JP2020115175A - Information processor, method for processing information, and program - Google Patents

Abstract

To provide an information processor, for example, which can appropriately estimate a learning situation.SOLUTION: An information processor 1 includes: an acquisition unit for acquiring answer information showing answers that a learner wrote by hand to questions; and a classification unit for classifying the handwriting of the answer shown in the answer information. The classification unit desirably uses a classifier which has learned the feature amount of the handwriting of an exemplary answer by a person, for the classification. Classifying the handwriting of an answer to a question allows an appropriate estimation of the way of tackling a problem of the learner in answering, that is, the situation of the learner.SELECTED DRAWING: Figure 1

Description

The present invention relates to an information processing device, an information processing method, and a program.

Various techniques for supporting learners have been proposed in fields such as education. For example, in Patent Document 1, a learning management device that presents a problem to a learner, stores a time zone and a place where the learner learned the problem in the past, and outputs the same so that the same problem can be learned in the same situation. A learning management device and the like are disclosed.

Patent No. 6398538

However, the invention according to Patent Document 1 considers the environment at the time of learning, and does not consider how the learner himself/herself approaches the problem, that is, the situation of the learner himself/herself.

In one aspect, an object is to provide an information processing device or the like that can appropriately estimate a learning situation.

In one aspect, the information processing device includes an acquisition unit that acquires answer information indicating an answer handwritten by the learner with respect to the question, and a classification unit that classifies the handwriting of the answer indicated by the answer information. It is characterized by

In one aspect, the learning situation can be properly estimated.

It is a schematic diagram which shows the structural example of a learning condition estimation system. It is a block diagram which shows the structural example of a server. It is explanatory drawing which shows an example of a record layout of learner DB and answer DB. It is explanatory drawing regarding a handwriting classification process. It is explanatory drawing regarding the production|generation process of a classifier. It is explanatory drawing regarding the output processing of a re-taking problem. It is a flow chart which shows a processing procedure of generation processing of a classifier. It is a flowchart which shows the procedure of a handwriting classification process. FIG. 9 is an explanatory diagram related to a generation process of a classifier according to the second embodiment. 9 is a flowchart showing a procedure of a classifier generation process according to the second embodiment. FIG. 13 is an explanatory diagram related to handwriting classification processing according to the third embodiment. 9 is a flowchart showing a procedure of handwriting classification processing according to the third embodiment. It is a functional block diagram which shows operation|movement of the server of the form mentioned above.

Hereinafter, the present invention will be described in detail with reference to the drawings showing an embodiment thereof.
(Embodiment 1)
FIG. 1 is a schematic diagram showing a configuration example of a learning situation estimation system. In the present embodiment, a learning situation estimation system that presents a question to a learner and estimates the learning situation based on the handwriting of an answer handwritten by the learner for the presented question will be described. The learning situation estimation system includes an information processing device 1 and a terminal 2. Each device is communicatively connected via a network N such as the Internet.

The information processing device 1 is an information processing device capable of various information processing and information transmission/reception, and is, for example, a server device, a personal computer, or the like. In the present embodiment, the information processing device 1 is assumed to be a server device, and is read as server 1 for simplicity. The server 1 outputs the question presented to the learner to the terminal 2, and acquires from the terminal 2 the answer information indicating the answer handwritten by the learner with respect to the output question. The answer information is, for example, image data obtained by optically reading a question paper (paper medium) in which the answer is written by hand, or stroke data including a stroke group electronically input by a means such as an electronic pen. In the following description, for convenience, the answer information will be described as image data.

The server 1 classifies the handwriting of the answer of the learner from the acquired answer information. For example, the server 1 classifies the handwritten answers according to whether or not the answers are scribbled. If it is scribbled, it is likely that the learner did not take the problem seriously and wrote the answer randomly. Therefore, the server 1 classifies whether or not the handwriting of the answer is scribbled so that the learning status can be estimated as to how seriously the learner was learning.

Specifically, as will be described later, the server 1 learns the feature amount of the handwriting of the answer example handwritten by a human by machine learning to generate a classifier (learned model), and uses the classifier. , Classify whether the learner's handwriting is scribbled. Further, the server 1 determines the learning efficiency of the learner according to the classification result of the handwriting, and outputs the re-taking problem to the terminal 2 according to the determined learning efficiency.

