KR102040400B1 - System and method for providing user-customized questions using machine learning - Google Patents

Abstract

According to the present invention, provided is a system to provide user-customized questions by using machine learning. The system includes: a communication module receiving a request for setting expected questions from a user terminal, and transmitting the expected questions to the user terminal in response to the request; a database storing textbook paragraph information, sample question information and expected question information; a grammar extraction module extracting key grammar information and subsidiary grammar information from the textbook paragraph stored in the database; a sample question analysis module analyzing the types and difficulty levels of the sample question information stored in the database; an expected question learning module generating an artificial neural network for inferring the expected question information based on the textbook paragraph information and the sample question information; and an expected question setting module extracting the expected questions based on the key grammar information and subsidiary grammar information extracted by the grammar extraction module and the types and difficulty levels of the sample question information analyzed by the sample question analyzing module.

Description

 SYSTEM AND METHOD FOR PROVIDING PROBLEMS USING MACHINE LEARNING {SYSTEM AND METHOD FOR PROVIDING USER-CUSTOMIZED QUESTIONS USING MACHINE LEARNING}

The present disclosure relates to a system and a method for providing a customized problem using machine learning, and more particularly, to analyze a user's learning ability, learning progress, and the like using an artificial neural network to improve the learning efficiency of the user. The present invention relates to a system and a method for providing a user with a personalized anticipated problem.

In general, the education providing method may be divided into offline education through an institute or visiting, or online education that can be learned using ICT (Information and Communication Technology) such as E-learning. These teaching methods provide learning materials and learning services that are generally tailored to the average learner's level in order to be provided to multiple learners (or users).

For example, in the case of academy education, since many students provide the same education content to a large number of learners regardless of their academic level, the tendency of the test of individual academic level or school Difficulty in learning at the level of difficulty. In addition, in the case of offline education, learners (or users) need to move to academy in order to be provided with learning services, and therefore, it takes considerable time and effort.

In recent years, due to the development of ICT, the provision of online education to solve the problem of offline education has become popular. For example, E-learning provides a environment in which a learner (or user) can learn at a desired time and place through the learner's terminal, and can solve an expected problem about the learning content. In addition, since learners can decide on the learning method and the progress of learning, it is possible to learn by level, regardless of time and place.

However, in the case of online education, even if the level of learning is possible, there is a limit in providing learning materials and learning services tailored to the test tendency or individual vulnerability of the school where the learner (or user) is attending. As shown in the national level academic achievement test, each school has a different average student's academic level, and the difficulty and type of test questions that are given by each school. Teachers make exam questions in consideration of the average academic level and environment of the students and the inclinations of their fellow teachers, so each school's past questions have their own inclination. Therefore, since the test tendency of each school is different from each other, it is difficult to provide the expected problem or learning service suitable for the difficulty of the test or the type of questions to be asked by each school. In addition, because each learner has a different vulnerability, it is difficult to improve learning efficiency through existing online education.

As described above, formal education provision methods have difficulty in providing individualized learning services to learners (or users). Therefore, there is a need to provide a user-specific prediction problem that can overcome the problems of existing education providing methods and improve the learning efficiency of learners.
[Preceding technical literature]
1. Korea Patent Registration No. 10-1775133 (2017.08.30)
2. Korean Patent Registration Publication No. 10-1129601 (2012.03.16)

Embodiments disclosed herein, to provide a user-specific problem providing system and method that can increase the learning efficiency and problem hit rate of the user by providing a personalized customized prediction problem using the artificial neural network.

According to an embodiment of the present disclosure, a user-customized problem providing system may include a communication module for receiving a question request of an expected problem from a user terminal and transmitting the expected problem to the user terminal in response to the question request, textbook fingerprint information, and problem information. And a database for storing expected problem information, a grammar extraction module for extracting key grammar information and additional grammar information from textbook fingerprint information stored in the database, and analyzing the problem type and difficulty of the existing problem information stored in the database. Focused grammar information and additional grammar information extracted from the prediction problem learning module and the grammar extraction module to generate an artificial neural network that can infer the expected problem information based on the previous problem analysis module, textbook fingerprint information, and previous problem information. Problem of previous problem information analyzed by problem analysis module Based on the type and degree of difficulty, it may comprise estimated problem questions module for extracting the estimated problem.

