Artificial intelligence-based learning system for detecting slow learner behavior and providing feedback

Abstract

The present invention relates to an artificial intelligence-based multiple intelligence analysis and slow learner behavior detection and response system. The above system may be characterized by comprising: a user terminal that presents items to a user to perform a multiple intelligence test and inputs a slow learner question; a main server that receives the user's multiple intelligence test results from the user terminal, generates the user's multiple intelligence test profile, and generates an answer to the slow learner question; wherein the main server includes: a generation module that analyzes the user's multiple intelligence test results using artificial intelligence to generate the user's multiple intelligence test profile; a peer comparison module that compares the user's multiple intelligence test profile with the multiple intelligence test profile of a peer group using artificial intelligence; a judgment module that extracts and splits the user's unit behavior from the user's slow learner question to determine whether it is a slow learner behavior; an augmented search module that generates a prompt using artificial intelligence-based augmented search from the user's unit behavior; and a slow learner answer module that receives the prompt and generates an answer to the slow learner question.

Description

Technology Field

[0001] The present invention relates to an artificial intelligence-based multiple intelligence analysis and slow learner behavior detection and response system. Background Technology

[0002] Since the human brain operates differently through the cooperation of various types of intelligence, it is necessary to accurately identify how the user's (tester's) strengths and weaknesses are combined to recommend a truly personalized career path.

[0003] On the other hand, since the existing MBTI test is a personality type test, it cannot provide information on exploring a user's aptitudes or talents, and IQ tests have limitations in that they only measure a very small portion of the diverse human intelligences, such as verbal and logical-mathematical intelligence.

[0004] The Multiple Intelligence Test is a testing method based on the multiple intelligence theory of H. Gardner, a professor at Harvard University's Graduate School of Education, and is a tool that can discover a user's innate talents and aptitudes, as well as their strengths and weaknesses in eight intelligence domains.

[0005] Meanwhile, there has recently been an increasing need for users' guardians (e.g., parents) to identify early whether a user is a slow learner and to receive feedback on learning methods to help them improve.

[0006] Whether a user is a slow learner can be determined by five behaviors: language (e.g., answering simple questions but unable to express oneself logically when faced with questions that require thought), memory (e.g., not remembering well even after repeated attempts), perception (e.g., frequently confused when reading similar letters or numbers), concentration (e.g., becoming distracted while doing tasks), and processing speed (e.g., learning speed slower than peers).

[0007] Artificial Intelligence (AI) is a subfield of computer science that seeks to artificially replicate human learning, reasoning, and perceptual abilities. It is also an infrastructure technology within the field of information engineering. It is a concept distinct from natural intelligence, which refers to the intelligence possessed by animals, including humans.

[0008] It is a computer system equipped with functions that mimic human intelligence, representing the artificial demonstration (implementation) of human intelligence in machines or similar devices. It is generally assumed to be applied to general-purpose computers. This term also refers to the scientific and technological field that researches methodologies and feasibility for creating such intelligence.

[0009] Therefore, research and development in this area is necessary, as processing and analyzing the results of multiple intelligence tests—which exist as vast amounts of big data—using artificial intelligence can provide users with credible, personalized career recommendations based on diversified indicators.

[0010] Furthermore, research and development are needed on technology that, when a guardian is wondering whether a user is a slow learner, utilizes an AI language model to analyze questions describing the user's suspected behaviors to determine if they are indeed slow learner behaviors; if so, uses augmented search to aggregate internal and external data to generate prompts, inputs these prompts into an LLM, and provides a response offering a diagnosis and guidance for improvement tailored to the user.

[0011] Prior Art 1: Korean Published Patent Application No. 10-2023-0107032 (Published July 14, 2023)

[0012] Prior Art 2: Korean Patent Publication No. 10-2639894 (Published Feb. 27, 2024) The problem to be solved

[0013] The problem that the present invention aims to solve is to provide an AI-based multiple intelligence analysis and slow learner behavior detection and response system.

