Learning support system and learning support method

Abstract

To provide a learning support system and others that present, even during answering, reference information useful to answerers and adjust difficulty of problems.SOLUTION: The present invention relates to a learning support system that learns through inputting an answer to a touch panel, and that includes: a prediction model which acquires, by an acquisition part, input information having been input by a previous answerer to the touch panel and an evaluation result of the input information, performs machine learning by a prediction model generation part based upon the input information and the evaluation result, and predicts, on the basis of change in an answer state, whether or not a correct answer may be reached; and a presentation part which applies the input information having been input by a current answerer to the touch panel and presents reference information corresponding to the answer state of the current answerer.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a learning support system, a learning support method, and a prediction model for use therein. [Background technology]

[0002] Handwriting input systems using fingers or touch pens are now being used on information terminals connected to the Internet. Until now, distance learning systems used at cram schools and at home have generally involved students writing answers to questions by handwriting them directly on a tablet, or converting handwritten characters into text data using online recognition technology, which is then sent to a question server where they are compared with sample answers and graded and evaluated. When the answers were sent, only the final results were used and discarded.

[0003] The learning support device and learning support program of Patent Document 1 are capable of providing a user with information useful for considering how to study. The device and learning support program include a learning need score calculation unit that calculates a learning need score SN, which indicates the level of learning required by a previous answerer to correctly answer each question; a current progress learning score calculation unit that calculates a current progress learning score SP, which indicates the level of learning required by a current answerer up to the present time for each question; a learning progress calculation unit that calculates a learning progress level, which indicates the progress the current answerer has made to correctly answer questions that the current answerer has answered incorrectly, based on the learning need score SN and the current progress learning score SP; and a learning progress presentation unit that presents the calculated learning progress level, and for questions that the current answerer has answered incorrectly, the learning progress level up to the correct answer is presented for each incorrect question, thereby providing reference information on how much more learning is required to correctly answer the questions.

[0004] The training data refining method and computer system of Patent Document 2 generate a training dataset that improves the prediction accuracy of a model. This discloses a training data refining method that includes the steps of connecting a computer to a DB that stores training datasets and validation datasets and generating multiple sample datasets from the validation dataset, calculating a score for each of the multiple training data representing the strength of the influence of the training data on the prediction accuracy of the model for one sample dataset, identifying harmful training data that adversely affects the prediction accuracy of the model for the sample dataset based on the score, determining whether or not to delete the harmful training data based on the score, and generating a refined training dataset in which the harmful training data has been deleted from the training dataset based on the determination result. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2019-191388 [Patent Document 2] Japanese Patent Publication No. 2021-033544 Summary of the Invention [Problem to be solved by the invention]

[0006] Learners are increasingly using touchscreen devices in educational settings. Conventional learning support systems, including those described in Patent Documents 1 and 2, only use the final result. This means that they can only evaluate whether the answer is correct or incorrect after the answer is completed. Furthermore, they have to wait for processing until the final answer is entered. However, in order for learners to learn more efficiently, it is considered better to present appropriate information according to the learning situation.

[0007] The present invention aims to provide a learning support system that can present useful reference information to a solver while he or she is solving a problem, and can adjust the difficulty level of the next problem. [Means for solving the problem]

[0008] The present inventors have conducted extensive research to solve the above problems and have found that the following inventions meet the above objectives, thereby completing the present invention.

[0009] <1> It is a learning support system that allows students to study by inputting answers into a touch panel. The input information entered into the touch panel by the previous answerer and the evaluation results of that input information are acquired, and machine learning is performed based on the input information and the evaluation results, and a prediction model is used to predict whether the correct answer will be reached based on changes in the answer situation. A learning support system that applies input information entered by a current answerer on a touch panel to the prediction model and presents reference information according to the answering status of the current answerer. <2> The learning support system as described above, wherein the presentation unit presents reference information when it is determined that the current answerer is unlikely to derive the correct answer. <3> The learning support system, wherein the input information includes the input content and the time it took for the content to be input. <4> This is a learning support method for learning by inputting answers into a touch panel. A learning support method having a presentation step of acquiring input information inputted into a touch panel by a current answerer, input information inputted into a touch panel by previous answerers, and evaluation results of the input information, performing machine learning based on the input information and the evaluation results, and applying the information to a prediction model created to predict whether the correct answer will be reached from changes in the answer situation, thereby presenting reference information according to the answer situation of the current answerer. <5> A prediction model that obtains the input information entered by the respondent into the touch panel and the evaluation results of that input information, performs machine learning based on the input information and the evaluation results, and predicts whether the correct answer will be reached based on changes in the answer situation. [Effects of the Invention]

