplary diagram showing the operating environment of an AI chatbot tutor providing server for personalized learning support based on complete learning according to an embodiment of the present invention.

Referring to FIG. 1, the AI chatbot tutor providing server 100 for supporting personalized learning based on complete learning (hereinafter referred to as 'server 100') learns from at least one external database (DB) 110, 111. Information and learner information can be received, and an AI chatbot tutor for personalized learning support based on complete learning can be provided to the learner through the learner terminal 10.

Here, the server 100 may be connected to at least one external database (DB) 110, 111 and the user terminal 10 through a wired or wireless network.

Here, the user terminal 10 is a communication capable desktop computer, a laptop computer, a laptop, a smart phone, a tablet PC, a mobile phone, Smart watch, smart glass, e-book reader, PMP (portable multimedia player), portable game console, navigation device, digital camera, DMB (digital multimedia broadcasting) player, It may be a digital audio recorder, digital audio player, digital video recorder, digital video player, PDA (Personal Digital Assistant), etc.

In addition, wired and wireless network methods include Bluetooth communication, BLE (Bluetooth Low Energy) communication, Near Field Communication (NFC), WLAN communication, Zigbee communication, Infrared Data Association (IrDA) communication, It may include various communication methods such as WFD (Wi-Fi Direct) communication, UWB (ultra-wideband) communication, Ant+ communication, WIFI communication, and RFID (Radio Frequency Identification) communication, but is not limited thereto.

Meanwhile, the server 100 and at least one external database (DB) 110, 111 may be implemented as one system by being combined with each other.

Referring to FIG. 1, the server 100 can collect learner information. Here, learner information includes age, grade, learning style (e.g. visual, auditory, reading/writing, physical activity), subject interests, academic history, learning goals, language skills, socioeconomic background, cultural background, major, subject preferences, This may include, but is not limited to, target occupation, research interest, group activity preference, online learning affinity, time management efficiency, learning speed, learning efficiency, leadership, study habits, main study time, and main study method.

The server 100 may classify the collected learner information into at least one category according to the properties of the information. Next, the server 100 may calculate a learner information score by combining learner information classified into at least one category.

When a learner starts learning through the user terminal 10, the server 100 may provide an interface that introduces the learning direction. For example, learning directions include complementary learning that complements the learning content already performed, iterative learning that re-learns the learning content that has already been performed, additional learning that learns additional content beyond the learning content that has already been performed, and content of the learning that has already been performed. There can be complete learning in which learning is repeated until the subject is fully understood, and the learning performance is measured by evaluating the completeness of the learning.

In addition, the server 100 can provide help on strategies and methods for the learning direction through an interface that introduces the learning direction, and the provided help is created in advance and stored in the external database (DB) 110, 111. Data is available.

Next, the server 100 may perform a diagnostic evaluation to determine the learner's current learning level before starting learning according to the learning direction.

Here, the diagnostic evaluation for determining the learner's learning level may infer the learner's learning level based on the received learner information and use diagnostic evaluation questions stored in the external database (DB) 110, 111. For example, the server 100 uses an artificial intelligence-based learning problem selection model to select diagnostic evaluation problems appropriate for the learner's level among diagnostic evaluation problems stored in the external database (DB) 110 and 111, and selects diagnostic evaluation problems appropriate for the learner's level through the user terminal. (10) can be provided to the learner to diagnose the learner's learning level.

At this time, the artificial intelligence-based learning problem selection models are deep neural network (DNN), convolutional neural network (CNN), recurrent neural network (RNN), restricted Boltzmann machine (RBM), deep belief network (DBN), and bidirectional recurrent deep neural network (BRDNN). ), a deep Q-network (DQN), or a pre-trained learning model using one of these methods, but is not limited to this.

Meanwhile, when the learner's learning diagnosis is completed, the server 100 may determine the learner's learning level based on the learner's diagnostic evaluation result and determine the learner's learning goal according to the determined learner's learning level. Here, the learner's learning goal may include learning subjects, learning time, learning progress, and learning completeness depending on the learner's learning level, but is not limited to this, and may mean a learning goal that the learner wishes to achieve through a learning session. .

Once the learner's learning goal is determined, the server 100 may start a learning session. Here, a learning session may mean including learning content and processes to achieve the determined learning goal of the learner, but is not limited thereto, and may include learning session issues as learning materials for evaluating the learner's learning level. there is.

The learning session problem may be provided through the user terminal 10 by receiving the learning session problem from an external database (DB) 110, 111 according to the learner's diagnostic evaluation result and the learner's goal.

In addition, when the learner's learning topic and learning progress change according to the learner's learning goal, the server 100 collects learning session problems from the external database (DB) 110, 111 according to the learner's learning topic and learning progress. It can be received and provided to the learner through the user terminal 10.

Through this, the server 100 can provide gradual learning for learners to achieve their learning goals and provide learning according to a previously determined learning direction.

Next, the server 100 can monitor real-time learning to answer questions about the learner's learning content and answers to learning session problems as the learner proceeds with learning according to the learning goal.

For example, when a learner performs learning through the user terminal 10, the server 100 provides responses generated through interaction with the user terminal 10, such as the learner's login, learning session, and learning material progress rate. , you can monitor a learner's learning session in real time through answers to questions or learning session issues, interactions with other learners (including questions and answers), learning time, and interaction history with the artificial intelligence-based chatbot. . In addition, the present invention is not limited to this, and as an interface of the user terminal 10, the student's learning session can be monitored in real time by analyzing touch, app execution, and use of learning tools.

Additionally, the server 100 may provide real-time feedback on the learner's learning session according to real-time monitoring of the learner's learning session.

For example, the server 100 may provide feedback such as providing learning content or providing correct answers according to the learner's questions or answers about the learning content, but is not limited to this, and may respond to learning session problems provided during the learning session. Hints can be provided. Here, the hint is a hint for a learning session problem and can be direct or indirect learning content for solving the learning session problem.

Additionally, when the learner completes all of the learning session problems, the server 100 provides correct answers and explanations for errors in solving the learning session problems, thereby confirming and correcting the learner's learning direction.

In other words, the server 100 can provide customized learning options for the learner by reflecting the learning situations that occur while the learner progresses through the learning session and the learning results according to the resolution of the learning session problems, and provides real-time monitoring and feedback to help the learner. Learning efficiency can be immediately improved.

Meanwhile, the server 100 can check the learner's learning progress according to the learner's learning direction and end the learning session.

For example, if the learner's learning direction is complete learning, the server 100 repeats learning until the learner is completely familiar with the content of the learning already performed, evaluates the completeness of the learning, and repeatedly measures the learning performance, so that the learner If you have a complete understanding of the learning topic, you can end the learning session. Here, complete learning may mean an understanding level above a threshold predetermined by the server 100. For example, if the correct answer rate of the learning session problem is 80% or more or is completely 100%, learning according to the learner's learning direction has been completed. You can decide that this is the case and end the learning session.

When a learning session ends, the server 100 may store the learner's learning progress status, learning contents, and learning session data that the learner progressed in the learning session as the learner's learning path.

The server 100 may analyze the learner's learning performance and create a learner profile. For example, the server 100 may create a learner profile by mapping the learning direction, learning session problems, learning situation, and learning path performed by the learner during the learning session with learner information and storing it.

