KR102659336B1 - Method, program, and device for quantifying correlation between units - Google Patents

Abstract

According to some embodiments of the present disclosure, a method for quantifying the relationship between units is disclosed. The method includes: predicting a correct answer rate for each of the second unit problem groups corresponding to some of the first unit problem groups; calculating the average value of the predicted correct rate for each problem group of the second unit; defining the average value as a correlation value of the second unit with respect to the first unit; and corresponding the correlation value to a point where the first unit on the vertical axis and the second unit on the horizontal axis intersect.

Description

{METHOD, PROGRAM, AND DEVICE FOR QUANTIFYING CORRELATION BETWEEN UNITS}

The present invention relates to a method for quantifying the correlation between units, and more specifically, to a method, program, and device for improving the ability to predict the probability of correct answer by quantifying the correlation between units.

Typically, methods of providing education can be divided into offline education through academies or visits, or online education that can be learned using ICT (Information and Communication Technology) such as e-learning. These educational methods generally provide learning materials and learning services tailored to the average learner's level in order to serve a large number of learners (or users).

For example, in the case of academy education, in most cases, the same educational content is provided to a large number of learners regardless of each individual's academic level, so the individual's academic level or the tendency of tests administered by each school It is difficult to receive learning that matches the level of difficulty.

Additionally, in the case of offline education, learners (or users) must travel to an academy to receive learning services, which requires considerable time and effort.

In recent years, due to the development of ICT, the provision of online education that can solve the problems of offline education has become common. E-learning provides an environment in which learners (or users) can learn at the desired time and place through the learner's terminal and solve expected problems about the learning content.

Additionally, because the learner can decide on the learning method and learning progress, learning by level is possible regardless of time and place.

However, there are limitations in providing learning materials and learning services tailored to tendencies or individual vulnerabilities. As can be seen in the national academic achievement evaluation, the average academic level of students at each school is different, and the difficulty and type of test questions for each school are different.

Because teachers write test questions taking into account the students' average academic level and environment, as well as the tendencies of their fellow teachers, each school's past questions have their own unique tendencies.

Therefore, because each school's propensity to present test questions is different, it is difficult to provide expected questions or learning services tailored to the difficulty level or type of test questions presented by each school. Additionally, because each learner has different vulnerabilities, it is difficult to improve learning efficiency through existing online education.

Therefore, there is a need to overcome the problems of existing education provision methods and provide user-customized problems that can improve the learning efficiency of learners, and there is a need to develop technology to calculate the probability of the user getting the correct answer in order to provide customized problems. This is a desperate situation.

Meanwhile, in the case of mathematics, compared to other subjects, units are often related in terms of content, and if learning for a specific unit is insufficient, learning for other units becomes difficult or problems are difficult to solve. In the case of subjects such as mathematics, where there is a high correlation between units, the prediction power of the correct answer rate can be improved by considering the characteristics of each unit, but it is still difficult to find research and development status or literature on this.

Republic of Korea Patent Registration No. 10-2245286 (registered on April 21, 2021)

The present invention was created to solve the above problems, and the purpose of the present invention is to provide a method, program, and device for quantifying and visualizing the relationship between units.

The technical problems to be achieved through the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned can be clearly understood by those skilled in the art from the description below. It could be.

A method of quantifying the inter-unit correlation according to some embodiments of the present disclosure includes: predicting the percentage of correct answers for each of the second unit problem groups corresponding to some of the first unit problem groups; calculating the average value of the predicted correct rate for each problem group of the second unit; defining the average value as a correlation value of the second unit with respect to the first unit; and corresponding the correlation value to a point where the first unit on the vertical axis and the second unit on the horizontal axis intersect.

According to some embodiments of the present disclosure, the step of predicting the correct answer rate may include selecting problem groups corresponding to a first characteristic among problem groups of the first unit stored in the storage of the device; Extracting problem groups n times using the problem groups selected in the first unit as samples; and calculating a predicted correct answer rate for each of the second unit problem groups corresponding to the extracted first unit problem group set.

According to some embodiments of the present disclosure, the selecting step includes checking the difficulty level of problem groups stored in the storage unit of the device, and the first feature may be a first difficulty section.

According to some embodiments of the present disclosure, the extraction may be restoration extraction.

According to some embodiments of the present disclosure, calculating the predicted correct rate includes defining the extracted problem group set of the first unit as a first sequence; and calculating a predicted correct answer rate for each problem group of the second unit using the first sequence as history.

According to some embodiments of the present disclosure, the predicted correct rate may be calculated using a knowledge tracking model.

According to some embodiments of the present disclosure, it further includes learning the knowledge tracking model, wherein the learning the knowledge tracking model includes defining an extracted problem group set as an input sequence; Calculating an attention information set including attention information by quantifying the degree of relationship between problem groups included in the input sequence; Applying guide masking to the attention information set; And it may include learning the knowledge tracking model using the attention information set to which the guide masking is applied.

According to some embodiments of the present disclosure, the guide masking is performed based on at least one of a preset association between units stored in the storage of the device, information about the problem groups, and/or learning records about the problem groups. It can be set based on

A computer program stored in a computer-readable storage medium according to some embodiments of the present disclosure, when executed on a processor of a device, performs steps of quantifying the relationship between units by the processor of the device, the steps comprising: Problem of the first unit Predicting the percentage of correct answers for each of the second unit problem groups corresponding to some of the groups; calculating the average value of the predicted correct rate for each problem group of the second unit; defining the average value as a correlation value of the second unit with respect to the first unit; and corresponding the correlation value to a point where the first unit on the vertical axis and the second unit on the horizontal axis intersect.

According to some embodiments of the present disclosure, the step of predicting the correct answer rate may include selecting problem groups corresponding to a first characteristic among problem groups of the first unit stored in the storage of the device; Extracting problem groups n times using the problem groups selected in the first unit as samples; and calculating a predicted correct answer rate for each of the second unit problem groups corresponding to the extracted first unit problem group set.

According to some embodiments of the present disclosure, the selecting step includes checking the difficulty level of problem groups stored in the storage unit of the device, and the first feature may be a first difficulty section.

According to some embodiments of the present disclosure, the extraction may be restoration extraction.

According to some embodiments of the present disclosure, calculating the predicted correct rate includes defining the extracted problem group set of the first unit as a first sequence; and calculating a predicted correct answer rate for each problem group of the second unit using the first sequence as history.

According to some embodiments of the present disclosure, the predicted correct rate may be calculated using a knowledge tracking model.

