Information processing device, information processing method, program, and model generation method

Abstract

To output, to a learner, information on learning required to achieve the goal for each predetermined condition.SOLUTION: An information processing device acquires a learned model that has learned a relationship between learning history data for each unit period indicating the learning content of a learner in a first period and learning object data for each unit period indicating learning content necessary for achieving the goal after the first period. In addition, the information processing device acquires learning history data corresponding to a condition in time series. Then, the information processing device inputs the acquired learning history data to the learned model corresponding to the condition, and outputs learning object data corresponding to the condition in time series. Further, the information processing device outputs, as recommendation question information, information about recommendation questions that the learner should work on in order to achieve the goal, based on the learning goal data output by a goal output unit.SELECTED DRAWING: Figure 14

Description

[Technical Field]

[0001] The present invention relates to a technique for outputting information related to learning. [Background technology]

[0002] In recent years, development of systems for providing learning support to learners has been progressing. For example, Patent Document 1 discloses a learning support program that supports learners in executing and recording their learning plans. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6572412 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the invention of Patent Document 1 merely stores the learner's study plan in advance, and is unable to automatically create a study plan necessary to achieve a goal.

[0005] The present invention has been made to solve, for example, the above-mentioned problems, and aims to provide an information processing device that outputs to a learner information regarding the learning necessary to achieve a goal for each specified condition. [Means for solving the problem]

[0006] In one aspect of the present invention, a model acquisition unit acquires a trained model that has learned the relationship between learning history data for each unit period that indicates the learning content of a learner in a first period and learning goal data for each unit period that indicates the learning content necessary to achieve a goal in the first period and thereafter; a history acquisition unit that acquires, in chronological order, learning history data associated with conditions including a learning level that indicates the academic ability of a learner; and inputs the learning history data acquired by the history acquisition unit into a trained model that corresponds to the conditions including the learning level, thereby study a goal output unit that outputs learning goal data in chronological order; and a question output unit that outputs, as recommended question information, information on recommended questions that the learner should tackle in order to achieve the learning goal, based on the learning goal data output by the goal output unit; The system includes a model generation unit that uses training data for each condition including the learning level to learn the relationship between learning history data for each unit period that indicates the learning content of the learner in the first period and learning goal data for each unit period that indicates the learning content necessary to achieve the goal in the first period and thereafter, and generates the trained model; and a training data generation unit that divides the learning history data of learners who correspond to the conditions including the learning level and have achieved the goal into unit periods, and generates training data by using the learning history data of the first half period as input data and the learning history data of the second half period as correct answer data, and the training data generation unit generates multiple sets of training data from the learning history data of one learner. According to this aspect, the information processing device can provide the learner with learning goal data according to the learner's academic ability. In addition, the information processing device can provide the learner with recommended questions that they should specifically work on. Furthermore, according to this aspect, the information processing device can use the training data to generate a trained model corresponding to any condition such as the academic ability of a learner, etc. Furthermore, the information processing device can generate multiple sets of training data from the learning history data of one learner who has achieved their goal.

[0008] In one aspect of the information processing device, the conditions include the field of the learning content. According to this aspect, the information processing device can provide the learner with learning objective data corresponding to each field, such as "trigonometric functions" or "calculus" in mathematics. The information processing device can also generate trained models corresponding to each field.

[0009] In one aspect of the information processing device, the conditions include the difficulty level of questions included in the learning content. According to this aspect, the information processing device can provide the learner with learning goal data for each difficulty level of the questions. Furthermore, the information processing device can generate trained models corresponding to each difficulty level of the questions.

[0010] In one aspect of the information processing device, the conditions include the field of the learning content and the difficulty level of the questions included in the learning content, and the device includes a question information storage unit that stores question information that associates the questions with the difficulty level of the questions, and a question identification unit that identifies the recommended questions from the question information storage unit for each of the field and the difficulty level of the questions based on the learning goal data, and the question output unit outputs the identified recommended questions for each of the field and the difficulty level of the questions. According to this aspect, the information processing device can provide the learner with recommended questions to specifically work on for each of the field and the difficulty level of the questions.

[0011] In one aspect of the information processing device, the problem information storage unit further stores a medium containing the problem and a location of the problem on the medium in association with each other, and the problem output unit outputs the recommended problem or a medium containing the recommended problem and the location of the recommended problem on the medium. According to this aspect, the information processing device can provide the learner with the medium on which the recommended problem is written and its location for each level of difficulty of the problem in each field. This allows the learner to easily recognize, for example, the name of the text and the page in the text on which the recommended problem is written.

[0012] In another aspect of the present invention, an information processing method executed by an information processing device includes: a training data generation process for generating multiple sets of training data from the learning history data of one learner by dividing the learning history data, which corresponds to conditions including the learning level that indicates the academic ability of the learner and indicates the learning content of the learner who has achieved the goal, into unit periods, and using the learning history data of the first half period as input data and the learning history data of the second half period as correct answer data; a model generation process for using the training data for each condition including the learning level to learn the relationship between the learning history data for each unit period that indicates the learning content of the learner in a first period and the learning goal data for each unit period that indicates the learning content necessary to achieve the goal in the first period and thereafter, and generating a trained model; a model acquisition step of acquiring a trained model; a history acquisition step of acquiring, in time series, learning history data associated with conditions including the learning level; and inputting the learning history data acquired by the history acquisition step into a trained model corresponding to the conditions including the learning level, study The system has a goal output process for outputting learning goal data in chronological order, and a problem output process for outputting, as recommended problem information, information relating to recommended problems that the learner should work on to achieve the goal, based on the learning goal data output by the goal output process. By executing this information processing method, the information processing device can generate a trained model that corresponds to any condition, such as the academic ability of a learner, using multiple sets of training data generated from the learning history data of one learner who has achieved a goal. By executing this information processing method, the information processing device Learners' academic ability, etc.Learning objective data corresponding to any condition can be provided to the learner. Furthermore, by executing this information processing method, the information processing device can provide the learner with specific recommended problems to work on.