In addition, the user-specific problem providing method according to an embodiment of the present disclosure, the communication module, the step of receiving the expected problem question from the user terminal through the network, by the grammar extraction module, based on the question request, the database Extracting the key grammar information and additional grammar information from the textbook fingerprint information stored in the; analyzing the problem type and difficulty of the textbook fingerprint information stored in the database by the previous problem analysis module; based on the textbook fingerprint information and the previous problem information Based on the problem type and difficulty of the core grammar information and additional grammar information extracted from the grammar extraction module and the previous problem information analyzed by the previous problem analysis module by the artificial neural network trained to infer predicted problem information By the step of extracting the expected problem and the communication module, It may comprise transmitting the problem.

According to various embodiments of the present disclosure, by providing the user with an expected problem extracted by the artificial neural network learned based on the previous problem information and the textbook fingerprint information for each school, the learning efficiency and problem hit rate of the user may be increased. . In addition, the artificial neural network can learn the user's vulnerability through the predicted problem inferred by learning, and can provide a more suitable predictive problem for the user according to the identified vulnerability.

The effects of the present disclosure are not limited to the effects mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description of the claims.

Embodiments of the present disclosure will be described with reference to the accompanying drawings, which are described below, wherein like reference numerals denote similar elements, but are not limited thereto.
1 is a schematic diagram of a user customized problem providing system according to an embodiment of the present disclosure.
2 is a block diagram illustrating a detailed configuration of a user-customized problem providing system according to an exemplary embodiment.
3 is a structural diagram showing an artificial neural network according to an embodiment of the present disclosure.
4 is a flowchart illustrating a user-customized problem providing method according to an exemplary embodiment of the present disclosure.
5 is an exemplary diagram illustrating a screen of a user interface for receiving a question request for an expected problem according to an exemplary embodiment of the present disclosure.
6 is an exemplary view illustrating a screen of an interface for receiving a question request for an expected problem according to another exemplary embodiment of the present disclosure.
7 is an exemplary diagram illustrating information on an expected problem provided to a user terminal according to an embodiment of the present disclosure.

Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. However, in the following description, when there is a risk of unnecessarily obscuring the subject matter of the present disclosure, a detailed description of well-known functions and configurations will be omitted.

In the accompanying drawings, the same or corresponding components are given the same reference numerals. In addition, in the following description of the embodiments, it may be omitted to duplicate the same or corresponding components. However, even if the description of the component is omitted, it is not intended that such component is not included in any embodiment.

According to an embodiment of the present disclosure, a system for providing a user-customized problem using machine learning may include a previous question based on unit information indicating a school, a grade, and a test range of a user in order to provide a personalized learning service to a user (or learner). Perform analysis of information and / or textbook fingerprint information. In addition, the system uses a supervised learning, which is one of the machine learning methods, to generate an artificial neural network capable of inferring predictive problem information, and extracts the predicted problem, thereby customizing the user's customized predicted problem and / or Learning services can be provided.

1 is a schematic diagram of a user customized problem providing system 100 according to an embodiment of the present disclosure. The system 100 shown in FIG. 1 receives information input by a user (or learner) from the user terminal 110, and uses the communication network 120 to present a question to a prediction problem server 130. You can request Here, the user terminal 110 may be an appropriate terminal device having a user interface capable of receiving the question request information of the expected problem from the user. For example, the user interface of the terminal 110 may include one or more of a touch display, a keyboard, a mouse, a touch pen or a stylus, a microphone, a motion recognition sensor, and the like, but is not limited thereto. The terminal 110 may be any one of a personal computer, a smart phone, a tablet computer, and the like, but is not limited thereto.

The prediction problem providing server 130 may read textbook fingerprint information and past problem information from a database, and by learning an artificial neural network based on the information read in this manner, based on the information input by the user, Infer and generate predicted problems that are appropriate for school-specific academic levels. Here, the problem providing server 130 is a computing device that can communicate with other devices through a wired or wireless network, and includes a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), or the like. It may be a computing device capable of performing a calculation operation using any one, but is not limited thereto.

