JP7080759B2 - Predicted score providing device, predicted score providing method and predicted score providing program - Google Patents

Description

The present invention relates to a predicted score providing device, a predicted score providing method, and a predicted score providing program.

In Patent Document 1 below, a language that advances language learning while bidirectionally transmitting and receiving video and audio signals between a terminal for a language instructor located in a call center and a mobile phone carried by a user. The learning system is disclosed. In this language learning acquisition system, when a user reserves a lesson to be taken, the terminal for a language instructor and the user's mobile phone are connected via a network server at the reserved time, and the lesson starts.

Japanese Unexamined Patent Publication No. 2009-272690

In the language learning acquisition system described in Patent Document 1, an assessment is performed based on the content and comprehension level of the language lesson, and the result of the assessment is compared with other students to predict the possibility of improving the ability of the student. There is no mechanism to do it.

The present invention has been made in view of such circumstances, and provides a predictive score providing device, a predictive score providing method, and a predicted score providing program capable of predicting and providing an increase in a student's ability. With the goal.

The predictive score providing device according to one aspect of the present invention decomposes the student's answers to a plurality of questions used in the language lesson into words for each question, and the question is based on the combination of the part words of the decomposed words. A grammar discriminator that discriminates the type of grammar assigned to, and a tabulation unit that aggregates the number of questions for each type of grammar for each student based on the type of grammar discriminated by the grammar discriminator. , The correctness judgment unit that determines the correctness of the answer for each answer, the number of questions for each type of the grammar in all the students who took the language lesson, the correctness of the answer in the whole student, the student The machine learning unit that constructs a machine learning model in which the word used in the answer as a whole and the previous score of the assessment in the whole student are used as explanatory variables, and the score fluctuation value of the assessment in the whole student is used as the objective variable. With respect to the explanatory variables of the constructed machine learning model, the number of questions for each type of grammar in the target student to be assessed, the correctness of the answer in the target student, and the correctness of the answer in the target student. The word used in the answer and the previous score of the assessment in the target student are set respectively, and the objective variable calculation unit for calculating the objective variable of the machine learning model and the objective variable calculation unit calculated by the objective variable calculation unit. A predictive score providing unit that provides a predictive score for the assessment in the target student based on the objective variable is provided.

In the above aspect, the machine learning model is at least a plurality of models among a linear regression model, a classification model by a random forest, and a classification model by a neural network, and the plurality of objective variables calculated by the objective variable calculation unit are included in the plurality of objective variables. Based on the above, a second objective variable calculation unit for calculating one objective variable corresponding to the target student is further provided, and the prediction score providing unit is the one calculated by the second objective variable calculation unit. Based on one objective variable, the predicted score of the assessment in the target student may be provided.

In the above embodiment, the machine learning unit uses the objective variables of the plurality of constructed machine learning models as the second explanatory variables, and the score fluctuation value of the assessment in the entire student as the second objective variable. The second linear regression model is further constructed, and the second objective variable calculation unit uses the plurality of objective variables calculated by the objective variable calculation unit to obtain the second linear regression model of the second linear regression model. The one objective variable may be calculated by setting each of the explanatory variables of 2 and calculating the second objective variable.

The method for providing a predicted score according to another aspect of the present invention is a method for providing a predicted score to be executed by a processor, in which the student's answers to a plurality of questions used in a language lesson are decomposed into words for each of the questions. Based on the grammar discrimination step that discriminates the type of grammar assigned to the question based on the combination of the part of speech of the decomposed word, and the grammar type discriminated in the grammar discrimination step, the above-mentioned for each student. A tabulation step for totaling the number of questions for each type of grammar, a correct / incorrect judgment step for determining the correctness of the answer for each answer, and the number of questions for each type of grammar among all the students who took the language lesson. The correctness of the answer in the entire student, the word used in the answer in the entire student, and the previous score of the assessment in the entire student are used as explanatory variables, and the score fluctuation value of the assessment in the entire student is used. The number of questions for each type of grammar in the target students to be assessed for the machine learning step for constructing the machine learning model as the objective variable and the explanatory variables of the constructed machine learning model, the target. The correctness of the answer in the student, the word used in the answer in the target student, and the previous score of the assessment in the target student are set respectively, and the objective variable calculation for calculating the objective variable of the machine learning model. Includes a step and a predictive score providing step that provides a predictive score for the assessment in the subject student based on the objective variable calculated in the objective variable calculation step.