[0013] In another aspect of the present invention, a program executed by a computer includes: a model acquisition unit that acquires a trained model that has learned the relationship between learning history data for each unit period that indicates the learning content of a learner in a first period and learning goal data for each unit period that indicates the learning content necessary for achieving a goal in the first period and thereafter; a history acquisition unit that acquires, in chronological order, learning history data associated with conditions including a learning level that indicates the academic ability of a learner; a goal output unit that inputs the learning history data acquired by the history acquisition unit into a trained model that corresponds to the conditions including the learning level, and outputs learning goal data that corresponds to the conditions in chronological order; and a question output unit that outputs, as recommended question information, information regarding recommended questions that are questions that the learner should work on to achieve the goal, based on the learning goal data output by the goal output unit; a model generation unit that uses training data for each condition including the learning level to learn the relationship between learning history data for each unit period that indicates the learning content of the learner in the first period and learning goal data for each unit period that indicates the learning content necessary to achieve the goal in the first period and thereafter, and generates the trained model; a training data generation unit that divides the learning history data of learners who correspond to the conditions including the learning level and have achieved the goal into unit periods, and generates training data by using the learning history data of the first half period as input data and the learning history data of the second half period as correct answer data; and causing the computer to function as The training data generation unit generates a plurality of sets of training data from the learning history data of one learner. By installing and executing this program on a computer, an information processing device according to the present invention can be configured. [Effects of the Invention]

[0015] According to the information processing device of the present invention, it is possible to output to a learner information relating to the learning required to achieve a goal for each predetermined condition. [Brief explanation of the drawings]

[0016] [Figure 1] 1 shows the configuration of a learning objective prediction system to which a server of the present invention is applied. [Figure 2] FIG. 2 is a block diagram showing the hardware configuration of a server. [Figure 3] 10 is an example of a data configuration of a training data DB. [Figure 4] FIG. 10 is a diagram illustrating a model management DB. [Figure 5] 10 is an example of a data configuration of a learning history DB. [Figure 6] 10 is an example of the data structure of a question DB. [Figure 7] FIG. 2 is a block diagram showing the functional configuration of a server. [Figure 8] FIG. 2 is a block diagram showing the hardware configuration of the terminal device. [Figure 9] FIG. 10 is a diagram illustrating a learning goal prediction model. [Figure 10] FIG. 10 is a diagram illustrating training data. [Figure 11] 10 is an example of a subject-specific goal screen. [Figure 12] 10 is an example of a field-specific goal screen. [Figure 13] 10 is a flowchart of a model generation process by a server. [Figure 14] 10 is a flowchart of a target screen display process. [Figure 15] 10 is a flowchart of a recommended question specification process performed by a server. DETAILED DESCRIPTION OF THE INVENTION

[0017] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. [Overall configuration] 1 shows the configuration of a learning goal prediction system to which the server of the present invention is applied. Learning goal prediction system 100 is a system that uses artificial intelligence (AI) to predict and output learning goal data for each unit period that indicates the learning content necessary to achieve goals in the first period and thereafter, based on learning history data for each unit period that indicates the learning content of a learner in the first period. Learning goal prediction system 100 includes a server 10 and a terminal device 20, which are configured to be able to communicate with each other via a network 5 such as the Internet.

[0018] The server 10 is an information processing device that processes, stores, and transmits and receives various information, such as a server device, a personal computer, or a general-purpose tablet PC (personal computer). The terminal device 20 is an information processing device that accepts and transmits learning history data and receives and displays learning goal data, such as a smartphone, a mobile phone, a wearable device such as an Apple Watch (registered trademark), a tablet, or a personal computer terminal.

[0019] The server 10 according to this embodiment acquires learning history data for each unit period in chronological order, which indicates the learning content of a learner during a first period. The server 10 inputs the acquired chronological learning history data into a trained model that has learned the relationship between the learning history data for each unit period during the first period and learning goal data for each unit period indicating the learning content necessary to achieve goals from the first period onward, and outputs the learning goal data for each unit period from the first period onward in chronological order. Furthermore, based on the learning goal data, the server 10 outputs recommended question information, which is information regarding recommended questions that the learner should work on to achieve their goals. The trained model will be described later. The server 10 transmits the output chronological learning goal data and recommended question information to the terminal device 20. Based on the received learning goal data and recommended question information, the terminal device 20 displays graphs and recommended questions based on the learning goal data.

[0020] [Server Configuration] 2 is a block diagram showing the hardware configuration of server 10. Server 10 includes a communication unit 11, a control unit 12, a storage unit 13, a recording medium 14, a display unit 15, and an input unit 16. These components, a training data database (hereinafter, "database" will be referred to as "DB") 31, a model management DB 32, a learning history DB 33, and a question DB 34 are interconnected via a bus 19.

[0021] The server 10 may be executed by a single computer, or may be executed by a plurality of computers in a distributed manner, or may be executed by a virtual machine in a distributed manner.

[0022] The communication unit 11 is a communication unit for communicating with the terminal device 20 via the network 5. Specifically, the communication unit 11 transmits learning objective data and recommended question information to the terminal device 20, and receives condition data (described later) from the terminal device 20.

[0023] The control unit 12 includes a central processing unit (CPU), a micro-processing unit (MPU), a graphics processing unit (GPU), and other arithmetic processing devices, and performs various information processing, control processing, and the like related to the server 10 by reading and executing programs stored in the storage unit 13. The programs can be deployed so that they are executed on a single computer, located at one site, or distributed across multiple sites and on multiple computers interconnected by a communications network. Although the control unit 12 is described in FIG. 2 as being a single processor, it may also be a multi-processor.

[0024] The storage unit 13 includes memory elements such as RAM (Random Access Memory) and ROM (Read Only Memory), and stores programs, data, etc. required for the control unit 12 to execute processing. The storage unit 13 also temporarily stores data, etc. required for the control unit 12 to execute arithmetic processing.

[0025] The recording medium 14 is a non-volatile, non-transitory recording medium such as a disk-shaped recording medium or a semiconductor memory, and is configured to be detachable from the server 10. The recording medium 14 records various programs executed by the control unit 12. When the server 10 executes the generation process or the recommended question display process, the programs recorded on the recording medium 14 are loaded into the memory unit 13 and executed by the control unit 12.

[0026] The display unit 15 is a liquid crystal display, an organic EL (electroluminescence) display, or the like, and displays various information in accordance with instructions from the control unit 12. The input unit 16 is an input device such as a mouse, keyboard, touch panel, or button, and outputs received operation information to the control unit 12.

[0027] The training data DB31 stores training data used to train a learning goal prediction model corresponding to various conditions. The learning goal prediction model is a predictor that predicts (outputs) learning goal data for each unit period after a first period based on learning history data for each unit period in the first period, and is a trained model generated by machine learning. The training data is data for constructing (generating) the learning goal prediction model.

[0028] 3 shows an example of the data structure of the training data DB 31. The training data DB 31 associates a training ID that identifies each training data, input data that is input in learning the learning goal prediction model, and correct answer data that corresponds to the input data. The input data is composed of condition data and learning history data, and the correct answer data is composed of learning goal data.

