WO2024029708A1 - Procédé de génération de modèle de fourniture de code de problème mathématique à base d'intelligence artificielle, et procédé et dispositif de génération de problème mathématique à l'aide dudit modèle - Google Patents

Procédé de génération de modèle de fourniture de code de problème mathématique à base d'intelligence artificielle, et procédé et dispositif de génération de problème mathématique à l'aide dudit modèle Download PDF

Info

Publication number
WO2024029708A1
WO2024029708A1 PCT/KR2023/007317 KR2023007317W WO2024029708A1 WO 2024029708 A1 WO2024029708 A1 WO 2024029708A1 KR 2023007317 W KR2023007317 W KR 2023007317W WO 2024029708 A1 WO2024029708 A1 WO 2024029708A1
Authority
WO
WIPO (PCT)
Prior art keywords
code block
code
math
block
math problem
Prior art date
Application number
PCT/KR2023/007317
Other languages
English (en)
Korean (ko)
Inventor
이재윤
유병용
이용현
Original Assignee
주식회사 프로키언
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from KR1020230066190A external-priority patent/KR102645590B1/ko
Application filed by 주식회사 프로키언 filed Critical 주식회사 프로키언
Publication of WO2024029708A1 publication Critical patent/WO2024029708A1/fr

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/10—Text processing
    • G06F40/103—Formatting, i.e. changing of presentation of documents
    • G06F40/111—Mathematical or scientific formatting; Subscripts; Superscripts
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/10—Text processing
    • G06F40/12—Use of codes for handling textual entities
    • G06F40/151—Transformation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/40—Processing or translation of natural language
    • G06F40/53—Processing of non-Latin text
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/40—Processing or translation of natural language
    • G06F40/58—Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Definitions