In the predictive score providing program according to another aspect of the present invention, the computer decomposes the student's answers to a plurality of questions used in the language lesson into words for each question, and is based on the combination of the part words of the decomposed words. Then, based on the grammar discriminating unit that discriminates the type of grammar assigned to the question and the type of grammar discriminated by the grammar discriminating unit, the number of questions for each type of grammar is totaled for each student. Aggregation unit, correctness judgment unit that judges the correctness of the answer for each answer, the number of questions for each type of grammar in all the students who took the language lesson, the correctness of the answer in the whole student, the attendance Machine learning to build a machine learning model in which the words used in the answer in the whole life and the previous score of the assessment in the whole student are used as explanatory variables, and the score fluctuation value of the assessment in the whole student is used as the objective variable. For the explanatory variables of the constructed machine learning model, the number of questions for each type of grammar in the target student to be assessed, the correctness of the answer in the target student, and the correctness of the answer in the target student. The objective variable calculation unit that sets the word used in the answer and the previous score of the assessment in the target student, and calculates the objective variable of the machine learning model, and the objective calculated by the objective variable calculation unit. Based on the variable, it functions as a predictive score providing unit that provides the predictive score of the assessment in the target student.

According to the present invention, it is possible to predict and provide an increase in the ability of a student.

It is a figure which shows one of the examples of the English learning system in an embodiment. It is a figure which illustrates the functional structure of the English learning system shown in FIG. It is a schematic diagram which illustrates the flow of the grammar discrimination processing executed in the English learning system shown in FIG. It is a schematic diagram which illustrates the flow of the correctness determination processing executed in the English learning system shown in FIG. It is a schematic diagram which illustrates the procedure at the time of constructing the machine learning model by the English learning system shown in FIG. FIG. 3 is a schematic diagram illustrating a procedure for calculating a predicted fluctuation value of a score, which is an objective variable, in the English learning system shown in FIG. 1. It is a figure which shows an example at the time of providing the fluctuation range and the reliability of a prediction score as a graph.

Hereinafter, a preferred embodiment of the English learning system including the function of the predictive score providing device according to the present invention will be described with reference to the accompanying drawings. In the embodiment, the case of learning English by way of example will be described, but the subject of learning is not limited to English, and the present invention can be applied to learning of other languages.

First, the English learning system according to the embodiment will be described with reference to FIG. The English learning system 1 can be constructed by using one or a plurality of server devices and databases. The English learning system 1 connects the student terminal 2 and the instructor terminal 3 which have been connected via the communication lines 4a and 4b, respectively, by using the telephone line 5. The English learning system 1 controls the progress of an English lesson that the student and the instructor proceed while talking by using the student terminal 2 and the instructor terminal 3 connected by the telephone line 5.

The student terminal 2 is a terminal carried by the student. The student terminal 2 corresponds to, for example, a mobile terminal having a call function such as a smartphone or a mobile phone. The instructor terminal 3 is a terminal used by the instructor. The instructor terminal 3 corresponds to, for example, a small terminal such as a personal computer, and the small terminal has a built-in call function using a telephone line. A plurality of instructor terminals 3 are arranged in the same facility such as a call center. The facility for accommodating the instructor terminal 3 may be provided in Japan or in a foreign country.

The instructor terminal 3 is not limited to small terminals arranged in the same facility, and may be, for example, a personal computer, a smartphone, a mobile terminal, or the like owned by the instructor.

The English learning system 1 includes, for example, a processor, a storage device, and a communication interface as physical configurations. The processor is composed of an arithmetic and logical operation unit and various registers, and realizes various functions described later by executing a computer program stored in a storage device.

The storage device is, for example, a computer-readable recording medium such as a disk drive or a semiconductor memory. A communication interface is a hardware module for connecting to a network and communicating with other terminals on the network.

The functional configuration of the English learning system 1 will be described with reference to FIG. The English learning system 1 has, for example, a homework providing function. The homework providing function includes, for example, a grammar determination unit 11, a totaling unit 12, a correctness determination unit 13, a machine learning unit 14, an objective variable calculation unit 15, and a prediction score providing unit 16. The function of the English learning system 1 is not limited to the homework providing function, and may have, for example, a lesson progress function, an assessment function, and the like.