[0029] The condition data is data related to conditions corresponding to the learning history data, and includes the subject, field, difficulty level, learning level, target level, etc. The subject is, for example, "Japanese" or "Mathematics," and is the subject that the learner has studied. The field is, for example, "Trigonometric Functions" or "Calculus," and is the field that the learner has studied within each subject. The difficulty level is expressed in stages, for example, using numbers from 1 to 5, and is the difficulty of the problems that the learner has studied. The learning level is a level that represents the learner's academic ability at a predetermined time (for example, one year before the exam), such as "standard deviation 45" or "standard deviation 50." The target level is a level that represents the learner's academic ability at a predetermined time (for example, the exam time), such as "standard deviation 60" or "standard deviation 65."

[0030] In this embodiment, the difficulty level is based on any concept such as the difficulty of the question or the rate of correct answers, and may be set by evaluating each question using Item Response Theory (IRT) or the like, or may be set artificially in advance.

[0031] The learning history data is learning history data for each unit period in the first period, and is a numerical value quantified based on the learner's learning amount (for example, the number of problems solved or the study time). For example, the learning history data in this embodiment is the monthly actual problem count, which indicates the number of problems (hereinafter also referred to as "actual problems") that the learner has worked on on a monthly basis. The learning goal data is learning goal data for each unit period after the first period, and, like the learning history data, is a numerical value quantified based on the learner's learning amount. For example, the learning goal data in this embodiment is the monthly target problem count, which indicates the number of problems (hereinafter also referred to as "target problems") that the learner needs to work on to achieve their goal on a monthly basis.

[0032] The unit period is not limited to a month, such as one month, but may be any unit such as one week.

[0033] The model management DB 32 stores learning objective prediction models in association with information related to the learning objective prediction models. Fig. 4 is a diagram illustrating the model management DB 32. As shown in Fig. 4(a), the model management DB 32 stores learning objective prediction models generated for each condition consisting of subject, field, difficulty level, learning level, target level, etc., in association with a model ID that identifies the learning objective prediction model.

[0034] For example, as shown in the figure, the model management DB 32 stores a learning objective prediction model 50a corresponding to the conditions of the subject "mathematics," the field "trigonometric functions," and the difficulty level "1," in association with the model ID "M001." The management DB 32 also stores a learning objective prediction model 50b corresponding to the conditions of the subject "mathematics," the field "trigonometric functions," and the difficulty level "5," in association with the model ID "M005." The model management DB 32 also stores a learning objective prediction model 50c corresponding to the conditions of the subject "mathematics," the field "calculus," and the difficulty level "1," in association with the model ID "M011." The model management DB 32 also stores a learning objective prediction model 50d corresponding to the conditions of the subject "mathematics," the field "calculus," and the difficulty level "5," in association with the model ID "M015."

[0035] 4(b) shows an example of the data configuration of the model management DB 32. Specifically, the model management DB 32 stores condition data and learning objective prediction models in association with model IDs. The condition data is data related to conditions corresponding to the learning objective prediction models, and includes subjects, fields, difficulty levels, learning levels, target levels, etc. The learning objective prediction models specifically store files of learned learning objective prediction models.

[0036] The learning history DB 33 stores learners in association with their learning history data. Fig. 5 shows an example of the data configuration of the learning history DB 33. The learning history DB 33 associates a learner ID that identifies the learner, the learner's learning history data, and condition data corresponding to the learning history data. The condition data is data related to conditions corresponding to the learning history data, and includes the subject, field, difficulty level, learning level, target level, etc. The learning history data is a numerical value quantified based on the learner's learning amount, and includes, for example, the number of problems solved per month and a problem ID that identifies the problem that the learner tackled.

[0037] In this embodiment, as an example, the terminal device 20 is linked in advance with the learner ID of the learner who uses it, and learning history data of the learner's learning using the terminal device 20 is automatically stored in the learning history DB 33 of the server 10. Furthermore, the condition data may be automatically acquired by the server 10 based on the learning history data, or may be manually input by the learner or the like to the server 10. The method of acquiring the learning history data and the condition data by the server 10 is not limited to these and can be set arbitrarily.

[0038] The question DB 34 stores information about questions in association with question IDs that identify the questions. FIG. 6 shows an example of the data configuration of the question DB 34. The question DB 34 associates question IDs that identify the questions with subjects, fields, difficulty levels, question set names, thumbnail images, pages, and question data. The subject, field, and difficulty level refer to the subject, field, and difficulty level of the question, respectively. The question set name is the name of the question set, such as a reference book or textbook, that contains the question. The thumbnail image is a thumbnail image that corresponds to the question, such as a cover image of the question set that contains the question. The page is the page on which the question is written in the question set. The question data is data for displaying the question.

[0039] The storage format of each DB described above is an example, and other storage formats may be used as long as the relationships between the data are maintained. Each DB is realized by a recording medium such as a hard disk drive (HDD) or a solid state drive (SSD). In this embodiment, the storage unit 13 and the various DBs may be configured as an integrated storage device, or may be separate storage devices. The various DBs may be external storage devices connected to the server 10, and the configuration thereof may be set arbitrarily.

[0040] 7 is a block diagram showing the functional configuration of the server 10. Functionally, the server 10 includes a training data acquisition unit 41, a learning goal prediction model identification unit 42, a learning goal prediction model generation unit 43, a learning history data acquisition unit 44, a learning goal prediction unit 45, a recommended question identification unit 46, a training data DB 31, a model management DB 32, a learning history DB 33, and a question DB 34.

[0041] The training data acquisition unit 41, learning goal prediction model identification unit 42, learning goal prediction model generation unit 43, learning history data acquisition unit 44, learning goal prediction unit 45, and recommended question identification unit 46 are realized by the control unit 12 executing a program.

[0042] The training data acquisition unit 41 acquires, from the training data DB 31, training data to be used for training the learning goal prediction model.

[0043] The learning objective prediction model specifying unit 42 refers to the model management DB 32 based on the condition data associated with the learning history data, and specifies a learning objective prediction model corresponding to the condition.

[0044] Specifically, the learning objective prediction model identifying unit 42 identifies conditions such as the subject, field, difficulty level, learning level, and target level based on condition data associated with the learning history data included in the training data, and identifies a learning objective prediction model corresponding to the conditions by referring to the model management DB 32. Furthermore, the learning objective prediction model identifying unit 42 identifies conditions such as the subject, field, difficulty level, learning level, and target level based on condition data associated with the learning history data acquired by the learning history data acquiring unit 44, which will be described later, and identifies a learning objective prediction model corresponding to the conditions by referring to the model management DB 32.

