KR20200012431A - Method for providing an information of a question analysis using big data - Google Patents

Abstract

The present invention relates to a method for providing analysis information for the predicted correct answer rate of a learner. The method comprises: a learning providing step of online providing learning contents including learning items which a learner accessing a service server can use; a learning information receiving step of receiving learning information including the contents of answers of the learner to the learning items; a learning information datafication step of datafying the information received in the learning information receiving step; a learning item analysis step of analyzing the learning items answered by the learner through the analysis of data stored in the learning information datafication step; and a learning item analysis information providing step of providing analysis information for the learning items analyzed in the learning item analysis step. The learning item analysis step includes a step of determining questions expected to be solved by guessing when the answers of the learner to the learning items are correct. The step of determining questions expected to be solved by guessing determines the learning items as questions expected to be solved by guessing when the response time to answer the learning items is shorter than a specific time. The present invention can help the learner to improve learning habits and skills thereof by providing analysis information for items which the learner has learned through online learning.

Description

METHOD FOR PROVIDING AN INFORMATION OF A QUESTION ANALYSIS USING BIG DATA}

The present invention relates to a method of providing information about learning to a learner, and more particularly, to a method of providing more detailed analysis information of an item learned by a learner.

As information and communication technology is developed, as a database is easily managed through a computer, a learning service that stores learning information in a database and provides learners using the Internet online is provided with various contents.

Learners can learn from the problems provided through websites that provide learning programs or through apps that can be accessed through mobile devices. Therefore, a number of service methods for providing learning information, directly performing learning, and receiving a result thereof have been developed.

The present invention also relates to a method of providing learning information using such online.

On the other hand, learning hardly does not necessarily follow performance.

In other words, it is well known that it is important not to work hard but how to work hard.

Especially, in learning to solve the presented learning questions, reviewing the results of learning after learning can be as helpful as learning.

For example, if a question that is solved is a question that can be fully answered, it can be corrected after solving it if it is wrong due to a learner's wrong habits. Complementing learning through other methods can help greatly in learning.

The feedback of the analysis and review of the learning items can be easily instructed by the instructor by the off-line learning, but in the case of the online learning, it provides only the learning information or problems, and further guidance on the learning method after solving. This is not possible unless the tutor visits.

The matters described in the background art are provided to help the understanding of the background of the invention, and may include matters that are not known to those skilled in the art to which the art belongs.

Korean Laid-Open Patent Publication No. 10-2010-0127367

The present invention has been made to solve the above-described problems, the present invention provides analysis information on the questions learned through online learning to analyze the learning items using big data that can help learners to improve their learning habits and skills The purpose is to provide a method of providing information.

Learning item analysis information providing method using the big data according to an aspect of the present invention, the learning providing step including the learning items available to learners connected to the service server online learning providing step, for the learning items The learner receives the learning information including the contents of the learner's answer, the learning information data step of data-setting the information received by the learning information receiving step, and the corresponding learner through data analysis stored by the learning information data-setting step. A learning item analysis step of analyzing the answered learning item and the learning item analysis information providing step of providing analysis information on the learning item analyzed by the learning item analysis step, wherein the learning item analysis step includes the learner's learning If the answer to the question is the correct answer, it is correct. Problem determination steps included a problem determination step is expected, and the dip expected to right with, if the answer to the response time for the learning item is shorter than the specified time is characterized in that determination of a problem that is expected to hit the dip.

More specifically, the learning item analysis step models a percentage of the correct answer of the item as a function of response time from the plurality of answer data of the corresponding item, and the specific time is the response rate per response time of the response time within 40% It is characterized by calculating the response time at which the change amount of (%) becomes the maximum as a specific time.

The specific time may be updated when the number of the plurality of answer data is added to a predetermined number or more.

In addition, the ratio of the correct answer and the wrong answer among the plurality of the plurality of answers data number is characterized in that the equivalent.