The English learning system 1 includes, for example, a student information database 181 and an answer history information database 182 and a question information database 183 as databases constructed in the storage device.

The student information database 181 is a database for storing student information about students. The student information database 181 has, for example, a student ID item, an e-mail address item, a name item, a telephone number item, a progress status item, a grade item, and the like as data items.

The progress item stores the progress of the lesson. As the progress of the lesson, for example, the latest module number of the curriculum reached by the student can be used. The details of the module will be described later. The grade item stores, for example, the level of ability of the student such as comprehension and conversation ability. This progress item and grade item are provided for each evaluation item described later.

The answer history information database 182 is a database that stores the answers of the students to the questions used in the lessons taken by the students. The answer history information database 182 records, for example, the date and time (time is the time zone) when the lesson was taken for each student ID, and records the answer content for each date and time.

The question information database 183 is a database that stores the contents of the questions used in the lesson. Questions can be provided, for example, for each layer, evaluation item, grade and module. Layers indicate the level of the lesson. As the layer, for example, a (1) "Basic" layer which is an elementary level, (2) an "Intermediate" layer which is an intermediate level, and (3) an "Advanced" layer which is an advanced level can be provided.

The evaluation items indicate the contents of the evaluation provided for each layer. For example, the following evaluation items can be provided for each of the above three layers. (1) "Basic" layer: Five evaluation items of "Attitude", "Speaking", "Listening", "Grammar" and "Vocabulary". (2) "Intermediate" layer: Four evaluation items: "Logical Speaking", "Assertiveness", "Active Listening" and "Relationship Building". (3) "Advanced" layer: Four evaluation items of "Management Communication", "Storytelling", "Negotiation" and "Facilitation".

Grade indicates the ability level for each evaluation item. As a grade, for example, six ability levels can be provided. In that case, grades can be set in order from the lowest ability level, such as "E", "D", "C", "B", "A", and "S".

A module is the smallest unit that makes up the curriculum, and students take lessons in module units. Students can move on to the next grade by clearing all the modules assigned to the current grade.

The grammar discriminating unit 11 shown in FIG. 2 discriminates the type of grammar assigned to the question used in the lesson of the target student. With reference to FIG. 3, a grammar discrimination process for discriminating the type of grammar will be described in detail.

First, the grammar discriminating unit 11 decomposes the student's answer to the first question among the plurality of questions used in the lesson of the target student into words (step S11). For example, if the student's answer to the first question is "I have been to America.", The grammar discriminator 11 will use the five words "I", "have", "been", "to", and "America." Disassemble into ".

Subsequently, the grammar discriminating unit 11 discriminates the part of speech of the word decomposed in step S11 (step S12). When discriminating a part of speech, the grammar discriminating unit 11 replaces the plurality of consecutive parts of speech with one part of speech when a plurality of consecutive parts of speech can be integrated into one part of speech. A plurality of consecutive parts of speech can be integrated into one part of speech, for example, there are cases where a "preposition" is followed by a "noun", and a "preposition" is followed by an "article" and a "noun".

Specifically, in "to America", "to" is a "preposition" and "America" is a "noun". In this case, "to" and "America" are integrated and their part of speech is replaced with "adverb". In "in the park", "in" is a "preposition", "the" is an article, and "park" is a noun. In this case, "in", "the" and "park" are integrated and their part of speech is replaced with "adverb".

Subsequently, the grammar determination unit 11 determines the type of grammar assigned to the question based on the combination of part of speech determined in step S12 (step S13). For example, when the combination of part of speech is noun + auxiliary verb (have) + verb (past participle), the type of grammar is determined to be "G1: present perfect". Similarly, when the combination of part of speech is noun + be verb + verb (past participle), the type of grammar is determined to be "G2: passive voice". When the combination of part of speech is noun + be verb + verb (current participle), the type of grammar is determined to be "G3: progressive tense".

Subsequently, the grammar discrimination unit 11 performs the processes of steps S11 to S13 for the next question among the plurality of questions used in the lesson of the target student. On the other hand, when the processes of steps S11 to S13 are completed for all the questions used in the lessons of the target students, the grammar discrimination process is terminated.