An apparatus for quantifying the relationship between units according to some embodiments of the present disclosure includes: a storage unit storing at least one program command; and a processor for executing the at least one program command, wherein the processor predicts a correct answer rate for each of the second unit problem groups corresponding to some of the first unit problem groups, and The average value of the predicted correct rate for each problem group is calculated, the average value is defined as the correlation value of the second unit with respect to the first unit, and the average value is defined as the correlation value of the second unit with respect to the first unit, and is located at the point where the first unit on the vertical axis and the second unit on the horizontal axis intersect. The above correlation value can be corresponded.

According to some embodiments of the present disclosure, in predicting the correct rate, problem groups corresponding to the first feature are selected from among the problem groups of the first unit stored in the storage of the device, and selected in the first unit. Using the selected problem groups as samples, problem groups can be extracted n times, and the predicted correct answer rate can be calculated for each of the second unit problem groups corresponding to the extracted first unit problem group set.

According to some embodiments of the present disclosure, when selecting the problem groups, the difficulty level of the problem groups stored in the storage unit of the device is checked, and the first feature may be a first difficulty section.

According to some embodiments of the present disclosure, the extraction may be restoration extraction.

According to some embodiments of the present disclosure, in calculating the predicted correct rate, the extracted problem group set of the first unit is defined as a first sequence, and each of the second unit is defined using the first sequence as a history. The predicted correct rate for problem groups can be calculated.

According to some embodiments of the present disclosure, the predicted correct rate may be calculated using a knowledge tracking model.

According to embodiments of the present disclosure, the correlation between units can be intuitively confirmed by quantifying and visualizing the correlation between units. Additionally, the predictive power of the knowledge tracking model can be improved based on the correlation between units.

The effects that can be obtained through the present disclosure are not limited to the effects mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the description below. will be.

Various embodiments of the present disclosure are described with reference to the drawings, where like reference numerals are used to collectively refer to like elements. In the following examples, for purposes of explanation, numerous specific details are set forth to provide a thorough understanding of one or more embodiments. However, it will be clear that such embodiment(s) may be practiced without these specific details.
1 is a block diagram illustrating a device for quantifying and visualizing the relationship between units according to some embodiments of the present disclosure.
FIG. 2 is a flowchart illustrating an example of a method of quantifying the correlation between units and visualizing the correlation between units using the numerical values, according to some embodiments of the present disclosure.
Figure 3 is a diagram for explaining an example of a method for predicting the percentage of correct answers in Figure 2.
FIG. 4 is a diagram illustrating an example of a method for predicting a correct rate and calculating an average value of the predicted correct rate according to some embodiments of the present disclosure.
FIG. 5 is a diagram illustrating an example of visualizing the relationship between units according to some embodiments of the present disclosure.
Figure 6 is a flowchart illustrating the steps of learning a knowledge tracking model according to some embodiments of the present disclosure.
FIG. 7 is a diagram illustrating an example of a method for learning the knowledge tracking model of FIG. 6.
FIG. 8 is a diagram illustrating an example of guide masking according to some embodiments of the present disclosure.

Hereinafter, various embodiment(s) of the device and the control method of the device according to the present disclosure will be described in detail with reference to the drawings. However, regardless of the reference numerals, identical or similar components are assigned the same reference numbers and overlapping The explanation will be omitted.

The purpose and effects of the present disclosure, and technical configurations for achieving them, will become clear by referring to the embodiments described in detail below along with the accompanying drawings. In describing one or more embodiments of the present disclosure, if it is determined that a detailed description of related known technology may obscure the gist of at least one embodiment of the present disclosure, the detailed description will be omitted.

The terms in this disclosure are terms defined in consideration of the functions in this disclosure, and may vary depending on the intention or custom of the user or operator. In addition, the attached drawings are only intended to facilitate understanding of one or more embodiments of the present disclosure, and the technical idea of the present disclosure is not limited by the attached drawings, and all changes included in the spirit and technical scope of the present disclosure are not limited. , should be understood to include equivalents or substitutes.

The suffixes “module” and “part” for components used in the following description are given or used interchangeably only considering the ease of preparation of the present disclosure, and do not have distinct meanings or roles in themselves.

Terms containing ordinal numbers, such as first, second, etc., may be used to describe various components, but the components are not limited by the terms. The above terms are used only for the purpose of distinguishing one component from another. Accordingly, the first component mentioned below may also be the second component within the technical spirit of the present disclosure.

Singular expressions include plural expressions unless the context clearly dictates otherwise. That is, unless otherwise specified or clear from context to indicate a singular form, the singular in this disclosure and claims should generally be construed to mean “one or more.”

In the present disclosure, terms such as “comprising,” “includes,” or “have” are intended to indicate the presence of features, numbers, steps, operations, components, parts, or combinations thereof described in the present disclosure. It should be understood that this does not exclude in advance the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

In the present disclosure, the term “or” should be understood not as “or” in the exclusive sense but as “or” in the connotative sense. That is, unless otherwise specified or clear from context, “X utilizes A or B” is intended to mean one of the natural implicit substitutions. That is, either X uses A; Either X uses B; Or, if X uses both A and B, “X uses A or B” can apply to either of these cases. Additionally, the term “and/or” as used in this disclosure should be understood to refer to and include all possible combinations of one or more of the related listed items.

As used in this disclosure, the terms “information” and “data” may be used interchangeably.

Unless otherwise defined, all terms (including technical and scientific terms) used in this disclosure may be used with meanings that can be commonly understood by those skilled in the art. Additionally, terms defined in commonly used dictionaries are not overly interpreted unless specifically defined.

However, the present disclosure is not limited to the embodiments disclosed below and may be implemented in various different forms. Only some embodiments of the present disclosure are provided to fully inform those skilled in the art of the present disclosure of the scope of the present disclosure, and the present disclosure is only defined by the scope of the claims. Therefore, the definition should be made based on the content throughout this disclosure.

The processor of the device of the present disclosure can quantify the correlation between different units, and use the quantified values to improve the ability to predict the percentage of correct answers to a problem. Hereinafter, a method of calculating and visualizing the correlation between different units will be described with reference to FIGS. 1 to 5.

1 is a block diagram illustrating a device for quantifying and visualizing the relationship between units according to some embodiments of the present disclosure.

Referring to FIG. 1, the device 100 that quantifies the relationship between units and visualizes it may include a storage unit 110, a processor 120, and a communication unit 130. The components shown in FIG. 1 are not essential for implementing the device 100, so the device 100 described in this disclosure may have more or less components than those listed above.

Each component of the device 100 of the present disclosure may be integrated, added, or omitted depending on the specifications of the device 100 that is actually implemented. That is, as needed, two or more components may be combined into one component or one component may be subdivided into two or more components. In addition, the functions performed by each block are for explaining the embodiments of the present disclosure, and the specific operations or devices do not limit the scope of the present invention.

The device 100 described in this disclosure may include all devices that perform at least one of transmitting and receiving data, content, services, and applications.