Additionally, the server 100 may update the generated learner profile based on the learning direction, learning session problems, learning situation, and learning path performed during the learner's learning session.

Through this, the server 100 can improve learning content according to the learner's learning response and learning preference, and provide a learner-friendly AI chatbot by providing a conversation style suitable for the learner's tendency.

Figure 2 is an example diagram showing an example of an AI chatbot tutor providing server for personalized learning support based on complete learning. Referring to Figure 2, the AI chatbot tutor providing server 200 for supporting personalized learning based on complete learning (hereinafter referred to as 'server 200') includes a learner information collection unit 210, a diagnostic evaluation unit 220, Learning session unit 230, learning performance analysis unit 240, learning path recommendation unit 250, learner profile management unit 260, learning data database (DB) 270, and learner data database (DB) 280. It can be included.

The learner information collection unit 210 can collect learner information. Here, learner information includes age, economic background, cultural background, language ability, grade, subjects of interest, academic performance, learning goals, major, preferred subjects, target occupation, research interests, and learning style (e.g. visual, auditory, reading/writing, Physical activity), group activity preference, online learning friendliness, time management efficiency, learning speed, learning efficiency, study habits, main study time, main study method may be included, but is not limited to these.

In addition, learner information may be stored in advance in the learner database (DB) 280, but is not limited to this, and if the learner information is not stored in the learner database (DB) 280, the learner information collection unit 210 A new learner profile can be created in the database (DB) 280 and learner information can be stored and managed.

The learner information collection unit 210 may classify the collected learner information into at least one category according to the properties of the information. For example, the learner information collection unit 210 may classify age, economic background, cultural background, and language ability into social background categories. The social background category may include factors that affect the learner's overall educational experience and approach, and the learner information included in the social background category is the learner's personal and basic disposition and is less likely to change over time. This may be information that can be used to determine the learner's needs and potential performance.

Additionally, the learner information collection unit 210 may classify grade, subject of interest, academic performance, learning goal, major, preferred subject, target occupation, and research interest into academic goal categories. The academic-oriented category refers to the learner's academic goals and target performance, and may include indicators related to long-term performance, and the learner information collection unit 210 collects learner information classified into the academic-oriented category to customize the learner's learning experience. You can utilize it.

In addition, the learner information collection unit 210 collects learning styles (e.g. visual, auditory, reading/writing, physical activity), group activity preference, online learning affinity, time management efficiency, learning speed, learning efficiency, study habits, Study time and main study methods can be classified into study habit categories. The learning habits category may include indicators that indicate the learner's method and ability to acquire and interact with learning content.

The learner information collection unit 210 may calculate a learner information score by combining learner information classified into at least one category. For example, the learner information collection unit 210 can calculate cultural economic indicators based on the average of economic background and cultural background as shown in the equation below.

At this time, the economic background can be determined in advance by the learner based on economic factors such as average family income, average household education level, and occupation, and can be quantified with a score of 0 to 1. Additionally, cultural background is a quantification of the learner's cultural background including race, country of origin, and religion, and can be quantified by the learner with a score of 0 to 1.

Meanwhile, the learner information collection unit 210 can calculate the language age integration index based on the learner's age and language ability as shown in the equation below.

At this time, the age conversion score can be calculated as the difference between the appropriate age of the learning topic the learner wants to learn and the learner's age, as shown in the equation below. For example, if the appropriate learning age for first-year university subject 'A' is 19 years old, and the learner's age is 18 years old, the age conversion score is calculated as 0.95 according to the equation below.

Additionally, the language ability may be normalized to 0 to 1, and may be received in advance by the learner information collection unit 210 before the learner begins learning. As an example, language ability may be determined on a predetermined scale such as beginner (0.3), intermediate (0.6), and advanced (0.9). For example, if language ability is intermediate (0.6) and the age conversion score is 0.95, language proficiency The integrated indicator is calculated as shown in the equation below.

At this time, the learner information collection unit 210 can calculate the social background score through a weighted sum of the cultural economic index and the language age integration index. At this time, when the predetermined weights w1 and w2 are each 0.5, the social background score is calculated as shown in the equation below.

The calculated social background score is a score that reflects the learner's basic background and cultural background. The larger the value, the higher the learner's potential learning growth ability, and the smaller the value, the more diverse the learner's improvement can be. It can mean.

In addition, the learner information collection unit 210 calculates the learning goal score based on the grade, subject of interest, academic grade, learning goal, major, preferred subject, target occupation, and research interest classified into the academic-oriented category as shown in the equation below. You can.

Here, the grade conversion score can be calculated as the difference between the appropriate grade of the learning topic the learner wants to learn and the learner's grade as shown in the equation below. For example, if a third-year university student takes the second-year university course 'A', the grade conversion score is calculated as 0.5 according to the equation below.

Additionally, the academic grade can be normalized to a value of 0 to 1 through maximum-minimum normalization. For example, if a student's academic grade is 3.0 and the maximum academic grade is 4.5, the academic grade can be normalized to 0.67.

Additionally, learning objectives can be normalized based on SMART. Here, the SMART criteria are Specific, Measurable, Achievable, Relevant, and Time-Bound, which represent a framework for achievable learning goals and measure specific goals. The feasibility criteria determine whether there is a realistic possibility of achievement taking into account available resources and constraints, are related to other important goals, have a clearly defined schedule, including a specific start date and schedule, and are predetermined. It can be normalized to 0 to 1 depending on the standard.

Additionally, research interest is the learner's level of interest in the field he or she is learning and can be normalized to 0 to 1 according to predetermined standards. For example, a learner who is not interested in 'differential calculus' may have a research interest of 0, and a learner who is highly interested may have a research interest of 1.

The learner information collection unit 210 includes learning styles classified into study habit categories (e.g., visual, auditory, reading/writing, physical activity), group activity preference, online learning affinity, time management efficiency, learning speed, learning efficiency, By combining study habits, main study time, and main study methods, the study habit score can be calculated as shown in the equation below.

Here, online learning affinity may mean affinity including the learner's adaptability to online learning, and may be predetermined as 0 to 1 by the evaluation of the instructor, parents, educational institution personnel, or the learner himself.

Additionally, time management efficiency indicates compliance with a predetermined learning period for a predetermined amount of learning, and can be calculated as 0 to 1. For example, if a one-week learning period is given for each of the 10 learning chapters, and the number of times the learner completes the learning within a designated one-week period is 7, time management efficiency can be determined as 0.7.

In addition, learning speed may refer to the rate at which a learner completes learning quickly compared to the average learning time of a group of learners learning the same learning topic. For example, the average learning completion time of a group of learners learning the same learning content is 1 hour. In this case, if the learner's learning completion time is less than 1 hour, the learner's learning speed can be judged to be fast. For example, if the learner's learning speed is judged to be faster 5 times out of 10 learning times, the learning speed is 0 to 0. It can be quantified from a value of 1 to 0.5.