On the other hand, the learning item analysis step, if the answer to the learner's answer to the answer is wrong, including the correct but wrong problem determination step and unknown problem determination step, the correct but wrong problem and the unknown problem It is characterized by determining based on the predicted correct answer rate of the learner.

Here, the predicted correct answer rate is calculated by the item response theory reflecting the inherent difficulty of the learning item and the estimated ability of the learner, the inherent difficulty (question difficulty) is a difficulty set by other learners for the item, The learner's ability (learner ability) is characterized in that the learner's ability compared to the other learners for the other questions in the past.

The predicted correct answer rate is calculated by the following equation.

On the other hand, the learner's ability value is characterized by being calculated by the Bayesian ability calculation algorithm.

As a result, the learning item analysis step, when the learner's predicted correct answer rate is more than 85%, characterized in that it can be determined as 'can be correct but wrong problem'.

And, the learning item analysis step, if the learner's predicted correct answer rate is less than 50%, characterized in that it is determined as a 'unknown problem wrong.'

In addition, when the predicted correct answer rate of the learner is more than 50% and less than 85%, according to the degree of the response time of the answer to the learning item characterized in that it can be determined as the 'correct but wrong problem' or the 'unknown problem wrong'. do.

In this case, the required response time may be determined by comparing the actual answer response time for the learner's learning item with an estimate of the predicted answer response time of the learner.

The predicted answer response time may be estimated based on an average answer response time of the learner and an average answer time of the learner of other learners who answered a correct answer to the learning question.

According to the method for providing learning item analysis information using the big data of the present invention, learners can worry about learning methods and habits by providing information that can be improved in detail for the questions solved by the learners. To help learners improve their true skills.

1 illustrates a method for providing learning item analysis information using big data according to the present invention.
2 illustrates a more detailed process of one step of the method for providing learning item analysis information using big data according to the present invention.
3 illustrates an example of calculating a specific time using big data.

In order to fully understand the present invention, the operational advantages of the present invention, and the objects achieved by the practice of the present invention, reference should be made to the accompanying drawings which illustrate preferred embodiments of the present invention and the contents described in the accompanying drawings.

In describing the preferred embodiment of the present invention, well-known techniques or repeated descriptions that may unnecessarily obscure the subject matter of the present invention will be shortened or omitted.

1 illustrates a method for providing learning item analysis information for learners according to the present invention, and FIG. 2 illustrates a more detailed process of one step of the method for providing learning item analysis information for learners according to the present invention.

Hereinafter, a method for providing learning item analysis information for learners according to an embodiment of the present invention will be described with reference to FIGS. 1 and 2.

The method of providing learning item analysis information according to the present invention is a service method for providing information for learning online and allowing a user connected to online to proceed with learning.

In other words, the service provider provides a learning service through an application implemented through an Internet website or a smart device, and the user of the service can access a website or application through a procedure such as logging in to perform learning. . Descriptions of technologies widely recognized for such a service method and a usage method will be omitted in the present invention.

In addition, the learning provided by the present invention is provided to allow the learner to perform the learning according to the progress provided by providing an appropriate amount of learning by classifying the cycles according to the cycles such as daily, weekly, monthly, etc. rather than one-time. .

Specifically, the learning item analysis information providing method according to the present invention includes a learning providing step, learning information receiving step, learning information data conversion step, learning item analysis step and learning item analysis information providing step.

In the learning providing step S10, the learning content available to the learner accessing the service is provided online by the service server.

The learning content provided by the learning providing step S10 is a learning item to be solved by the learner, and also includes learning content.

The learning contents may be provided by dividing each round at regular intervals, and the learning contents provided in each round may include learning contents together with learning questions, and only learning contents may be provided.

The learning items are provided with a plurality of learning items according to a predetermined number and difficulty levels.

For example, 10 questions may be provided in a batch to allow a learner to scroll and solve the problem, or to provide a small amount of one or two items so that they can be checked over the screen.