The tabulation unit 12 shown in FIG. 2 totals the number of questions for each type of grammar for each student based on the type of grammar discriminated by the grammar discriminating unit 11. For example, if the number of questions used in a student's lesson is 10, the present perfect is 4 questions, the passive voice is 4 questions, and the progressive tense is 2 questions.

The correctness determination unit 13 determines whether the answer is correct or incorrect for each student's answer. With reference to FIG. 4, the correctness determination process for determining the correctness of the answer will be described in detail.

First, the correctness judgment unit 13 excludes the auxiliary verb "have" from the student's answer "I have been to America." And matrixes "I am to America." With the verb "been" as the present tense. (Step S21). By excluding the auxiliary verb and making the verb the present tense, the judgment target can be simplified.

Subsequently, the correctness determination unit 13 determines whether the part of speech of each word before and after the be verb “am” is a noun or an adjective, and generates a 2-by-2 matrix (step S22). ). For example, since "I" before "am" is a noun, "1" is set in 1 row and 1 column of the matrix, and "0" is set in 2 rows and 1 column of the matrix. Since "to America" after "am" is an adverb, "0" is set in 1st row 2nd column and 2nd row 2nd column of the matrix, respectively. If there are multiple words before and after the verb, the judgment results of each are added and set in the matrix.

Subsequently, the correctness determination unit 13 compares the matrix generated in step S22 with the matrix of the correct pattern registered in advance, and based on the sentence pattern of the correct pattern matching the generated matrix and the verb to be determined. , Determine whether the student's answer is correct (step S23). For example, when the sentence pattern of the correct pattern that matches the generated matrix is the first sentence pattern, the verb "am" to be determined is a verb that can be used as the verb of the first sentence pattern. Therefore, it is judged that the student's answer is correct. It is preferable to prepare in advance a table that defines the sentence patterns that can be used for each verb, and to refer to the table to determine whether or not the verb can be used for the target sentence pattern.

Here, when the answer includes a plurality of verbs, the processing of steps S21 to S23 is performed for each verb, and when it is determined that all the verbs are correct, the student's answer is given. Judge as correct. On the other hand, if it is determined that one of the verbs is incorrect, the student's answer is determined to be incorrect.

The machine learning unit 14 shown in FIG. 2 constructs a machine learning model by machine learning the contents of language lessons and the answers of students. In this embodiment, as a machine learning model, a linear regression model, a classification model by a random forest (hereinafter, also referred to as “random forest model”) and a classification model by a neural network (hereinafter, also referred to as “neural network model”) are used. Use.

Specifically, the machine learning department 14 includes the number of questions for each type of grammar in all the students who took the target language lesson, the correctness of the answer in the whole student, the words used in the answer in the whole student, and the course. The first linear regression model, Random Forest, is created by machine learning using a machine learning algorithm that uses the previous score of the assessment for the entire life as an explanatory variable and the fluctuation value of the assessment score for the entire student as the objective variable. Build a model and a neural network model respectively.

Further, the machine learning unit 14 uses the outputs (objective variables) from the first linear regression model, the random forest model, and the neural network model constructed above as explanatory variables, respectively, and aims at the score fluctuation value of the assessment for the entire student. A second linear regression model is constructed by machine learning using a machine learning algorithm as a variable.

The procedure for constructing the machine learning model will be described in detail with reference to FIG.

Ia in FIG. 5 is input data for each student, and specifically includes all scripts of the subject lessons to be machine-learned and the results of continuous assessment. The script is composed of, for example, one sentence in English. The results of successive assessments are, for example, the results of previous and current assessments.

Ta is a table showing the number of questions for each type of grammar aggregated by the aggregation unit 12. Tb is a table showing the correctness of the answer determined by the correctness determination unit 13. La is a list showing the total number of words used in the answer, the unique words used in the answer, and the like, which are analyzed based on the words used in the student's answer. Sa is the score of the previous assessment, and Sb is the fluctuation value of the score of the current assessment. The fluctuation value Sb of the score of the current assessment can be calculated by subtracting the score Sa of the previous assessment from the score of the current assessment.

First, the aggregation unit 12 and the correctness determination unit 13 generate the tables Ta and Tb based on the input data Ia, and further generate the list La, the score Sa and the variation value Sb of the score (step S31).