[0014] The problems that the present invention aims to solve are not limited to those mentioned above, and other unmentioned problems will be clearly understood by a person skilled in the art from the description below. means of solving the problem

[0015] The system of the present invention for solving the aforementioned problem may be characterized by comprising: a user terminal that presents questions to a user to perform a multiple intelligence test and inputs a slow learner question; a main server that receives the user's multiple intelligence test results from the user terminal, generates the user's multiple intelligence test profile, and generates an answer to the slow learner question, wherein the main server comprises: a generation module that analyzes the user's multiple intelligence test results using artificial intelligence to generate the user's multiple intelligence test profile; a peer comparison module that compares the user's multiple intelligence test profile with the multiple intelligence test profile of a peer group using artificial intelligence; a judgment module that extracts the user's unit behavior from the user's slow learner question, splits it, and determines whether it is a slow learner behavior; an augmented search module that generates a prompt using artificial intelligence-based augmented search from the user's unit behavior; and a slow learner answer module that receives the prompt and generates an answer to the slow learner question. Effects of the invention

[0016] The system of the present invention has the advantage of enabling career recommendations based on an individual's innate talents and aptitudes.

[0017] In addition, the system of the present invention has the advantage of being able to identify clear strengths and weaknesses by comparing and analyzing a user's multiple intelligence test profile with a peer group.

[0018] In addition, the system of the present invention has the advantage of being able to determine the similarity of the user's intelligence to the desired job group by comparing and analyzing the user's multiple intelligence test profile with a group of experts.

[0019] In addition, the system of the present invention has the advantage of being able to recommend a customized job to the user by harmonizing the user's strengths and weaknesses.

[0020] In addition, the system of the present invention has the advantage of being able to determine whether a user is a slow learner when a guardian inputs suspicious behavior of the user.

[0021] Furthermore, the system of the present invention has the advantage of utilizing augmented search to generate an integrated report containing the user's major problem areas, recommendations, and learning guides, and providing this as an answer when the user is a slow learner.

[0022] The effects of the present invention are not limited to those mentioned above, and other unmentioned effects will be clearly understood by a person skilled in the art from the description below. Brief explanation of the drawing

[0023] Figure 1 is a schematic diagram of the system of the present invention. FIG. 2 is a conceptual diagram showing the configuration of the main server of the system of the present invention, a graph comparing a user and a peer group by a peer comparison module, a graph comparing a user and an expert by an expert comparison module, and a career recommendation module that combines strengths in the user's multiple intelligences to recommend a career. Figure 3 is a conceptual diagram illustrating the system of the present invention identifying a user's slow learner from a guardian's question and providing an answer thereto. Specific details for implementing the invention

[0024] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the present invention, and the present invention is defined only by the scope of the claims.

[0025] The terms used in this specification are for describing embodiments and are not intended to limit the invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. The terms "comprises" and / or "comprising" used in this specification do not exclude the presence or addition of one or more other components in addition to the components mentioned. Throughout the specification, the same reference numerals refer to the same components, and "and / or" includes each of the mentioned components and all combinations of one or more. Although terms such as "first," "second," etc., are used to describe various components, these components are not limited by these terms. These terms are used merely to distinguish one component from another. Therefore, the first component mentioned below may be the second component within the technical scope of the invention.

[0026] Unless otherwise defined, all terms used herein (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.

[0027] Hereinafter, in describing the system of the present invention, the natural language processing model may be interpreted as an artificial intelligence model capable of learning text data to understand and generate language, and may be, for example, an LLM, sLLM, SLM, etc., but is not limited thereto and should be broadly interpreted as a comprehensive concept that includes all engines that perform various natural language processing (NPL) tasks.

[0029] Hereinafter, the system of the present invention will be described with reference to the drawings. FIG. 1 is a schematic diagram of the system of the present invention, FIG. 2 is a configuration diagram of the main server of the system of the present invention, a graph comparing a user with a peer group by a peer comparison module, a graph comparing a user with an expert by an expert comparison module, and a conceptual diagram showing a career recommendation module that combines strengths in the user's multiple intelligences. FIG. 3 is a conceptual diagram showing the system of the present invention identifying a user as a slow learner from a guardian's question and providing an answer thereto.