  • the present invention provides a method for generating an artificial intelligence model trained to learn mathematical problems and output problem code blocks to automatically generate various mathematical problems that can be derived or derived from the fingerprint of a specific mathematical problem, and the artificial intelligence model This is a method and device for generating mathematical problems using.
  • the present invention proposes a method for generating various mathematical problems using computer program codes.
  • the method of creating math problems with program codes requires not only mathematical knowledge but also an understanding of programming languages, and because mathematical concepts are expressed in codes, detailed settings are required, so a method to improve productivity is needed.
  • the present invention is intended to solve the above-mentioned problems, and uses an artificial intelligence model trained to learn mathematical problems and output problem code blocks to automatically generate various mathematical problems that can be derived or derived from the fingerprint of a specific mathematical problem.
  • One technical task is to provide a method and device for generating mathematical problems using the corresponding artificial intelligence model.
  • the embodiment according to the first aspect of the present invention includes learning target math problem data, code blocks required to generate the learning target math problem data, and annotation data for the code blocks.
  • Disclosed is a method for generating a model for providing a mathematical problem code including generating an artificial intelligence model trained to output at least one of the main code block clusters including a plurality of problem code blocks for the mathematical problem.
  • an embodiment according to the second aspect of the present invention includes at least one processor, a model generating program that provides a mathematical problem code that is electrically connected to the processor, and is performed in the processor. It provides a model generation device that provides a mathematical problem code containing memory to be stored. The memory is trained by performing a preprocessing process in which the processor converts learning target math problem data, code blocks required to generate the learning target math problem data, and annotation data for the code blocks into an embedding vector.
  • Generate a data set and through learning of the training data set, receive a fingerprint of a specific mathematical problem and select a problem code block for the mathematical problem and a cluster of main code blocks including a plurality of problem code blocks for the mathematical problem.
  • an embodiment according to the third aspect of the present invention includes the steps of obtaining a fingerprint of a math problem and converting the fingerprint into an embedding vector, and preprocessing the math problem data to be studied.
  • a fingerprint of a specific math problem is input, and a problem code block for the math problem and the above
  • a problem code block for the math problem based on the fingerprint converted to an embedding vector using an artificial intelligence model trained to output at least one of the main code block clusters including a plurality of problem code blocks for the math problem, and Disclosing a method for generating a mathematical problem comprising the steps of obtaining at least one of the main code block clusters including a plurality of problem code blocks for the mathematical problem, and generating a new mathematical problem using the obtained code block. do.
  • an embodiment according to the fourth aspect of the present invention includes at least one processor, electrically connected to the processor, and storing a mathematical problem generation program executed by the processor.
  • a mathematical problem providing device including a memory.
  • the memory includes a code block necessary for the processor to obtain a fingerprint for a math problem, convert the fingerprint into an embedding vector, preprocessed math problem data to be learned, and generate the math problem data to be learned, and the code block.
  • a fingerprint of a specific mathematical problem is input and a problem code block for the mathematical problem and a cluster of major code blocks including a plurality of problem code blocks for the mathematical problem are selected.
  • a main code block cluster including a problem code block for the math problem and a plurality of problem code blocks for the math problem based on the fingerprint converted to an embedding vector using an artificial intelligence model trained to output at least one Obtain at least one of the following, and store a code that causes a new mathematical problem to be created using the obtained code block.
  • a main code block cluster including a problem code block for the mathematical problem and a plurality of problem code blocks for the mathematical problem is generated from the fingerprint of a specific mathematical problem through an artificial intelligence model. It can be easily derived.
  • understanding of programming language is achieved by receiving a math problem fingerprint from a user who wants to create a math problem, outputting a problem code block for the fingerprint, and creating a math problem based on the problem code block. Even without , users can easily create math problems through mathematical knowledge.
  • Figure 1 is a block diagram showing the configuration of a mathematical problem code providing model generating device according to an embodiment of the present invention.
  • FIG. 2 is a diagram illustrating an example of training an artificial intelligence model using the model generating device for providing mathematical problem codes shown in FIG. 1.
  • Figures 3 and 4 are diagrams showing examples of problem code blocks and comments.
  • Figures 5 to 7 are diagrams showing examples of training data sets.
  • Figure 8 is an operation flowchart explaining a method for generating a model for providing mathematical problem codes according to another embodiment of the present invention.
  • FIG. 9 is a flowchart showing detailed steps included in the steps of the method for generating a mathematical problem code providing model shown in FIG. 8.
  • Figure 10 is a diagram illustrating a mathematical problem generating device connected to communication with a user terminal according to another embodiment of the present invention.
  • FIG. 11 is a block diagram showing the configuration of the mathematical problem generating device shown in FIG. 10.
  • FIG. 12 is a diagram illustrating an example of generating a math problem using the math problem generating device shown in FIG. 11.
  • Figure 13 is an operation flowchart explaining a method for generating a math problem according to another embodiment of the present invention.
  • first, second, etc. used in this specification are used only for the purpose of distinguishing one component from another component and do not limit the order or relationship of the components.
  • first component of the present invention may be named a second component, and similarly, the second component may also be named a first component.
  • singular forms of expression should be construed to also include plural forms of expression, unless the contrary is clearly indicated.
  • the communication module described below may include a device including hardware and software necessary to transmit and receive signals such as control signals or data signals through wired or wireless connections with other network devices.