The database includes a textbook database (DB) 132, in which textbook fingerprint information by publisher is classified and stored by school, grade, and unit, and a preliminary problem DB (134), in which previous problem information is classified and stored by school, grade, and unit. The problem may include an expected problem DB 136 that is classified and stored by type and difficulty. Here, the textbook DB 132 and the previous question DB 134 are periodically or non-generated from the plurality of school servers 140_1 to 140_n and / or the plurality of publisher servers 150_1 to 150_n through the communication network 120. Periodically (eg, in real time), the textbook fingerprint information and / or update information of the previously asked question information may be received and stored.

In an embodiment, the expected problem providing server 130 may read textbook fingerprint information and past problem information from the textbook DB 132 and the past problem DB 134. In addition, the prediction problem providing server 130 may extract grammar information from the textbook fingerprint information and analyze the problem type and difficulty of the previous problem information. Based on the textbook fingerprint information and the previous problem information, the prediction problem providing server 130 may learn the artificial neural network in a supervised learning method. The prediction problem providing server 130 may infer and generate the prediction problem information based on the unit information indicating the school, grade, and test range of the user input through the user terminal 110 using the learned artificial neural network. Can be.

The prediction problem providing server 130 may provide the generated prediction problem to the terminal 110 through the communication network 120, and sorts the prediction problem thus generated by type and difficulty to the prediction problem DB 136. Can be stored. In FIG. 1, the databases 132, 134, and 136 are shown to be included in the prediction problem providing server 130. However, the present disclosure is not limited thereto, and the database may be separately located outside the prediction problem providing server 130 according to an embodiment. It may be connected to the communication network 120.

2 is a block diagram illustrating a detailed configuration of a user-customized problem providing system 100 according to an exemplary embodiment of the present disclosure. Hereinafter, the components of the user-customized problem providing system 100 of the present disclosure will be described in more detail. As shown in FIG. 2, the problem providing system 100 includes a communication module 220, a grammar extraction module 242, a previous problem analysis module 244, a predicted problem learning module 246, and a predicted problem question module ( 248 and database 230. The problem providing system 100 illustrated in FIG. 1 may include the functions or components of the problem providing system 100 illustrated in FIG. 2. In addition, among the functions or components of the problem providing system 100, for the components having the same reference numerals or names as those described in FIG. 1, the detailed description may be omitted to avoid repetition. Only changes or additional parts can be explained.

The database 230 may include a textbook DB 132, a past problem DB 134, and a predicted problem DB 136. The prediction problem providing server 130 may receive textbook fingerprint information for each publisher from the plurality of publisher servers 150_1 to 150_n. Textbook fingerprint information may be classified by school, grade and unit, and may be stored in the textbook DB 132, and the grammar and additional grammar extracted from the textbook fingerprint information may be classified and stored for each unit. The previous question DB 134 may store the previous question information classified by grade, unit and type, and receive the previous question information from the plurality of school servers 140_1 to 140_n. For example, the technical problem information may be classified into multiple-choice problem information, reading problem information, blank reasoning problem information, and the like, and may be stored in a database according to the type thereof. In addition, the expected problem DB 136 may classify and store the expected problems by type and difficulty.

The communication module 220 may receive a question request of an expected problem from the user terminal 110, and in response to the question request of an expected problem, the communication module 220 may extract the expected problem extracted by the predicted question question module 248. ) Can be sent. The question of the expected question may include unit information indicating a school, a grade, and a test range of the user, input by the user (or learner) using the user terminal 110. The communication module 220 may transmit a question request for an expected problem including the user input information to the processor 240.

According to an embodiment of the present disclosure, the user may input user information through the user terminal 110 to request a question of an expected problem. For example, when a user inputs a AA high school, first grade, and unit 1 using the user terminal 110, and requests a question of an expected problem, the communication module 220 may provide information about the school, grade, and unit of the user. (Eg, “AA”, “1”, “1”) may be transmitted to the processor 240 of the prediction problem providing server 130.

The grammar extraction module 242 may read textbook fingerprint information stored in the textbook DB 132 based on user input information received by the communication module 220. In addition, the grammar extraction module 242 may extract the central grammar information and the additional grammar information from the textbook fingerprint information. The grammar extraction module 242 may extract key grammar information and additional grammar information based on grammatical features (eg, vocabulary, modifier, tense, image and calligraphy) of words and sentences included in textbook fingerprint information. . In addition, the grammar information extracted by the grammar extraction module 242 may tag the tag information for each grammar category and / or difficulty, such as past completion, same name, passive voice, current injection, comparative grade, and the like. . The grammar extraction module 242 may arrange and store the important grammar information and the additional grammar information classified by tag information (eg, past completed, present completed, passive voice, etc.) in the textbook DB 132.