The device 100 of the present disclosure is capable of pairing or connecting to other devices, external servers, etc. through a wired/wireless network, and can transmit/receive predetermined data through this. You can. In this case, data transmitted/received through the device 100 may be converted before transmission/reception.

The device 100 of the present disclosure includes, for example, a standing device such as a server, a personal computer (PC), a microprocessor, a mainframe computer, a digital processor, a device controller, a smart phone, All mobile devices (mobile devices or handheld devices) such as tablet PCs, laptops, etc. may be included. However, the present disclosure is not limited to this.

In this disclosure, “server” refers to a device or system that supplies data to or receives data from various types of user terminals, that is, clients. A server, for example, a web server or portal server that provides web pages, web content or web services, or an advertisement that provides advertising data. Advertising server, content server providing content, SNS server providing SNS (Social Network Service), service server provided by the manufacturer, VoD (Video on Demand) or It may include an MVPD (Multichannel Video Programming Distributor) for providing streaming services, a service server for providing paid services, etc.

In the present disclosure, when the name device 100 is used, it means a server depending on the context, but may also mean a fixed device or a mobile device, and may be used to include all of them unless specifically mentioned.

The storage unit 110 may store data supporting various functions of the device 100. The storage unit 110 may store a plurality of application programs (application programs or applications) running on the device 100, data for operation of the device 100, commands, and at least one program command. . At least some of these applications can be downloaded from an external server through wireless communication. Additionally, at least some of these application programs may exist on the device 100 from the time of shipment for the basic functions of the device 100. Meanwhile, the application program may be stored in the storage unit 110, installed on the device 100, and driven by the processor 120 to perform the operation (or function) of the device 100.

The storage unit 110 may store any type of information generated or determined by the processor 120 and any type of information received through the communication unit 130.

The storage unit 110 has a flash memory type, a hard disk type, a Solid State Disk type, an SDD type (Silicon Disk Drive type), and a multimedia card micro type. type), card-type memory (e.g. SD or XD memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), EEPROM (electrically erasable programmable memory) It may include at least one type of storage medium among read-only memory (PROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, and optical disk. The device 100 may be operated in connection with web storage that performs the storage function of the storage unit 110 on the Internet.

In the present disclosure, the storage unit 110 may store a plurality of problems. According to some embodiments of the present disclosure, all problems included in the storage unit 110 may be grouped according to certain criteria, and problem groups formed accordingly may be stored in the storage unit 110. Accordingly, each of the problem groups may include at least one problem.

Here, a difficulty value may be assigned to each of the plurality of problems. Additionally, which unit of the plurality of problems each problem corresponds to may be assigned to each of the plurality of problems. Accordingly, problem groups can be grouped according to unit and difficulty value. In other words, one problem group may include problems with substantially the same level of difficulty among problems included in the same unit. However, the present disclosure is not limited to this.

Problems included in the problem group in this disclosure may be problems related to one of the subjects such as elementary school, middle school, and high school. For example, problems may include elementary math problems, middle school math problems, high school math problems, elementary Korean language problems, middle school Korean language problems, etc. However, the present disclosure is not limited to this.

In the present disclosure, a unit may refer to a learning unit centered around one topic or content. For example, if the problem is an advanced math problem, the unit may consist of a first unit learning addition, subtraction, multiplication, and division of polynomials, a second unit learning polynomials and identities, etc. However, this is an example, and the present disclosure is not limited thereto. In the present disclosure, each of the first unit and the second unit is not limited to meaning a specific unit as above, but may be used as a term to represent any two different units among all units.

In the present disclosure, the difficulty value assigned to a problem group is a fixed value, and when a plurality of problem groups are stored in the storage unit 110, it may be assigned to each of the plurality of problem groups and stored together with the plurality of problem groups. . However, the present disclosure is not limited to this.

According to some embodiments of the present disclosure, a plurality of problem groups may be stored in various program languages (eg, HTML, etc.). When a plurality of problem groups are stored in a program language as in the present disclosure, when the size of the display unit of the user terminal changes, the number of texts contained in one line can be adjusted accordingly.

In the present disclosure, the difficulty value of a problem group may be a value calculated by analyzing a data set provided from a preset server. However, the present disclosure is not limited to this. The preset server may be a server of a preset institution, or may be a server of an education platform different from the preset institution. However, the present disclosure is not limited to this.

Students can enter the correct answers they wrote on the test into the education platform they are using. Therefore, the server of the education platform may store the correct answer rate for each problem in the storage unit 110, and since the correct answer rate for each of a plurality of problems is required to calculate the difficulty value of the problem provided by the preset institution, the preset correct answer rate for each problem may be stored in the storage unit 110. It is appropriate to think of the server as an education platform server.

The data set includes multiple problems, points allocated to each of multiple problems, problem type of multiple problems, percentage of correct answers to each of multiple problems, raw scores by grade of multiple users who solved multiple problems, and multiple users who solved multiple problems. Standard scores for each level of users, cumulative ratio for each level of multiple users who solved multiple problems, average score of multiple users who solved multiple problems, and standard score of multiple users who solved multiple problems May contain deviations. However, the present disclosure is not limited to this.

The processor 120 may typically control the overall operation of the device 100 in addition to operations related to application programs. The processor 120 can provide or process appropriate information or functions by processing signals, data, information, etc. input or output through the components of the device 100 or by running an application program stored in the storage unit 110. there is.

The processor 120 may control at least some of the components of the device 100 to run an application program stored in the storage unit 110. Furthermore, the processor 120 may operate in combination with at least two or more of the components included in the device 100 to run an application program.

The processor 120 may consist of one or more cores and may be any of a variety of commercial processors. For example, the processor 120 may include a central processing unit (CPU), a general purpose graphics processing unit (GPUGP), a tensor processing unit (TPU), etc. You can. However, the present disclosure is not limited to this.

The processor 120 of the present disclosure may be configured with a dual processor or other multiprocessor architecture. However, the present disclosure is not limited to this.

The processor 120 can quantify the correlation between units. Specifically, the processor 120 predicts the correct answer rate for each of the second unit problem groups corresponding to some of the first unit problem groups among the problems stored in the storage unit 110, and the second unit problem groups The average value of the predicted correct rate for each can be calculated. The processor 120 may quantify the correlation between units by defining the average value as a correlation value of the second unit with respect to the first unit. When the processor 120 calculates the predicted correct rate, the processor 120 may calculate the predicted correct rate using a knowledge tracking model.

In the present disclosure, the correlation between units may mean the degree of correlation between two different units. For example, if the quantified correlation value between the first unit and the second unit is large, it can be expected that the percentage of correct answers in the first unit and the percentage of correct answers in the second unit will be highly correlated (i.e., proportional), and the When the quantified correlation value between the first unit and the second unit is small, the correlation between the correct answer rate of the first unit and the correct answer rate of the second unit can be expected to be low. However, the present invention is not limited to this.