The terminal 2 is a terminal device operated by a learner (for example, a student) or another user (for example, a teacher), and is, for example, a personal computer or a tablet terminal. As described above, the server 1 acquires the answer information from the terminal 2. For example, a terminal 2a such as a personal computer used by a teacher acquires image data obtained by reading a question sheet from a scanner (not shown) and transmits it to the server 1 as answer information. When the stroke data is used as the answer information, for example, the terminal 2b such as the tablet terminal used by the learner himself/herself displays the question, accepts the handwriting input of the answer by the electronic pen or the like, and transmits the stroke data to the server 1. .. The server 1 classifies the handwriting based on the answer information acquired from the terminal 2 and determines whether the handwriting has high learning efficiency.

FIG. 2 is a block diagram showing a configuration example of the server 1. The server 1 includes a control unit 11, a main storage unit 12, a communication unit 13, and an auxiliary storage unit 14.
The control unit 11 has an arithmetic processing unit such as one or more CPUs (Central Processing Units), MPUs (Micro-Processing Units), and GPUs (Graphics Processing Units), and stores the programs P stored in the auxiliary storage unit 14. By reading and executing, various information processing, control processing, and the like are performed. The main storage unit 12 is a temporary storage area such as an SRAM (Static Random Access Memory), a DRAM (Dynamic Random Access Memory), and a flash memory, and temporarily stores data necessary for the control unit 11 to execute arithmetic processing. Remember. The communication unit 13 is a communication module for performing processing related to communication, and transmits/receives information to/from the outside.

The auxiliary storage unit 14 is a non-volatile storage area such as a large-capacity memory or a hard disk, and stores the program P and other data necessary for the control unit 11 to execute processing. The auxiliary storage unit 14 also stores a learner DB 142 and an answer DB 143. The learner DB 142 is a database that stores information about each learner. The answer DB 143 is a database that stores information about answers by learners.

The auxiliary storage unit 14 may be an external storage device connected to the server 1. Moreover, the server 1 may be a multi-computer including a plurality of computers, or may be a virtual machine virtually constructed by software.

Further, in the present embodiment, server 1 is not limited to the above configuration, and may include, for example, an input unit that receives an operation input, a display unit that displays an image, and the like. Further, the server 1 includes a reading unit that reads a portable storage medium P1 such as a CD (Compact Disk)-ROM and a DVD (Digital Versatile Disc)-ROM, and reads the program P from the portable storage medium P1 and executes the program P. You can Alternatively, the server 1 may read the program P from the semiconductor memory P2.

FIG. 3 is an explanatory diagram showing an example of a record layout of the learner DB 142 and the answer DB 143.
The learner DB 142 includes a learner ID column, a name column, an affiliation column, and an age column. The learner ID column stores learner IDs for identifying each learner. The name column, the affiliation column, and the age column respectively store the learner's name, affiliation, and age in association with the learner ID.

The answer DB 143 includes an answer ID column, a learner ID column, an answer date/time column, a question column, a field column, a correct/wrong column, and a handwriting column. The answer ID column stores the answer ID for identifying the answer by the learner. The learner ID column, the answer date/time column, the question column, the field column, the correct/wrong column, and the handwriting column are each associated with the answer ID, and the learner ID of the learner who has answered, the date and time of the answer (answer time), the question, The problem field (for example, subject), correctness of the answer, and handwriting data of the answer are stored.

FIG. 4 is an explanatory diagram regarding the handwriting classification processing. In FIG. 4, an image area corresponding to an answer (hereinafter referred to as an “answer area” for convenience) is extracted from the image data obtained by reading the question paper, the handwriting is classified, and the learning efficiency of the learner is determined. The outline is schematically illustrated. The outline of the present embodiment will be described with reference to FIG.

As described above, when the server 1 acquires the image data obtained from the terminal 2 using the image data of the question paper (paper medium) in which the answer is written by hand as the answer information, the server 1 uses the acquired image data as a blank sheet. And the answer area, which is the difference between the two, is extracted.

The server 1 inputs the extracted answer area to the classifier 141 and classifies the handwriting of the answer entered by the learner. As described above, the classifier 141 is a learned model in which the characteristics of human handwriting have been learned by machine learning, and is, for example, a neural network related to CNN (Convolution Neural Network).