Here, the grammar pointed out means a grammar that focuses on the unit corresponding to the learning objective of each unit designated for each textbook. On the other hand, the additional grammar can be said to be the remaining grammar that can be extracted from the fingerprint except the main grammar. Table 1 below shows examples of key grammar information and additional grammar information that can be extracted from high school first grade English textbook fingerprint information of each publisher. In Table 1, the red numbers represent the units of the grammar and the black numbers represent the units of the grammar.

For example, if the learning goal of Unit 2 of the English textbook of "A Publisher" is the subjective relative pronoun, the subjective pronoun can be classified as the central grammar, and the grammar extracted from the rest of the fingerprint can be called the additional grammar. .

The question problem analysis module 244 reads the question information stored in the question problem DB 134 based on the user input information received by the communication module 220, and the problem type of the question indicated by the question information. And difficulty. Specifically, the question problem analysis module 244, the number of questions and the frequency of the questions by the difficulty of the questions on the previous question information to analyze the question frequency, and the questions asked in the order of high-frequency questions to the question questions DB (134) school It can be stored sorted by class, grade, and unit. In addition, tag information may be tagged according to a problem type and a difficulty level of the previous problem information.

The predictive problem learning module 246 may generate an artificial neural network capable of inferring predictive problem information based on textbook fingerprint information and past problem information. For example, the artificial neural network may infer predicted problem information through supervised learning. The neural network-based supervised learning method can be used to infer the previous problem pattern of each school, and based on this, it can be generated by predicting the predicted problem of similar difficulty and type.

The input variable input to the input layer of the artificial neural network may be a vector representing the key grammar information and the additional grammar information extracted by the grammar extraction module 242. For example, the categories of the central grammar information and the additional grammar information may be configured as one vector data element.

On the other hand, the output variable output from the output layer of the artificial neural network, according to one embodiment, may be a vector representing the problem information type of the past analysis analyzed by the problem analysis module 244, such as reading, multiple choice multiple choice, blank inference. For example, in the case of reading questions related to reading, the selection of sentences that are not related to the flow, the order of passages to follow, and the completion of the summary, and the number of questions on the reading questions related to reading may be data elements of the vector. . The predictive problem learning module 246 may train the artificial neural network using these input variables and output variables.

The expected problem question module 248 is based on the problem type and difficulty level of the grammar extraction module 242 extracted from the grammar extraction module 242 and the problem information of the previous problem information analyzed by the question analysis module 244. The predicted problem may be extracted using the neural network generated by the problem learning module 246. When it is determined that there is no similar part by comparing the similarity between the extracted expected problem and the previous problem information stored in the previous problem DB 134, the expected problem may be transmitted to the user terminal through the communication module. Meanwhile, the extracted predicted problem may be stored in the predicted problem DB 136.

According to some embodiments, the similarity between the extracted predicted question and the previous question information related to the unit information indicating the school, grade, and test range of the user may be determined by comparing words and / or sentences. For example, if the type of extracted expected problem information is a multiple choice multiple choice, it can be compared with a backed multiple choice DB of the previous problem information. In this case, when the words and / or sentences are the same, the weights associated with the words and / or sentences are applied, and the sum of the frequency of the weighted words and / or sentences is equal to or greater than a preset threshold. It can be determined that the problem and the previous problem information are the same, and the inferred expected problem can be deleted. On the other hand, if the frequency of the words and / or sentences summed by the weight is applied or less than a predetermined threshold value, the prediction problem may be stored in the prediction problem DB.

3 is a structural diagram showing an artificial neural network 300 according to an embodiment of the present disclosure. The artificial neural network 300 is a machine learning (Machine Learning) technology and cognitive science, a statistical learning algorithm implemented based on the structure of a biological neural network or a structure for executing the algorithm. That is, the artificial neural network 300 is a node that is an artificial neuron that forms a network by synaptic coupling as in a biological neural network iteratively adjust the weight of the synapse, the correct output corresponding to a specific input and the inferred output By learning so that the error between them is reduced, it represents a machine learning model with problem solving ability.