It will be provided with a means for the learner to check the correct answer to solve the question.

In addition, a means of selecting the next item is presented so that the learner can skip the question presented later for solving another item. Only the next page may be selected by the touch input method.

In addition, the entire question of the round may be scored in a batch or may be provided to proceed with the scoring for each item provided on a page.

In the learning providing step S10, when the learner checks the answer and selects the grading, the result of the correct answer or the wrong answer by the grading is also provided.

On the other hand, in the case of an incorrect answer may be provided to re-answer the question.

In addition, the learning content may be a text or video format of a lecture on the content to be known in the corresponding round, and may be text or an image or a hint of a commentary about each item.

If there is a learner's answer to the learning item among the learning contents provided by the learning providing step S10, the service server receives the learner's answer by the learning information receiving step S20.

The received learning information may also be an answer to a learning item, or may be a result of learning content.

That is, the learning information on the learning item may include the learner's answer to the item, the response time from the time of providing the item to the answer, and the number of attempts to answer if the learner's answer is incorrect.

And, the result related to the learning content relates to whether or not using the learning content content provided in S10.

That is, it is information about whether or not the lecture, commentary, hints, etc. provided in the corresponding session are confirmed.

In this way, the answer to the learner's answer to the learning item in the learning information receiving step (S20) and the result of the learning content is data by the service server in the learning information data step (S30).

The number of learners is infinite due to the nature of the online service. In the learning data forming step (S30), the learning information by a large number of learners is made into big data, and the detailed analysis of the questions learned through the learning item analysis step (S40) to be described later. Information is provided by the learning item analysis information providing step (S50).

In the step of providing learning item analysis information (S50), not only the learning item analysis information by S40, but also information about general learning results such as the overall correct answer rate per session, the total solve time, the grade, the difficulty, and the comparison with the learner average, to be.

In the learning item analysis step (S40), the learner solves the learning item solved by the learner by the learner's learning information received by the learning information receiving step (S20) and the information held by the big data by the learning information data making step (S30). By analyzing the feedback information that can be helpful to the learner.

The analysis of the learning questions is based on S41, which can be correct but wrong problem determination (S41), unknown problem determination (S42), and problem determination (S43) that are expected to be corrected. To S43, and items which do not belong to this category are classified as unclassified items.

The determination process of S41 to S43 does not mean the order of item analysis, but determination is performed individually or entirely by each process.

Although the correct problem determination (S41) and do not know the wrong problem determination (S42) by the learning information receiving step (S20) receives the correct answer / incorrect answer and the answer response time, the learner if the corresponding question is incorrect Based on the predicted correct answer rate and predicted solving time, the learner can determine whether the answer can be corrected due to an item or a mistake that the learner has a high probability of not knowing.

In S41, it is determined whether the learner's predicted correct answer rate for the corresponding item is 85% or more, and if the predicted correct answer rate is 85% or more and the wrong answer is correct, it is determined as 'correct but wrong problem'.

On the other hand, in S42, it is determined whether the learner's predicted correct answer rate for the corresponding item is less than 50%, and if the predicted correct answer rate is less than 50% and the wrong answer is incorrect, it is determined as 'unknown problem'.

However, when the predicted correct answer rate of the learner is 50% or more and less than 85%, it is determined differently based on the predicted correct answer rate and the degree of the response time.

That is, it is determined by determining whether the degree of the real answer response time of the learner is fast, slow or medium.

Thus, if the predicted correct answer rate is 50% or more and less than 85%, but the response response time is not slow, fast or medium, it is determined by S41 as a 'correct but wrong problem'.

In addition, when the predicted correct answer rate is 50% or more and less than 85%, but the response response time is fast, it is determined by S42 as a 'not known problem'.

The predicted correct answer rate and the predicted answer response time serving as the discrimination criteria of S41 and S42 described above will be described.