In addition, the main learning time quantifies the time when the learner studies the most. Depending on the learner's learning pattern, it has a value of 1 when the standard deviation of the learner's main learning time is 0, and the standard deviation is greater than 0 and less than 2. In this case, a value of 0.5 can be given, and if the standard deviation is greater than 2 and less than 4, a value of 0.2 can be given. In other words, the time of the day when the learner can study is divided into time areas in 2-hour increments (e.g. 12-14:00, 14-16:00, 16-18:00, ??), and the divided time areas are used for each learning period. If it varies from time to time, the main learning time may be judged to be irregular, and the score given to the main learning time may be calculated low.

The learner information collection unit 210 can calculate the learner information score through the weighted sum of the calculated social background score, learning goal score, and study habit score.

For example, when the calculated social background score is 0.7, the learning goal score is 0.6, and the study habit score is 0.7, and the predetermined weights w1 to w3 are 0.2, 0.5, and 0.3, respectively, the learner information score is calculated as follows.

The learner information score can have a value of 0 to 1. A higher learner information score means that the social and economic background in which the learner can learn is flexible, and the learner's learning goal is clear and the learner's activeness in learning is high. This means that the learner's learning habits are a suitable environment for learning by applying a personalized AI-based chatbot tutor.

Additionally, the learner information collection unit 210 may create a learner group by clustering learners whose calculated learner information scores are included within a predetermined error range.

The diagnostic evaluation unit 220 may perform a diagnostic evaluation to evaluate the learner's learning level in order to create a learning session appropriate for the learner's learning level. The diagnostic assessment unit 220 can provide a report by analyzing the learner's learning strengths and weaknesses as a result of the performed diagnostic evaluation, and can be used as basic data to determine the learner's learning performance.

For example, the diagnostic evaluation unit 220 performs a diagnostic evaluation to determine the learner's learning level by inferring the learner's learning level based on the received learner information and determining the diagnostic evaluation problem stored in the learning data database (DB) 270. You can use it. For example, the diagnostic evaluation unit 220 uses an artificial intelligence-based learning problem selection model to select diagnostic evaluation problems appropriate for the learner's level among diagnostic evaluation problems stored in the external learning data database (DB) 270, You can diagnose the learner's learning level by providing it to the learner through the user terminal.

At this time, the artificial intelligence-based learning problem selection models are deep neural network (DNN), convolutional neural network (CNN), recurrent neural network (RNN), restricted Boltzmann machine (RBM), deep belief network (DBN), and bidirectional recurrent deep neural network (BRDNN). ), a deep Q-network (DQN), or a pre-trained learning model using one of these methods, but is not limited to this.

Meanwhile, the diagnostic evaluation unit 220 may provide the same diagnostic evaluation problem to learners clustered into the same learner group based on the analysis of learner information by the learner information collection unit 210. Through this, it is possible to conduct a diagnostic evaluation at an appropriate level for learners of the same level without prior information about the learner.

In addition, the diagnostic evaluation unit 220 may select diagnostic evaluation problems so that the problems presented to the learner are composed of various levels of difficulty from beginner to advanced as an evaluation to evaluate the learner's learning level. For example, the diagnostic evaluation unit 220 may select a problem using a quiz table.

As an example, the diagnosis convenience unit 220 may apply an adaptive quiz method (Adaptive Quiz). The adaptive quiz method initially presents questions of medium difficulty and adjusts the difficulty level according to the learner's responses. For example, at the beginning of the diagnostic evaluation, a problem of medium difficulty can be presented to the learner, if the learner's answer is correct, a problem of more difficult difficulty can be presented, and if the learner has given an incorrect answer, a problem of easier difficulty can be presented. . This adaptive quiz method can accurately assess the learner's current level by quickly identifying the learner's level and presenting questions appropriate to the learner's level.

Additionally, the diagnostic evaluation unit 220 may perform diagnostic evaluation through a category-based quiz. The classification-based quiz method presents problems for each learning topic or field, allowing learners to identify their strengths and weaknesses by topic. The diagnostic evaluation unit 220 presents several basic problems for each learning topic at the start of the diagnostic evaluation, and if the learner shows weakness in a problem of a specific topic according to the learner's response, it presents additional problems of that topic and helps the learner If a learner shows strength in a specific topic, the learner is evaluated by moving on to another topic, allowing the learner's strengths and weaknesses to be accurately identified by topic.

To provide this diagnostic evaluation method, the diagnostic evaluation unit 220 can add a problem difficulty column to the quiz table and classify problems by difficulty level. In addition, the diagnostic evaluation unit 220 can classify which topic the problem belongs to by connecting the quiz table and the section table, and can identify weaknesses and strengths for each topic by analyzing the learner's response.

Additionally, problems for diagnostic evaluation may be pre-classified by difficulty level and learning category by the learning data database (DB) 270, which may be received from an external database. In addition, the problem is not limited to this, and the diagnostic evaluation unit 220 uses learning data learned from each learner's diagnostic evaluation session to provide diagnostic evaluation problems to learners with learner information similar to each learner. You can classify them directly. At this time, the similarity of the learner information can be determined using cosine similarity or Euclidean similarity, and for learners with similar learner information, the diagnostic evaluation problem stored in the learning data database (DB) 270 is the same. can be provided. Through this, the diagnostic evaluation unit 220 can cluster learners with similar learner information and store and manage learners' relative diagnostic evaluation information.

The diagnostic evaluation unit 220 may collect the learner's diagnostic evaluation data and store the learner's response pattern. For example, the learner's response pattern may include correct answers, incorrect answers, no input, response time, etc., and if the learner writes a subjective answer, the content can be analyzed. In the case of analysis of subjective answers, latent semantic analysis (LSA), Latent Dirichlet Allocation (LDA), BERT (Bidirectional Encoder Representations from Transformers), Long Short-Term Memory (LSTM), Siamese Networks, etc. Using a semantic analysis algorithm, you can compare the learner's answer with the correct answer to determine whether it is correct.

In addition, the diagnostic evaluation unit 220 can identify the learner's strengths and weaknesses learning areas by calculating the percentage of correct answers by subject and level of difficulty. For example, if the percentage of correct answers in the humanities learning area is 0.8, while the percentage of correct answers in the astrophysics learning area is 0.4, the humanities learning area can be identified as a strength and the astrophysics learning area can be identified as a weakness.

In addition, the diagnostic evaluation unit 220 can analyze the time the learner spent solving problems to determine which topic or level of difficulty the learner spent more time on, and use this to identify the learner's strengths and weaknesses. there is. For example, if the learner's response time in the humanities learning area is 12 seconds on average, while the average response time in the astrophysics learning area is 60 seconds, the humanities learning area can be identified as a strength and the astrophysics learning area can be identified as a weakness.

The diagnostic evaluation unit 220 may use the learner's diagnostic evaluation results to create a user interface displayed on the learner's user terminal. For example, the diagnostic evaluation unit 220 provides the results of the learner's diagnostic evaluation, such as correct and incorrect questions, percentage of correct answers, materials visualizing the learner's strengths and weaknesses in graphs or charts, and recommended learning materials or topics for the learner's strengths or weaknesses. You can create a user interface that represents . Additionally, the diagnostic evaluation unit 220 may generate a diagnostic evaluation report based on the generated user interface. The diagnostic evaluation report is the result of the learner's diagnostic evaluation at the time of performing the diagnostic evaluation, and can serve as standard data for comparison with the learner's learning performance in the future.