The communication unit 130 may include one or more modules that enable wired/wireless communication between the device 100 and a wired/wireless communication system, between the device 100 and another device, or between the device 100 and an external server. there is. Additionally, the communication unit 130 may include one or more modules that connect the device 100 to one or more networks.

The communication unit 130 refers to a module for wired/wireless Internet access and may be built into or external to the device 100. The communication unit 130 may be configured to transmit and receive wired/wireless signals.

The communication unit 130 uses technical standards or communication methods for mobile communication (e.g., Global System for Mobile communication (GSM), Code Division Multi Access (CDMA), Code Division Multi Access 2000 (CDMA2000), and EV-DO ( Enhanced Voice-Data Optimized or Enhanced Voice-Data Only), Wideband CDMA (WCDMA), High Speed Downlink Packet Access (HSDPA), High Speed Uplink Packet Access (HSUPA), Long Term Evolution (LTE), Long Term (LTE-A) It is possible to transmit and receive wireless signals with at least one of a base station, an external terminal, and a server on a mobile communication network built according to (Evolution-Advanced), etc.).

Wireless Internet technologies include, for example, WLAN (Wireless LAN), Wi-Fi (Wireless-Fidelity), Wi-Fi (Wireless Fidelity) Direct, DLNA (Digital Living Network Alliance), WiBro (Wireless Broadband), and WiMAX (Worldwide). There may be Interoperability for Microwave Access), High Speed Downlink Packet Access (HSDPA), High Speed Uplink Packet Access (HSUPA), Long Term Evolution (LTE), and Long Term Evolution-Advanced (LTE-A). However, the communication unit 130 can transmit and receive data according to at least one wireless Internet technology, including Internet technologies not listed above.

In addition, the communication unit 130 may be configured to transmit and receive signals through short range communication. The communication unit 130 includes Bluetooth (Bluetooth??), RFID (Radio Frequency Identification), Infrared Data Association (IrDA), UWB (Ultra-Wideband), ZigBee, NFC (Near Field Communication), and Wi-Fi (Wireless). -Fidelity), Wi-Fi Direct, and Wireless USB (Wireless Universal Serial Bus) technologies can be used to perform short-distance communication. The communication unit 130 may support wireless communication through wireless area networks. Local area wireless networks may be wireless personal area networks.

The device 100 according to some embodiments of the present disclosure may be connected to a test device and a wired/wireless network through the communication unit 130. Here, “wired/wireless network” collectively refers to a communication network that supports various communication standards or protocols for pairing or/and data transmission/reception between the device 100 and the test device. These wired/wireless networks include all communication networks that are currently or will be supported in the future according to the standard, and can support one or more communication protocols for them.

In the present disclosure, the device 100 may be connected to a user terminal for communication through the communication unit 130. Here, the user terminal may refer to a student terminal that performs learning by solving a plurality of problems stored in the storage unit 110 of the device 100, and the user terminal may be a mobile phone, smart phone, or tablet PC ( It may be a mobile device such as a tablet PC, ultrabook, etc. However, the present disclosure is not limited to this.

In the present disclosure, the processor 120 of the device 100 may check the storage unit 110 for information necessary for predicting the correct answer rate for each of the second unit problems corresponding to some of the problem groups of the first unit. At this time, the necessary information may mean information about the problem group, the user's learning record, etc. Information about the problem group may include problem number ID, difficulty value, etc., which are unique values for each problem group, and the user's learning record may include problem number ID, which is a unique value of the problem group solved by the user, and the time when the user solved the problem. , may include whether or not there is an incorrect answer to the user's problem. However, the present disclosure is not limited to this.

Meanwhile, according to some embodiments of the present disclosure, the processor 120 may quantify the correlation between a plurality of units stored in the storage unit 110 and visualize it. At this time, the processor 120 may calculate a correlation value that quantifies the correlation between units, and express the correlation between units in a heatmap format based on the correlation value. Heat map is a word that combines heat, which means heat, and map, which means map. It displays various information that can be expressed in color as a visual graphic in the form of heat distribution on a certain image. means that This is explained in detail with reference to FIGS. 2 to 5.

FIG. 2 is a flowchart illustrating an example of a method of quantifying the correlation between units and visualizing the correlation between units using the numerical values, according to some embodiments of the present disclosure. Figure 3 is a diagram for explaining an example of a method for predicting the percentage of correct answers in Figure 2. FIG. 4 is a diagram illustrating an example of a method for predicting a correct rate and calculating an average value of the predicted correct rate according to some embodiments of the present disclosure.

Referring to FIG. 2, the processor 120 can quantify the correlation between units and visualize the correlation between units using the numerical value.

In step S201, the processor 120 may predict a correct answer rate for each of the problem groups of the second unit corresponding to some of the problem groups of the first unit among the problem groups stored in the storage unit 110. The step of calculating the predicted correct rate for each of the problem groups of the second unit that corresponds to some of the problem groups of the first unit may be as described below.

Referring further to FIGS. 3 and 4 , in step S301, the processor 120 may select problem groups corresponding to the first characteristic among problem groups of the first unit stored in the storage unit 110. At this time, the first feature may be a first difficulty section. Specifically, the processor 120 may check the difficulty level of problem groups stored in the storage unit 110. Thereafter, the processor 120 may select problem groups corresponding to the first difficulty level in the first unit.

At this time, the first difficulty section may be a section in which the difficulty value assigned to the problem group is greater than the preset first value (x1) and less than the preset second value (x2). According to some embodiments of the present disclosure, the first difficulty section may be a section including a difficulty value corresponding to a median value.

If the first difficulty section includes extreme values rather than the middle value, the reliability of the predicted correct answer rate for each of the problem groups in the second unit may decrease. In the first unit, for problem groups with low difficulty values, a relatively high correct rate will be calculated, and for problem groups with high difficulty values, a relatively low correct rate will be calculated. Therefore, when calculating the predicted correct rate, in addition to the correlation between units, the difficulty level will be calculated. There may be an additional variable called . Therefore, to ensure reliability, the first difficulty section may be a section containing a difficulty value corresponding to the median value. However, the present disclosure is not limited to this.

For example, referring to FIG. 4, when the range of difficulty values of problem groups stored in the storage unit 110 is 0 to 900, the first difficulty section has a difficulty value greater than 400 (e.g., x1) and 500 ( For example, it may be a section smaller than x2). That is, the first difficulty section may be a section including 450, which is the median value. The difficulty value of problem group 25 is 422, the difficulty value of problem group 26 is 176, the difficulty value of problem group 27 is 483, the difficulty value of problem group 28 is 563, and the difficulty value of problem group 29 is 277. You can. Accordingly, among all problem groups in the first unit, problem groups No. 25 and problem group No. 27, whose difficulty values are included in the first difficulty range, may be selected. However, since this is an example, the present disclosure is not limited thereto.