[0030] The system of the present invention may include a user terminal (10), a network (20), and a main server (30).

[0031] The user terminal (10) is a terminal of a user that performs multiple intelligence tests, slow learner behavior analysis, and answers, and may be a terminal that provides a web page, an app page (app list), a program, or an application related to the system of the present invention.

[0032] In this case, the user terminal (10) can be implemented as a computer that can access and connect to a remote main server (30) through a network (20).

[0033] The user terminal (10) is a wireless communication device and may include all types of handheld-based wireless communication devices such as navigation, PCS (Personal Communication System), GSM (Global System for Mobile communications), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) terminal, smartphone, smartpad, tablet PC, etc., as well as wired communication devices such as a general desktop PC, but is not limited thereto.

[0034] A network (20) refers to a connection structure capable of exchanging information between each node, such as a user terminal (10) and a main server (30). Examples of such networks include, but are not limited to, RF, 3GPP (3rd Generation Partnership Project) networks, LTE (Long Term Evolution) networks, 5GPP (5th Generation Partnership Project) networks, WIMAX (World Interoperability for Microwave Access) networks, the Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network), Bluetooth networks, NFC networks, satellite broadcasting networks, analog broadcasting networks, and DMB (Digital Multimedia Broadcasting) networks.

[0035] The main server (30) is a server for an administrator operating the system of the present invention, and may be a server that provides web pages, app pages, programs, or applications related to the system of the present invention.

[0037] The user terminal (10) may be a terminal that presents questions for a multiple intelligence test to the user. The user can perform the multiple intelligence test through the user terminal (10) and transmit the multiple intelligence test results (answer data) to the main server (30).

[0038] In addition to receiving multiple intelligence test results from the user terminal (10) and generating the user's multiple intelligence test profile, the main server (30) can compare the user's multiple intelligence test profile with the multiple intelligence test profile of a peer group, compare the user's multiple intelligence test profile with the multiple intelligence test profile of a professional group, and recommend a career path by combining strengths in the user's multiple intelligence test profile.

[0039] To this end, the main server (30) may include a generation module (31) that generates a user's multiple intelligence test profile by analyzing the user's multiple intelligence test results using artificial intelligence, a peer comparison module (32) that compares the user's multiple intelligence test profile with a peer group's multiple intelligence test profile using artificial intelligence, an expert comparison module (33) that compares the user's multiple intelligence test profile with an expert group's multiple intelligence test profile using artificial intelligence, and a career recommendation module (34) that recommends a career by combining strengths in the user's multiple intelligence test profile using artificial intelligence.

[0040] The generation module (31) can generate multiple intelligence test results of a highly reliable user using artificial intelligence.

[0041] The multiple intelligence test is a testing method based on the theory of multiple intelligences by H. Gardner, a professor at Harvard University’s Graduate School of Education. The results of the multiple intelligence test are derived as specific values ​​(e.g., natural numbers) for each of the eight categories, such as spatial, logical-mathematical, interpersonal, bodily-kinesthetic, verbal, musical, intrapersonal, and naturalistic. Accordingly, the user’s multiple intelligences can be evaluated, but are not limited to this.

[0042] On the other hand, regarding user convenience, it is preferable to limit the number of items, and the reliability of the test results may be an issue due to various variables such as the influence of each item on the evaluation results or users' incorrect answers.

[0043] To address this, in the system of the present invention, items for performing a multiple intelligence test may be divided into a first item and a second item. In this case, the first item and the second item may each have different content and may be items capable of independently deriving multiple intelligence test results.

[0044] Therefore, when a user performs a multiple intelligence test by answering questions provided on a user terminal (10), a first multiple intelligence test result for the first question and a second multiple intelligence test result for the second question can be obtained.

[0045] In this case, the generation module (31) can utilize artificial intelligence to increase the reliability of the user's multiple intelligence test results.

[0046] Artificial Intelligence (AI) is a field of computer engineering and information technology that realizes human learning, reasoning, perception, and natural language understanding capabilities through computer programs. In particular, supervised learning, which is utilized in the widest range of fields within AI, is a method that predicts future values ​​using training data that includes correct answers (labels).