For example, diagnostic evaluation results may be provided as shown in the table below.

subject Total number of questions Number of correct answers Correct answer rate (%) X 5 3 60 Y 5 2 40 Z 5 5 100

Additionally, the diagnostic evaluation unit 220 may analyze the diagnostic evaluation as follows.

difficulty Total number of questions Number of correct answers Correct answer rate (%) basic 10 9 90 middle 5 2 100 advanced 5 One 20

Through this, the diagnostic evaluation unit 220 can analyze the diagnostic evaluation of a specific learner and determine that the learner has a strength in learning topic Z and a weakness in topic Y, and has a low percentage of correct answers in problems with high difficulty. , it is possible to create a solution that can provide a learning method that solves high-difficulty problems.

Meanwhile, the diagnostic evaluation unit 220 may determine the learner's learning goals and recommended learning topics based on the diagnostic evaluation results. Here, the learning goal may refer to the step of presenting the direction of learning based on the learner's personal needs and the results of the diagnostic evaluation.

At this time, the diagnostic evaluation unit 220 may analyze the learner profile. For example, the diagnostic evaluation unit 220 may analyze the learner information stored in the learner data database (DB) 280 and the learner profile created based on the learner information. For example, the diagnostic evaluation unit 220 may confirm the learner's past learning profile by checking the learner's past learning performance or quiz or test results to determine the learner's basic knowledge level. Additionally, the diagnostic evaluation unit 220 can determine the learner's learning style (visual learner, auditory learner, etc.) and learning preferences or interests through the learner profile.

Next, the diagnostic evaluation unit 220 identifies the area in which the learner gave a large number of incorrect answers or the learning area identified as a weakness in the learner's diagnostic evaluation, and selects the area in which the learner gave a large number of correct answers or the learning area identified as a strength. can be identified.

Next, the diagnostic evaluation unit 220 may present learning goals for learning topics or areas that require improvement based on the learner's weaknesses. For example, the diagnostic evaluation unit 220 can set the learner's learning goal by applying the SMART principle. For example, you can set a specific goal such as “By next week (within 7 days), solve 10 problems on topic

Meanwhile, the diagnostic evaluation unit 220 can calculate a learning goal score for the learning goal set by the learner as shown in the equation below.

Here, the learning goal score is calculated for each specific learner by the diagnostic evaluation unit 220, and specificity is less than 1 if the learner's weaknesses, strengths, learning fields, and recommended learning field items are calculated specifically, and less than 1 otherwise. It can be set to a value greater than 0, and the measurability is expressed as 1 in the case of a quantifiable learning method such as '10 problems' or '1 chapter', and the achievability is a value of 0 to 1, and is expressed as 1 by the instructor. , can be determined in advance by the judgment of parents or learners themselves.

In addition, relevance is determined by how relevant the learner's learning content is to the learning goal, and this can be determined in advance by the judgment of the instructor, parents, or the learner himself.

The time limit is determined by the diagnostic evaluation unit 220 and is automatically determined according to the difficulty level of each learning material and problem stored in the learning data database (DB) 270, for example, a predetermined maximum learning time for each learning topic. It can be calculated as the ratio of a given learning period to the period. For example, if the maximum predetermined learning period for each learning topic is 7 days, and the learning period given by the diagnostic evaluation unit 220 is 5 days, the time limit can be calculated as 0.71.

At this time, the diagnostic evaluation unit 220 may determine a recommended learning topic provided to the learner. As an example, the diagnostic evaluation unit 220 may determine a weakness-based learning topic by recommending learning topics or materials according to the learner's weakness area based on the diagnostic evaluation results, and may determine a learning topic based on the learner's learning preferences or interests. Materials can be recommended, and at this time, the diagnostic evaluation unit 220 can induce the learner's interest and maximize the learning effect by providing various types of learning materials such as videos, texts, quizzes, and exercises.

Additionally, the diagnostic evaluation unit 220 may provide a step-by-step learning plan based on the diagnostic evaluation. For example, even when learning about the same topic, the diagnostic evaluation unit 220 can present a step-by-step learning plan. For example, the diagnostic evaluation unit 220 creates a step-by-step learning plan such as “Week 1: Learning the basic concepts of Topic can be presented.

By providing a step-by-step learning plan, learners can perform performance-based learning according to their own learning performance and clarify the direction of learning based on diagnostic evaluation. Through this, the AI chatbot tutor provides learners with a more personalized learning experience. can be provided.

The learning session unit 230 may proceed with learning according to the learning goal and learning plan determined by the diagnostic evaluation unit 220.

For example, the learning session unit 230 may provide learning materials for each learning step according to a step-by-step learning plan determined by the diagnostic evaluation unit 220. At this time, the learning session unit 230 may extract learning materials and problems suitable for the learner's learning from the learning data database (DB) 270 and provide them to the learner.

Meanwhile, the learning session unit 230 may select learning materials based on learning data in the learning data database (DB) 270. For example, the learning session unit 230 may collect learning materials that fit the learner's learning goals and recommended learning topics from the learning data database (DB) 270. Next, the learning session unit 230 can provide learning materials needed by the learner based on the learner's weaknesses and strengths. At this time, if the learner requests separate materials, the learning session unit 230 may provide the learning materials needed by the learner. Learning materials can be given priority to learners by prioritizing the provision of learning materials.

In addition, the learning session unit 230 receives learning materials from the learning data database (DB) 270, sorts the learning materials according to the learner's learning goal, learning plan, and learning difficulty level, and provides the learning materials to the learner to provide a learning session. You can proceed.

For example, the learning session unit 230 may modularize the learning materials received from the learning data database (DB) 270 step by step. As an example, when learning materials are received from the learning data database (DB) 270 for learning field “A”, the learning session unit 230 is classified according to the difficulty level of the content and learning method (e.g., visual, auditory, You can modularize learning materials by clustering them by activity, etc.) and learning content (e.g., problem solving, learning by progress, intermediate diagnostic evaluation, etc.).

Additionally, the learning session unit 230 may provide the learner with a chatbot module to ask questions about learning materials or problems and provide answers to them while the learning progresses. For example, the learning session unit 230 can ensure learning efficiency by immediately responding and providing learning support when a learner raises a question or error about learning materials.

At this time, the learning session unit 230 uses natural language processing (NLP), natural language understanding (NLU), natural language generation (NLG), machine learning, reinforcement learning, and emotion analysis algorithms to provide chatbot learning support to the learner. Real-time interaction can be provided, and chatbots for learner support can be implemented through, for example, Dialogflow, Microsoft Bot Framework, and Rasa.

Next, the learning session unit 230 can track the learner's learning progress. For example, the learning session unit 230 tracks the learner's learning progress in real time and can check the learning progress. Additionally, if the learner does not continue to progress in specific learning content, the learning session unit 230 may determine that part as a progress barrier and provide additional support for progressing the learning. For example, the learning session unit 230 may receive additional material for the corresponding learning content from the learning data database (DB) 270 and provide it to the learner, and may provide low-difficulty problems to aid understanding. there is.