The learner's predicted correct answer rate is the probability that the learner is expected to answer the question, which is analyzed by setting the inherent difficulty of the question and estimating the learner's ability.

The inherent difficulty of the question is to estimate the relative difficulty set by other learners for the question through the information datad by the learning information data step S30.

In other words, it is estimated based on the correct answer rate of all learners for the corresponding question and the number of answer attempts required to correct the correct answer.

The average value of the inherent difficulty of an item is 0, and the difficulty is estimated by correcting (+) or (-) according to the correct answer rate of all learners and the number of attempts to answer the correct answer.

This calibration is corrected by the updated value after each response by Bayesianism.

And the learner's ability (learner ability) is calculated by comparing the other learners with other learners.

The learner's ability value is estimated by setting the average value to 0, and assigning a weight of (+) or (-) to the learner's correct answer rate according to whether the learner answers a plurality of questions correctly or incorrectly.

This weight is also obtained by the updated value after each response by the Bayesian capability estimation algorithm (Bayesianism).

Furthermore, the weight according to the correct or incorrect answer may be set higher for the result of the most recently answered question.

The predicted correct answer rate for an item is calculated based on the item response theory (IRT) by the inherent difficulty of the item and the learner's ability.

That is, the higher the learner ability or the lower the intrinsic difficulty, the greater the exp (question difficulty + learner ability) value, resulting in a larger P value and a maximum value of 1.

In addition, even if a learner solves a problem with a high inherent difficulty, the learner with a low learner ability value solves the problem.

Meanwhile, in order to calculate the learner's predicted correct answer rate according to the learner's ability and the inherent difficulty, the learner's ability may be divided into 2N + 1 (N is an integer) and 2M + 1 (M is an integer).

For example, if the range of the inherent difficulty of the item and the range of learner abilities are set to ± 4 (N, M is 4), the predicted correct answer rate for the item can be calculated as shown in the following table according to the above equation. The learner's ability can have a value that increases or decreases with an arithmetic progression, and furthermore, the adjustment of the inherent difficulty of the item according to the learner's ability can be adjusted by increasing or decreasing it according to the table below.

% Predicted Answer Rate Inherent difficulty level of an item -4 -3 -2 -One 0 One 2 3 4 Learner Stats -4 50 37.5 25 12.5 0 0 0 0 0 -3 62.5 50 37.5 25 12.5 0 0 0 0 -2 75 62.5 50 37.5 25 12.5 0 0 0 -One 87.5 75 62.5 50 37.5 25 12.5 0 0 0 100 87.5 75 62.5 50 37.5 25 12.5 0 One 100 100 87.5 75 62.5 50 37.5 25 12.5 2 100 100 100 87.5 75 62.5 50 37.5 25 3 100 100 100 100 87.5 75 62.5 50 37.5 4 100 100 100 100 100 87.5 75 62.5 50

The predicted correct answer rate (P) for the item calculated by the equation is in the range of 0.5 to 1, and if it is less than 0.5, it is calculated on the assumption that the item has a linearity with respect to the inherent difficulty of the item to obtain the result shown in the above table. Can be.

And, as shown in the table, as the learner's ability increases, the predicted correct rate may increase linearly, and as the intrinsic difficulty decreases, the predicted correct rate may increase linearly.

In addition, if the learner's ability and inherent difficulty are the same in the table for each section, P becomes 0.5. As the learner's ability increases along 2N + 1 intervals, the predicted answer rate may increase linearly. As the inherent difficulty decreases along 2M + 1 intervals, the predicted answer rate may increase linearly.

Whether it is determined in S41 or S42 is determined according to the predicted correct answer rate calculated as described above.

Next, the predicted answer response time is divided into fast, slow or medium speed of the answer to the corresponding questions of the learner, and accordingly determine whether or not in S41 or S42.

The speed of the learner's response to the question is determined by comparing the learner's actual answer response time with the learner's predicted answer response time.