[0045] Based on the acquired training data, the learning objective prediction model generation unit 43 generates, for each condition, a learning objective prediction model that takes learning history data as input and outputs learning objective data. Specifically, the learning objective prediction model generation unit 43 learns the learning objective prediction model identified by the learning objective prediction model identification unit 42 based on the acquired training data.

[0046] For example, the learning goal prediction model is a neural network related to Seq2Seq, and based on the training data, it inputs learning history data in a time series on a monthly basis for a predetermined period and outputs learning goal data in a time series on a monthly basis for the period after the predetermined period. The server 10 compares the output monthly learning goal data with the learning goal data included in the correct answer data of the training data, and optimizes various parameters of the intermediate layer so that the two are similar to each other, thereby generating the learning goal prediction model.

[0047] The learning objective prediction model generated and learned by the learning objective prediction model generation unit 43 is stored in the model management DB 32.

[0048] The learning history data acquisition unit 44 acquires learning history data from the learning history DB 33 .

[0049] The learning goal prediction unit 45 uses the learning goal prediction model identified by the learning goal prediction model identification unit 42 to predict the learning content required for the learner to achieve the goal from the learning history data acquired by the learning history data acquisition unit 44, and outputs learning goal data.

[0050] Based on the learning goal data output by the learning goal prediction unit 45, the recommended question identification unit 46 refers to the question DB 34 and identifies recommended questions that the learner should work on to achieve their goal. Specifically, the recommended question identification unit 46 first identifies the target number of questions based on the learning goal data. Furthermore, based on the learning history data acquired by the learning history data acquisition unit 44, the recommended question identification unit 46 refers to the learning history DB 33 and identifies the question IDs of proven questions that the learner has already worked on. Then, the recommended question identification unit 46 identifies questions from the question DB 34 that meet the conditions but are not proven questions that the learner has already worked on as recommended questions for each month, so as to achieve the target number of questions.

[0051] Furthermore, the recommended question identification unit 46 acquires information about recommended questions from the question DB 34 and outputs it as recommended question information. Specifically, the recommended question identification unit 46 acquires the question set name, thumbnail image, page, question data, etc. of the recommended questions from the question DB 34 and outputs it as recommended question information.

[0052] In the above configuration, the learning goal prediction model generation unit 43, learning history data acquisition unit 44, and learning goal prediction unit 45 of the server 10 are examples of the model generation unit, history acquisition unit, and goal output unit of the present invention, respectively. Also, the recommended question identification unit 46 of the server 10 is an example of the question output unit and question identification unit of the present invention.

[0053] [Terminal device configuration] 8 is a block diagram showing the hardware configuration of terminal device 20. Terminal device 20 includes a communication unit 21, a control unit 22, a storage unit 23, a display unit 25, and an input unit 26. These components are interconnected via a bus 29.

[0054] The communication unit 21 is a communication unit for communicating with the server 10 via the network 5. Specifically, the communication unit 21 transmits condition data to the server 10 and receives learning objective data and recommended question information from the server 10.

[0055] The control unit 22 includes an arithmetic processing unit such as a CPU, an MPU, a GPU, etc., and performs various information processing, control processing, etc. related to the terminal device 20 by reading and executing programs stored in the storage unit 23. Note that although the control unit 22 is described in Fig. 8 as being a single processor, it may be a multiprocessor.

[0056] The storage unit 23 includes memory elements such as RAM and ROM, and stores programs, data, etc. required for the control unit 22 to execute processing. The storage unit 23 also temporarily stores data, etc. required for the control unit 22 to execute arithmetic processing.

[0057] The display unit 25 is a liquid crystal display, an organic EL display, or the like, and displays various information in accordance with instructions from the control unit 22. The input unit 26 is an input device such as a mouse, keyboard, touch panel, or button, and outputs received operation information to the control unit 22.

[0058] [Learning goal prediction model] 9 is a diagram illustrating a learning objective prediction model. In this embodiment, the server 10 learns the relationship between the learning history data for each unit period in the first period and the learning objective data for each unit period after the first period, and generates a learning objective prediction model.

[0059] Specifically, the learning goal prediction model learns the relationship between monthly learning history data for a specified period and monthly learning goal data from the specified period onwards during the learning period from April when the school year begins to February of the following year when exams are taken.

[0060] The learning goal prediction model is, for example, a neural network generated by deep learning, and Seq2Seq (Sequence to Sequence), a type of RNN (Recurrent Neural Network), can be applied. The learning goal prediction model has an input layer that accepts input of monthly learning history data for a specified period, an intermediate layer that performs calculations based on the input data, and an output layer that outputs monthly learning goal data for the specified period and thereafter based on the calculation results in the intermediate layer.

[0061] The input layer has a plurality of neurons that receive input of learning history data on a monthly basis for a predetermined period, with each neuron receiving the input of learning history data on a monthly basis in chronological order.

[0062] The intermediate layer is composed of multiple neurons called LSTM (Long Short-Term Memory) blocks. Each neuron in the intermediate layer performs calculations based on the learning history data input to the corresponding neuron in the input layer. The LSTM block performs calculations while referencing the calculation results of the previous LSTM block (neuron), i.e., the calculation results based on the learning history data that appears immediately before. This allows the learning goal prediction model to predict learning goal data for each month from a specified period onwards based on chronological order.

[0063] The output layer has multiple neurons that sequentially output monthly learning goal data from a predetermined period onward in chronological order. Each neuron predicts the monthly learning goal data to be output based on the calculation results of the corresponding neuron in the intermediate layer.

[0064] As shown in Figure 9, the neural network for Seq2Seq is divided into a network part that functions as an encoder and a network part that functions as a decoder.<EOS(End of String)> is a reserved word that indicates the end of data during learning or prediction processing. In Figure 9, it is assumed that the learning history data for April, May, June, and July are input to the Encoder in chronological order.

[0065] The Encoder is a network unit that receives input of learning history data that represents the amount of learning in a time series in monthly units for a predetermined period and outputs features. For example, as shown in FIG. 9, the Encoder outputs the amount of learning in April to <eos>as input and outputs features. The Decoder is a network unit that outputs learning goal data representing the amount of learning on a monthly basis from the specified period onwards in chronological order based on the features output from the Encoder. For example, as shown in Figure 9, the Decoder outputs the amount of learning from August to February based on the features output from the Encoder.

[0066] The Decoder predicts monthly learning goal data for a specified period and beyond based on the chronological learning history data input to the Encoder. For example, if learning history data from April to July is input to the Encoder in chronological order, the Decoder first predicts the learning goal data for August, which is after July. The Decoder predicts the learning goal data for August and inputs the predicted August learning goal data into the next LSTM block to predict the learning goal data for September. The Decoder repeats this process, sequentially predicting learning goal data from October onwards, and finally predicts monthly learning goal data. For example, if learning history data for each of April to July is input into the learning goal prediction model, it can output learning goal data for each of August to February of the following year.