FIG. 5 is an explanatory diagram regarding the generation processing of the classifier 141. FIG. 5 conceptually illustrates how the classifier 141 is generated by machine learning. The processing content of machine learning will be described with reference to FIG.

As described above, the server 1 as the classifier 141 generates a neural network related to CNN by machine learning. The classifier 141 includes an input layer that receives input of image data (for example, a pixel value of each pixel) of the answer region, an intermediate layer (hidden layer) that extracts a feature amount of the image data input to the input layer, and an intermediate layer. An output layer that classifies the handwriting on the basis of the feature amount extracted in and outputs the classification result. Each layer is composed of one or more neurons. The intermediate layer has a configuration in which a convolution layer and a pooling layer are alternately connected, and extracts the image feature amount while compressing the pixel value of each pixel input from the input layer. The output layer classifies the handwriting of the answer based on the image feature amount extracted by the intermediate layer, and outputs the classification result. The classification result output from the output layer may be a continuous probability value (for example, a value from “0” to “1”), or a discrete value (for example, “0” or “1”). Value).

Although the classifier 141 is described as a CNN in the present embodiment, the classifier 141 may be a learned model in which the characteristics of human handwriting have been learned. For example, a neural network other than CNN, SVM( It may be a trained model generated by another learning algorithm such as Support Vector Machine) or random forest.

In the present embodiment, the server 1 performs machine learning for learning the feature amount of a character string (handwriting) written in scribble, and classifies whether or not the character string entered as an answer is a scribble. 141 is generated. Specifically, the server 1 associates the correct answer value of the classification result when the image data (answer information) showing the answer example handwritten by the human is classified as scribbled. A classifier 141 is generated using the data. The server 1 inputs each image data included in the teacher data to the classifier 141, and acquires the classification result obtained by classifying whether or not it is scribbled as an output. The server 1 compares the obtained classification result with the correct answer value, and optimizes parameters such as the weight between the neurons and the coefficient of the activation function so that they approximate each other. As a result, the server 1 generates the classifier 141.

Note that, in the present embodiment, for convenience of description, the server 1 will be described as performing both the machine learning process for generating the classifier 141 and the handwriting classification process using the classifier 141. The processing subject on the ware may be different. For example, the server 1 may perform machine learning to generate the classifier 141, install it in the terminal 2, and the terminal 2 may classify handwriting. In this way, the processing entity that executes machine learning may be different from the processing entity that classifies handwriting by using the product of machine learning.

Further, in the present embodiment, the handwriting is classified using the image data obtained by reading the paper medium (problem paper), but as described above, the handwriting is classified from the stroke data electronically input by the electronic pen or the like. May be. In this case, for example, the server 1 causes the classifier 141 to learn the teacher data of strokes that represent the coordinate values of the electronic pen moving on the touch panel, the moving time, etc., and the classifier 141 that has performed the learning strokes of the learner's answer. You can enter data to classify handwriting. As described above, the classifier 141 may be a learned model in which the feature amount of human handwriting has been learned, and the feature amount to be handled is not limited to the image feature amount.

Further, in the present embodiment, the classification is performed according to whether or not the handwriting is scribbled, but the present embodiment is not limited to this, and the handwriting may be classified according to other classification criteria. Good. For example, the server 1 classifies whether or not it is a handwriting of an answer that is likely to be confident by using teacher data including an answer example that the answerer answers with confidence and an answer example when the answerer is not confident ( It is also possible to generate a classifier 141 for prediction and classify whether or not the handwriting is likely to be confident using the classifier 141. As described above, whether or not it is scribble is an example of the correct answer value labeled in the example answer, and the classification standard of the handwriting is not particularly limited.

Returning to FIG. 4, the description will be continued. The server 1 classifies the handwriting of the answer entered by the learner using the classifier 141 generated above. That is, the server 1 inputs the answer area extracted from the image data of the question paper into the classifier 141, and acquires the classification result obtained by classifying the handwriting related to the answer as an output. For example, the server 1 acquires as a classification result a probability value indicating whether or not the handwriting of the learner's answer is scribbled and the degree of scribbled.