Subsequently, the machine learning unit 14 uses the values of the tables Ta, Tb, the list La, and the score Sa generated in step S31 as explanatory variables, and the machine learning algorithm using the fluctuation value Sb of the score as the objective variable. The first linear regression model, the random forest model, and the neural network model are constructed by machine learning (step S32).

Subsequently, the machine learning unit 14 uses the outputs from the first linear regression model, the random forest model, and the neural network model constructed in step S32 as explanatory variables, and the fluctuation value Sb of the score as the objective variable. A second linear regression model is constructed by machine learning using an algorithm (step S33).

The objective variable calculation unit 15 shown in FIG. 2 calculates the predicted fluctuation value of the score by setting the data of the target students to be assessed as the explanatory variables of the machine learning model and calculating the objective variable. ..

Specifically, the objective variable calculation unit 15 asks the target students the number of questions for each type of grammar for the explanatory variables of the first linear regression model, the random forest model, and the neural network model constructed by the machine learning unit 14. , The correctness of the answer in the target student, the word used in the answer in the target student, and the previous score of the assessment in the target student, respectively, and the purpose of the first linear regression model, random forest model, and neural network model. Calculate each variable.

In addition, the objective variable calculation unit (second objective variable calculation unit) 15 sets the three calculated objective variables as explanatory variables of the second linear regression model, respectively, and calculates the objective variable to take the target course. Calculate one objective variable corresponding to raw.

With reference to FIG. 6, the procedure for calculating the predicted fluctuation value of the score, which is the objective variable, will be described in detail.

Ib in FIG. 6 is input data of the target student, and specifically, includes all scripts of the target period of the target student and the result of the previous assessment. The target period can be arbitrarily set in consideration of the period during which the data required for predicting the score of this assessment can be secured based on the result of the previous assessment. Since the table Ta, Tb, the list La, the score Sa, and the fluctuation value Sb of the score are the same as those in FIG. 5, the description thereof will be omitted.

First, the aggregation unit 12 and the correctness determination unit 13 generate the tables Ta and Tb based on the input data Ib, and further generate the list La and the score Sa (step S41).

Subsequently, the objective variable calculation unit 15 uses the values of the tables Ta, Tb, list La, and score Sa generated in step S41 as explanatory variables of the first linear regression model, random forest model, and neural network model. Each is set and three objective variables are calculated (step S42).

Subsequently, the objective variable calculation unit 15 sets the three objective variables calculated in step S42 as explanatory variables of the second linear regression model, and calculates one objective variable (step S43). This one objective variable is the predicted fluctuation value of the score of the target student.

The prediction score providing unit 16 shown in FIG. 2 provides the prediction score of the assessment in the target student based on one objective variable calculated by the objective variable calculation unit 15. Specifically, the predicted score providing unit 16 calculates the predicted score by adding the predicted fluctuation value of the score calculated as one objective variable to the score Sa of the previous assessment.

When providing the predicted score, the predicted score providing unit 16 may provide the fluctuation range and the reliability of the predicted score. For example, when the predicted score is "280 points", the predicted score providing unit 16 may have "280 points ± 30 points (reliability 70%)", "280 points ± 50 points (reliability 90%)", or the like. It may be to provide information. When providing the fluctuation range and the reliability of the predicted score, it may be provided as character information as described above, or may be provided as a graph as illustrated in FIG. 7. Ga in FIG. 7 is a graph showing a predicted score (280 points), Gb is an index showing a range of predicted score (280 points) ± 30 points (reliability 70%), and Gc is a predicted score (reliability 70%). It is an index showing the range of 280 points) ± 50 points (reliability 90%).

As described above, according to the English learning system 1 in the embodiment, the grammar solved by the target student according to the type of grammar found from the combination of the part of speech of the word included in the answer of the target student to be assessed. Since it is possible to determine the number of questions for each type and the correctness of the answer to each question, it is possible to grasp the degree of understanding of the lessons and the language ability of the target students for each type of grammar.