[0067] Note that Seq2Seq is an example of a learning goal prediction model, and the server 10 may generate other neural networks such as a Transformer as the learning goal prediction model. Furthermore, the learning goal prediction model is not limited to a neural network, and may be a model based on other learning algorithms, such as a decision tree, a random forest, or an SVM (Support Vector Machine).

[0068] [Training data] The server 10 generates a learning goal prediction model using, for example, training data stored in the training data DB 31. The training data is a combination of monthly learning history data for a predetermined period and monthly learning goal data for the period after the predetermined period, in association with each other. The training data is generated based on a large amount of learning history data collected from successful learners. The training data may also be data created separately and manually.

[0069] FIG. 10 is a diagram illustrating training data. FIG. 10(a) is an example of a learning history graph 52. The learning history graph 52 is a graph showing monthly learning history data of a learner (Mr. A) for a learning period from April, when the school year begins, to February of the following year, when the exam is taken. The horizontal axis of the graph indicates the learner's learning period (for example, April to February of the following year). The vertical axis of the graph indicates the number of actual questions, which is one type of learning history data. Note that the learning history data is not limited to the number of actual questions, and can be set arbitrarily to any other data such as study time, score, or a combination thereof.

[0070] Furthermore, the learning history data may be quantified by combining the number of questions completed and the level of difficulty. The level of difficulty may be set, for example, for each level of correct answer rate for each question. For example, a question with a correct answer rate of 0-20% may be assigned a level of difficulty of "5," a question with a correct answer rate of 21-40% may be assigned a level of difficulty of "4," a question with a correct answer rate of 41-60% may be assigned a level of difficulty of "3," a question with a correct answer rate of 61-80% may be assigned a level of difficulty of "2," and a question with a correct answer rate of 81-100% may be assigned a level of difficulty of "1."

[0071] The learning history data can be quantified, for example, by multiplying the difficulty level by the number of questions completed. If the number of questions completed is set to 10 for a difficulty level of 5, the learning history data is 50 (5 x 10), and if the learning amount is set to 10 for a difficulty level of 4, the learning history data is 40 (4 x 10). The learning history data may be quantified based on any calculation formula.

[0072] Based on the learner's learning history data, training data can be created by combining the period and the number of questions completed. As shown in Figure 10(a), the learning history data of successful candidate A from April to February of the following year is displayed in learning history graph 52. Ten sets of training data can be created from A's learning history data from April to February of the following year. Specifically, as shown in Figure 10(b), the first set of training data is created using the learning history data from April as input data and the learning history data from May to February of the following year as correct answer data. Next, the second set of training data is created using the learning history data from April and May as input data and the learning history data from June to February of the following year as correct answer data. Based on these combinations of learning history data, a tenth set of training data is finally created using the learning history data from April to January of the following year as input and the learning history data from February of the following year as correct answer data.

[0073] The training data may include, as input data, condition data such as the subject, field, and difficulty of the problem tackled by the learner, the learner's learning level, and target level.

[0074] In this way, multiple sets of training data can be generated from the learning history data of one successful candidate (a learner who has achieved their goal). Therefore, training data for learning to be used in a learning goal prediction model can be generated from the learning history data of many successful learners.

[0075] Next, returning to Figure 9, the process of generating a learning goal prediction model will be described in detail. The server 10 inputs learning history data (input data) for each month during a predetermined training period into the learning goal prediction model in chronological order, and obtains learning goal data (output data) for each month after the predetermined period from the output layer. The server 10 compares the correct answer data included in the training data with the output data output from the learning goal prediction model, and optimizes various parameters such as weights used in calculations in the intermediate layer so that the two approximate each other.

[0076] As a result, the server 10 generates a learning objective prediction model. The server 10 stores the generated learning objective prediction model in the model management DB 32. In this embodiment, the server 10 predicts monthly learning objective data using the above learning objective prediction model, and outputs monthly learning objective data for a predetermined period and thereafter in chronological order, in response to monthly learning history data for the predetermined period.

[0077] [Target screen] The goal screen is a screen that displays the learning content necessary to achieve a goal. First, the subject-specific goal screen will be described. FIG. 11 is an example of a subject-specific goal screen. The subject-specific goal screen is a screen that is displayed on the terminal device 20, and includes a condition selection combo box 60, a graph display field 61, and a question display field 62. The condition selection combo box 60 is a combo box for selecting a subject or field. The graph display field 61 is a display field that displays a graph showing learning history data and learning goal data corresponding to the selected condition. The question display field 62 is a display field that displays information about recommended questions.

[0078] When the terminal device 20 receives a subject selection operation in the condition selection combo box 60, it transmits to the server 10 a learner ID that identifies the learner using the terminal device 20 and condition data that indicates the selected subject. The server 10 acquires learning history data for a predetermined period (e.g., learning history data from April to May) from the learning history DB 33 based on the learner ID and the condition data that indicates the subject received from the terminal device 20. The server 10 identifies a learning objective prediction model that corresponds to the selected subject by referring to the model management DB 32 based on the received condition data.

[0079] The server 10 inputs the learning history data acquired from the learning history DB 33 in chronological order into the identified learning goal prediction model, and outputs the learning goal data from a predetermined period onward (for example, learning goal data from June to February of the following year) in chronological order. The server 10 creates a graph showing the actual number of questions and the target number of questions based on the output learning goal data and the learning history data acquired from the learning history DB 33. The graph showing the actual number of questions may be created based on the latest learning history data stored in the learning DB 33, including, for example, the learning history data for June. Furthermore, the server 10 identifies recommended questions based on the output learning goal data by referring to the question DB 34 so as to achieve the target number of questions, and acquires recommended question information related to the recommended questions. The server 10 then transmits the graph information related to the created graph and the recommended question information to the terminal device 20.

[0080] The terminal device 20 receives the graph information sent from the server 10 and displays a graph showing the actual number of questions and the target number of questions in the graph display field 61. As shown in the figure, a vertical bar graph showing the relationship between the number of questions and the period is displayed in the graph display field 61. The horizontal axis of the vertical bar graph indicates the period (for example, April to February of the following year). The vertical axis of the vertical bar graph indicates the number of questions that quantify the amount of learning. Based on the learning history data included in the received graph information, the terminal device 20 creates a graph of the actual number of questions, represented by solid black, for each month and displays it in the graph display field 61. Based on the learning goal data included in the received graph information, the terminal device 20 also creates and displays a graph of the target number of questions, represented by diagonal lines, for each month.