[0047] For reference, machine learning can be defined as a methodology that enables computers to learn autonomously without being explicitly programmed; it is a method in which a program learns patterns from data on its own. Machine learning is classified into supervised learning and unsupervised learning depending on whether correct answers are specified in the data required for learning. Furthermore, depending on the purpose of use, it is categorized into methodologies such as classification, which divides data into a finite number of categories; regression, which maps data to continuous values; clustering, which groups similar data; and dimensionality reduction, which maps multidimensional data to a representative lower dimension.

[0048] Deep learning is a machine learning technique that has dramatically improved the performance of machine learning, which had been stagnant for some time. Deep learning is a methodology based on Artificial Neural Network (ANN) algorithms that mimic the superposition of synapses in the structure of the human brain. Deep learning structures can include Deep Neural Networks (DNNs), which have multiple hidden layers between the input layer and the output layer; Convolutional Neural Networks (CNNs), which place filters necessary for factor extraction in front of the hidden layers and learn the filters together; and Recurrent Neural Networks (RNNs), which can process time-series data by stacking artificial neural networks at each time step. Here, the high performance of deep learning models is explained in two ways. First, artificial neural networks are universal approximators capable of approximating all types of functions through the superposition of weighted sums of functions in each layer; therefore, if sufficiently general data is provided, they can simulate data with high accuracy. Second, in order to effectively classify data, it is important to appropriately extract representative factors; this can be achieved by utilizing spiral neural networks and learning filters to extract optimal factors. Furthermore, deep learning is an advanced form of the artificial intelligence model known as the neural network, characterized by a structure in which hidden layers within the hierarchical neural network are arranged in multiple stages. Recent deep learning models have increased the number of hidden layers, with the number of weights (representing connection strength) connecting nodes reaching up to billions.

[0049] For example, the generation module (31) can select a first user who answered the same as the user (the user who performed the current test) in the first question and a second user who answered the same as the user in the second question among past users. For reference, the past users are different individuals from the user and may be users who performed the test in the past.

[0050] Next, the generation module (31) can train the first artificial intelligence using the first user's answers to the first and second questions as feature values ​​and the average value of the first user's first multiple intelligence test results and the second user's second multiple intelligence test results as label values, and can train the second artificial intelligence using the second user's answers to the first and second questions as feature values ​​and the average value of the second user's first multiple intelligence test results and the second user's second multiple intelligence test results as label values.

[0051] Accordingly, the first artificial intelligence may be an artificial intelligence trained on the first question using data from past users identical to the user, and the second artificial intelligence may be an artificial intelligence trained on the second question using data from past users identical to the user.

[0052] Next, the generation module (31) can set the average value of the user's actual first multiple intelligence test result (a test result derived from a test method based on multiple intelligence theory from a value entered by the user in the first question) and the actual second multiple intelligence test result (a test result derived from a test method based on multiple intelligence theory from a value entered by the user in the second question) as the first value, set the result value of the multiple intelligence test output by inputting the user's actual answer to the first question and the answer to the second question into the first artificial intelligence as the second value, and set the result value of the multiple intelligence test output by inputting the user's actual answer to the first question and the answer to the second question into the second artificial intelligence as the third value.

[0053] Next, if the absolute value of the difference between the first value and the second value is greater than or equal to the reference value and the absolute value of the difference between the first value and the third value is greater than or equal to the reference value, the generated module (31) may determine that the learned artificial intelligence has low reliability due to a lack of training data, etc., and may set the first value derived by the test method based on the multiple intelligence theory, which is obtained by the actual user inputting the first and second questions, as the value of the user's multiple intelligence test profile.

[0054] In contrast, the generation module (31) can view the reliability of the learned artificial intelligence as intermediate when either the absolute value of the difference between the first value and the second value or the absolute value of the difference between the first value and the third value is less than the reference value.