That is, it is determined that the learner's actual answer response time is located in a section higher than 1.28 due to the standard deviation than the predicted answer response time and is longer than other 90% learners.

And, it is determined that the learner's actual answer response time is located in a section lower than 1.28 due to the standard deviation than the predicted answer response time, and is shorter than other learners of 90%.

Slow and fast are not intermediate speeds.

The learner's actual response response time is the time actually taken when the learner first answered the question.

However, the learner can check the question through the viewer and answer the other questions first without answering the question immediately, and may repeat this behavior several times.

In such a case, the longest time of staying on the screen where the question is presented is the actual response response time of the learner.

The learner's predicted answer response time is the time expected when the learner solves the question, which may vary from question to question and may vary from learner to learner.

The learner's predicted answer response time is a combination of the average answer time based on the result of the learner continuously answering the question based on the average answer response time of the other learners who answered the correct answer to the question. Estimate

Next, the method of determining 'problem determination (S43) that is expected to be beaten' in addition to S41 and S42 will be described.

Although the learner's answer is the correct answer, it is identified as a 'problem expected to be beaten' if the learner responds in a time that does not determine that he or she has had any time to think.

Determination of the photographing behavior is determined based on the response time to the question. This is because the rate of correct answers drops sharply when the response response time is shorter than a certain time in most questions, and it is determined whether the answer response time is shorter than the specific time. It is.

This specific time is calculated using the big data information through the learning information data step (S30), and in the learning item analysis step (S40), it is determined whether the answer to the learning item corresponds to the taking action based on the calculated specific time. Done.

3 illustrates an example of calculating a specific time using big data, and a method of calculating a specific time will be described with reference to the drawing. In FIG. 3, the first mathematics, the first first language, the first Korean, and the first apple refer to a subject name.

The calculation of a specific time is a value that can vary from item to item, and the value that can be actively changed as the amount of data increases. The initial setting is 1,000 ms.

When a certain number of answers to the corresponding question are dataized, the specific time is updated by adjusting the initial setting (CALIBRATION).

For example, the answer to the question may be set to adjust when 50 correct answers and 50 incorrect answers are obtained.

The adjustment calculation method first uses a lognormal function with four coefficients to model each percent correct as a function of ln (response time).

As a result of the modeling, a graph as shown in FIG. 3 is derived, and the response time (slope on the graph) in which the amount of change in the percentage of correct answers per response time is the maximum among the response time below the 40th percentile is calculated as a specific time. do.

In the example of the graph, as a result of the calculation, in the case of the first math, 3 seconds of the maximum slope of the 40th percentile or less is calculated as a specific time.

In big data analysis, the larger the response time, the higher the correct answer rate, but there is a limit.

However, there is an exceptional case that the response rate is higher than the response time of 3 seconds, whereas the response rate is higher as the response time is larger.

This is a very familiar problem like the nature of the speed quiz, so the response time is short, and it is necessary to calculate a specific time by excluding such a case.

To do this, the specific time is calculated by applying the reference point of -1, with the exception that the correct answer rate of the 10th fastest response of the percentile is greater than or equal to (the percentage of correct answers during the response time corresponding to the 40th to 60th percentile? 0.9). . The reference point -1 corresponds to a data tag that is separately processed so that the correct answer to the fast response time is not separated by a photograph.

Next, in the case of an item that is not classified by S41 to S43 and treated as an unclassified item, all of the correct answerers of the item fall under the reference value.

For example, the predicted correct answer rate and predicted answer response time cannot be estimated for questions with less than 10 correct answerees.

As described above, when the learning item is analyzed, the learning item information is provided through the learning item information providing step S50 so that the learner can check it.

This may be provided with information such as general ratings for the session.

In addition, each item may be displayed or the number of items and information of the items may be collectively presented.