In step S303, the processor 120 may extract problem groups n times (where n is a natural number of 1 or more) using the problem groups selected in the first unit as samples. That is, the processor 120 may extract problem groups n times using problem groups corresponding to the first difficulty level in the first unit as samples. At this time, the extraction may be restoration extraction. Therefore, already extracted problem groups may be extracted repeatedly. In contrast, in the case of non-restored extraction, the number of extractions is limited to the number of selected problem groups. In other words, if the number of selected problem groups is less than n, which is the number of extractions, the number of extractions cannot be satisfied with non-restorative extraction. By extracting n times with restoration extraction, the processor 120 can satisfy the number of extractions regardless of the number of selected problem groups. However, the present disclosure is not limited to this.

For example, referring again to FIG. 4, the processor 120 may extract problem group No. 25 and problem group No. 27 selected in the first unit five times through restoration extraction. Accordingly, problem groups 25-27-27-25-27 can be extracted in order. However, the number of restoration extractions (5) and the extraction order are illustrative, and the present disclosure is not limited thereto.

In step S305, the processor 120 may calculate a predicted correct answer rate for each of the problem groups of the second unit corresponding to the extracted problem group set of the first unit. The processor 120 may define the extracted problem group set of the first unit as the first sequence. The first sequence may refer to a learning record for the extracted problem group set of the first unit. At this time, the learning record may mean all records related to the problem group and all records related to the user who solved the problem group. For example, the learning record may include a problem number ID, which is a unique value of the problem group solved by the user, the order in which the user solved the problem group, the time when the user solved the problem, whether the user gave an incorrect answer to the problem, etc.

The processor 120 may calculate the predicted correct answer rate for each problem group in the second unit using the first sequence as history. Using the first sequence as history may mean randomly generating a virtual learning record based on the users' learning records stored in the storage unit 110. That is, the processor 120 may generate a virtual learning record for the selected problem groups of the first unit and calculate the predicted correct answer rate for all problem groups of the second unit.

According to some embodiments of the present disclosure, the processor 120 may calculate a predicted correct answer rate for each problem group of the second unit assuming that all problem group sets of the first unit of the first sequence are correct. According to another embodiment, the processor 120 may calculate the predicted correct answer rate (or incorrect answer rate) for each problem group of the second unit, assuming that all problem group sets of the first unit of the first sequence are incorrect. .

If only part of the problem group set of the first unit of the first sequence is correct or partially incorrect, the degree of influence of the first unit on the second unit may not be clearly derived. Only when the problem group set of the first unit of the first sequence is all correct or all incorrect can the degree of influence of the first unit on the second unit be clearly derived, thereby improving the reliability of the relationship between the first unit and the second unit. . In other words, when creating a virtual learning record for the selected problem groups of the first unit, it is desirable to assume that all problem group sets of the first unit of the first sequence are correct. However, the present disclosure is not limited to this, and the processor 120 may generate a virtual learning record that satisfies the conditions for deriving the relationship between the first unit and the second unit. For example, the predicted correct rate can be calculated assuming that the problem group set of the first unit of the first sequence is partially correct or partially incorrect. However, the following description will be made assuming that all problem group sets of the first unit of the first sequence are answered correctly.

For example, referring again to FIG. 4, the processor 120 may define the extracted problem group set of the first unit, No. 25-27-27-25-27, as the first sequence. The processor 120 can calculate the predicted correct answer rate for each problem group in the second unit, assuming that all numbers 25-27-27-25-27 of the first sequence are correct. At this time, the predicted correct answer rate for each of problem groups 72 to 76, which are all problem groups included in the second unit, can be calculated. For example, the predicted percentage correct for problem group 72 is approximately 0.622, the predicted percentage correct for problem group 73 is approximately 0.602, the predicted percentage correct for problem group 74 is 0.297, the predicted percentage correct for problem group 75 is 0.902, The predicted correct answer rate for problem group 76 can be derived as 0.855.

At this time, the processor 120 may calculate the predicted correct rate through the knowledge tracking model 400. The method by which the processor 120 learns the knowledge tracking model 400 that calculates the predicted correct rate is as described in FIG. 6 described later.

Referring again to FIG. 2, in step S203, the processor 120 may calculate the average value of the predicted correct answer rate for each of the problem groups of the second unit. By calculating the average value of the predicted correct rate of problem groups of the second unit with respect to the first unit, the processor 120 can generalize the correlation of the second unit with the first unit.

For example, referring again to FIG. 4, the processor 120 calculates the average value of the predicted correct answer rate for all problem groups of the second unit calculated in step S201 (e.g., problem groups 72 to 76). You can. The processor 120 calculates the predicted correct rate for problem group No. 72, which is about 0.622, the predicted correct rate for problem group No. 73, which is about 0.602, the predicted correct rate for problem group No. 74, which is about 0.297, and the predicted correct rate for problem group number 75, which is about 0.297. About 0.902, and about 0.696, which is the average value for the predicted correct rate for problem group 76 of about 0.855, can be calculated.

In step S205, the processor 120 may define the average value as a correlation value of the second unit with respect to the first unit. That is, through this, the processor 120 can quantify the correlation between units.

For example, if the correlation value of the second unit with respect to the first unit is large, the average value is large, and if the problem groups included in the first unit are guessed correctly, the probability that the problem groups included in the second unit are also correct is high. will be. That is, in this case, it can be predicted that the correlation between the first unit and the second unit is relatively large. If the correlation value of the second unit with respect to the first unit is small, the average value is small, and even if the problem groups included in the first unit are correct, there is a high probability that the problem groups included in the second unit are incorrect. That is, in this case, it can be predicted that the correlation between the first unit and the second unit is relatively small.

The correlation value can have values between 0 and 1. For example, the greater the correlation between the second unit and the first unit, the closer the correlation value of the second unit to the first unit may be to 1. Conversely, the smaller the correlation between the second unit and the first unit, the closer the correlation value of the second unit to the first unit may be to 0. However, the present disclosure is not limited to this.

The processor 120 may calculate the overall correlation value of one unit among the units to which all problem groups included in the storage unit 110 belong. That is, the processor 120 can calculate correlation values equal to the square of the total number of units.

The processor 120 can visualize the correlation between units based on the correlation values. The processor 120 can use the correlation values to express the correlation between units in a heatmap format.

FIG. 5 is a diagram illustrating an example of visualizing the relationship between units according to some embodiments of the present disclosure.