[0055] In this case, if the absolute value of the difference between the first value and the second value is less than the reference value, the reliability of the first artificial intelligence is considered to be higher than the reliability of the second artificial intelligence, and the average value of the first value and the second value can be set as the value of the user's multiple intelligence test profile. If the absolute value of the difference between the first value and the third value is less than the reference value, the reliability of the second artificial intelligence is considered to be higher than the reliability of the first artificial intelligence, and the average value of the first value and the third value can be set as the value of the user's multiple intelligence test profile.

[0056] In contrast, the generation module (31) can view the reliability of the learned artificial intelligence as high when the absolute value of the difference between the first value and the second value and the absolute value of the difference between the first value and the third value are both less than the reference value, and it is desirable to supplement the reliability of the test result by utilizing only the result value of the artificial intelligence, and the average value of the second value and the third value can be set as the value of the user's multiple artificial intelligence test profile.

[0057] In this case, since the first value, the second value, and the third value are set for each of the independent multiple categories (space, logical-mathematical, human-friendly, bodily-kinesthetic, language, music, self-reflection, nature), if the number of categories in which the absolute value of the difference between the first value and the second value or the first value and the third value is greater than or equal to the reference value is greater than or equal to the majority, it can be determined that the absolute value of the difference between the first value and the second value or the first value and the third value is greater than or equal to the reference value, and if it is less than the majority, it can be determined that the absolute value of the difference between the first value and the second value or the first value and the third value is less than the reference value.

[0058] Accordingly, the value of the user's final multi-AI test profile can have a value in which the actual test result and the prediction by AI represent a trade-off between the reliability of the test and the reliability of the AI.

[0059] The values ​​of the multiple artificial intelligence test profiles may be specific values ​​(e.g., natural numbers) set for each category of spatial, logical-mathematical, interpersonal, bodily-kinesthetic, linguistic, musical, intrapersonal, and natural, but are not limited thereto.

[0060] A peer comparison module (32) that compares a user's multiple intelligence test profile with a peer group's multiple intelligence test profile using artificial intelligence can compare the user's multiple intelligence test profile with a peer group's multiple intelligence test profile, and in this case, an artificial intelligence similarity judgment can be applied to extract a sample of the same group from the user and the population.

[0061] For example, a user can transmit user information to a main server (30) through a user terminal (10), and the user information may include the user's gender and age, address, parents' educational background and parents' income level, etc., and the user's information and past user information can be analyzed by similarity judgment using artificial intelligence to set past users with high similarity as a homogeneous group, and comparative analysis can be performed by comparing the value of the user's multiple intelligence test profile with the value of the homogeneous group's multiple intelligence test profile and displaying it as a percentile graph.

[0062] Meanwhile, similarity can be calculated by assigning weights to specific information among user information using algorithms such as Euclidean distance, Dynamic Time Warping (DTW), and Longest Common Subsequence (LCSS), but is not limited to these.

[0063] Furthermore, even without using artificial intelligence, it is possible to determine past users who have content identical to the content of each item of user information as belonging to the same group as the current user.

[0064] The expert comparison module (33), which compares the user's multiple intelligence test profile with the expert group's multiple intelligence test profile using artificial intelligence, can match the user with the specific job group by classifying the user into a specific job group, setting the average value of the multiple intelligence test profile within the specific job group as the multiple intelligence test profile of that job group, and then comparing the pattern of the graph representing the value of the multiple intelligence test profile of each job group with the graph representing the value of the user's multiple intelligence test profile (x-axis category, y-axis value) to determine the job group with the highest similarity.

[0065] In this case, similarity can be calculated using the Euclidean distance algorithm, the Dynamic Time Warping (DTW) algorithm, the Longest Common Subsequence (LCSS) algorithm, etc., but is not limited to these.

[0066] It may include a career recommendation module (34) that recommends a career path by combining strengths from the user's multiple intelligence test profile using artificial intelligence.

[0067] In this case, the career recommendation module (34) may have appropriate values ​​of multiple categories of multiple intelligence test results mapped to each of multiple careers (e.g., figure skater, soccer player, etc.).