[0081] Furthermore, the terminal device 20 receives recommended question information transmitted from the server 10, and displays the name of the question set, thumbnail images, pages, etc. related to the recommended questions in the question display field 62.

[0082] Although the graph displayed in the graph display field 61 is in the form of a vertical bar graph, it is not limited to this and may be in other graph forms.

[0083] Furthermore, the goal screen may be updated and displayed with the learning content necessary to achieve the goal at any timing, such as weekly or monthly. For example, if the learning content necessary to achieve the goal is updated monthly on the subject-specific goal screen described above, the server 10 inputs the learning history data from April to June into the identified learning goal prediction model in chronological order at any timing when the next month arrives, and outputs the learning goal data from July to February of the following year. In other words, by inputting the updated learning history data, new learning goal data is output. The server 10 and the terminal device 20 create and display a subject-specific goal screen with the updated learning content necessary to achieve the goal based on the updated learning history data and the new learning goal data.

[0084] Next, the field-specific goal screen will be described. Fig. 12 is an example of the field-specific goal screen. The field-specific goal screen is a screen displayed on the terminal device 20, and includes a condition selection combo box 60, graph display fields 61a to 61c for each level of difficulty, and question display fields 62a to 62c for each level of difficulty. The condition selection combo box 60 is a combo box for selecting a subject and a field. The graph display fields 61a to 61c are display fields that display graphs showing learning history data and learning goal data corresponding to the selected subject and field, sorted by level of difficulty. The question display fields 62a to 62c are display fields that display information about recommended questions, sorted by level of difficulty.

[0085] When the terminal device 20 receives a selection operation for a subject and a field in the condition selection combo box 60, it transmits to the server 10 a learner ID that identifies the learner using the terminal device 20 and condition data that indicates the selected subject and field. The server 10 acquires learning history data for a predetermined period (e.g., learning history data from April to May) by difficulty level from the learning history DB 33 based on the learner ID and the condition data that indicates the subject and field received from the terminal device 20. The server 10 identifies learning objective prediction models that correspond to the subjects, fields, and each level of difficulty by referring to the model management DB 32 based on the subjects, fields, and each level of difficulty indicated in the received condition data.

[0086] Specifically, when the subject and field indicated by the received condition data are "mathematics" and "calculus," respectively, the server 10 refers to the model DB 32 to identify the learning objective prediction model 50c corresponding to the subject "mathematics," the field "calculus," and the difficulty level "1," as shown in Fig. 4(a). The server 10 also identifies a learning objective prediction model corresponding to the subject "mathematics," the field "calculus," and the difficulty level "2," a learning objective prediction model corresponding to the subject "mathematics," the field "calculus," and the difficulty level "3," a learning objective prediction model corresponding to the subject "mathematics," the field "calculus," and the difficulty level "4," and a learning objective prediction model 50d corresponding to the subject "mathematics," the field "calculus," and the difficulty level "5."

[0087] The server 10 inputs the learning history data of each level of difficulty acquired from the learning history DB 33 in chronological order into the learning goal prediction model corresponding to the subject, field, and level of difficulty, and outputs the learning goal data from a predetermined period onward (for example, learning goal data from June to February of the following year) in chronological order.

[0088] Specifically, the server 10 inputs learning history data corresponding to the subject "Mathematics," the field "Calculus," and the difficulty level "1" obtained from the learning history DB 33 into the learning goal prediction model 50c corresponding to the subject "Mathematics," the field "Calculus," and the difficulty level "1," and outputs learning goal data from a predetermined period onward. Similarly, the server 10 inputs learning history data corresponding to the difficulty level "2" into the learning goal prediction model corresponding to the difficulty level "2," and outputs the learning goal data. The server 10 also inputs learning history data corresponding to the difficulty level "3" into the learning goal prediction model corresponding to the difficulty level "3," and outputs the learning goal data. The server 10 also inputs learning history data corresponding to the difficulty level "4" into the learning goal prediction model corresponding to the difficulty level "4," and outputs the learning goal data. The server 10 also inputs learning history data corresponding to the difficulty level "5" into the learning goal prediction model 50d corresponding to the difficulty level "5," and outputs the learning goal data.

[0089] The server 10 then creates a graph showing the number of actual questions and the number of target questions for each level of difficulty based on the output learning goal data for each level of difficulty and the learning history data for each level of difficulty acquired from the learning history DB 33. The graph showing the number of actual questions may be created based on the latest learning history data stored in the learning DB 33.

[0090] Furthermore, the server 10 identifies recommended questions for each level of difficulty so as to achieve the target number of questions for each level of difficulty by referring to the question DB 34 based on the output learning goal data for each level of difficulty, and acquires recommended question information for each level of difficulty.The server 10 then transmits to the terminal device 20 the graph information for each level of difficulty relating to the graphs created for each level of difficulty and the recommended question information for each level of difficulty.

[0091] . The terminal device 20 receives the graph information for each level of difficulty transmitted from the server 10, and displays graphs representing the number of completed questions and the target number of questions for each level of difficulty in the graph display fields 61a to 61c, respectively. The format of the graphs is the same as that of the subject-specific goal screen, so a description thereof will be omitted. The terminal device 20 also receives the recommended question information for each level of difficulty transmitted from the server 10, and displays the recommended questions for each level of difficulty in the question display fields 62a to 62c, respectively.

[0092] Although the question display columns 62a to 62c display recommended questions, they may also display the names of problem sets, thumbnail images, and pages related to the recommended questions, or may display them together with the recommended questions, as in the subject-specific goal screen. The configuration of the question display columns 62a to 62c can be set arbitrarily.

[0093] Furthermore, the graphs displayed in the graph display fields 61a to 61c are in the form of vertical bar graphs, but are not limited to this and may be in other graph forms.

[0094] Furthermore, like the subject-specific goal screen, the field-specific goal screen may also be updated and displayed with the learning content required to achieve the goal at any time.

[0095] [Model generation process] Next, we will explain the model generation process for performing machine learning to generate the learning goal prediction model 50. Fig. 13 is a flowchart of the model generation process by the server 10. This process is realized by the server 10 executing a program prepared in advance.

[0096] The server 10 acquires training data in which input data and correct answer data are associated with each other from the training data DB 31 (step S101). The input data includes condition data including a subject, field, difficulty level, learning level, and target level, and learning history data including the number of actual questions by month.

[0097] Based on the acquired input data, the server 10 generates a learning objective prediction model for each condition included in the condition data, which uses the learning history data as input and the learning objective data as output (step S102).