Referring further to FIG. 5, in step S207, the processor 120 may correspond to a correlation value at a point where the first unit on the vertical axis and the second unit on the horizontal axis intersect. That is, the processor 120 may generate a table in which the number of the unit to be extracted is set on the vertical axis, and the number of the unit for which the predicted percentage of correct answers is calculated is set on the horizontal axis. For example, the number of the first unit may be set on the vertical axis, and the number of the second unit may be set on the horizontal axis. Accordingly, the processor 120 may generate a table in which the correlation values of the units on the horizontal axis with respect to the units on the vertical axis correspond to the points where the units on the vertical axis and the units on the horizontal axis intersect.

In step S209, the processor 120 may express the correlation value in a corresponding color. The processor 120 may display each cell included in the table in a different color depending on the size of the correlation value. For example, the processor 120 can display a brighter color as the correlation value approaches 1, and display it in a darker color as the correlation value approaches 0. That is, through this, the processor 120 can express the relationship between units in a heatmap format. However, the present disclosure is not limited to this.

Because the same units are highly related to each other, the colors can be expressed brightly in the diagonal direction. In the heat map, different colors for each vertical line can indicate the level of difficulty for each unit. In other words, the unit on the horizontal axis to which the dark vertical line corresponds has a low percentage of correct answers overall for all units, so it can be predicted that it is a unit with relatively high difficulty. In addition, the units on the horizontal axis to which the bright vertical lines correspond have a high overall percentage of correct answers for all units, so it can be predicted that they are units with relatively low difficulty. However, the present disclosure is not limited to this.

In addition, although only the method of quantifying the association of the second unit with the first unit has been described above, the method of quantifying the association of the first unit with the second unit can also be performed separately. That is, the relevance of the second unit to the first unit and the correlation of the first unit to the second unit may not be the same. However, the present disclosure is not limited to this, and the processor 120 may quantify the association between the first unit and the second unit by only performing a method of quantifying the association of the second unit with the first unit.

According to some embodiments of the present disclosure, the processor 120 can quantify the correlation between units and visualize the correlation between units using the numerical value. Accordingly, the connection between units can be intuitively confirmed. In addition, the accuracy of the predicted correct rate can be improved by applying the correlation between units derived from this to predict the correct rate. In other words, the predictive power of the knowledge tracking model can be improved by using the correlation between units. In addition, the heatmap format that quantifies and visualizes the correlation between units can also be used as a measure to measure whether the learning of the knowledge tracking model described later has been performed well.

In step S201 of FIG. 2, the knowledge tracking model 400 that predicts the correct answer rate for each of the problem groups in the second unit can be learned as follows.

Figure 6 is a flowchart illustrating the steps of learning a knowledge tracking model according to some embodiments of the present disclosure. FIG. 7 is a diagram illustrating an example of a method for learning the knowledge tracking model of FIG. 6.

The knowledge tracking model may include multiple layers. Here, the knowledge tracking model may include at least one attention layer. For example, at least one attention layer may be provided in plural numbers within the knowledge tracking model, and a plurality of attention layers may be respectively arranged between different layers within the knowledge tracking model.

Referring to FIG. 6, in step S601, the processor 120 may randomly extract some of the problem groups stored in the storage unit 110 of the device. However, the present disclosure is not limited to this, and the processor 120 may select some of the problem groups based on preset criteria.

For example, further referring to FIG. 7, the entire problem groups stored in the storage unit 110 of the device are problem groups 25 to 29 included in the first unit and problem groups 72 to 76 included in the second unit. , it may be numbers 101 to 105 included in the third section. Here, the processor 120 can extract numbers 76, 27, 25, 75, 102, 72, and 105.

In step S603, the processor 120 may define the extracted problem group set as an input sequence. For example, the processor 120 may define the problem group set 76-27-25-75-102-72-105 as an input sequence.

In step S605, the processor 120 may calculate the attention information set 700 including attention information by quantifying the degree of relationship between problem groups included in the input sequence.

In other words, the input sequence is input to the knowledge tracking model, and the attention information set 700 can be calculated from each attention layer included in the knowledge tracking model based on the input sequence. Accordingly, the attention information set 700 may include attention information that is a numerical value of the degree of relationship between problem groups included in the input sequence. At this time, the degree of relevance may mean the degree of similarity between problem groups or the degree to which another problem group influences one problem group. However, the present disclosure is not limited to this.

In step S607, the processor 120 may individually apply guide masking 701 and 702 to the attention information set 700. When applying guide masking 701 and 702, the processor 120 may set the masking portion differently for each attention information set 700. That is, according to the guide masking 701 and 702, the processor 120 can set the masking portion differently for the attention information set 700 calculated from each attention layer.

Guided masking 701 and 702 may refer to a masking operation in which the masking portion is determined differently for each attention layer (i.e., attention information set 700) based on a standard set by the operator.

Specifically, the guide masking (701, 702) may include basic masking (701) that prevents attention from being performed on information corresponding to future locations when predicting the correct answer rate of a problem group. Basic masking 701 may be a masking operation applied to prevent the predicted correct answer rate from being calculated based on problems that the user has not yet solved. Accordingly, the attention information set 700 to which the basic masking 701 is applied can be expressed as a black triangle in the upper right corner, as shown in Figure 7 8, and the attention information is 0. Basic masking 701 can be applied to the entire attention layer in the knowledge tracking model. However, the present disclosure is not limited to this.

According to some embodiments of the present disclosure, the guide masking 701 and 702, in addition to the basic masking 701, differently mask the portion for each attention layer (i.e., attention information set 700) based on a preset standard. You can set it. For example, guided masking is masking for each attention information set 700 based on at least one of a preset association between units stored in the storage unit 110, information about problem groups, and learning records about problem groups. This may mean setting the parts differently. At this time, the masked part in the attention information set 700 may mean a part to which attention information is applied as 0. In other words, it may mean a part of the attention information set 700 that is considered irrelevant.

The attention layer may be an attention block, one of various blocks included in the knowledge tracking model. Within the attention layer, guide masking may be applied at the encoding stage or at the decoding stage.

Here, the preset relationship between units may mean the relationship between one unit and another unit among all units set by an expert. Experts may refer to teachers, professors, lecturers, researchers, etc. in the relevant field. Additionally, information about problem groups and learning records about problem groups may refer to all information stored in the storage unit 110 when a user solves a problem. For example, information about problem groups may include eigenvalues for each problem group, units containing each problem group, level of difficulty corresponding to each problem group, etc., and learning records for problem groups can be recorded by the user. It may include information on whether the user got the problem right or wrong (e.g., whether the answer was incorrect), the order in which the user solved the problem, and the time the user solved the problem.