[0068] The career recommendation module (34) can select a career path from among multiple career paths that has a high value of similarity to the value of the user's multiple intelligence test profile and recommend it to the user.

[0069] In this case, similarity can be calculated using the Euclidean distance algorithm, the Dynamic Time Warping (DTW) algorithm, the Longest Common Subsequence (LCSS) algorithm, etc., but is not limited to these.

[0070] In contrast, the career recommendation module (34) can recommend a career to the user by an arithmetic method, and among multiple careers, it can recommend to the user the career with the smallest average absolute value of the difference between the user's multiple intelligence test profile value and the appropriate value in each of multiple categories.

[0071] Whether using the aforementioned artificial intelligence tool or an arithmetic method, the system of the present invention has the advantage of being able to recommend a trade-off career path to the user by reflecting strengths and weaknesses in each of the multiple categories of the multiple intelligence test profile.

[0073] The user terminal (10) may be a terminal into which a question regarding a slow learner is entered. That is, if the user's guardian (e.g., a parent) suspects whether the user's behavior corresponds to that of a slow learner, they may enter a question regarding this through the user terminal (10).

[0074] The main server (30) can generate an answer to a slow learner question entered from the user terminal (10) and transmit it to the user terminal (10). In this case, the answer may be provided in the form of a report containing a result of determining whether the user is a slow learner in response to the slow learner question and a guide for correction.

[0075] To this end, the main server (30) may include a judgment module (41) that extracts and splits the user's unit behavior from the user's slow learner question to determine whether it is a slow learner behavior, an augmented search module (42) that uses AI-based augmented search from the user's unit behavior to input into a natural language processing model to generate a prompt for generating an answer, and a slow learner answer module (43) that receives the prompt and generates an answer to the slow learner question.

[0076] The judgment module (41) may be a module that analyzes slow learner questions to determine whether the user is a slow learner. In this case, the slow learner questions may be text about behavioral patterns observed by a guardian of the user, but are not limited thereto.

[0077] Slow learner behaviors can be defined by five types of behaviors: language (e.g., answering simple questions but unable to express oneself logically on questions that require thought), memory (e.g., not remembering well even after repeated attempts), perception (e.g., frequently confusing similar letters or numbers), concentration (e.g., becoming distracted while doing tasks), and processing speed (e.g., learning speed slower than peers).

[0078] For example, a question from a slow learner could be, "I am worried because I am a boy born in 2017 who cannot watch educational animations repeatedly with concentration, cannot logically explain the content of the educational animations he has watched repeatedly, and cannot perform simple arithmetic operations such as 1+8 that are repeatedly explained in the educational animations," but it is not limited to this.

[0079] The judgment module (41) may be a natural language processing model capable of extracting (splitting) a user's unit behavior through text splitting in a slow learner question and determining whether the user's unit behavior corresponds to a slow learner behavior by comparing it with a slow learner behavior.

[0080] To this end, the judgment module (41) may have a natural language processing model learned from correlated slow learner question samples and unit action samples embedded therein, and accordingly, when a slow learner question is input, it may output a unit action.

[0081] For example, the judgment module (41) can extract unit behaviors from the slow learner questions depending on the context and situation, such as 1) being unable to watch educational animations with concentration, 2) being unable to logically explain the content of educational animations watched repeatedly, and 3) being unable to perform simple arithmetic operations learned repeatedly.

[0082] Additionally, the judgment module (41) can use an artificial intelligence module to compare the user's unit behavior with slow learner behavior to make a judgment. To this end, text regarding the user's unit behavior is indexed, embedded, vectorized, and stored, and the user's unit behavior can be classified as slow learner behavior by similarity judgment with slow learner behavior that is pre-configured or collected from external sources, indexed, embedded, and vectorized. In this case, keyword analysis, multi-turn methods, etc., may be utilized, but are not limited thereto.