As described above, the server 1 uses the classifier 141 to classify whether the handwriting of the learner's answer is scribble. If the answer was written in scribble, it is probable that the learner did not concentrate on the problem and the learning efficiency had decreased. Therefore, the server 1 determines the learning efficiency of the learner at the time of answering by using the classification result of the handwriting as an index, and presents the re-taking problem to the learner according to the determined learning efficiency.

FIG. 6 is an explanatory diagram related to the output processing of the re-taking question. FIG. 6 conceptually illustrates how the learning efficiency is determined based on the classification result of the handwriting, and the re-taking problem is presented (output) to the learner according to the determined learning efficiency. The output process of the re-taking problem will be described with reference to FIG.

First, the server 1 stores (stores) the classification result of the handwriting in the answer DB 143 in association with the learner ID of the learner who wrote the answer each time the answer information is acquired from the terminal 2 and the handwriting is classified. .. Specifically, in addition to the learner ID, the server 1 associates the answer time when the learner answered the question, the field of the question answered by the learner, and the like with the classification result of the handwriting (scribing level). , And the correctness of the answer are stored in association with each other.

The answer time is, for example, a time period when the solver solves the problem. The field in question is, for example, a subject. The correctness of the answer is information indicating whether or not the answer is correct. Note that, for example, the server 1 may also acquire the scoring results scored by a learner, a teacher, or the like when acquiring the answer information from the terminal 2, and the correctness is automatically determined by means such as character recognition. You may judge. The server 1 acquires the above-mentioned various information from the terminal 2 as answer information and stores it in the answer DB 143 in association with the classification result of the handwriting.

When the server 1 newly obtains answer information from the terminal 2 and classifies the handwriting, the server 1 determines whether or not the handwriting has high learning efficiency as compared with the past handwriting of the learner stored in the answer DB 143. judge. For example, the server 1 refers to the handwriting data in which the answer time (time zone) and the problem field are common among the handwritings of the learner's past answers stored in the answer DB 143. The server 1 searches the answer DB 143 for answer data of the handwriting of the answer newly classified this time and the handwriting of which the classification result by the classifier 141 is the same or similar. For example, the server 1 retrieves the handwriting data of the answer newly classified this time and the handwriting data in which the probability value indicating the degree of the scribble approximates within a predetermined numerical range.

The server 1 refers to the correctness of the answer associated with the retrieved handwriting and determines whether or not the handwriting has high learning efficiency. For example, the server 1 determines the learning efficiency to be higher as it is closer to the handwriting of the answer at the time of correct answer, and the learning efficiency is lower as it is closer to the handwriting of the answer at the time of incorrect answer. In this way, the server 1 determines that the learning efficiency is higher as the handwriting is closer to the correct answer.

For example, the server 1 determines whether the determined learning efficiency is less than or equal to a predetermined threshold value to determine whether the handwriting has high learning efficiency. When it is determined that the handwriting has high learning efficiency, a re-taking question for the learner to re-learn is output to the terminal 2. For example, the server 1 outputs to the terminal 2 another question that has the same field as the question answered by the learner this time as a re-taking question. Through the above-described processing, the server 1 preferentially re-learns a problem in a field where learning efficiency is poor.

As described above, according to the present embodiment, the server 1 can properly estimate the learning situation by classifying the handwriting of the answer of the learner. Particularly, in the present embodiment, it is possible to promote efficient learning by estimating the learning efficiency from the classification result of the handwriting and presenting the re-taking problem.

FIG. 7 is a flowchart showing the processing procedure of the generation processing of the classifier 141. The processing content of the generation processing of the classifier 141 will be described based on FIG. 7.
The control unit 11 of the server 1 acquires the teacher data in which the answer information group indicating the answer example manually written by a human and the correct answer value of the classification result obtained by classifying the handwriting of each answer information are associated with each other (step S11). .. The answer information is, for example, image data obtained by optically reading a paper medium (question paper) in which a person manually writes an answer. The correct answer value is, for example, a parameter indicating whether or not it is scribbled.