According to an embodiment of the present disclosure, an expert may set the relationship between one unit and another unit as one of a plurality of steps. The plurality of stages may mean strong, medium, weak, 1st stage, 2nd stage, 3rd stage, 4th stage, 5th stage, etc. However, the present disclosure is not limited to this.

The processor 120 can quantify this according to the correlation set by the expert. For example, the processor 120 can quantify the largest value when the correlation between units is strong, the medium value when the correlation between units is weak, and the smallest value when the correlation between units is weak. In other words, the larger the numerical value, the more related the units are to each other. Therefore, the processor 120 can apply attention information so that highly related units can influence each other when calculating the predicted correct rate (i.e., attention information masking may not be applied). Accordingly, the frequency of applying masking to the attention information included in the attention information set 700 can be reduced between problem groups included in the unit. Conversely, the smaller the numerical value, the less correlation between units, so the processor 120 may not apply attention information to each other so as not to influence each other when calculating the predicted correct rate (i.e., masking is applied to attention information). applicable). Accordingly, the frequency of applying masking to the attention information included in the attention information set 700 can be increased among problem groups included in the unit.

In step S609, the processor 120 may learn a knowledge tracking model using attention layers to which guide masking 701 and 702 are respectively applied. That is, the processor 120 can learn a knowledge tracking model using the attention information set 700 to which the guide masking 701 and 702 are applied.

When the relationship between units is taught to the knowledge tracking model using guide masking (701, 702) as above, the knowledge tracking model can be learned based on the relationship between units set according to the operator's intention. In other words, when the knowledge tracking model is learned, the operator can more clearly guide the relationship between units through guide masking (701, 702). Therefore, compared to a general knowledge tracking model, the knowledge tracking model learned in this way can calculate the predicted correct rate by considering the characteristics of the relationship between units. Accordingly, the prediction ability of the knowledge tracking model's correct rate can be improved, and the accuracy and reliability of the calculated predicted correct rate can be improved.

FIG. 8 is a diagram illustrating an example of guide masking according to some embodiments of the present disclosure.

For example, FIG. 8 may be a diagram showing attention information sets 801, 802, 803, and 804 to which different guide maskings 801', 802', 803', and 804' are applied.

According to an embodiment of the present disclosure, the guide masking (801', 802', 803', and 804') may be masking set based on a preset relationship between units. For example, the association between the first and second units is set to medium, the association between the first and third units is set to weak, and the association between the second and third units is set to strong. It can be. However, the present disclosure is not limited to this.

Referring to FIG. 8, the processor 120 may apply only basic masking among guide masking 801' to the attention information set 801 in the first attention layer. Accordingly, attention information corresponding to future locations from the first attention layer may be masked so that they do not affect each other.

The processor 120 may apply a guide masking 802' different from the guide masking 801' applied to the first attention layer to the attention information set 802 in the second attention layer. The processor 120 may mask attention information corresponding to the degree of relevance between the first and third units with the weakest correlation in the second attention layer. Accordingly, the problem groups of the first unit and the problem groups of the third unit may not affect each other starting from the second attention layer.

The processor 120 may apply the guide masking (801', 802') applied to the first and second attention layers and another guide masking (803') to the attention information set (803) in the third attention layer. . The processor 120 may mask attention information corresponding to the degree of relationship between the first unit and the second unit, which have a medium (next weakest) relationship in the third attention layer. Accordingly, the problem groups of the first unit and the problem groups of the second unit may not affect each other starting from the third attention layer.

The processor 120 applies guide masking (801', 802', 803') applied to the first to third attention layers and another guide masking (804') to the attention information set 804 in the fourth attention layer. can do. The processor 120 may mask attention information corresponding to the degree of relevance between the second and third units with the strongest correlation in the fourth attention layer. Accordingly, the problem groups of the second unit and the problem groups of the third unit may not affect each other starting from the fourth attention layer. In the fourth attention layer, only attention information corresponding to the degree of relevance between the same units can be applied without masking.

At this time, the knowledge tracking model may be learned differently depending on which attention layer masking is applied to the attention information. In other words, the knowledge tracking model may be learned differently depending on the location of the attention layer where masking begins to be applied to attention information. For example, the portion where masking is applied to the attention information sets 801 and 802 from the relatively front attention layer may have less influence on each other, and accordingly, the predicted correct answer rate calculated through the knowledge tracking model may vary. there is. On the other hand, the degree to which masking is applied to the attention information sets (803, 804) starting from the relatively later attention layer, or where masking is not applied at all, may increase the degree of influence on each other, and accordingly, the The predicted correct rate may vary. Accordingly, in the guide masking 801', 802', 803', and 804', the position of the attention layer where masking for each attention information starts may vary depending on the preset correlation between units.

The above content is illustrative and the present disclosure is not limited thereto.

According to another embodiment of the present disclosure, the guide masking (801', 802', 803', and 804') may be masking set based on information about problem groups. For example, the more similar the difficulty values are to each problem group, the more masking can be applied to the attention information set starting from the attention layer at the back. In the case of two problem groups with completely different difficulty values, it can be set so that incorrect answers between the two problem groups do not significantly affect each other. This is because even if you guess a group of problems with low difficulty, it is not predicted that you can guess a group of problems with high difficulty. This is the same as saying that even if you get a group of high-difficulty problems wrong, it is not expected that you can get a group of low-difficulty problems right. However, if masking is not applied, an error may occur that increases the predicted correct rate for the high-difficulty problem group if the low-difficulty problem group is correct, and if the high-difficulty problem group is incorrect, the predicted correct rate for the low-difficulty problem group may increase. This lowering error may occur. Accordingly, in guide masking 801', 802', 803', and 804', the position of the attention layer where masking for each attention information starts may vary depending on the difficulty value.

According to another embodiment of the present disclosure, the guide masking (801', 802', 803', and 804') may be masking set based on learning records for problem groups. For example, as the time information at which the user solved each problem group is similar, masking can be applied to the attention information set from the attention layer at the back. In the case of two problem groups with completely different time information, it can be set so that incorrect answers between the two problem groups do not significantly affect each other. If it has been a long time since the problem was solved, the user's skill at that time may be different from the user's skill at the current time. Therefore, the larger the difference in time information, the lower the probability that the user's previous correct answer will affect the current correct answer. At this time, if masking is not applied, the current incorrect answer is determined depending on the past incorrect answer, which may cause an error in the current predicted correct answer rate. Accordingly, the guide masking 801', 802', 803', and 804' can mask part of the attention information set so as not to affect the calculation of the predicted correct response rate for learning records that have passed a certain amount of time.

The above-described guide masking may be performed simultaneously or may be performed individually.