In addition, according to the English learning system 1, the explanatory variables of the first linear regression model, the random forest model, and the neural network model, which were machine-learned using the data of all the students, were classified according to the type of grammar in the target students. The number of questions, the correctness of the answer, the word used for the answer, and the previous score of the assessment are set respectively to calculate the three objective variables, and from the first linear regression model, random forest model, and neural network model. Three objective variables can be set for the explanatory variables of the second linear regression model machine-learned using the output data, and the objective variable, which is the predicted fluctuation value of the score, can be calculated. Therefore, it is possible to analyze the fluctuation tendency of the score of the entire student in combination with the degree of understanding and ability, apply the data of the target student to the analysis result, and calculate the predicted fluctuation value of the score of the target student. Become.

Further, according to the English learning system 1, it is possible to calculate and provide the predicted score of the assessment of the target student based on the predicted fluctuation value of the score of the target student.

Therefore, according to the English learning system 1 in the embodiment, it is possible to predict and provide the growth of the student's ability.

[Modification example]
The present invention is not limited to the above-described embodiment, and can be implemented in various other forms without departing from the gist of the present invention. For this reason, the above embodiments are merely exemplary in all respects and are not to be construed in a limited manner. For example, the above-mentioned processing steps can be arbitrarily changed in order or executed in parallel within a range that does not cause a contradiction in the processing contents.

In the above-described embodiment, machine learning is performed using a linear regression model, a random forest model, and a neural network model as the machine learning model, but the present invention is not limited thereto. For example, machine learning may be performed using any one or two machine learning models of the linear regression model, the random forest model, and the neural network model. When there is only one machine learning model, the second linear regression model that combines a plurality of objective variables into one can be omitted.

Further, in the above-described embodiment, the student terminal 2 and the instructor terminal 3 are connected via the telephone line 5, but the configuration is not limited to this. For example, the student terminal 2 and the instructor terminal 3 may be connected via a communication line without using the telephone line 5. The communication line corresponds to, for example, a communication line to which VoIP (Voice over Internet Protocol) technology including Skype (registered trademark) is applied. In this case, the identification ID assigned to the student terminal 2 as a VoIP account can be used as the student ID of the student information.

1 ... English learning system, 2 ... Student terminal, 3 ... Instructor terminal, 4a, 4b ... Communication line, 5 ... Telephone line, 11 ... Grammar discrimination unit, 12 ... Aggregation department, 13 ... Correctness judgment unit, 14 ... Machine learning department, 15 ... Objective variable calculation department, 16 ... Prediction score providing department, 181 ... Student information database, 182 ... Answer history information database, 183 ... Question information database

Claims (9)