[0010] According to the present invention, it is possible to present useful reference information to the answerer while he or she is answering the questions, which is expected to enable the answerer to study more efficiently. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a schematic diagram of a learning support system according to the present invention; [Figure 2] 1 is a schematic diagram for explaining a situation in which the learning support system of the present invention is used; [Figure 3] FIG. 10 is a diagram for explaining input information for creating a model of the learning support system. [Figure 4] FIG. 10 is a diagram for explaining a state in which reference information is displayed by the learning support system. DETAILED DESCRIPTION OF THE INVENTION

[0012] The following describes in detail an embodiment of the present invention, but the following description of the constituent elements is one example (typical example) of an embodiment of the present invention, and the present invention is not limited to the following content unless the gist of the present invention is changed. Note that when the expression "to" is used in this specification, it is used as an expression that includes the numerical values ​​before and after it.

[0013] [Learning support system of the present invention] The learning support system of the present invention is a learning support system for learning by inputting answers into a touch panel, and has: a prediction model that acquires input information input into the touch panel by previous answerers and evaluation results of the input information, performs machine learning based on the input information and the evaluation results, and predicts whether the correct answer will be reached from changes in the answer situation; and a presentation unit that applies input information input into the touch panel by a current answerer to the prediction model and presents reference information according to the answer situation of the current answerer.

[0014] [Learning support method of the present invention] The learning support method of the present invention is a learning support method for learning by inputting answers onto a touch panel, and includes a presentation step of acquiring input information input onto the touch panel by a current answerer, input information input onto the touch panel by previous answerers, and an evaluation result of the input information, and performing machine learning based on the input information and the evaluation result, applying the information to a prediction model created to predict whether the correct answer will be reached from changes in the answer situation, and presenting reference information according to the answer situation of the current answerer.

[0015] [Prediction model of the present invention] The predictive model of the present invention obtains input information entered by the answerer into the touch panel and the evaluation results of that input information, performs machine learning based on the input information and the evaluation results, and predicts whether the correct answer will be reached based on changes in the answer situation.

[0016] According to the present invention, it is possible to present useful reference information suitable for the solver even while the solver is solving a problem. In this application, the prediction model of the present invention can be used in the learning support system of the present invention and the learning support method of the present invention. In this application, the corresponding configurations can be used mutually.

[0017] FIG. 1 is a schematic diagram of the learning support system of the present invention. FIG. 2 is a schematic diagram for explaining a situation in which the learning support system of the present invention is used. Answerers 11 and 12 input their answers on touch panels 21 and 22. The input answers are saved as ink data in answerer ink data databases (DBs) 41 and 42, which are personal cache databases (personal cache DBs). In addition, the answerers' learning histories are saved in answerer learning history databases (DBs) 51 and 52. The personal data is processed by a question / answer / supplemental information generation engine 61. The information handled by the question / answer / supplemental information generation engine 61, the processed information, the evaluation information to be compared with the input information, and the evaluation results are saved in a database 7 as appropriate.

[0018] [Learning Support System] The present invention relates to a learning support system. The learning support system is a system that supports learning. The learning subject can be any subject, such as various subjects or various qualifications for which exams are conducted. The system can handle learning subjects from elementary schools, junior high schools, high schools, universities, graduate schools, various vocational schools, preparatory schools, schools for qualification exams, and liberal arts schools. For example, the system can handle learning subjects such as Japanese language, arithmetic, mathematics, science, chemistry, social studies, history, geography, foreign languages, English, physical education, art, and music. The system is particularly useful for multiple-choice questions or questions that require appropriate description of the process of consideration as you solve the problem, rather than questions that can be answered immediately.

[0019] [Answer] A solver is someone who solves the problems that are presented. Solvers include previous solvers and current solvers. Previous solvers are people who have solved similar problems that are the target of support in the past. Current solvers are people who are solving the problem that is the target of support. The current solver and previous solvers may be the same person, or different people. When a current solver solves a problem that they solved in the past as a previous solver, the same person may be both the current solver and a previous solver. Previous solvers provide sample data for machine learning, so the more previous solvers there are, the better, and current solvers may also be added as information on previous solvers after scoring is complete, and machine learning may continue.

[0020] [Touch panel] The learning support system of the present invention supports learning using input information entered on a touch panel. The touch panel can be integrated into various highly portable electronic computers such as smartphones, tablet terminals, and laptop terminals, or an input terminal that is appropriately connected to the main body of the electronic computer. Input to the touch panel is performed using an electronic pen or a finger.

[0021] [Input information] FIG. 3 is a diagram explaining the input information used as training data for creating a model for the learning support system. A respondent may write down various considerations on a touch panel of a terminal or the like before entering their answer. The input information is the information that the respondent enters on the touch panel before giving their answer. The input information is entered as ink data. Specifically, depending on the type of problem, characters, symbols, figures, etc. that describe the considerations for solving the problem are entered.