In general, an artificial neural network is implemented as a multilayer perceptron (MLP) composed of multiple nodes and a connection therebetween. The artificial neural network 300 according to the present embodiment may be implemented using one of various artificial neural network structures including an MLP. As shown in FIG. 3, the artificial neural network 300 may include an input layer 320 that receives input signals or data 310 and 312 from an external source, and output signals or data 350, 352, and 354 corresponding to the input data. ) N hidden layers 330_1 to 330_n positioned between the output layer 340, the input layer 320, and the output layer 340, which receive signals from the input layer 320, extract characteristics, and transmit the characteristics to the output layer 340. It is composed of Here, the output layer 340 receives a signal from the hidden layers 330_1 to 330_n and outputs the signal to the outside.

In general, the learning method of the artificial neural network 300 includes a supervised learning method for learning to be optimized for solving a problem by inputting a teacher signal (correct answer), and an unsupervised learning method that does not require the teacher signal ( Unsupervised Learning). The system for providing a user-customized problem using machine learning according to the present disclosure uses supervised learning to provide a personalized learning service to a user (or learner). Performs an analysis on, and learns the artificial neural network 300 can infer the expected problem information. The learned artificial neural network 300 may provide a learning service suitable for an individual or school level of education by inferring and generating an expected problem in response to a user's expected problem request.

According to one embodiment, as shown in Figure 3, the input variable of the artificial neural network 300, which can infer the expected problem information, may be the key grammar information and additional grammar information extracted by the grammar extraction module. As described above, the input variable input to the input layer 320 of the artificial neural network 300 is composed of one vector data element each of categories of the central grammar information and the additional grammar information extracted by the grammar extraction module. It may be a grammar vector 310 and an additional grammar vector 312.

For example, if you use the English textbook of "B Publisher" at the school you are in, and the unit of the test ranges in units of 6 and 7, the grammar corresponding to the core grammar of unit 6 and 7 is indirect question, respectively. Assume that and is presently complete. In this case, a value of 1 may be assigned to the indirect question and the current completion elements among the data elements of the key grammar vector 310, and a value of 0 may be assigned to the remaining elements. Also, among the data elements of the additional grammar vector 312, a value of 1 is assigned to elements corresponding to the types of grammars extracted by the grammar extraction module from the textbook fingerprint information of units 6 and 7, and the remaining elements. You can assign a value of 0 to.

On the other hand, the output variable output from the output layer 340 of the artificial neural network 300, may be a vector representing the pre-issue problem type information analyzed by the pre-question problem analysis module. According to an exemplary embodiment, the output variable may include a read vector 350, a multiple-choice vector 352, and a blank inference vector 354. For example, the reading vector 350 may include the number of questions of the past questions that have been asked for each reading related question type, such as selecting a sentence unrelated to flow, ordering of subsequent articles, and completion of a summary sentence. Can be included as an element. In the present disclosure, the output variable of the artificial neural network 300 is not limited to the three types described above. It can show various types of questions related to past questions and the number of questions of each type.

In this way, the input layer 320 and the output layer 340 of the neural network 300 are matched with a plurality of output variables corresponding to the plurality of input variables, respectively, and thus the input layer 320, the hidden layers 330_1 to 330_n and the output layer ( By adjusting the synaptic value between nodes included in 340, it is possible to learn to infer the correct output corresponding to a specific input. Through this learning process, it is possible to grasp the characteristics hidden in the input variable of the artificial neural network 300, and between the nodes of the artificial neural network 300 to reduce the error between the output variable and the target output calculated based on the input variable. You can adjust the synaptic value (or weight). Using the learned neural network 300, in response to the user's input of the expected question, the user can extract the personalized expected problem.