  • the memory may store at least one of information and data input to the communication module, information and data required for functions performed by the processor, and data generated according to execution of the processor. Memory should be interpreted as a general term for non-volatile storage devices that retain stored information even when power is not supplied, and volatile storage devices that require power to maintain stored information. Memory may include magnetic storage media or flash storage media in addition to volatile storage devices that require power to maintain stored information, but the scope of the present invention is not limited thereto.
  • a processor may include various types of devices that control and process data.
  • a processor may refer to a data processing device built into hardware that has physically structured circuitry to perform functions expressed by codes or instructions included in a program.
  • the processor may include a microprocessor, central processing unit (CPU), processor core, multiprocessor, application-specific integrated circuit (ASIC), or field programmable gate (FPGA). array), etc., but the scope of the present invention is not limited thereto.
  • Figure 1 is a block diagram showing the configuration of a mathematical problem code providing model generating device 100 according to an embodiment of the present invention
  • Figure 2 shows an artificial intelligence model trained by the mathematical problem code providing model generating device 100. This is a drawing showing an example of what to do.
  • the mathematical problem code providing model generating device 100 includes at least one processor 130 and a memory 140, and further includes a communication module 110 and a database 120. can do.
  • the communication module 110 may transmit and receive information with an external device or server to transmit and receive data necessary for generating a model for providing a mathematical problem code.
  • the database 120 may be a place where data necessary for creating a model for providing mathematical problem codes is stored.
  • the database 120 may be built in a portion of the memory 140 or may be implemented as separate hardware.
  • the processor 130 performs operations according to the code stored in the memory 140.
  • the memory 140 is electrically connected to the processor 130 and stores a model generation program that provides mathematical problem codes performed by the processor 130.
  • the memory 140 stores code that, when executed through the processor 130, causes the processor 130 to perform the following functions and procedures.
  • the memory 140 contains the learning target math problem data 210, the code block 220 required to generate the learning target math problem data 210, and the annotation data 230 for the code block.
  • a training data set 300 is generated by performing a preprocessing process including.
  • the code block 220 includes at least one of a variable setting code block, a variable condition code block, a result condition code block, and a result generation code block.
  • the variable setting code block is a code block that sets the variables of the math problem data to be studied.
  • a variable condition code block is a code block that sets the conditions of a variable.
  • the result condition code block is a code block that sets conditions for the correct and incorrect answers of the math problem data to be studied.
  • the result generation code block is a code block that generates one correct answer that satisfies the conditions of the correct answer and one or more incorrect answers that satisfy the conditions of the incorrect answer, based on the variables created through the variable setting code block and the variable condition setting code block.
  • the annotation 230 includes a description of at least one block among a variable setting code block, a variable condition code block, a result condition code block, and a result generation code block.
  • the memory 140 stores a code that causes code similarity to be calculated by comparing a specific code block included in the training data set 300 with a different code block.
  • the memory 140 stores a code that causes a code block cluster to be created by clustering a specific code block and a code block different from the specific code block based on code similarity. If the code similarity is greater than a preset value, the memory 140 stores a code that causes a code block cluster to be generated by clustering a specific code block and a code block different from the specific code block.
  • Code similarity includes at least one of variable setting code similarity, variable condition code similarity, result condition code similarity, and result generation code similarity.
  • Variable setting code similarity is a similarity derived by comparing a variable setting code block included in a specific code block with a variable setting code block included in a code block different from the specific code block.
  • Variable condition code similarity is a similarity derived by comparing a variable condition code block included in a specific code block with a variable condition code block included in a code block different from the specific code block.
  • Resulting condition code similarity is the similarity derived by comparing a result condition code block included in a specific code block with a result condition code block included in a code block different from the specific code block.
  • Result generation code similarity is a similarity derived by comparing a result generation code block included in a specific code block with a result generation code block included in a code block different from the specific code block.
  • the memory 140 stores a code that causes the code block cluster to match the math problem data to be learned included in the training data set 300 corresponding to the code block cluster and the annotations included in the training data set 300.
  • the training data set 300 may be a data table formed by matching math problem data to be learned converted into an embedding vector, code blocks, annotations, and code block clusters.
  • the code block cluster includes at least one of a variable setting code cluster, a variable condition code cluster, a result condition code cluster, and a result generation code cluster.
  • Variable setting code clustering clusters variable setting code blocks included in a specific code block included in the training data set 300 and variable setting code blocks included in a code block different from the specific code block, based on variable setting code similarity. It is a cluster of codes.
  • the variable condition code cluster is a code cluster that clusters variable condition code blocks included in a specific code block and variable condition code blocks included in a code block different from the specific code block, based on variable condition code similarity.
  • the result condition code cluster is a code cluster that clusters result condition code blocks included in a specific code block and result condition code blocks included in a code block different from the specific code block, based on result condition code similarity.
  • the result generation code cluster is a code cluster that clusters result generation code blocks included in a specific code block and result generation code blocks included in a code block different from the specific code block, based on result generation code similarity.
  • the memory 140 receives the fingerprint of a specific math problem through learning of the training data set 300, and contains a main code block cluster containing a problem code block for a specific math problem and a plurality of problem code blocks for a specific math problem.
  • a code that causes the artificial intelligence model 400 to be trained to output at least one of the following is stored.