[0022] Before determining the answer to a problem, the solver may conduct various investigations. For example, in the case of a math problem, the solver may write down notes such as intermediate calculation formulas or case distinctions using symbols and figures to understand the question. There are also problems where there are multiple possible investigation processes and solutions. In addition, the input information for each investigation process may be entered in a short amount of time, or the answer may be reached immediately by mental calculation without inputting anything, so the investigation process leading up to the answer can also be important.

[0023] If the process of these considerations is appropriate, you will eventually find the correct answer. On the other hand, if you start making the wrong considerations during the consideration process, or if you are doodling and inputting information that is unrelated to the problem and that distracts you, or if you do not consider anything at all, you are likely to get the wrong answer.

[0024] The tendency to arrive at a correct or incorrect answer is determined by machine learning based on the time it takes to input information that accompanies such various considerations. This allows for a prediction model that can determine whether the input information and its time are likely to lead to a correct answer.

[0025] By analyzing the input information in this way, it is possible to grasp the trend of the answer even during the process of answering the question. It is also possible to classify the direction and tendency of errors. The predictive model can also be used while improving by registering the input information of new answerers in the database.

[0026] [Ink Data] Input information can be ink data. Dot sequence data obtained from an input device such as a pen can also be expressed as ink data. By acquiring ink data along with changes over time, position information and pressure of the electronic pen, time series information of the dot sequence, xy coordinates, and writing speed can be obtained, and ink data is generated based on these.

[0027] The elements that make up ink data may change depending on the type of stationery the electronic pen is set to use as, for example, stationery information such as writing implement, eraser, or pointing implement, and within writing implements, types such as ballpoint pen, brush, or marker.

[0028] In addition, you can set the attribute information of the writing implement, such as the thickness and color of the lines drawn with the implement, or the size of the erased area in the case of an eraser. Information about the time of writing, the writing speed, and unique information about specific points (starting point, ending point, bending point, point of change in writing speed, etc.) are also recorded along with the trajectory information of the electronic pen on the tablet. Input information can also include the input content and the time it took to input the content.

[0029] This allows the system to analyze the relationship between the most recent ink data and past learning records by focusing on instantaneous recorded information, such as the time it took to arrive at the answer, information that was written several times and then erased with an eraser tool, and private information leading up to the answer.

[0030] The system of the present invention may use an acquisition unit. The acquisition unit acquires input information entered on the touch panel by previous solvers and evaluation results for that input information. The acquisition unit may also acquire information other than the input information entered on the touch panel. For example, the input information may be obtained through a questionnaire survey on the solver's attributes, number of answers, time of answering, etc. The evaluation results are the result of comparing the input information with data for evaluating each answer to a question stored in an evaluation database, such as the correctness or incorrectness of the question and the score. A question, answer, and supplementary information database 7 may be used as the evaluation database. Answers to questions by multiple solvers (learners, scribes) have traditionally been collected as final score sequence information and recorded in a database. This information is then analyzed, and scoring and personal information registration are completed based on the analysis results. According to the present invention, written information can be acquired and used even during operation. In the configuration of FIG. 1, the question, answer, and supplementary information generation engine 61 can function as this acquisition unit.

[0031] The system of the present invention may include a creation unit. The creation unit performs machine learning based on the input information and evaluation results acquired by the acquisition unit to create a prediction model that predicts changes in the answer status over time. Machine learning can be performed, for example, using a type of algorithm that classifies using supervised learning. The creation unit performs machine learning using training data such as what input information was entered appropriately, at what timing (time), whether the result of the input information was "correct" or "incorrect" for the question, or the score, to create a prediction model. The input information for such machine learning may include the input content and the time until the content was entered. In the configuration of FIG. 1, the question / answer / supplementary information generation engine 61 can also function as this creation unit.

[0032] [Prediction model] The prediction model is a model created by machine learning based on the input information and evaluation results acquired by the acquisition unit. The prediction model is created by an appropriate means depending on the means of collecting training data and machine learning. For example, the prediction model may be created by the question / answer / supplementary information generation engine 61, or may be created by a configuration using other networks or clouds not shown in FIG. 1. A model created by machine learning in this way predicts whether a correct answer will be reached using changes in the answer situation as input information. Changes in the answer situation refer to what input information is entered by the answerer for which question, at what timing (time), and changes in the results (correctness or incorrectness) of that input information and the score. In this way, the prediction model predicts the likelihood of a correct answer or an incorrect answer, the expected score, etc., based on what information is entered into the touch panel by what time.