4 is a flow chart illustrating a user-customized problem providing method 400 according to an embodiment of the present disclosure. The user-customized problem providing method 400 may be initiated by the communication module, in step S410, receiving a request for an expected problem question from a user terminal through a network. The prediction problem providing server may receive a prediction question entry request inputted by a user (or a learner) from a user terminal, and store input information included in the prediction problem entry request in a database for each user. Here, the request for a question for an expected question may include unit information indicating a school, a grade, and a test range of a user. In an example, referring to FIG. 2, the communication module 220 may receive a request for a question of an expected problem from the user terminal 110. For example, when a user inputs a AA high school, first grade, and unit 1 using the user terminal 110, and requests a question of an expected problem, the communication module 220 may provide information about the school, grade, and unit of the user. (Eg, “AA”, “1”, “1”) may be transmitted to the processor 240 of the prediction problem providing server 130.

After receiving the expected problem question, based on the question, the core grammar information and the additional grammar information may be extracted from the textbook fingerprint information stored in the database (S420). The grammar extraction module may read textbook fingerprint information stored in a textbook DB based on a question request, and extract key grammar information and additional grammar information from the read textbook fingerprint information. In an example, referring to FIG. 2, the grammar extraction module 242 may read textbook fingerprint information stored in the textbook DB 132 based on user input information received by the communication module 220. In addition, the grammar extraction module 242 may extract the central grammar information and the additional grammar information from the textbook fingerprint information.

For example, if the input information included in the question is AA high school, first grade, and unit 1, the information on the unit 1 textbook of the first grade textbook of AA high school may be read. As described above, the key grammar information refers to a grammar that is mainly studied in the unit corresponding to the learning objective of each unit designated for each textbook, and the additional grammar is extracted from the fingerprint except the key grammar. It can be the rest of the grammar you can do. Imported grammar information and additional grammar information can be extracted based on grammatical qualities (eg, vocabulary, modifiers, tense, prize, etc.) of words and sentences from textbook fingerprint information read. The extracted key grammar information and the additional grammar information may be stored in a database by allocating and classifying tags by category and difficulty.

Thereafter, the question type and difficulty of the question information stored in the database may be analyzed by the question analysis module (S430). In this step (S430), based on the user input information (for example, AA high school, first grade, unit 1) included in the received question request can read the question information stored in the question questions DB. From the previous problem information, it is possible to analyze the problem type of the previous problem (for example, the sequence of articles to be followed, the completion of a summary, etc.) and the difficulty level (for example, upper, middle, and lower). In an example, referring to FIG. 2, the question problem analysis module 244 may read the question information stored in the question question DB 134 based on the user input information received by the communication module 220. In addition, the problem type and difficulty of the previous problem can be analyzed from the previous problem information. The question frequency and the frequency of questions by difficulty of the questions in the previous question information are counted to analyze the question frequency, and the questions in the order of high-frequency questions can be stored in the question questions DB 134 sorted by school, grade and unit.

In FIG. 4, the step of extracting grammar information (S420) is shown to be performed before the step of analyzing the question information (S430), but is not limited thereto. The step of analyzing the question information (S430) proceeds first. Alternatively, the step S420 of extracting grammar information and the step S430 of analyzing the question information may be performed in parallel at the same time.

In the next step (S440), the artificial neural network can be trained to infer the expected problem information based on the textbook fingerprint information and the previous problem information. The artificial neural network can be trained using the problem type and difficulty level of the grammar information and the additional grammar information extracted in the previous step (S430) and the analyzed problem information. According to an embodiment of the present invention, the neural network is trained using a supervised learning method so as to infer a pre-school problem pattern for each school, and to predict and generate a predicted problem of similar difficulty and type. Referring to FIG. 2, in an example, the predictive problem learning module 246 may generate an artificial neural network capable of inferring predictive problem information based on textbook fingerprint information and past problem information.

For example, the input layer of the artificial neural network consists of extracted key grammar information and additional grammar information (for example, indirect interrogation, current completion, etc.) as one vector data element, and the output layer has a pre-issue problem. An output variable (eg, multiple-choice multiple choice, reading, blank reasoning, etc.) representing the question type information analyzed by the analysis module may be repeatedly learned. As such, by matching a plurality of output variables corresponding to the plurality of input variables to the input layer and the output layer of the neural network, respectively, by adjusting the synaptic values between the nodes included in the input layer, the hidden layer, and the output layer, corresponding to a specific input You can learn to infer the correct output.