  • the problem code block includes the variable setting code block that sets the variables of the math problem, the variable condition code block that sets the conditions for the variables, the result condition code block that sets the conditions for the correct and incorrect answers in the math problem, and the variable setting code block and variable conditions. Based on the variables created through the code block, it includes at least one of a result generation code block that generates one correct answer that satisfies the conditions of the correct answer and one or more incorrect answers that satisfy the conditions of the incorrect answer.
  • the artificial intelligence model 400 learns the code block position relationship between the variable setting code block, variable condition code block, result condition code block, and result generation code block included in the code block of the training data set.
  • the artificial intelligence model 400 calculates fingerprint similarity by comparing the fingerprint of a specific math problem with the learning target math problem data included in the training data set 300 and at least one of the annotations included in the training data set 300. .
  • the artificial intelligence model 400 selects a code block cluster corresponding to the learning target math problem data included in the training data set 300 or the annotation included in the training data set 300, based on fingerprint similarity. If the fingerprint similarity is more than a preset value, the artificial intelligence model 400 selects a code block cluster corresponding to the learning target math problem data included in the training data set 300 or the annotation included in the training data set 300. .
  • the artificial intelligence model 400 is an artificial intelligence trained to output code blocks that satisfy preset criteria among code blocks included in the code block cluster as problem code blocks for mathematical problems, or to output the code block cluster as the main code block cluster. It is an intelligence model.
  • the artificial intelligence model may be a supervised learning model. More specifically, the artificial intelligence model may be an artificial intelligence model based on the Omikuji algorithm.
  • 3 and 4 are diagrams illustrating an example of a code block 220 and annotation 230.
  • the code block 220 includes at least one of a variable setting code block 221, a variable condition code block 222, a result generation code block 223, and a result condition code block 224. Includes more.
  • the annotation 230 includes the variable setting annotation 231 for the variable setting code block 221, the variable condition annotation 232 for the variable condition code block 222, and the result generation annotation for the answer generation code block 223 ( (not shown) and a correct answer condition annotation (not shown) for the result condition code block 224.
  • the variable condition code block 222 may be a code block in which variable conditions are set, such as the condition that the x value and the y value must be prime or that they must not be 0.
  • the resulting condition code block 224 may be a code block in which conditions such as a condition that the value of b must not be 0, a condition that it must be a positive number, etc. are set.
  • the result generation code block 223 is one or more that satisfies the correct answer and incorrect answer conditions set by the correct answer generation code block 223 based on the variables created through the variable setting code block 221 and the variable condition code block 222. This is a code block set up to generate incorrect and correct answers.
  • Figures 5 to 7 are diagrams showing an example of the training data set 300.
  • Figure 5 is a diagram showing a training data set in the form of a data table in which mathematical problems, annotations, and code blocks are matched to each other
  • Figure 6 is a diagram showing another type of training data set in which mathematical problems and annotations are matched.
  • FIG. 7 is a diagram showing an example of converting the training data set shown in FIG. 6 into an embedding vector.
  • the training data set shown in FIGS. 5 and 6 is a diagram showing a training data set that has not been converted to an embedding vector.
  • the training data set 300 includes learning target math problem data 510, a variable setting code block included in the code block 530 for the learning target math problem data 510, a variable condition code block, This is a data table that matches the annotations 520 for each of the answer condition code block and the answer generation code block.
  • the training data set 300 includes math problem data to be learned 610, code blocks (not shown) for the math problem data to be learned 610, and annotation data 620 for the code blocks. ) may be a data table that matches each other. If the data included in the data table shown in FIG. 6 is converted into an embedding vector, it may appear as shown in FIG. 7.
  • FIG. 8 is an operation flowchart explaining a method for generating a mathematical problem code providing model according to another embodiment of the present invention
  • FIG. 9 is a flowchart showing detailed steps included in the steps of the method for generating a mathematical problem code providing model.
  • a method for generating a model for providing a mathematical problem code will be described.
  • Each step of the method for generating a mathematical problem code providing model to be described below may be performed by the mathematical problem code providing model generating apparatus 100 previously described with reference to FIGS. 1 to 7 . Accordingly, the contents of the embodiments of the present invention previously described with reference to FIGS.
  • the method of generating a model for providing a mathematical problem code includes a training data set creation step (S1100) and an artificial intelligence model training step (S1200).
  • the training data set generation step (S1100) performs a preprocessing process including converting the learning target math problem data, the code blocks required to generate the learning target math problem data, and the annotation data for the code blocks into an embedding vector to generate training data. This is the step of creating a set.
  • the artificial intelligence model training step (S1200) based on the training data set, at least one of the main code block clusters including a problem code block for the mathematics problem and a plurality of problem code blocks for the mathematics problem is extracted from the fingerprint of a specific mathematics problem. This is the step of creating an artificial intelligence model that has been trained to output.
  • the artificial intelligence model calculates fingerprint similarity by comparing the fingerprint of the math problem with at least one of the target math problem data included in the training data set and annotations included in the training data set.
  • the artificial intelligence model selects a cluster of code blocks corresponding to the math problem data to be learned or annotations included in the training data set based on fingerprint similarity.
  • the artificial intelligence model is an artificial intelligence model that is trained to output code blocks that meet preset criteria among code blocks included in a code block cluster as problem code blocks for math problems, or output code block clusters as main code block clusters. .
  • the training data set generation step (S1100) includes a code similarity derivation step (S1110), a code block cluster generation step (S1120), and a data matching step (S1130).
  • the code similarity derivation step (S1210) is a step of calculating code similarity by comparing a specific code block included in the training data set with a code block different from the specific code block included in the training data set.