It is not limited to this, and the learning session unit 230 may collect feedback from learners to expand and improve learning materials. For example, the learner's answers or the conversation log created while the learner talks with the chatbot can be analyzed and applied to the learning of learners with the same or similar learner information. Through this, it is possible to flexibly overcome learners' progress barriers and improve the quality and diversity of learning materials.

Additionally, the learning session unit 230 can provide appropriate feedback and hints in real time according to the learner's questions or answers that occur while learning. For example, if the learner derives an incorrect answer through a solution method that is different from the correct solution method, the learning session unit 230 may not immediately expose the correct answer, but may guide the learner to reach the correct answer by suggesting a new solution method. .

At this time, the learning session unit 230 can periodically check the learner's progress. Here, the learning session unit 230 can determine whether the learner has completely understood all content in the current learning stage.

For example, the learning session unit 230 may check the learner's understanding through various forms of evaluation after the learner learns a specific topic. For example, if the learner's learning direction is complete learning, it can be determined whether complete learning has been performed through accuracy, application ability, depth, and self-evaluation of the learner's learning content.

For example, in accuracy-based evaluation, the learning session unit 230 evaluates the learner's responses in real time and provides feedback while analyzing the results of problem solving based on the learner's learning content to determine whether the learner accurately understands the learning content. You can. Although this accuracy-based evaluation is simple in the old way, it has the advantage of being able to quantitatively and directly determine the learner's achievement.

In addition, the learning session unit 230 can measure the learner's level of understanding by analyzing the learner's learning progress speed and number of repetitions and measuring the learning time and number of repetitions required for the learner to understand a specific topic or content. This can be used as an indicator to identify the learner's strengths or weaknesses during the learning process.

In addition, the learning session unit 230 can determine the learner's level of understanding by presenting questions about learning content based on the AI-based tutor provided to the learner and analyzing the responses. This can be implemented by using an algorithm that can semantically analyze the learner's answers, such as the natural language processing (NLP) mentioned above.

Additionally, the learning session unit 230 may evaluate the learner's level of understanding through the learner's self-evaluation. As this is a learner's self-evaluation, it can be an indicator that can identify the difference between the learner's self-perceived ability and actual performance.

Additionally, the learning session unit 230 can analyze the learner's level of understanding and learning pattern by comprehensively analyzing the learner's learning log, response pattern, response time, and AI tutor usage time.

Meanwhile, the learning session unit 230 may provide customized learning options to the learner according to the learner's learning pattern. For example, the learning session unit 230 may provide customized learning materials to the learner based on the learner's learning progress status and level of understanding.

For example, if a learner learns with text-based materials and experiences frequent progress barriers, video or audio-based learning materials can be provided. In addition, if a learner has difficulty progressing in a topic in which he or she is weak or experiences a progress barrier, additional explanations or materials about the learning content can be provided based on an AI chatbot, and the learner can be evaluated based on the evaluation questions provided compared to the learner's learning performance. If a number of incorrect answers are given, the difficulty level can be adjusted to provide a lower evaluation problem.

Meanwhile, if it is determined that the learner fully understands the provided learning materials or is deeply interested in a specific topic, the learning session unit 230 may suggest additional activities, such as providing other topics related to the topic or recommending related books. You can.

Accordingly, the learning session unit 230 may provide learning options to the learner and provide learning options according to the learner's selection.

The learning session unit 230 can determine whether the learner understands the content currently being studied and track the learner's progress. For example, the learning session unit 230 presents evaluation problems related to the content of a specific session and chapter for the learning content that the learner receives from the learning data database (DB) 270, and allows the learner to solve the problem. If you solve it, you can judge that you have completed your understanding of the session and chapter. At this time, evaluation questions to determine the learner's understanding may consist of various types of questions, such as subjective, multiple choice, short answer, and multiple choice questions.

At this time, if it is determined that the learner fully understands the topic, the learning session unit 230 may generate a notification recommending that the learner move on to the next learning topic. On the other hand, if the learner does not completely understand the topic, You can create alerts that provide additional material or explanation, or encourage you to re-study the topic.

At this time, the learning session unit 230 may store or update the learner's learning progress confirmation details and evaluation problem solving details in the learner profile stored in the learner data database (DB) 280. Through this, the learning session unit 230 enables the learner to continue learning, tracks the learner's learning progress and understanding, and provides additional support, if necessary, without additional procedures.

The learning performance analysis unit 240 can analyze the learning performance according to the progress of the learner's learning session. For example, the learning performance analysis unit 240 may analyze the learner's learning performance based on the result of the learning session unit 230's determination of the learner's level of understanding. For example, the learning performance analysis unit 240 is based on accuracy-based evaluation, analysis of learning progress speed and number of repetitions, evaluation of questions about learning content based on the AI-based tutor provided to the learner, learner self-evaluation, and learning pattern evaluation. Learner performance can be analyzed.

At this time, the learning performance analysis unit 240 may determine the learner's level of understanding differently by assigning different weights depending on the type of evaluation question (e.g., multiple choice, subjective, short answer, etc.) for determining the learner's level of understanding. For example, the learner's level of understanding can be determined by assigning a weight of 0.3 for multiple choice, 0.5 for short answer, and 0.9 for subjective. For example, if a learner accurately answers a complex subjective question or a short answer question, it can be determined that the learner has a deep understanding of the subject.

Additionally, the learning performance analysis unit 240 may additionally determine the learner's level of understanding based on the learner's evaluation problem solving time. For example, if you complete an evaluation problem faster than the predetermined time, it may be judged that you did not spend enough time thinking about the problem or are unsure about the topic, and it may not be reflected in the evaluation.

Additionally, the learning performance analysis unit 240 can evaluate the learner by comparing the learner's past evaluation history with the learner's profile in the learner data database (DB) 280. For example, even for the same learning topic, the learner's previous evaluation problem results can be compared with the current evaluation problem solution. For example, if the learner's grades are continuously improving, it can be determined that the learner's understanding of the topic is increasing. You can. In addition, the learning performance analysis unit 240 is integrated into the AI chatbot tutor's algorithm to improve the extraction accuracy of evaluation questions and provide appropriate feedback to learners.

The learning performance analysis unit 240 may store the learner's learning path including learner information, learner's learning topic, learning progress, evaluation problems, feedback, and learning patterns. This allows learners to continue learning in the same learning environment in the next learning session and provides a personalized learning experience.

For example, the learning performance analysis unit 240 may determine learner information, the learner's learning topic, learning progress, evaluation problems, feedback, and learning patterns as a learning path, and model the learner's learning information.

For example, the learning performance analysis unit 240 may collect the learner's learning history that occurs during the progress of the learning session by the learning session unit 230 to monitor it in real time. In this way, the learning path containing the learner's learning pattern is saved for each learner, and when the learner resumes learning, learning is provided according to the pre-stored learning path for each learner, thereby providing the learner with a learning experience in the same environment. And, learners can ensure continuity of learning by resuming learning from the last state of the previous learning session.

The learning performance analysis unit 240 may analyze the learner's learning tendency based on the learner's learning path. The learning performance analysis unit 240 can analyze the learning path data to identify the learner's strengths, weaknesses, learning style, etc., and based on this, the "AI chatbot tutor" can effectively store and manage the learner's learning path. It can provide learners with a personalized learning experience.