Based on the problem type and difficulty of the grammar extraction module extracted from the grammar extraction module and the problem information and the difficulty of the previous problem information analyzed by the previous problem analysis module, the predicted problem can be inferred and extracted by the learned artificial neural network. (S450). In this step (S450), the artificial neural network can infer and generate the expected problem based on the problem type and difficulty of the key grammar information and additional grammar information, and the previous problem information analyzed by the previous problem analysis module. Referring to FIG. 2, in one example, the expected problem question module 248 includes the key grammar information and the additional grammar information extracted from the grammar extraction module 242, and the written problem information analyzed by the written problem analysis module 244. Based on the problem type and the difficulty of the, the prediction problem can be extracted using the artificial neural network generated by the prediction problem learning module 246.

The generated expected problem may be compared with similarity with the previous problem information stored in the previous problem DB. By comparing the expected problem with the previous problem information, if it is determined that there is no similar part, the expected problem providing server may transmit the expected problem to the user terminal using the communication module to provide to the user (S460). Here, the transmitted expected problem can be stored in the expected problem DB.

5 is an exemplary diagram illustrating a screen 500 of a user interface for receiving a question request for an expected problem according to an embodiment of the present disclosure, and FIG. 6 illustrates an input of a question for an expected problem according to another embodiment of the present disclosure. An example diagram showing a screen 600 of a receiving interface. The user may input the question of the expected problem through a user interface (eg, a touch display, a keyboard, a mouse, a touch pen or a stylus, a microphone, a gesture recognition sensor, etc.). The question request for an expected question may include unit information indicating a school, a grade, and a test range of a user, which a user (or learner) inputs using a user terminal, and the input information is stored and sorted by user in a database. Can be.

In an embodiment, in order to input a question of an expected question, as shown in FIG. 5, divided into tabs 510 representing two sub-items, such as school, grade, and test range, to receive corresponding information. have. A screen 500 of a preset interface for inputting a school and a grade may be configured and provided to a user through a user terminal. For example, the user may input school information by selecting a check box 512 of a corresponding school item (eg, an elementary school), searching for and selecting a school name of the school at which the user is attending (514). . In addition, the user may click the corresponding grade (eg, second grade) 516 to input grade information. The inputted school and grade information may be input from the user terminal by finally determining the school and grade corresponding to the user by clicking the selection button 518.

Then, when the next button is clicked, the screen 600 of the preset interface for inputting the test range may be output to the user terminal. The user may input information on a test range using the screen 600 of the provided interface. For example, as shown in FIG. 6, a user may input a test range by selecting a check box (for example, Chapter 1 and Chapter 2) of a section corresponding to the test range. The user may finally input the test range by clicking the Done button, and send a request for a question for prediction, which includes information input from the user terminal (for example, school, grade and unit information), to the prediction problem providing server. Can be. Here, in the case where the user terminal is a mobile communication terminal (for example, a smartphone or a tablet computer), the software for requesting an expected problem may be provided in an application form, and the user terminal may be a computer terminal (for example, Personal computers, laptops, etc.) may be provided on the web on the Internet.

FIG. 7 is an exemplary diagram illustrating information 700 about an expected problem provided to a user terminal according to an embodiment of the present disclosure. The user terminal may provide a prediction problem to the user (or learner) by outputting the prediction problem received from the prediction problem providing server. According to some embodiments, the expected problem may be classified and provided as a problem related to each unit information (eg, Chapter 1 to Chapter N) corresponding to the test range.

For example, as illustrated in FIG. 7, a system and method for providing a customized problem is described in Chapter. The expected problem related to 1 can be displayed on the screen of the user terminal. The problem may include information related to the problem, such as difficulty and type. In the case of the predicted problem shown in FIG. 7, the difficulty 710 of the predicted problem corresponds to the "medium" difficulty 710, and the type 720 corresponds to "phrase". The user can click the Next button to receive an expected problem from the user terminal related to another unit information (eg, Chapter 2) corresponding to the test range. In the present disclosure, the predicted problem associated with each unit information is not limited to the one predicted problem described above, and the user-customized problem providing system and method includes a plurality of predicted objects having various types and difficulty associated with each unit information. Can provide a problem.

Although the present disclosure has been described herein in connection with some embodiments, various modifications and changes may be made without departing from the scope of the present disclosure to those skilled in the art. Also, such modifications and variations are intended to fall within the scope of the claims appended hereto.