  • the code block cluster generation step (S1120) is a step of generating a code block cluster by clustering a specific code block and a code block different from the specific code block based on code similarity.
  • the data matching step (S1130) is a step of matching the code block cluster with the learning target math problem data included in the training data set corresponding to the code block cluster and the annotations included in the training data set.
  • Figure 10 is a diagram showing a mathematical problem generating device 1000 connected to communication with a user terminal 1100 according to another embodiment of the present invention.
  • the math problem generating device 1000 is interconnected with the user terminal 1100 through a wired or wireless communication network.
  • a communication network is any wired network such as a Local Area Network (LAN), Wide Area Network (WAN), or Value Added Network (VAN), a mobile radio communication network, or a satellite communication network. It can be implemented as any type of wireless network.
  • the math problem generating device 1000 may be formed as a cloud computing server such as Software as a Service (SaaS), Platform as a Service (PaaS), or Infrastructure as a Service (IaaS).
  • the user terminal 1100 is, for example, a laptop equipped with a web browser, a desktop, a laptop, a wireless communication device or smartphone that ensures portability and mobility, and a tablet including a touchpad. It can refer to all types of handheld-based wireless communication devices such as PCs, etc.
  • the user terminal 1100 includes a communication module, an input/output module, at least one processor, and a memory.
  • the communication module of the user terminal 1100 may transmit and receive information with the math problem generating device 1000.
  • the input/output module of the user terminal 1100 may receive information or data transmitted to the user terminal 1100 from the outside, or may output information or data held by the user terminal 1100 to the outside.
  • the input/output module of the user terminal 1100 may include a display, a touch pad, a speaker, and a microphone.
  • the processor of the user terminal 1100 is electrically connected to the memory and operates by executing the code stored in the memory.
  • the memory of the user terminal 1100 is electrically connected to the processor and stores at least one code executed by the processor.
  • the memory of the user terminal 1100 stores code that, when executed through the processor, causes the processor to perform the following functions and procedures.
  • a code that causes a fingerprint for a math problem to be transmitted to the math problem generating device 1000 is stored in the memory of the user terminal 1100 .
  • the memory of the user terminal 1100 receives a problem code block for a math problem or a main code block cluster including a plurality of problem code blocks for a math problem from the math problem generating device 1000 and causes it to be output to the input/output module.
  • the code is saved.
  • the memory of the user terminal 1100 stores a code that causes a problem code block for a math problem or a selection input for a specific problem code block among a plurality of problem code blocks for a math problem to be transmitted to the math problem generating device 1000. do.
  • a code that causes the user terminal 1100 to receive a math problem generated through a problem code block or a specific problem code block corresponding to a selection input from the math problem generating device 1000 is stored in the memory of the user terminal 1100 .
  • FIG. 11 is a block diagram showing the configuration of the math problem generating device 1000
  • FIG. 12 is a diagram showing an example of generating a math problem by the math problem generating device 1000.
  • the math problem generating device 1000 includes at least one processor 1030 and a memory 1040, and may further include a communication module 1010 and a database 1020. .
  • the communication module 1010 can transmit and receive data necessary for creating math problems by exchanging information with an external device or server.
  • the database 1020 may be a place where data necessary for creating math problems is stored.
  • the database 1020 may be built in a portion of the memory 1040 or may be implemented as separate hardware.
  • the processor 1030 performs operations according to the code stored in the memory 1040.
  • the memory 1040 is electrically connected to the processor 1030 and stores a math problem generation program performed by the processor 1040.
  • the memory 1040 stores code that, when executed through the processor 1030, causes the processor 1040 to perform the following functions and procedures.
  • the memory 1040 stores a code that causes the fingerprint 1200 for a math problem to be obtained and the fingerprint to be converted into an embedding vector. More specifically, a code that causes a fingerprint 1200 for a math problem to be received from the user terminal 1100 is stored in the memory 1040 . Memory 1040 stores a code that causes the fingerprint to be converted into an embedding vector.
  • the memory 1040 includes a problem code block 1400 for a math problem and a plurality of problem code blocks for a math problem based on a fingerprint converted to an embedding vector using the math problem code provision model 1300.
  • Code that causes at least one of the code block clusters to be obtained is stored.
  • the math problem code provision model 1300 learns a specific training data set including preprocessed math problem data to be learned, code blocks required to generate the math problem data to be learned, and annotation data for the code blocks. It is trained to receive a fingerprint of a math problem as an input and output at least one of a problem code block for the math problem and a major code block cluster containing a plurality of problem code blocks for the math problem.
  • the math problem code providing model 1300 calculates fingerprint similarity by comparing the fingerprint of the math problem with at least one of the learning target math problem data included in the training data set and the annotations included in the training data set, and calculates the fingerprint similarity. Based on this, a code block cluster corresponding to the learning target math problem data included in the training data set or annotations included in the training data set is selected, and a preset standard is applied among the code blocks included in the code block cluster. It can be trained to output a satisfying code block as a problem code block for the math problem, or to output the code block cluster as the main code block cluster.
  • the problem code block is a variable setting code block that sets the variables of a math problem, a variable condition code block that sets conditions for variables, and a result condition code block and variable setting code block that sets conditions for correct and incorrect answers in a math problem. It includes at least one of a result generation code block that generates one correct answer that satisfies the conditions of the correct answer and one or more incorrect answers that satisfy the conditions of the incorrect answer based on the variables created through the condition code block.
  • the math problem code provision model 1300 is based on a training data set that preprocesses the math problem data to be learned, the code blocks needed to generate the math problem data to be learned, and the annotation data for the code blocks, and is based on a training data set that preprocesses the math problem data to be learned, and the annotation data for the code blocks required to generate the math problem data to be learned. It is an artificial intelligence model trained to output at least one of the problem code blocks for math problems and major code block clusters that include multiple problem code blocks for specific math problems.