Although the present invention as described above has been described with reference to the illustrated drawings, it is not limited to the described embodiments, it can be variously modified and modified without departing from the spirit and scope of the present invention is common knowledge in the art Self-evident to those who have Therefore, such modifications or variations will have to belong to the claims of the present invention, the scope of the invention should be interpreted based on the appended claims.

S10: step of providing learning
S20: step of receiving learning information
S30: learning information data step
S40: Learning Item Analysis Stage
S50: step of providing learning item analysis information

Claims (13)

A learning providing step of providing learning contents online including learning items available to a learner accessing the service server online;
Learning information receiving step of receiving learning information including contents of a learner's answer to the learning item;
A learning information data step of converting the information received by the learning information receiving step into data;
A learning item analysis step of analyzing a learning item answered by the corresponding learner by analyzing the data stored by the learning information data conversion step; And
And providing learning item analysis information providing analysis information on the learning item analyzed by the learning item analyzing step.
The learning item analyzing step includes a problem determination step that is expected to be correct by taking a picture when the answer to the learning item of the learner is a correct answer,
The problem determination step that is expected to be corrected,
Characterized in that if the response time for the answer to the learning item is shorter than a specific time to determine that the problem is expected to take the shot,
Method for providing learning analysis information using big data. The method according to claim 1,
The learning item analysis step may model the percent correct percentage of the question as a function of response time from the plurality of response data of the corresponding question, wherein the specific time is the percentage of the percent correct answer per response time of the response time within 40%. Characterized in that the response time to the maximum amount of change is calculated as a specific time,
Method of providing learning item analysis information using big data. The method according to claim 2,
The specific time is characterized in that is updated when the number of the plurality of answer data is added a predetermined number,
Method of providing learning item analysis information using big data. The method according to claim 3,
The ratio of the correct answer and the wrong answer among the plurality of the plurality of answer data number, characterized in that the equal,
Method for providing learning analysis information using big data. The method according to claim 2,
The learning item analysis step, if the answer to the learner of the learner is the wrong answer,
Correct but incorrect problem determination step; And
Include the wrong problem determination steps,
The correct but wrong problem and the unknown wrong problem is characterized by determining based on the learner's predicted correct answer rate,
Method of providing learning item analysis information using big data. The method according to claim 5,
The predicted correct answer rate is calculated by the item response theory reflecting the inherent difficulty of the learning item and the estimated ability of the learner,
The unique difficulty (question difficulty) is a difficulty set by other learners for the question,
The learner's ability (learner ability) is characterized in that the learner's ability compared to the other learners for the other questions in the past,
Method of providing learning item analysis information using big data. The method according to claim 6,
The predicted correct answer is characterized in that calculated by the following equation,
Method of providing learning item analysis information using big data.

The method according to claim 7,
The learner's ability value is obtained by a Bayesian ability estimation algorithm,
Method of providing learning item analysis information using big data. The method according to claim 5,
The learning item analysis step,
When the predicted correct answer rate of the learner is more than 85%, characterized in that it is determined as 'can be correct but wrong problem',
Method for providing learning analysis information using big data. The method according to claim 9,
The learning item analysis step,
When the learner's predicted correct answer is less than 50%, characterized in that it is determined as the 'unknown problem wrong,'
Method of providing learning item analysis information using big data. The method according to claim 9 or 10,
The learning item analysis step,
When the predicted correct answer rate of the learner is more than 50% and less than 85%, according to the degree of the response time of the answer to the learning item, characterized in that it can be determined as the 'correct but wrong problem' or 'the wrong problem because it does not know,'
Method of providing learning item analysis information using big data. The method according to claim 11,
The required response time may be determined by comparing the actual answer response time for the learner's learning item with an estimate for the predicted answer response time of the learner.
Method of providing learning item analysis information using big data. The method according to claim 12,
The predicted answer response time is estimated based on the average answer response time of the learner and the average answer time of the learner of the other learners who answered the correct answer to the learning item,
Method for providing learning analysis information using big data.

Images