A grammar discriminator that decomposes students' answers to multiple questions used in language lessons into words for each question, and determines the type of grammar assigned to the question based on the combination of part of speech of the decomposed words. When,
Based on the type of grammar determined by the grammar discrimination unit, a tabulation unit that aggregates the number of questions for each type of grammar for each student, and a tabulation unit.
For each of the answers, a correctness determination unit that determines the correctness of the answer, and
The number of questions for each type of grammar in all the students who took the language lesson, the correctness of the answer in the whole student, the words used in the answer in the whole student, and the previous assessment in the whole student The machine learning department that builds a machine learning model that uses the score as an explanatory variable and the score fluctuation value of the assessment for the entire student as the objective variable.
For the explanatory variables of the constructed machine learning model, the number of questions for each type of grammar in the target student to be assessed, the correctness of the answer in the target student, and the answer in the target student. The objective variable calculation unit for calculating the objective variable of the machine learning model by setting the words used in the above and the previous score of the assessment in the target student, respectively.
Based on the objective variable calculated by the objective variable calculation unit, the predictive score providing unit that provides the predicted score of the assessment in the target student, and the predictive score providing unit.
Predictive score providing device. The machine learning model is at least a plurality of models among a linear regression model, a classification model by a random forest, and a classification model by a neural network.
A second objective variable calculation unit for calculating one objective variable corresponding to the target student based on the plurality of objective variables calculated by the objective variable calculation unit is further provided.
The prediction score providing unit provides a prediction score for assessment in the target student based on the one objective variable calculated by the second objective variable calculation unit.
The predictive score providing device according to claim 1. The machine learning unit uses the objective variables of the plurality of constructed machine learning models as the second explanatory variables, and the score fluctuation value of the assessment in the entire student as the second objective variable. Further build a linear regression model,
The second objective variable calculation unit sets a plurality of the objective variables calculated by the objective variable calculation unit as the second explanatory variables of the second linear regression model, respectively, and the second objective variable calculation unit. By calculating the objective variable, the one objective variable is calculated.
The predictive score providing device according to claim 2. It is a method of providing a predicted score that is executed by the processor.
A grammar discrimination step that decomposes the student's answers to multiple questions used in a language lesson into words for each question, and determines the type of grammar assigned to the question based on the combination of the part of speech of the decomposed word. When,
Based on the type of grammar determined in the grammar determination step, a tabulation step for totaling the number of questions for each type of grammar for each student, and an aggregation step.
For each of the answers, a correct / incorrect judgment step for determining the correctness of the answer, and
The number of questions for each type of grammar in all the students who took the language lesson, the correctness of the answer in the whole student, the words used in the answer in the whole student, and the previous assessment in the whole student A machine learning step for constructing a machine learning model in which the score is used as an explanatory variable and the score fluctuation value of the assessment for the entire student is used as the objective variable.
For the explanatory variables of the constructed machine learning model, the number of questions for each type of grammar in the target student to be assessed, the correctness of the answer in the target student, and the answer in the target student. The objective variable calculation step for calculating the objective variable of the machine learning model by setting the words used in the above and the previous score of the assessment in the target student, respectively.
A predictive score providing step that provides a predictive score for an assessment in the target student based on the objective variable calculated in the objective variable calculation step, and a predictive score providing step.
Predictive score providing method including. The machine learning model is at least a plurality of models among a linear regression model, a classification model by a random forest, and a classification model by a neural network.
A second objective variable calculation step for calculating one objective variable corresponding to the target student based on the plurality of objective variables calculated in the objective variable calculation step is further included.
The predicted score providing step provides a predicted score for assessment in the target student based on the one objective variable calculated in the second objective variable calculation step.
The method for providing a predicted score according to claim 4. In the machine learning step, the objective variable of the plurality of constructed machine learning models is used as the second explanatory variable, and the score fluctuation value of the assessment in the entire student is used as the second objective variable. Further build a linear regression model,
In the second objective variable calculation step, the plurality of objective variables calculated in the objective variable calculation step are set as the second explanatory variables of the second linear regression model, respectively, and the second objective variable is set. By calculating the objective variable, the one objective variable is calculated.
The method for providing a predicted score according to claim 5. Computer,
A grammar discriminator that decomposes students' answers to multiple questions used in language lessons into words for each question, and determines the type of grammar assigned to the question based on the combination of part of speech of the decomposed words. ,
A tabulation unit that aggregates the number of questions for each type of grammar for each student based on the type of grammar determined by the grammar discrimination unit.
A correct / incorrect judgment unit that determines the correctness of the answer for each answer.
The number of questions for each type of grammar in all the students who took the language lesson, the correctness of the answer in the whole student, the words used in the answer in the whole student, and the previous assessment in the whole student The machine learning department that builds a machine learning model with the score as the explanatory variable and the score fluctuation value of the assessment for the entire student as the objective variable.
For the explanatory variables of the constructed machine learning model, the number of questions for each type of grammar in the target student to be assessed, the correctness of the answer in the target student, and the answer in the target student. The objective variable calculation unit, which sets the previous score of the assessment in the target student and the word used in the above, and calculates the objective variable of the machine learning model.
A predictive score providing unit that provides a predicted score for an assessment in the target student based on the objective variable calculated by the objective variable calculation unit.
A predictive score providing program that functions as. The machine learning model is at least a plurality of models among a linear regression model, a classification model by a random forest, and a classification model by a neural network.
The computer is further made to function as a second objective variable calculation unit that calculates one objective variable corresponding to the target student based on the plurality of objective variables calculated by the objective variable calculation unit.
The prediction score providing unit provides a prediction score for assessment in the target student based on the one objective variable calculated by the second objective variable calculation unit.
The predictive score providing program according to claim 7. The machine learning unit uses the objective variables of the plurality of constructed machine learning models as the second explanatory variables, and the score fluctuation value of the assessment in the entire student as the second objective variable. Further build a linear regression model,
The second objective variable calculation unit sets a plurality of the objective variables calculated by the objective variable calculation unit as the second explanatory variables of the second linear regression model, respectively, and the second objective variable calculation unit. By calculating the objective variable, the one objective variable is calculated.
The predictive score providing program according to claim 8.

Images