[0083] For example, the judgment module (41) can classify the user's unit behavior "1) unable to watch educational animations with concentration" into "concentration" which has high similarity among slow learner behaviors by splitting the text, tokenizing and vectorizing it, and classify the user's unit behavior "2) unable to logically explain the content of educational animations watched repeatedly" into "language" and "memory" which have high similarity among slow learner behaviors by splitting the text, tokenizing and vectorizing it, and classify the user's unit behavior "3) unable to perform simple arithmetic operations learned repeatedly" into "processing speed" which has high similarity among slow learner behaviors by splitting the text, tokenizing and vectorizing it.

[0084] Meanwhile, the judgment module (41) may terminate the consultation as "not applicable" if the input slow learner question is not classified as slow learner behavior, or, conversely, may request additional questions if it is difficult to determine whether the input slow learner question corresponds to slow learner behavior based only on the input slow learner question.

[0085] The augmented search module (42) can generate prompts by utilizing AI-based augmented search from the unit actions of a user classified as a slow learner behavior. That is, since the general guardian's question is not professional or specific, AI-based augmented search is utilized to generate prompts that are input into a natural language processing model for the answer.

[0086] In this case, the augmented search module (42) can collect data in real time through public and private data, search for relevant documents based on search augmentation generation technology (RAG), and generate prompts through a natural language processing model that contextually combines the content of relevant documents with slow learner questions.

[0087] For example, the augmented search module (42) can collect in real time, via API or similar methods, public data including discovery of children with borderline intellectual disabilities classified into five behaviors and academic evidence papers in each category area, academic research materials related to medicine, psychology, and health, binary data of children with borderline intellectual disabilities behaviors, and behavior extension data, as well as private data including observation logs and detailed observation data of actual slow learners and care data of actual children with borderline intellectual disabilities.

[0088] Next, the augmented search module (42) can search for and select related documents that match slow learner behavior within the same behavior category. In this case, the augmented search module (42) can index and embed documents of public data and private data to vectorize them, and select related documents related to slow learner behavior through similarity judgment and keyword analysis between the embedding data of slow learner behavior and the embedding data of public data and private data.

[0089] For example, regarding the user's unit behavior "1) unable to watch educational animations with concentration," one can select the first document with the highest text similarity from public and private data classified as "concentration" (for example, a paper on animation viewing and slow actors with a lack of concentration); regarding the user's unit behavior "2) unable to logically explain the content of educational animations watched repeatedly," one can select the second document with the highest text similarity from public and private data classified as "language" and "memory" (a paper on slow actors who cannot logically explain content watched repeatedly); and regarding the user's unit behavior "3) unable to perform simple arithmetic operations learned through repeated training," one can select the third document with the highest text similarity from public and private data classified as "processing speed" (a log of observation results of slow actors who cannot perform arithmetic operations learned through repeated training).

[0090] Next, the augmented search module (42) can generate a prompt by contextually combining the behavior of a slow learner with keywords from related literature. In this case, the keywords from related literature may be words that are repeatedly exposed in related literature, words that are repeatedly exposed in other literature that is highly similar to related literature, or text that is highly similar to the behavior of a slow learner in the title or summary of related literature, but are not limited thereto.

[0091] For example, the augmented search module (42) can generate a professional prompt that says, "You cannot concentrate on watching educational animations, cannot logically explain the content, and especially cannot perform simple arithmetic operations. This is a behavior that occurs repeatedly in 70% of existing slow learners, and in particular, 80% of related papers judged it as borderline intellectual disability. Please refer to this and provide a report in the form of a guide on whether you are a slow learner and an improvement guide."

[0092] The slow learner response module (43) is a generative artificial intelligence (e.g., sLLM) trained with a language bundle dedicated to slow learners, and when a prompt is entered, it can output an answer related to the slow learner in the form of a report.

[0093] To this end, the generative artificial intelligence of the slow learner response module (43) is trained on behavior (characteristic) data of borderline intellectual disability through hospital doctor interviews, opinion data of doctors and specialists with borderline intellectual disability, interviews with parents of those with borderline intellectual disability, behavior pattern extraction, education / learning data, and care guide data, and can generate an answer in the form of a report when a prompt related to a slow learner is entered.