Meanwhile, the learning path recommendation unit 250 collects learning path data generated during the learner's learning progress, creates a recommended learning path for the learner, and provides it to the learner.

For example, learning path data may specifically include: number of interactions per learning session, average duration per interaction, assessment score, percent correct, percent incorrect, response time, performance trends, total time spent in learning sessions, and average duration of learning sessions. It can include time, number of accessed learning materials, time spent for each learning material, number of feedbacks, and feedback time.

At this time, the learning path recommendation unit 250 may classify the learning path data into at least one category according to its properties.

For example, the learning path recommendation unit 250 may determine the number of interactions per learning session, the average duration per interaction, the total time spent in a learning session, the average time spent in a learning session, the number of accessed learning materials, and the time spent for each learning material. It can be classified into learning participation indicators. Learning engagement indicators can reflect the learner's level of participation and interaction with learning materials in learning using an AI tutor.

Additionally, the learning path recommendation unit 250 may classify evaluation scores, percentage of correct answers, percentage of incorrect answers, response time, and grade trends as learning ability indicators. Learning ability indicators may include indicators that may indicate a learner's academic achievement, understanding of material, or learning strengths or weaknesses.

Additionally, the learning path recommendation unit 250 may classify the number of feedback and feedback time into learning feedback index items. Learning feedback indicator items represent the learner's responsiveness and interaction with the system in terms of providing input or response to the learning experience, and may include the learner's feedback on the learning content.

The learning path recommendation unit 250 may calculate a learning path recommendation score by combining the learning path data. For example, the learning path recommendation unit 250 calculates the product of the number of interactions per learning session and the average duration per interaction, the product of the number of accessed learning materials and the time required for each learning material, and the total time spent in the learning session. The learning participation score can be calculated by dividing it by the sum of the average time spent in each learning session and using the equation below.

Here, the number of interactions per learning session refers to the number of interactions the learner performs to complete the learning session, the average duration per interaction refers to the average duration of individual interactions, and the number of accessed learning materials refers to the number of interactions the learner performs to complete the learning session. It refers to the total number of learning materials accessed during a learning session, and the time required for each learning material may refer to the time the learner spends on the accessed learning materials. In addition, the total learning session time refers to the total time taken to complete the learning session in which the learner is learning, and the average session time refers to the average time taken by other learners to complete the learning session in which the learner is learning. Additionally, interaction can refer to any activity that involves interacting with the system using the AI chatbot tutor, such as clicking on learning materials, asking questions, responding to answers, exploring materials, participating in discussions, taking notes, performing searches, and viewing audio-visual materials. This can include playback and navigation, bookmarking, data downloading, etc.

For example, the number of interactions per learning session is 25, the predetermined maximum number of interactions per learning session is 100, the average duration per interaction is 4 minutes, the predetermined maximum number of interactions per learning session is 60 minutes, and the approach learning The number of materials is 10, the predetermined maximum number of accessible learning materials is 15, the time required for each learning material is 6 minutes, the predetermined maximum time required for each learning material is 60 minutes, and the total learning session time is 90 minutes. When the maximum total learning session time is 120 minutes, the average session time is 60 minutes, and the predetermined average session time is 90 minutes, the learning participation score is calculated as follows.

In other words, analyzing the calculated learning participation scores can suggest that although learners interact frequently with a variety of materials, the time spent per interaction and material is relatively short, which means that learners explore a wide range of learning materials but do not participate deeply in learning. It can be seen that the learning path recommendation unit 250 recommends the learner to spend more time on each material for a deeper understanding and provides activities that require longer participation to improve comprehension and memory. can be recommended.

The learning path recommendation unit 250 can calculate a learning ability score by combining the percentage of correct answers, percentage of incorrect answers, response time, and grade trend classified as learning ability indicators, as shown in the equation below.

Here, the evaluation score may refer to the percentage of correct answers in the evaluation questions to measure the learner's performance in the learning session, and the incorrect answer rate may refer to the percentage of incorrect answers in the evaluation questions to measure the learner's performance in the learning session. Response time can refer to the average time a learner takes to respond to an assessment question, and performance trend is a measure of the increase or decrease in the learner's percentage of correct answers over time and can be measured by calculating the slope through linear regression. there is.

For example, if the learner's correct answer rate is 0.8, the incorrect answer rate is 0.2, the response time is 45 seconds, the predetermined maximum response time is 60 seconds, and the grade trend is 0.05, the learning ability score is calculated as shown in the equation below.

The calculated learning ability score can be used as an indicator to analyze the learner's strengths and weaknesses, and a value of 0.89 indicates that the learning ability is high and the learning trend is upward, indicating that the learner's learning performance is continuously increasing. Additionally, in this case, a learning path can be recommended to improve response time by providing high-difficulty evaluation problems.

Additionally, the learning path recommendation unit 250 may combine the number of feedbacks and feedback times classified as learning feedback indicators to calculate a learning feedback score as shown in the equation below.

Here, the number of feedbacks may refer to the number of feedbacks submitted by the learner while using the AI chatbot tutor, where feedback refers to responses to assignments or evaluation problems, evaluation of learning content, interactive responses using the AI chatbot tutor, discussions, and surveys. Learning-related feedback generated while conducting a learning session using an AI chatbot tutor, such as research, problem solving, and learning logs, may be included. Additionally, feedback time may refer to the average time spent on each feedback activity. For example, when the number of feedbacks for a specific learning session is 15, the predetermined maximum number of feedbacks is 30, the feedback time is 6 seconds, and the predetermined maximum feedback time is 60 seconds, the learning feedback score is calculated as follows.

At this time, the calculated number of feedback may indicate that the learner is providing high feedback, but does not spend much time on feedback, and the learning path recommendation unit 250 may provide a learning path that can provide more detailed learning feedback to the learner. can be recommended.

Meanwhile, the learning path recommendation unit 250 may calculate a learning path recommendation score by combining the “learning participation score,” the “learning ability score,” and the “learning feedback score.”

For example, when the learning participation score is 0.6, the learning ability score is 0.5, and the learning feedback score is 0.4, and the predetermined weights w1 to w3 are 0.5, 0.3, and 0.2, respectively, the learning path recommendation score is calculated as follows.

In other words, the learning participation score indicates the learner's degree of participation, the learning ability score indicates understanding of learning materials and effective learning methods, and the learning feedback score indicates whether the learner actively provides feedback. Therefore, a high learning path recommendation score can indicate the learning tendency according to the learner's learning profile, and the higher the learning path recommendation score, the more likely it is to recommend a learning path that provides high-level learning materials in the learner's strengths.

Additionally, the learning path recommendation unit 250 may generate a learner path group by clustering learners whose calculated learning path recommendation scores fall within a predetermined error range.

That is, the learning path recommendation unit 250 determines that learners with similar learning path recommendation scores are effective in the same learning path, and can determine the learner's learning path to be similar to that of other learners. Through this, learners can experience effective learning according to a learning path appropriate for their learning abilities.

The learner profile management unit 260 analyzes the learner's learning session performance history, creates a profile to provide to the learner, and creates or updates the learner profile in the learner data database (DB) 280. For example, the learner profile management unit 260 can update the learner profile if a learner profile has been created, and can create and manage a new learner profile if the learner profile has not been created.