The control unit 11 performs machine learning for learning the feature amount of the handwriting relating to the answer information using the teacher data, and generates the classifier 141 for classifying the handwriting of the answer information (step S12). Specifically, the control unit 11 inputs the answer information to the classifier 141, extracts the feature amount of the handwriting, and estimates whether or not it is scribbled. The control unit 11 compares the estimation result with the correct value of the teacher data, optimizes the weight between the neurons so that they approximate each other, and generates the classifier 141. The control unit 11 ends the series of processes.

FIG. 8 is a flowchart showing the procedure of handwriting classification processing. The content of the process of classifying the handwriting of the answer of the learner and determining the learning efficiency will be described with reference to FIG.
The control unit 11 of the server 1 acquires, from the terminal 2, answer information indicating the answer handwritten by the learner for the question given to the learner (step S31). The answer information is, for example, image data obtained by optically reading a paper medium (question paper) in which the learner writes the answer by hand. The control unit 11 compares the acquired image data with the image data of the blank question paper, and extracts the image area corresponding to the answer entered by the learner (step S32).

The control unit 11 inputs the extracted image area of the answer to the classifier 141 and classifies the handwriting of the answer (step S33). For example, the control unit 11 estimates the degree of the scribble, whether or not the learner has entered the answer in the scribble. The control unit 11 stores the classification result of the handwriting in step S33 and the correctness of the answer in the answer DB 143 in association with the answer timing (for example, time zone) of the answer by the learner, the field of the answered question, and the like. (Step S34).

The control unit 11 refers to the answer DB 143 based on the classification result of the handwriting classified in step S33, and determines whether the handwriting has high learning efficiency (step S35). For example, the control unit 11 refers to the handwriting data in which the answer time and the field of the question are common, among the handwriting data (classification result) of the learner stored in the answer DB 143. The control unit 11 determines the learning efficiency based on the correctness of the answer associated with the handwriting whose classification result is the same as or similar to the handwriting classified this time in step S33. For example, the control unit 11 determines that the learning efficiency is higher as the handwriting is closer to the correct answer and the learning efficiency is lower as the handwriting is closer to the incorrect answer. The control unit 11 compares the determined learning efficiency with a predetermined threshold to determine whether the handwriting has high learning efficiency.

The control unit 11 determines whether or not to give a re-taking question, depending on whether or not the handwriting has high learning efficiency (step S36). When it is determined that the repeatable question is to be given (S36: YES), the control unit 11 outputs the repeatable question corresponding to the question answered by the learner this time to the terminal 2 (step S37). After executing the process of step S37, or if NO in step S36, the control unit 11 ends the series of processes.

As described above, according to the first embodiment, the learning situation can be appropriately estimated by classifying the handwriting of the answer entered by the learner.

Further, according to the first embodiment, the learning situation can be appropriately estimated from the image data obtained by optically reading the paper medium on which the answer is written.

Further, according to the first embodiment, the learning situation can be estimated also from the stroke data electronically input by the electronic pen or the like.

Further, according to the first embodiment, it is possible to accurately classify the handwriting by using the classifier 141 that has learned the feature amount of the handwriting of the answer example handwritten by a human.

Further, according to the first embodiment, it is possible to generate the classifier 141 with higher accuracy by using the teacher data in which the correct answer value of the classification result of the handwriting is associated with the example answer.

Further, according to the first embodiment, by storing the classification result of the handwriting in the answer DB 143 in association with the correctness of the answer, it is possible to construct a database capable of determining the level of learning efficiency.

Further, according to the first embodiment, by referring to the answer DB 143 and determining whether or not the handwriting has high learning efficiency, it is possible to determine how much concentration the learner has tackled the problem and the learning status. It can be properly estimated.

Further, according to the first embodiment, the learning situation can be estimated more appropriately by determining the learning efficiency in consideration of the answer time when the learner answered the question and the field of the question.

Further, according to the first embodiment, it is possible to promote efficient learning by outputting the re-taking question according to the determined learning efficiency.

(Embodiment 2)
In the present embodiment, a form in which the classifier 141 is generated by semi-supervised learning will be described. In addition, the same reference numerals will be given to the same contents as those in the first embodiment and the description thereof will be omitted.
FIG. 9 is an explanatory diagram regarding the generation processing of the classifier 141 according to the second embodiment. The outline of the present embodiment will be described with reference to FIG.