[0094] Accordingly, the system of the present invention has the effect of providing a report based on actual interview results and care guides by applying augmented search technology to combine the content of professional related documents (papers, etc.), even if a guardian generates and inputs questions regarding user behavior from a non-professional perspective.

[0096] The method for performing the system of the present invention described above may be implemented as a program (or application) to be executed in combination with a server, which is hardware, and stored on a medium.

[0097] The aforementioned program may include code encoded in a computer language such as C, C++, JAVA, or machine language, which can be read by the computer's processor (CPU) through the computer's device interface, in order for the computer to read the program and execute the methods implemented in the program. Such code may include functional code related to functions that define the necessary functions for executing the methods, and may include control code related to execution procedures necessary for the computer's processor to execute the functions according to a predetermined procedure. Additionally, such code may further include memory reference code regarding where (address) additional information or media necessary for the computer's processor to execute the functions should be referenced in the computer's internal or external memory. In addition, if the processor of the computer needs to communicate with any other computer or server located remotely in order to execute the above functions, the code may further include communication-related code regarding how to communicate with any other computer or server located remotely using the communication module of the computer, and what information or media to transmit or receive during communication.

[0098] The above-mentioned storage medium refers to a medium that stores data semi-permanently and is readable by a device, rather than a medium that stores data for a short period of time, such as a register, cache, or memory. Specifically, examples of the above-mentioned storage medium include, but are not limited to, ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage device. That is, the above-mentioned program may be stored on various recording media on various servers that the computer can access, or on various recording media on the user's computer. Additionally, the above-mentioned medium may be distributed across networked computer systems, and computer-readable code may be stored in a distributed manner.

[0099] The steps of the method or algorithm described in connection with embodiments of the present invention may be implemented directly in hardware, implemented as a software module executed by hardware, or implemented by a combination thereof. The software module may reside in RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable ROM), EEPROM (Electrically Erasable Programmable ROM), Flash Memory, a hard disk, a removable disk, a CD-ROM, or any form of computer-readable recording medium well known in the art to which the present invention belongs.

[0100] Although embodiments of the present invention have been described above with reference to the attached drawings, those skilled in the art will understand that the present invention may be implemented in other specific forms without altering its technical concept or essential features. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive.

Claims

Claim 1 A user terminal that performs a multiple intelligence test by presenting a first question and a second question, each composed of different content and independently deriving a multiple intelligence test result, to the user, and inputs a slow learner question; and a main server that receives a first multiple intelligence test result corresponding to the first question and a second multiple intelligence test result corresponding to the second question from the user terminal, generates a user's multiple intelligence test profile, and generates an answer to the slow learner question, wherein the main server includes: a generation module that generates the user's multiple intelligence test profile by analyzing the user's multiple intelligence test results using artificial intelligence; a judgment module that extracts and splits the user's unit behavior from the user's slow learner question to determine whether it is a slow learner behavior; and an augmented search module that generates a prompt using artificial intelligence-based augmented search from the user's unit behavior.It includes a slow learner answer module that generates an answer to the above slow learner question upon input of a prompt, and the generation module selects a first user who answered identically to the user in the first question and a second user who answered identically to the user in the second question among past users, trains a first artificial intelligence using the first user's answers to the first and second questions as feature values ​​and the average value of the first user's first multiple intelligence test results and second multiple intelligence test results as label values, trains a second artificial intelligence using the second user's answers to the first and second questions as feature values ​​and the average value of the second user's first multiple intelligence test results and second multiple intelligence test results as label values, sets the average value of the user's first multiple intelligence test results and second multiple intelligence test results as the first value, inputs the user's answers to the first and second questions into the first artificial intelligence and sets the output multiple intelligence test results as the second value, and the user's A system characterized by inputting answers to the first and second questions into the second artificial intelligence to set the output multiple intelligence test result as the third value, and, based on the result of comparing the absolute value of the difference between the first value and the second value and the absolute value of the difference between the first value and the third value with a reference value, selecting and setting any one of the first value, the average value of the first value and the second value, the average value of the first value and the third value, and the average value of the second value and the third value as the value of the user's multiple intelligence test profile.;

Images