For example, the learner's learning performance after the learner's learning session can be analyzed as follows.

division item detail note One Total learning time 2 hours 15 minutes 2 Learning sessions completed Chapter 1, Section 1 Learning Content: 100%Chapter 1, Section 2 Learning Content: 50% 3 Quiz and Practice Problem Results Chapter 1, Section 1: 95% AccuracyChapter 1, Section 2: 70% Accuracy 4 strength Good at grasping and applying basic concepts Excellent problem solving ability 5 Weakness Errors occur in some topics at intermediate level Requires an approach to complex problems 6 Recommended study materials and methods Intermediate Topic Tutorials Guide to Strategies for Solving Complex Problems

Such a learner's learning history can be organized in a learner profile, and using a deep learning-based learning model, a learner profile can be created by clustering the learner's learning content into the same category.

Additionally, the learner profile management unit 260 can collect the learner's learning data step by step. For example, collect learner responses, questions, feedback and scores on assessment questions in real time during a learning session, learner's preferred learning style (e.g. visual learning, auditory learning, etc.) and user interface (UI) preferences. You can conduct surveys or questions to find out.

Next, the learner profile management unit 260 analyzes the collected data to analyze the learner's current learning level, weaknesses and strengths, and learning style, and identifies repetitive errors or difficulties with specific topics to identify areas of the learner's weaknesses. there is.

Additionally, the learner profile management unit 260 may periodically update the learner profile based on the analyzed data (e.g., when feedback occurs/when interaction occurs/every 3 minutes, etc.). Through this, the learner profile management unit 260 can optimize and provide learning paths, recommended learning materials, and learning methods.

Figure 3 is an exemplary algorithm configuration diagram for providing accurate answers in a learning session through RAG (Retrieval-Augmented Generation).

Referring to Figures 1 and 3, the server 100 can provide an AI-based chatbot tutor that can answer learners' questions in real time. For example, Retriever-Augmented Generation (RAG) technology can build chatbot systems specifically designed to answer learner questions with high accuracy and relevance, and can leverage the two main components of natural language processing: a retrieval model and a generative model. It can be derived by combining the advantages. For example, if a learner asks a content-related question, server 100 may search a large corpus of text using the LangChain framework to retrieve the most relevant phrase or clause that may contain an answer to the learner's question. You can. At this time, the server 100 may search for documents similar in context to the question from a database storing data related to answers, and for example, a model such as DPR (Dense Passage Retrieval) may be used.

Additionally, the server 100 may generate an answer appropriate for the situation from the searched document. At this time, the server 100 can synthesize the received information using the context-based language provided by the language model (LLM) and generate an answer appropriate to the learner's question. For example, the server 100 may use a transformer-based model such as BART or T5.

Figure 4 is a flowchart showing an example operation of an AI chatbot tutor providing server for personalized learning support based on complete learning according to an embodiment of the present invention.

Referring to Figures 1 and 4, the server 100 can collect learner information (S410). Additionally, when the server 100 collects learner information, it can classify the collected learner information into at least one category according to the properties of the information. Next, the server 100 may calculate a learner information score by combining learner information classified into at least one category.

Next, the server 100 may perform a diagnostic evaluation to evaluate the learner's learning level in order to create a learning session appropriate for the learner's learning level (S420). The server 100 can provide a report by analyzing the learner's learning strengths and weaknesses as a result of the performed diagnostic evaluation, and can be used as basic data to determine the learner's learning performance.

For example, the server 100 may perform a diagnostic evaluation to determine the learner's learning level by inferring the learner's learning level based on the received learner information and use diagnostic evaluation questions stored in the learning data database (DB). For example, the server 100 uses an artificial intelligence-based learning problem selection model to select diagnostic evaluation problems appropriate for the learner's level among diagnostic evaluation problems stored in an external learning data database (DB), and selects diagnostic evaluation problems appropriate for the learner's level through the user terminal. You can diagnose the learner's learning level by providing it to

Next, the server 100 may determine the learner's learning goal and recommended learning topic based on the diagnostic evaluation results (S430). Here, the learning goal may refer to the step of presenting the direction of learning based on the learner's personal needs and the results of the diagnostic evaluation.

The server 100 may proceed with learning according to the learning goal and learning plan determined by the diagnostic evaluation unit 220 (S440).

For example, the server 100 may provide learning materials for each learning step according to a step-by-step learning plan determined by the diagnostic evaluation results. At this time, the server 100 may extract learning materials and problems suitable for the learner's learning from the learning data database (DB) and provide them to the learner.

Meanwhile, the server 100 may select learning materials based on learning data in the learning data database (DB). For example, the server 100 may collect learning materials that fit the learner's learning goals and recommended learning topics from the learning data database (DB). Next, the server 100 can provide learning materials needed by the learner based on the learner's weaknesses and strengths. At this time, if the learner requests separate materials, the server 100 may provide the learning materials. Learning materials can be provided to learners first.

In addition, the server 100 may receive learning materials from the learning data database (DB), then arrange the learning materials according to the learner's learning goal, learning plan, and difficulty of learning, and provide the learning session to the learner (S450 ).

Next, the server 100 can track the learner's learning progress (S460). For example, the server 100 tracks the learner's learning progress in real time and can check the learning progress. Additionally, if the learner does not continue to progress in specific learning content, the server 100 may determine that part as a progress barrier and provide additional support for progressing the learning. For example, the server 100 may receive additional material for the corresponding learning content from a learning data database (DB) and provide it to the learner, and may provide low-difficulty problems to aid understanding.

At this time, if it is determined that the learner fully understands the topic, the server 100 may generate a notification recommending that the learner move on to the next learning topic and end the learning session (S470). On the other hand, if the learner does not fully understand the topic, you can provide additional material or explanation, or create a reminder to encourage the learner to study the topic again.

At this time, the server 100 may store or update the learner's learning progress confirmation history and evaluation problem solving history in the learner profile stored in the learner data database (DB). Through this, the server 100 enables the learner to continue learning, tracks the learner's learning progress and level of understanding, and provides additional support, if necessary, without additional procedures.

The server 100 can analyze the learning performance according to the progress of the learner's learning session (S480). For example, the server 100 may analyze the learner's learning performance based on the result of previously determining the learner's level of understanding. For example, the server 100 evaluates the learner's performance based on accuracy-based evaluation, analysis of learning progress speed and number of repetitions, evaluation of questions about learning content based on the AI-based tutor provided to the learner, learner self-evaluation, and learning pattern evaluation. can be analyzed.

Next, the server 100 can analyze the learner's learning session performance history, create a profile to provide to the learner, and create or update the learner profile in the learner data database (DB). For example, if a learner profile has been created, the server 100 can update the learner profile, and if a learner profile has not been created, the server 100 can create and manage a new learner profile. Additionally, the server 100 may periodically update the learner profile based on the analyzed data (e.g., when feedback occurs/when interaction occurs/every 3 minutes, etc.). Through this, the server 100 can optimize and provide learning paths, recommended learning materials, and learning methods.

Figure 5 is a hardware configuration diagram of an AI chatbot tutor providing server for complete learning-based personalized learning support according to 1.