As described above, in the present embodiment, the server 1 generates the classifier 141 by the semi-supervised learning. Specifically, the server 1 uses the supervised data in which the correct answer value of the classification result of the handwriting is known and the unsupervised data in which the correct answer value is unknown among the plurality of answer examples used as the teacher data. To generate.

For example, the server 1 first performs learning by using supervised data in which correct answer values are associated with some answer examples among a plurality of answer examples used as teacher data. In this case, the server 1 inputs the image data according to each answer example into the classifier 141, classifies the handwriting, and compares the classification result with the correct answer value to optimize various parameters such as weights, as in the first embodiment. Turn into.

The server 1 uses the parameters set above as initial values for the classifier 141, and then performs learning using unsupervised data. In this case, for example, the server 1 performs, as unsupervised learning, clustering for classifying each answer example included in the unsupervised data into a plurality of classes according to the handwriting. For example, the server 1 performs a two-class determination that classifies an answer example of unsupervised data into two classes, a class in which handwriting corresponds to scribble and a class in which handwriting does not correspond to scribble. The server 1 inputs the image data according to each answer example to the classifier 141 to perform classification, and classifies which class it belongs to.

The server 1 optimizes various parameters such as weights based on the classification result of the teacher data. The parameter optimization method is not particularly limited, but for example, the server 1 optimizes using the bootstrap method. Specifically, the server 1 updates the classification boundary of each class in the feature space according to the reliability of the classification result, and optimizes each parameter. Thereby, the server 1 finally generates the classifier 141.

The method of semi-supervised learning is not limited to the bootstrap method, and graph-based semi-supervised learning or the like may be used. That is, the server 1 only needs to be able to generate the classifier 141 from supervised data and unsupervised data, and its learning method is not particularly limited.

FIG. 10 is a flowchart showing the procedure of the generation process of the classifier 141 according to the second embodiment.
The control unit 11 of the server 1 acquires the teacher data of the answer information group indicating a plurality of answer examples (step S201). In the teacher data according to the present embodiment, among the plurality of answer examples, the teacher answer data in which the correct answer value of the classification result of the handwriting is associated with some answer examples, and the correct answer value is not associated Including unsupervised data.

First, the control unit 11 performs machine learning using supervised data and determines initial values of various parameters such as weights set in the classifier 141 (step S202). Subsequently, the control unit 11 adds the unsupervised data to the learning data set, optimizes various parameters, and finally generates the classifier 141 (step S203). For example, the control unit 11 optimizes various parameters using the bootstrap method or the like. The control unit 11 ends the series of processes.

As described above, according to the second embodiment, the classifier 141 can be generated by semi-supervised learning, and the cost of machine learning can be reduced.

(Embodiment 3)
In the present embodiment, a form in which the handwriting of the answer is classified on a rule basis will be described.
FIG. 11 is an explanatory diagram related to handwriting classification processing according to the third embodiment. The outline of the present embodiment will be described with reference to FIG.

In the present embodiment, the server 1 does not use the classifier 141 and classifies whether or not the answer is scribble on a rule basis. For example, the server 1 performs character recognition on the image of the answer area extracted from the question paper (paper medium), and further divides the answer area into images for each character.

The server 1 determines (classifies) for each character from each divided image whether each character corresponds to scribble. The rule used as the criterion is not particularly limited, but the server 1 performs classification based on the three criterion, for example. The first criterion is whether or not the lines forming the character are straight lines. If it is not a straight line, the server 1 determines that it corresponds to scribble. Further, the second criterion is the inclination of the character (the inclination of the line in the character). The server 1 determines that it corresponds to scribble when the inclinations of the characters are not uniform. The third criterion is the centerline of the character. The server 1 determines whether it corresponds to scribble, depending on whether the center line of the character passes through the center of the image.

As described above, the server 1 determines (classifies) on a character-by-character basis whether or not each criterion is met. The server 1 determines the degree of learning efficiency of the entire answer composed of each character based on the determination result of each criterion for each character. For example, as shown in FIG. 11, the server 1 aggregates whether or not each criterion is met for the entire character string of the answer, and calculates the learning efficiency according to the aggregated result. The server 1 outputs a re-taking question according to the calculated learning efficiency.