110: user terminal 120: communication network
130: expected problem server 132: textbook DB
134: Previous Questions DB 136: Expected Questions DB
140_1 to 140_n: school server 150_1 to 150_n: publisher server
220: communication module 230: database
240: processor 242: grammar extraction module
244: Problem Analysis Module 246: Predictive Problem Learning Module
248: Question Questions 300: Artificial Neural Network
310: grammar vector 312: grammar vector
320: input layer 330_1 to 330_n: hidden layer
340: output layer 350: reading vector
352: multiple choice vector 354: blank reasoning vector
510: Details 512: checkbox
514: search 516: grade
518: selection button 710: difficulty
720: type

Claims (10)

In the system for providing customized problem using machine learning,
A communication module for receiving a question request for an expected problem including unit information indicating a school, a grade, and a test range of a user from a user terminal, and transmitting the expected problem to the user terminal in response to the question request;
Textbook database where textbook fingerprint information by publisher is classified and stored by school, grade, and unit, the previous problem database, where the previous problem information is classified by school, grade, and unit, and the expected problem classified by type and difficulty A database including a problem database;
A grammar extraction module for extracting key grammar information and additional grammar information from the textbook fingerprint information stored in the database based on the question asked;
A past problem analysis module for analyzing a problem type and difficulty level of the past problem information stored in the database;
An expected problem learning module for generating an artificial neural network capable of inferring expected problem information based on the textbook fingerprint information and the previously asked problem information; And
An expected problem question module that extracts the expected problem based on a problem type and a difficulty level of the key grammar information and the additional grammar information extracted from the grammar extraction module and the previous problem information analyzed by the previous problem analysis module. Including,
The grammar extraction module extracts the key grammar information and the additional grammar information based on grammatical features of words and sentences included in the textbook fingerprint information stored in the textbook database.
The predicted problem learning module learns to infer the predicted problem information based on the extracted key grammar information, the additional grammar information, and the analyzed previous problem information of the generated artificial neural network,
The predicted question question module generates the predicted question by the artificial neural network based on a problem type and a difficulty level of the key grammar information and the additional grammar information, and the previous question information.
When it is determined that there is no similar part by comparing the similarity level with the previous problem information stored in the previous problem database for the expected problem, the stored problem is stored in the expected problem database, and the expected problem is transmitted to the user terminal through the communication module. , system.
delete The method of claim 1,
The textbook database and the past problem database,
And receiving and storing update information of the textbook fingerprint information and the past problem information in real time from a plurality of school servers or a plurality of publisher servers through a communication network.
delete The method of claim 1,
The question problem analysis module,
The question frequency of each question type and difficulty of the question information stored in the question question database is counted to analyze the question frequency.
And sorting and storing the questions on the questions in the order of high school questions in order of school, grade and unit.
delete In the method of providing a customized problem using machine learning,
Receiving, by the communication module, an expected question question request including unit information indicating a school, a grade, and a test range of a user from a user terminal through a network;
The textbook fingerprint information is stored by a grammar extraction module from a database in which textbook fingerprint information and technical problem information for each publisher classified by school, grade, and unit, and expected problem information classified by type and difficulty are stored, based on the question asked. Extracting key grammar information and additional grammar information based on the qualities of the included words and sentences;
Analyzing, by a problem solving module, a question type and difficulty level of the question information stored in the database;
The core grammar information and the additional grammar information extracted from the grammar extraction module by an artificial neural network trained to infer predicted problem information based on the textbook fingerprint information and the previously asked problem information, and the previously asked problem analysis module Extracting the expected problem based on a problem type and a difficulty level of the previously asked question information analyzed by;
If it is determined that there is no similar part by comparing the similarity level with the previous problem information stored in the previous problem database for the expected problem, storing in the expected problem database; And
Sending, by the communication module, the expected problem to the user terminal.
delete The method of claim 7, wherein
Extracting the key grammar information and the additional grammar information,
Extracting a vocabulary, a modifier, a tense, an image, and a calligraphy from the sentences included in the textbook text information and classifying them into the grammar and the additional grammar;
And assigning and classifying a tag according to a category and a difficulty level with respect to the grammar and the additional grammar, and storing the tag in the database. A computer readable storage medium storing a program comprising instructions for causing a computer to execute the method of claim 7.

Images