Referring to FIG. 5, the AI chatbot tutor providing server 500 for supporting personalized learning based on complete learning includes at least one processor 510 and instructs the at least one processor 510 to perform at least one step. It may include a memory 520 that stores instructions.

Here, the at least one processor 510 may mean a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor on which methods according to embodiments of the present invention are performed. You can. Each of the memory 520 and the storage device 560 may be comprised of at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory 520 may be comprised of at least one of read only memory (ROM) and random access memory (RAM).

In addition, the AI chatbot tutor providing server 500 for supporting personalized learning based on complete learning may include a transceiver 530 that communicates through a wireless network. In addition, the AI chatbot tutor providing server 500 for supporting personalized learning based on complete learning may further include an input interface device 540, an output interface device 550, a storage device 560, etc. Each component included in the complete learning-based AI chatbot tutor providing server 500 for personalized learning support is connected by a bus 570 and can communicate with each other.

Methods according to the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. Computer-readable media may include program instructions, data files, data structures, etc., singly or in combination. Program instructions recorded on a computer-readable medium may be specially designed and constructed for the present invention or may be known and usable by those skilled in the computer software art.

Examples of computer-readable media may include hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. Examples of program instructions may include machine language code such as that created by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The above-described hardware device may be configured to operate with at least one software module to perform the operations of the present invention, and vice versa.

Additionally, the above-described method or device may be implemented by combining all or part of its components or functions, or may be implemented separately.

Although the present invention has been described above with reference to preferred embodiments, those skilled in the art may make various modifications and changes to the present invention without departing from the spirit and field of the present invention as set forth in the claims below. You will understand that you can do it.

Claims (5)

As an AI chatbot tutor providing server for personalized learning support based on complete learning,
at least one processor; and
a memory storing instructions that instruct the at least one processor to perform at least one step;
The at least one step is,
Collecting learner information;
performing a diagnostic assessment to evaluate the learner's current learning level to create a learning session appropriate for the learner's learning level;
determining the learner's learning goal and recommended learning topic based on the diagnostic evaluation results;
Proceeding with learning according to the determined learning goals and learning plan;
Tracking the learner's learning progress in real time, and if the learner does not continue to progress in specific learning content, determining that part as a progress barrier and providing additional support for the learning progress;
If it is determined that the learner fully understands the topic, generating a notification encouraging the learner to move on to the next learning topic and terminating the learning session; and
Comprising: analyzing the learner's learning performance based on the result of determining the learner's level of understanding, and updating the learner's profile;
Collecting learning path data generated from the learner's learning progress;
Classify the collected learning path data into at least one category according to attributes, including the number of interactions per learning session, average duration per interaction, total time spent on learning sessions, average time spent on learning sessions, number of accessed learning materials, etc. Classifying the time required for each learning material as a learning participation indicator, classifying the evaluation score, percentage of correct answers, percentage of incorrect answers, response time, and grade trend as learning ability indicators, and classifying the number of feedback and feedback time as learning feedback indicator items;
The product of the number of interactions per learning session and the average duration per interaction, the product of the number of accessed learning materials and the time required for each learning material, the sum of the total time spent in a learning session and the average time spent in a learning session. Calculating a learning participation score by dividing by;
calculating a learning ability score by dividing the sum of the correct answer rate and the grade trend by the sum of the incorrect answer rate and response time;
Calculating a learning feedback score from the average of the number of feedbacks and feedback times;
calculating a learning path recommendation score based on a weighted sum of the calculated learning participation score, the learning ability score, and the learning feedback score; and
A step of creating a learner path group by clustering learners whose calculated learning path recommendation score falls within a predetermined error range,
The collected learner information is classified into at least one category according to attributes, with age, economic background, cultural background, and language ability classified into social background categories, grade, subject of interest, academic performance, learning goals, major, and preferred subjects. , target occupation, and research interests are classified into academic goal categories, and learning style, group activity preference, online learning affinity, time management efficiency, learning speed, learning efficiency, study habits, main study time, and main study method are categorized into study habit categories. It further includes a step of classifying,
A cultural economic index is calculated from the average of the economic background and the cultural background classified into the social background category, and a language age integration index is calculated from the average of the language ability and age conversion score. The age conversion score is calculated when the learner It is determined from the difference between the appropriate age of the learning topic to be learned and the learner's age. If the appropriate age of the learning topic the learner wants to learn is greater than or equal to the learner's age, the age conversion score is ( ), and if the appropriate age of the learning topic that the learner wants to learn is less than the learner's age, the age conversion score is ( ) and calculating a social background score from the weighted sum of the cultural economic index and the language-age integration index calculated;
A learning goal score is calculated from the grade conversion score, the academic performance, the learning goal, and the average of the research interests, and the grade conversion score is determined by the difference between the appropriate grade of the learning topic the learner wants to learn and the learner's grade, The grade conversion score is (if the appropriate grade of the learning topic that the learner wants to learn is greater than or equal to the learner's grade) ), and if the appropriate grade of the learning topic that the learner wants to study is less than the learner's grade, the grade conversion score is ( ) step of determining;
Calculating a learning habit score from a weighted sum of the online learning affinity, the time management efficiency, the learning speed, the learning efficiency, and the main learning time;
calculating a learner information score from a weighted sum of the calculated social background score, the learning goal score, and the learning habit score;
Creating a learner group by clustering learners whose calculated learner information score is within a predetermined error range; and
It further includes providing the same diagnostic assessment problem to learners clustered into the same learner group,
The grade trend is a measure of the increase or decrease in the learner's percentage of correct answers over time, and is measured by calculating the slope through linear regression,
From the product of the number of interactions per learning session and the average duration per interaction, the interaction is an interaction activity with the system using the AI chatbot tutor, such as clicking on learning materials, asking questions, responding to answers, exploring materials, participating in discussions, and taking notes. Includes creating, performing searches, playing and browsing audiovisual materials, bookmarking, and downloading materials;
The online learning affinity represents the learner's affinity including adaptability to online learning, the time management efficiency represents compliance with a predetermined learning period for a predetermined amount of learning, and the learning speed refers to the learner's ability to learn online. Characterized in that it represents the ratio of the learner completing learning quickly compared to the average learning time of a group of learners learning the same learning topic, and the main learning time is a quantification of the learner's most learning time,
AI chatbot tutor providing server for personalized learning support based on complete learning.
In claim 1,
Including, generating a user interface (UI) that provides a report analyzing the learner's learning performance, learning strengths and weaknesses, according to the diagnostic evaluation and progress of the learning session.
AI chatbot tutor providing server for personalized learning support based on complete learning.
In claim 1,
The step of performing the diagnostic evaluation is,
Diagnostic evaluation to determine the learner's learning level infers the learner's learning level based on the received learner information and extracts problems from the diagnostic evaluation problems stored in the learning data database (DB), using an artificial intelligence-based learning problem selection model. Characterized in that, among the diagnostic evaluation problems stored in the learning data database (DB), diagnostic evaluation problems appropriate for the learner's level are selected and provided to the learner through the user terminal,
The diagnostic evaluation is characterized in that the diagnostic evaluation is performed using an adaptive quiz method or a classification-based quiz method.
AI chatbot tutor providing server for personalized learning support based on complete learning. delete delete

Images