FIG. 12 is a flowchart showing the procedure of the handwriting classification process according to the third embodiment. After acquiring the answer information from the terminal 2 (step S31) and extracting the answer area (step S32), the control unit 11 of the server 1 executes the following processing.
The control unit 11 performs a process of classifying the answer handwriting from the image of the answer area (step S301). Specifically, the control unit 11 performs character recognition on the answer area, divides the image into character units, and classifies whether or not each character corresponds to scribble from each divided image based on a predetermined rule. To do. Although the rule is not particularly limited, for example, the control unit 11 performs classification according to the line, inclination, central axis, etc. of the character. The control unit 11 determines whether the handwriting has high learning efficiency based on the classification result of each character (step S302), and moves the process to step S36.

As described above, according to the third embodiment, the learning situation can be estimated without using the classifier 141.

(Embodiment 4)
FIG. 13 is a functional block diagram showing the operation of the server 1 having the above-described form. When the control unit 11 executes the program P, the server 1 operates as follows.
The acquisition unit 131 acquires answer information indicating an answer handwritten by the learner with respect to the question. The classification unit 132 classifies the handwriting of the answer indicated by the answer information.

The fourth embodiment is as described above, and the other points are the same as those of the first to third embodiments, and therefore, corresponding parts will be denoted by the same reference numerals and detailed description thereof will be omitted.

The embodiments disclosed this time are to be considered as illustrative in all points and not restrictive. The scope of the present invention is shown not by the above meaning but by the scope of the claims, and is intended to include meanings equivalent to the scope of the claims and all modifications within the scope.

1 server (information processing device)
11 Control Section 12 Main Storage Section 13 Communication Section 14 Auxiliary Storage Section P Program 141 Classifier 142 Learner DB
143 Answer DB

Claims (12)

An acquisition unit that acquires answer information indicating the answer handwritten by the learner for the problem,
An information processing apparatus, comprising: a classification unit that classifies the handwriting of the answer indicated by the answer information. The answer information is image data obtained by optically reading a paper medium on which the answer is written,
The information processing apparatus according to claim 1, wherein the classification unit classifies the handwriting based on the image data. The answer information is stroke data of the answer consisting of a stroke group input by handwriting,
The information processing apparatus according to claim 1, wherein the classification unit classifies the handwriting based on the stroke data. The said classification|category part classifies the said handwriting using the classifier which has learned the feature-value of the handwriting of the answer example handwritten by the human being. The claim 1 characterized by the above-mentioned. Information processing equipment. The classifier is a learned model generated using teacher data in which correct answer values of the classification result of the handwriting are associated with each of the plurality of answer examples. Information processing device. The classifier includes supervised data in which correct answer values of the classification result of the handwriting are associated with some of the answer examples among the plurality of answer examples, and the part of the plurality of answer examples. An unprocessed model generated using unsupervised data composed of the example answer other than the example answer described above. The information processing apparatus according to claim 4, wherein the model is a learned model. The information processing apparatus according to claim 1, further comprising: a storage unit that stores the classification result of the handwriting and the correctness of the answer corresponding to the handwriting in association with each other. When the classification unit classifies the handwriting, whether the handwriting has high learning efficiency based on the correctness of the answer associated with the handwriting having the same or similar classification result as the handwriting with reference to the storage unit. The information processing apparatus according to claim 7, further comprising a determination unit that determines whether or not the information processing apparatus is included. The storage unit stores the answer time at which the learner answered the question, or the classification result of the handwriting and the correctness of the answer for each field of the question,
When the classification unit classifies the handwriting, the determination unit refers to a classification result of the handwriting in which the answer corresponding to the handwriting and the answer time or field are common, and whether the handwriting has high learning efficiency. The information processing apparatus according to claim 8, wherein the information processing apparatus determines whether or not it is. The information processing apparatus according to claim 8 or 9, further comprising an output unit that outputs a problem that causes the learner to re-learn according to a determination result by the determination unit. Acquire answer information indicating the answer handwritten by the learner for the problem,
An information processing method comprising causing a computer to execute a process of classifying the handwriting of the answer indicated by the answer information. Acquire answer information indicating the answer handwritten by the learner for the problem,
A program for causing a computer to execute a process of classifying the handwriting of the answer indicated by the answer information.

Images