[0098] For example, the learning goal prediction model is a neural network related to Seq2Seq, which takes monthly time-series learning history data for a predetermined period as input and monthly time-series learning goal data for the period after the predetermined period as output. Server 10 compares the output monthly learning goal data with the learning goal data included in the correct answer data, and optimizes various parameters in the intermediate layer so that the two are similar to each other, thereby generating the learning goal prediction model.

[0099] Specifically, the server 10 generates a learning objective prediction model 50a corresponding to the subject "mathematics", the field "trigonometric functions", and the difficulty level "1" as shown in Fig. 4(a) by performing learning using training data whose condition data includes the conditions of the subject "mathematics", the field "trigonometric functions", and the difficulty level "1". Similarly, the server 10 generates a learning objective prediction model 50d corresponding to the subject "mathematics", the field "calculus", and the difficulty level "5" as shown in Fig. 4(a) by performing learning using training data whose condition data includes the conditions of the subject "mathematics", the field "calculus", and the difficulty level "5".

[0100] Then, the server 10 stores the generated learning objective prediction model in the model management DB 32 (step S103), and ends the series of processes. Specifically, the server 10 assigns a model ID to the generated learning objective prediction model, and stores the file of the learning objective prediction model as one record in the model management DB 32 in association with the assigned model ID.

[0101] [Recommended question display processing] Next, we will explain the recommended question display process, which displays a graph showing the number of completed questions and the number of target questions, and information about recommended questions on the target screen. Figure 14 is a flowchart of the recommended question display process. This process is realized by the server 10 and the terminal device 20 each executing a program prepared in advance.

[0102] First, the recommended question display process when only a subject is selected in the condition selection combo box 60 on the target screen will be described. The learner selects a subject in the condition selection combo box 60 on the target screen displayed on the terminal device 20. When the terminal device 20 accepts a subject selection operation in the condition selection combo box 60, it transmits a learner ID that identifies the learner and condition data that indicates the selected subject to the server 10 (step S301). The server 10 receives the learner ID and condition data from the terminal device 20 (step S302).

[0103] The server 10 acquires learning history data for a predetermined period from the learning history DB 33 based on the learner ID and condition data received from the terminal device 20 (step S303). Then, the server 10 identifies a learning objective prediction model corresponding to the subject by referring to the model management DB 32 based on the subject indicated by the condition data (step S304).

[0104] The server 10 inputs the learning history data acquired from the learning history DB 33 in chronological order into the learning goal prediction model corresponding to the subject, and acquires the learning goal data from a predetermined period onward by outputting it in chronological order (step S305).The server 10 then creates a graph showing the actual number of questions and the target number of questions based on the output learning goal data and the learning history data acquired from the learning history DB 33, and acquires the graph information (step S306).

[0105] Next, the server 10 performs a recommended question identification process to identify recommended questions based on the learning objective data (step S400). FIG. 15 is a flowchart of the recommended question identification process. In the recommended question identification process, the server 10 identifies the target number of questions based on the output learning objective data (step S401). The server 10 also identifies the question IDs of the proven questions that the learner has already tackled based on the acquired learning history data (step S402). The server 10 then identifies, as recommended questions for each month from the question DB 34, questions that correspond to the subject but are not proven questions that the learner has already tackled, so that the target number of questions is met (step S403). The server 10 then acquires recommended question information related to the recommended questions from the question DB 34 (step S404). This causes the server 10 to end the recommended question identification process and proceed to step S307 in the recommended question display process.

[0106] The server 10 transmits the acquired graph information and recommended question information to the server 10 (step S307). The terminal device 20 receives the graph information and recommended question information transmitted from the server 10, and displays on the goal screen, as shown in Fig. 11, a graph indicating the number of completed questions and the target number of questions, as well as the name of the question set, thumbnail images, and pages related to the recommended questions (step S308). This causes the server 10 to end the recommended question display process.

[0107] Next, a description will be given of the recommended question display process when a subject and field are selected in the condition selection combo box 60 on the target screen. The learner selects a subject and field in the condition selection combo box 60 on the target screen displayed on the terminal device 20. When the terminal device 20 accepts the selection operation of the subject and field in the condition selection combo box 60, it transmits a learner ID that identifies the learner and condition data that indicates the selected subject and field to the server 10 (step S301). The server 10 receives the learner ID and condition data from the terminal device 20 (step S302).

[0108] The server 10 acquires learning history data for a predetermined period by difficulty level from the learning history DB 33 based on the learner ID and condition data received from the terminal device 20 (step S303).Then, the server 10 identifies a learning goal prediction model corresponding to each subject, field, and difficulty level by referring to the model management DB 32 based on the subject, field, and each difficulty level indicated by the condition data (step S304).

[0109] Server 10 acquires the learning goal data for each difficulty level acquired from learning history DB 33 by inputting the learning history data for each difficulty level in chronological order into a learning goal prediction model corresponding to the subject, field, and each difficulty level, and outputs the learning goal data for each difficulty level after a predetermined period in chronological order (step S305). Server 10 then creates a graph showing the number of actual questions and the number of target questions for each difficulty level based on the output learning goal data for each difficulty level and the learning history data for each difficulty level acquired from learning history DB 33, and acquires graph information for each difficulty level (step S306).

[0110] Next, the server 10 performs a recommended question identification process to identify recommended questions based on the learning objective data (step S400). In the recommended question identification process, the server 10 identifies the target number of questions for each difficulty level based on the output learning objective data for each difficulty level (step S401). The server 10 also identifies, for each difficulty level, the question IDs of the proven questions that the learner has already tackled based on the acquired learning history data for each difficulty level (step S402). The server 10 then identifies, for each difficulty level and for each month, questions that correspond to the subject and field and are not proven questions that the learner has already tackled, from the question DB 34, so as to meet the target number of questions (step S403). The server 10 then acquires recommended question information related to recommended questions for each difficulty level from the question DB 34 (step S404). The server 10 then terminates the recommended question identification process and proceeds to step S307 in the recommended question display process.

[0111] The server 10 transmits the acquired graph information for each level of difficulty and recommended question information to the server 10 (step S307). The terminal device 20 receives the graph information for each level of difficulty and the recommended question information transmitted from the server 10, and displays a graph indicating the number of completed questions and the target number of questions, and recommended questions, for each level of difficulty on the target screen, as shown in Fig. 12 (step S308). This causes the server 10 to end the recommended question display process.

[0112] In the recommended question display process described above, the subject and field are selected in the condition selection combo box 60 on the goal screen, and graphs and recommended questions are displayed for each level of difficulty, but the present invention is not limited to this, and any of the subjects, fields, difficulty levels, learning levels, and goal levels listed as conditions may be selected. Furthermore, the method for displaying graphs and recommended questions can be set arbitrarily.