  • the memory 1040 stores a specific problem code block selected among a plurality of problem code blocks included in the main code block cluster and a code that causes a math problem to be generated based on at least one of the problem code blocks. More specifically, in the memory 1040, when a main code block cluster is obtained by the mathematical problem code provision model 1300, a list is generated in which a plurality of problem code blocks included in the main code block cluster are displayed and selected from the list. Upon obtaining a selection input for a specific problem code block, code that causes the math problem to be generated (1500) based on the specific problem code block is stored.
  • the memory 1040 stores a code that, when the problem code block 1400 is obtained by the mathematical problem code provision model 1300, generates a mathematical problem 1500 based on the problem code block. More specifically, the memory 1040 stores a code that causes code blocks included in the problem code block 1400 or a specific problem code block to be arranged in order using an artificial intelligence model.
  • the memory 1040 stores a problem code block 1400 in which code blocks are arranged in order or a code that causes a math problem to be generated 1500 based on a specific problem code block.
  • the artificial intelligence model is based on the code block position relationship between the variable setting code block, variable condition code block, result condition code block, and result generation code block included in the code block required to generate the math problem data to be studied, and the problem code to be analyzed. It is an artificial intelligence model trained to place the variable setting code block, variable condition code block, answer condition code block, and answer generation code block included in the block in order.
  • the artificial intelligence model may be a model that uses the Extreme Multi-classification algorithm.
  • Figure 13 is an operation flowchart explaining a method for generating a math problem according to another embodiment of the present invention.
  • a method for generating a math problem will be described.
  • Each step of the math problem generation method to be described below may be performed by the math problem generation device 1000 previously described with reference to FIGS. 10 to 12 . Therefore, the contents of the embodiments of the present invention previously described with reference to FIGS. 10 to 12 can be equally applied to the embodiments to be described below, and any content that overlaps with the description described above will be omitted below.
  • the steps described below do not necessarily have to be performed in order, the order of the steps may be set in various ways, and the steps may be performed almost simultaneously.
  • the math problem generation method is a method in which each step is performed by a processor, and includes a vector conversion step (S2100), a code provision step (S2200), and a math problem creation step (S2300).
  • the vector conversion step (S2100) is a step of obtaining a fingerprint for a math problem and converting the fingerprint into an embedding vector.
  • the code provision step (S2200) uses a math problem code provision model to select a problem code block for a math problem based on a fingerprint converted to an embedding vector and a cluster of main code blocks including a plurality of problem code blocks for a math problem. This is the step of acquiring at least one thing.
  • the math problem code provision model is designed to provide a specific math problem through learning of a training data set that includes preprocessed math problem data to be learned, code blocks needed to generate the math problem data to be learned, and annotation data for the code blocks. It is trained to receive a fingerprint as an input and output at least one of the main code block clusters including a problem code block for a math problem and a plurality of problem code blocks for a math problem.
  • the fingerprint similarity is calculated by comparing the fingerprint of the mathematical problem in the mathematical problem code providing model with at least one of the learning target mathematical problem data included in the training data set and the annotations included in the training data set, and the fingerprint similarity is based on Select a code block cluster that corresponds to the learning target math problem data included in the training data set or annotations included in the training data set, and code that satisfies preset criteria among the code blocks included in the code block cluster. It can be trained to output blocks as problem code blocks for math problems, or to output clusters of code blocks as clusters of major code blocks.
  • the problem code block includes the variable setting code block that sets the variables of the math problem, the variable condition code block that sets the conditions for the variables, the result condition code block that sets the conditions for the correct and incorrect answers in the math problem, and the variable setting code block and variable conditions. Based on the variables created through the code block, it includes at least one of a result generation code block that generates one correct answer that satisfies the conditions of the correct answer and one or more incorrect answers that satisfy the conditions of the incorrect answer.
  • the math problem code provision model is based on a training data set that preprocesses the math problem data to be learned, the code blocks needed to generate the math problem data to be learned, and the annotation data for the code blocks, and is based on a training data set that preprocesses the math problem data to be learned, and the annotation data for the code blocks, from the fingerprint of the specific math problem. It is an artificial intelligence model trained to output at least one of the major code block clusters, including problem code blocks for and multiple problem code blocks for specific math problems.
  • the math problem generation step (S2300) is a step of generating the math problem based on at least one of a specific problem code block and a problem code block selected among a plurality of problem code blocks included in the main code block cluster. More specifically, in the mathematical problem generation step (S2300), when a major code block cluster is obtained by a mathematical problem code provision model, a list displaying a plurality of problem code blocks included in the major code block cluster is generated and a specific problem code block from the list is generated. When a selection input for a problem code block is obtained, generating the math problem based on a specific problem code block, and when a problem code block is obtained by a math problem code provision model, generating a math problem based on the problem code block. am.
  • Computer-readable media can be any available media that can be accessed by a computer and includes both volatile and non-volatile media, removable and non-removable media. Additionally, computer-readable media may include computer storage media. Computer storage media includes both volatile and non-volatile, removable and non-removable media implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules or other data.
  • the present invention is a technology for automatically generating various mathematical problems and can be used in the education industry, so it has industrial applicability.
  • Model generation device providing mathematical problem code