If guided masking is applied as above, the performance of the knowledge tracking model can be improved. In other words, by calculating the predicted correct answer rate for the next problem group by reflecting the relationship between units, the correct answer rate can be predicted more accurately. In addition, when the predicted correct rate is calculated using a knowledge tracking model learned using guided masking, and the correlation between units is quantified and visualized through this, the correlation between units can be quantified and visualized more clearly.

However, the above-described examples are merely examples and the present disclosure is not limited to the above-described examples.

In the present disclosure, the device 100 is not limited to the configuration and method of some of the embodiments described above, but all or part of the embodiments are selectively combined so that various modifications can be made. It may be configured.

Various embodiments described in this disclosure may be implemented, for example, in a recording medium readable by a computer or similar device using software, hardware, or a combination thereof.

According to hardware implementation, some embodiments described herein include application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), and field programmable gate arrays (FPGAs). , may be implemented using at least one of processors, controllers, micro-controllers, microprocessors, and other electrical units for performing functions, as described in the present disclosure. Some embodiments may be implemented in a processor.

According to software implementation, some embodiments, such as procedures and functions described in this disclosure, may be implemented as separate software modules. Each of the software modules may perform one or more functions, tasks, and operations described in this disclosure. Software code can be implemented as a software application written in an appropriate programming language. Here, the software code may be stored in the storage unit 110 and executed by the processor 120. That is, at least one program command is stored in the storage unit 110, and at least one program command can be executed by the processor 120.

A method of calculating a probability value of a user correcting a problem related to one unit performed by a processor of a device according to some embodiments of the present disclosure may include recording a recording medium readable by at least one processor provided in the device and at least one processor. It is possible to implement it as readable code. A recording medium readable by at least one processor includes all types of recording devices storing data that can be read by at least one processor. Examples of recording media that can be read by at least one processor include Read Only Memory (ROM), Random Access Memory (RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage devices.

Meanwhile, although the present disclosure has been described with reference to the accompanying drawings, these are only examples and are not limited to specific embodiments, and various modifications that can be made by those skilled in the art in the technical field to which the invention pertains are also included in the claims. falls within the scope of rights. Additionally, such modified implementations should not be understood individually from the technical spirit of the present invention.

Claims (20)

A method of quantifying the correlation between units by a processor of a device, comprising:
Predicting a correct answer rate for each of the second unit problem groups based on the incorrect answer history of each problem group included in the first sequence restored and extracted for a selected portion of the first unit problem groups;
calculating the average value of the predicted correct rate for each problem group of the second unit;
defining the average value as a correlation value of the second unit with respect to the first unit; and
Comprising: corresponding to the correlation value at the point where the first unit on the vertical axis and the second unit on the horizontal axis intersect,
method. According to paragraph 1,
The step of predicting the correct answer rate is
selecting problem groups corresponding to a first characteristic among problem groups of the first unit stored in the storage of the device;
Extracting problem groups n times using the problem groups selected in the first unit as samples;
defining a problem group set of the extracted first unit as the first sequence; and
Comprising: calculating a predicted correct answer rate for each problem group of the second unit based on the incorrect answer history of each problem group included in the first sequence,
method. According to paragraph 2,
The selecting step includes checking the difficulty level of problem groups stored in the storage of the device,
The first feature is the first difficulty section,
method. According to paragraph 2,
The extraction is restoration extraction,
method. delete According to paragraph 2,
The predicted correct rate is calculated using a knowledge tracking model,
method. According to clause 6,
Further comprising the step of learning the knowledge tracking model,
The step of learning the knowledge tracking model is,
defining a set of extracted problem groups as an input sequence;
Calculating an attention information set including attention information by quantifying the degree of relationship between problem groups included in the input sequence;
Applying guide masking to the attention information set; and
Comprising the step of learning the knowledge tracking model using the attention information set to which the guide masking is applied,
method. In clause 7,
The guide masking is,
Set based on at least one of a preset association between units stored in the storage unit of the device, information on the problem groups, and/or learning records for the problem groups,
method. A computer program stored on a computer-readable storage medium, wherein the computer program, when executed on a processor of a device, performs steps of quantifying the relationship between units by the processor of the device, the steps comprising:
Predicting a correct answer rate for each of the second unit problem groups based on the incorrect answer history of each problem group included in the first sequence restored and extracted for a selected portion of the first unit problem groups;
calculating the average value of the predicted correct rate for each problem group of the second unit;
defining the average value as a correlation value of the second unit with respect to the first unit; and
Comprising: corresponding to the correlation value at the point where the first unit on the vertical axis and the second unit on the horizontal axis intersect,
A computer program stored on a computer-readable storage medium. According to clause 9,
The step of predicting the correct answer rate is
selecting problem groups corresponding to a first characteristic among problem groups of the first unit stored in the storage unit of the device;
Extracting problem groups n times using the problem groups selected in the first unit as samples;
defining a problem group set of the extracted first unit as the first sequence; and
Comprising: calculating a predicted correct answer rate for each problem group of the second unit based on the incorrect answer history of each problem group included in the first sequence,
A computer program stored on a computer-readable storage medium. According to clause 10,
The selecting step includes checking the difficulty level of problem groups stored in the storage of the device,
The first feature is the first difficulty section,
A computer program stored on a computer-readable storage medium. According to clause 10,
The extraction is restoration extraction,
A computer program stored on a computer-readable storage medium. delete According to clause 10,
The predicted correct rate is calculated using a knowledge tracking model,
A computer program stored on a computer-readable storage medium. In the device for quantifying the relationship between units, the device:
a storage unit storing at least one program command; and
a processor executing the at least one program instruction;
Including,
The processor,
Predicting the correct answer rate for each of the second unit problem groups based on the incorrect answer history of each problem group included in the first sequence restored and extracted for a selected portion of the problem groups of the first unit,
Calculating the average value of the predicted correct rate for each of the problem groups in the second unit,
Defining the average value as a correlation value of the second unit with respect to the first unit,
The correlation value corresponds to the point where the first unit on the vertical axis and the second unit on the horizontal axis intersect,
Device. According to clause 15,
In performing the prediction of the correct answer rate,
Selecting problem groups corresponding to the first characteristic among the problem groups of the first unit stored in the storage of the device,
Using the problem groups selected in Unit 1 as samples, problem groups are extracted n times,
Defining the extracted problem group set of the first unit as the first sequence,
Calculating a predicted correct answer rate for each problem group of the second unit based on the incorrect answer history of each problem group included in the first sequence,
Device. According to clause 16,
In carrying out the selection of the above problem groups,
Check the difficulty level of problem groups stored in the storage of the device,
The first feature is the first difficulty section,
Device. According to clause 16,
The extraction is restoration extraction,
Device. delete According to clause 16,
The predicted correct rate is calculated using a knowledge tracking model,
Device.