[0033] FIG. 4 is a diagram illustrating the state in which reference information is displayed by the learning support system. The presentation unit applies the input information of the current answerer to the prediction model created by the creation unit and performs processing to present reference information according to the answering status of the current answerer. The reference information to be presented can be extracted and presented from the question / answer / supplementary information DB7 based on the results of processing by the question / answer / supplementary information generation engine 61 in FIG. 1. For example, if the current answerer is likely to get the correct answer as is, there is little need to present reference information. For this reason, it is preferable to present reference information when it is determined that the current answerer is unlikely to derive the correct answer or when it is determined that the current answerer is likely to receive a low score.

[0034] [Reference information] Reference information is information that is displayed after a question is asked, depending on the learning settings. Examples of reference information include advice (so-called hints) that are helpful in solving the problem, encouragement, notification of the passage of time, warnings, and notification of the evaluation status of the likelihood of getting the answer correct. Reference information can be displayed by any means. For example, it may be displayed on the answerer's touch panel monitor, notified by audio, or displayed on a display means such as a terminal separate from the one used by the answerer.

[0035] By notifying the answerer of reference information so that they can see it directly, the answerer can solve the problem independently.In addition, by displaying the reference information on another device and managing this other device, teachers or parents can use it as a reference for when to support the answerer.

[0036] In this way, a predictive model that uses input information from previous test takers can identify areas where test takers are likely to get stuck based on their tendencies. Furthermore, while learning efficiency is poor if the answer remains difficult and takes a long time, the present invention makes it easy to manage the time spent working on the question. Meanwhile, conventional methods that only evaluate answers do not provide insight into the test taker's progress.

[0037] [Difficulty of the question] The learning support system of the present invention can also evaluate the answering status and adjust the difficulty of the next question. It can provide appropriate learning materials individually based on the analysis of the digital information at the time of writing, according to the level of the answerer, which cannot be evaluated only by the correctness of the answer.

[0038] The present invention provides a pen interface equipped with a learning support device that predicts the timing of providing hints to lead to the answer and provides the answerer (learner) with learning material data with added value, such as supplementary information and related information. This invention can also contribute to reducing the burden on teachers and improving efficiency in line with learning progress.

[0039] This invention uses an ink tool (pen interface) as a digital input device, and is significantly different from conventional methods in that it focuses on the digital data information at the time of handwriting, which has been used for writer recognition, handwritten text input, gesture input, and command operation.

[0040] This time-series information was not used when the character recognition results were output. In this invention, past handwriting data (ink data) is stored in a database as an ink file, and online information when the pen interface is used is recorded. Then, information on the correctness of past questions, the time required to answer, recorded information on repeated corrections, etc. are comprehensively used to make predictions and judgments using machine learning and other methods.

[0041] It can also provide feedback on the setting of supplementary information based on individual learning effects and the time required for input operations from past accumulated data, as well as on the format design, such as the size of the input interface entry box, and information on the placement and order of questions.

[0042] The touch panel records handwriting data, chronological information for each point, the time of touching, GPS information showing the location of the tablet, and other information, and the recorded information linked to each individual's questions and answers is stored in a database.

[0043] Using machine learning and other methods, the system analyzes the chronological information of each learner's writing, such as the time it took to arrive at the answer and note information, to determine the level of understanding during the learning process and the process information needed to arrive at the correct answer.

[0044] A learning support system is provided that can respond to individual users by using a device that searches for and predicts the difficulty level of questions, the order in which they are presented, and related supplementary information, and presenting the results as hints and supplementary information. [Industrial Applicability]

[0045] The present invention can be used as a learning support system and is industrially useful. [Explanation of symbols]

[0046] 11, 12 Answerer 21, 22 Touch panel 3 Reference information 41, 42 Answerer Ink Data Database 51, 52 Answerer learning history database 61 Question, Answer and Supplementary Information Generation Engine 7. Database for generating questions, answers, and supplementary information

Claims

1. It is a learning support system that allows students to study by inputting answers into a touch panel. The system acquires input information including the content entered into the touch panel by the previous answerer, the time it took for the content to be entered, and the evaluation results of the input information, and performs machine learning based on the input information and the evaluation results, using a prediction model that predicts whether the correct answer will be reached based on changes in the answer situation. A learning support system that applies input information entered by a current answerer on a touch panel to the prediction model and presents reference information according to the answering status of the current answerer.

2. 2. The learning support system according to claim 1, wherein the presentation is to present reference information when it is determined that the current answerer is unlikely to derive a correct answer.

3. This is a learning support method for learning by inputting answers into a touch panel. A learning support method having a presentation step of acquiring input information including the content entered into the touch panel by the current answerer and the time it took for that content to be entered, input information entered into the touch panel by previous answerers, and an evaluation result of that input information, and performing machine learning based on the input information and the evaluation result, applying this to a prediction model created to predict whether the correct answer will be reached from changes in the answer situation, and presenting reference information according to the answer situation of the current answerer.

Images