Abstract

Un mode de réalisation de la présente invention concerne un procédé de génération d'un modèle de fourniture de code de problème mathématique, le procédé comprenant les étapes de : la génération d'un ensemble de données d'entraînement par l'intermédiaire d'un processus de prétraitement comprenant une opération de conversion, en un vecteur d'incorporation, de données de problème mathématique à apprendre, d'un bloc de code requis pour générer les données de problème mathématique à apprendre et de données d'annotation pour le bloc de code ; et la génération d'un modèle d'intelligence artificielle qui est entraîné avec l'ensemble de données d'entraînement pour recevoir le texte d'un problème mathématique particulier et délivrer en sortie au moins l'un d'un bloc de code de problème pour le problème mathématique et d'un groupe principal de blocs de code comprenant de multiples blocs de code de problème pour le problème mathématique.
PCT/KR2023/007317 2022-08-03 2023-05-26 Procédé de génération de modèle de fourniture de code de problème mathématique à base d'intelligence artificielle, et procédé et dispositif de génération de problème mathématique à l'aide dudit modèle WO2024029708A1 (fr)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
KR20220096633 2022-08-03
KR10-2022-0096633 2022-08-03
KR1020230066190A KR102645590B1 (ko) 2022-08-03 2023-05-23 인공지능 기반의 수학 문제 코드 제공 모델 생성 방법과 해당 모델을 이용한 수학 문제 생성 방법 및 장치
KR10-2023-0066190 2023-05-23

Publications (1)

Publication Number Publication Date
WO2024029708A1 true WO2024029708A1 (fr) 2024-02-08

Family

ID=89849156

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2023/007317 WO2024029708A1 (fr) 2022-08-03 2023-05-26 Procédé de génération de modèle de fourniture de code de problème mathématique à base d'intelligence artificielle, et procédé et dispositif de génération de problème mathématique à l'aide dudit modèle

Country Status (1)

Country Link
WO (1) WO2024029708A1 (fr)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090017427A1 (en) * 2007-07-12 2009-01-15 Microsoft Corporation Intelligent Math Problem Generation
KR20110078073A (ko) * 2009-12-30 2011-07-07 동국대학교 산학협력단 문제 선택 및 성취도 평가 방법 및 장치
KR20120063442A (ko) * 2010-12-07 2012-06-15 에스케이 텔레콤주식회사 수학문장의 시맨틱거리 추출 및 시맨틱거리에 의한 수학문장의 분류방법과 그를 위한 장치 및 컴퓨터로 읽을 수 있는 기록매체
KR20170105969A (ko) * 2016-03-11 2017-09-20 김병섭 유사 수학문제 검색장치 및 컴퓨터 프로그램
JP2021530066A (ja) * 2018-09-26 2021-11-04 杭州大拿科技股▲ふん▼有限公司 暗算問題に対する問題添削方法、装置、電子機器及び記憶媒体