[0113] For example, if a learner selects the learning level for April and the target level at the time of the exam in addition to the subject and field, the server 10 can use a learning goal prediction model corresponding to the subject, field, learning level, and target level to output learning goal data for each level of difficulty so that a learner whose learning level in April is "standard deviation 50" will reach the target level "standard deviation 60" at the time of the exam, and provide this to the learner along with recommended questions.

[0114] Furthermore, in the above recommended question display process, the server 10 acquires graph information based on the output learning objective data and transmits the graph information to the terminal device 20, but the present invention is not limited to this, and the learning objective data may be transmitted to the terminal device 20. In this case, the terminal device 20 may display any graph based on the learning objective data it receives.

[0115] The learning goal prediction system 100 of this embodiment can provide learners with the amount of study required to achieve their goals and recommended questions for each subject area. Furthermore, the learning goal prediction system 100 can provide learners with the amount of study required to achieve their goals and recommended questions for each difficulty level of the subject area. In other words, by generating and using a learning goal prediction model corresponding to any condition, the learning goal prediction system 100 can provide learners with the amount of study required to achieve their goals and recommended questions for each condition. This allows learners to easily recognize, for example, the type and amount of problems they should work on for each difficulty level of the "trigonometric functions" subject area. In this way, the learning goal prediction system 100 of this embodiment can visualize and quantify intuitive concepts in learning, such as "it's important to repeat basic problems" and "gradually work on applied problems," by providing each learner with the amount of study required and recommended questions for each difficulty level of the subject area.

[0116] The method of this embodiment may also be used to generate learning goal prediction models for other tests, including language tests such as TOEIC (registered trademark), certification tests (e.g., Test in Practical English Proficiency), qualification tests, or promotion tests. [Explanation of symbols]

[0117] 5. Network 10 Servers 20 Terminal equipment 31 Training Data DB 32 Model Management DB 33 Learning History DB 34 Problem DB 41 Training data acquisition section 42 Learning Objective Prediction Model Identification 43 Learning Objective Prediction Model Generation Unit 44 Learning history data acquisition unit 45 Learning Objective Prediction Section 46 Recommendation problem identification department 100 Learning Objective Prediction System< / eos>

Claims

1. a model acquisition unit that acquires a trained model that has learned the relationship between learning history data for each unit period that indicates the learning content of the learner in a first period and learning goal data for each unit period that indicates the learning content necessary for achieving the goal from the first period onwards; a history acquisition unit that acquires, in chronological order, learning history data associated with conditions including a learning level that indicates the academic ability of a learner; a goal output unit that inputs the learning history data acquired by the history acquisition unit into a trained model corresponding to a condition including the learning level, and outputs the learning goal data in time series; a problem output unit that outputs, as recommended problem information, information about recommended problems that the learner should tackle in order to achieve the goal, based on the learning goal data output by the goal output unit; a model generation unit that uses training data to learn, for each condition including the learning level, the relationship between learning history data for each unit period that indicates the learning content of the learner in the first period and learning goal data for each unit period that indicates the learning content necessary to achieve the goal from the first period onwards, and generates the trained model; a training data generation unit that generates training data by dividing the learning history data of a learner who meets the conditions including the learning level and has achieved the goal into unit periods, and treating the learning history data of the first half period as input data and the learning history data of the second half period as correct answer data; The training data generation unit is an information processing device that generates multiple sets of training data from learning history data of one learner.

2. The information processing device according to claim 1 , wherein the conditions include a field of study content.

3. The information processing device according to claim 1 , wherein the conditions include a difficulty level of questions included in the learning content.

4. The conditions include the field of the learning content and the difficulty level of the questions included in the learning content, a question information storage unit that stores question information that associates the questions with the difficulty levels of the questions; a problem identification unit that identifies the recommended problems for each of the fields and difficulty levels of the problems from the problem information storage unit based on the learning objective data, The information processing device according to claim 3 , wherein the question output unit outputs the recommended questions identified for each of the fields and the difficulty levels of the questions.

5. the question information storage unit further stores a medium containing the question and a location of the question on the medium in association with each other; The information processing device according to claim 4 , wherein the question output unit outputs the recommended question, or a medium containing the recommended question and a location of the recommended question in the medium.

6. An information processing method executed by an information processing device, a training data generation process for dividing learning history data, which corresponds to conditions including a learning level that indicates the academic ability of the learner and indicates the learning content of the learner who has achieved the goal, into unit periods, and by using the learning history data of the first half period as input data and the learning history data of the second half period as correct answer data, thereby generating multiple sets of training data from the learning history data of one learner; a model generation process for generating a trained model by using training data to learn the relationship between learning history data for each unit period indicating the learning content of the learner in a first period and learning goal data for each unit period indicating the learning content necessary for achieving the goal in the first period and thereafter, for each condition including the learning level; a model acquisition step of acquiring the trained model; a history acquisition step of acquiring, in chronological order, learning history data associated with conditions including the learning level; a goal output step of inputting the learning history data acquired by the history acquisition step into a trained model corresponding to a condition including the learning level, and outputting the learning goal data in time series; a problem output step of outputting, as recommended problem information, information on recommended problems that the learner should tackle in order to achieve the goal, based on the learning goal data output by the goal output step; An information processing method comprising:

7. A program executed by a computer, a model acquisition unit that acquires a trained model that has learned the relationship between learning history data for each unit period that indicates the learning content of the learner in a first period and learning goal data for each unit period that indicates the learning content necessary for achieving the goal from the first period onwards; a history acquisition unit that acquires, in chronological order, learning history data associated with conditions including a learning level that indicates the academic ability of the learner; a goal output unit that inputs the learning history data acquired by the history acquisition unit into a trained model corresponding to a condition including the learning level, and outputs learning goal data corresponding to the condition in time series; a problem output unit that outputs, as recommended problem information, information about recommended problems that the learner should work on to achieve the goal, based on the learning goal data output by the goal output unit; a model generation unit that uses training data to learn, for each condition including the learning level, the relationship between learning history data for each unit period that indicates the learning content of the learner in the first period and learning goal data for each unit period that indicates the learning content necessary for achieving the goal in the first period and thereafter, and generates the trained model; making the computer function as a training data generation unit that generates training data by dividing the learning history data of a learner who meets the conditions including the learning level and has achieved the goal into unit periods, and using the learning history data of the first half period as input data and the learning history data of the second half period as correct answer data; The training data generation unit is a program that generates multiple sets of training data from the learning history data of one learner.

Images