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090017427A1 (en) * 2007-07-12 2009-01-15 Microsoft Corporation Intelligent Math Problem Generation
KR20110078073A (ko) * 2009-12-30 2011-07-07 동국대학교 산학협력단 문제 선택 및 성취도 평가 방법 및 장치
KR20120063442A (ko) * 2010-12-07 2012-06-15 에스케이 텔레콤주식회사 수학문장의 시맨틱거리 추출 및 시맨틱거리에 의한 수학문장의 분류방법과 그를 위한 장치 및 컴퓨터로 읽을 수 있는 기록매체
KR20170105969A (ko) * 2016-03-11 2017-09-20 김병섭 유사 수학문제 검색장치 및 컴퓨터 프로그램
JP2021530066A (ja) * 2018-09-26 2021-11-04 杭州大拿科技股▲ふん▼有限公司 暗算問題に対する問題添削方法、装置、電子機器及び記憶媒体

Similar Documents

Publication Publication Date Title
WO2018151464A1 (fr) Système de codage et procédé de codage utilisant la reconnaissance vocale
WO2022088671A1 (fr) Procédé et appareil de réponse automatique à des questions, dispositif et support de mémoire
EP3629253A1 (fr) Procédé et appareil de génération de données d'apprentissage pour système vqa et support
WO2018088664A1 (fr) Dispositif de détection automatique d'erreur de corpus d'étiquetage morphosyntaxique au moyen d'ensembles approximatifs, et procédé associé
WO2018212396A1 (fr) Procédé, dispositif et programme informatique pour analyser des données
WO2018212584A2 (fr) Procédé et appareil de classification de catégorie à laquelle une phrase appartient à l'aide d'un réseau neuronal profond
WO2023279692A1 (fr) Procédé et appareil de traitement de données basés sur une plateforme questions-réponses, et dispositif associé
WO2022145829A1 (fr) Système de recommandation de contenu d'apprentissage pour prédire une probabilité d'obtention d'une réponse correcte d'un utilisateur à l'aide d'un filtrage collaboratif basé sur un facteur latent, et son procédé de fonctionnement
WO2024029708A1 (fr) Procédé de génération de modèle de fourniture de code de problème mathématique à base d'intelligence artificielle, et procédé et dispositif de génération de problème mathématique à l'aide dudit modèle
WO2024054090A1 (fr) Dispositif électronique et système de commande d'une pluralité de haut-parleurs afin de tester une vitesse et une directionnalité de réponse auditive
WO2024005413A1 (fr) Procédé et dispositif basés sur l'intelligence artificielle pour extraire des informations d'un document électronique
WO2014209051A1 (fr) Procédé permettant de publier les réponses de personnes interrogées dans un questionnaire et serveur de service de questionnaire qui prend en charge le procédé associé
WO2022270840A1 (fr) Système de recommandation de mots basé sur un apprentissage profond pour prédire et améliorer la capacité de vocabulaire d'un élève de langue étrangère
WO2022177372A1 (fr) Système de fourniture de service de tutorat à l'aide d'une intelligence artificielle et son procédé
WO2023022406A1 (fr) Procédé d'évaluation de capacité d'apprentissage, dispositif d'évaluation de capacité d'apprentissage et système d'évaluation de capacité d'apprentissage
WO2022154376A1 (fr) Appareil et procédé de fourniture d'un modèle d'analyse de style intérieur d'un utilisateur sur la base d'un texte sns
WO2022235073A1 (fr) Procédé de guidage d'amélioration des compétences de lecture et d'écriture, et dispositif associé
WO2022102966A1 (fr) Système de recommandation de questions d'apprentissage pour recommander des questions qui peuvent être évaluées au moyen de l'unification de types de distribution de probabilité de score, et son procédé de fonctionnement
KR102645590B1 (ko) 인공지능 기반의 수학 문제 코드 제공 모델 생성 방법과 해당 모델을 이용한 수학 문제 생성 방법 및 장치
WO2021256578A1 (fr) Appareil et procédé de génération automatique de légende d'image
WO2017122872A1 (fr) Dispositif et procédé permettant de générer des informations concernant une publication électronique
WO2023042987A1 (fr) Appareil, système et procédé pour analyser le sens d'un texte à l'aide d'une analyse de code source, et procédé pour fournir un service de questions-réponses d'agent conversationnel les utilisant
WO2024029946A1 (fr) Procédé et dispositif de fourniture de concept mathématique personnalisé par analyse d'image à résolution de problème
WO2023101235A1 (fr) Procédé et système de fonctionnement de plateforme fournissant un problème personnalisé sur la base d'informations de comportement d'élève
WO2023054829A1 (fr) Appareil, système et procédé pour fournir un service de questions et de réponses comprenant une explication de code source, et procédé pour fournir un service conversationnel l'utilisant

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23850219

Country of ref document: EP

Kind code of ref document: A1