A learning training program generation device based on AI

Abstract

The present invention relates to an AI-based learning training program generation device, which belongs to the field of artificial intelligence. The system includes: a program generator receives digitally processed knowledge points to obtain a solution template; and upon receiving a user's program execution instruction, the corresponding solution template is called and the next step node is inferred; at the same time, the execution of each step node is provided to the user with explanation content; the instance database receives and stores the instance data of each completed program to generate a set of instances to be learned; the inductive learning module selects the positive output instance data from the set of instances to be learned for labeling and clustering to obtain the knowledge points learned by induction; the user interaction module displays an interactive interface to the user so that the user can interact through the interactive interface to promote the execution of the program. The device described in the present invention uses artificial intelligence technology to solve learning problems in a short time, accurately push resources, and the solution process is controllable.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence, and in particular relates to a learning and training program generation device based on AI. Background Art

[0002] At present, in the process of classroom teaching, due to individual differences among students, each student will have different knowledge points that they do not know or are not firm in. How to solve these knowledge points after class is a common problem. There are currently two solutions, as follows:

[0003] 1. Active learning: Students can choose what videos to watch, when to watch them, what questions to do, how many questions to do, and when to do them on learning assistance applications. This is a supermarket-style self-selection model.

[0004] The disadvantages and problems of active learning are: the learning effect is not good or even no effect. The reason is that students cannot find resources that suit them, there are too many resources to choose from, and they waste time on useless work.

[0005] 2. Passive learning: Following a study plan that you don’t know where you got it from, you set fixed steps, such as: first watch a video, then do 5 exercises, then watch a PPT, and then do a set of assessment papers. This is a passive training mode.

[0006] The disadvantage and problem of passive learning is that some students have good learning effects, while others have poor or even no effects at all, wasting time on useless work. The reason is that there is no professional intervention knowledge, teaching research knowledge and teaching knowledge, plus the individual differences of students, different background knowledge, and different stages. It is not appropriate to use a fixed set of prescribed actions, the same resource type, and the same resources for all students.

[0007] Therefore, how to solve students’ problems with a certain knowledge point efficiently, individually, accurately and effectively is a technical problem that needs to be solved in current education and teaching.

[0008] In recent years, artificial intelligence technology has made rapid progress, and various applications have continuously refreshed people's imagination. From the ImageNet competition, where computer vision surpassed human eye recognition, to AIGC (AI-Generated Content) technology represented by ChatGPT, artificial intelligence has gradually become a new tool to promote the development of productivity.

[0009] Almost in sync with the rapid development of artificial intelligence, artificial intelligence education in primary and secondary schools has also attracted more and more attention. Artificial intelligence education in education and teaching will also move towards a "new generation of artificial intelligence education" represented by deep learning technology. Therefore, how to better integrate artificial intelligence into the education and teaching of students, help students solve learning problems in a short time and accurately push resources, etc., is one of the main problems that need to be solved in the current field of artificial intelligence education. Summary of the invention

[0010] In view of the above deficiencies in the prior art, the purpose of the present invention is to provide an AI-based learning training program generation device, which requires little human intervention, can solve learning problems in a short time, accurately push resources, and the solution process is controllable.

[0011] The present invention proposes an AI-based learning training program generation device, comprising: a program generator, an instance database, an inductive learning module and a user interaction module, wherein:

[0012] The solution generator is used to receive digitally processed knowledge points, and store them in the internal knowledge base in the form of corresponding matching solution templates according to the content of the knowledge points; and when receiving the user's solution execution instruction, it calls the corresponding solution template and infers the next step node, simulating expert intervention until the solution execution is completed; at the same time, during the solution execution process, it provides the user with explanation content for the execution of each step node to assist the user's learning and understanding process.

[0013] The instance database is used to receive and store instance data of each executed solution from the solution generator, and generate a set of instances to be learned;

[0014] The inductive learning module is used to filter out the instance data of positive output from the set of instances to be learned, mark and cluster them, obtain the knowledge points learned by induction, and provide them to the knowledge inputter for query, so that the knowledge inputter can further input the knowledge points learned by induction into the solution generator; wherein the instance data of the positive output is the instance data corresponding to the input node score higher than the output node score;

[0015] The user interaction module is used to display an interactive interface to the user so that the user can interact through the interactive interface, and send data during the interaction process to the solution generator to promote the execution of the solution.

[0016] Furthermore, the solution generator includes: a knowledge entry module, a knowledge base, an inference engine, a global database and an interpreter, wherein:

[0017] The knowledge entry module receives knowledge points from external knowledge entry users, and organizes and analyzes the abstract knowledge in the knowledge points to generate corresponding solution templates;

[0018] The knowledge base stores the knowledge points from the knowledge entry module in the form of solution templates;

[0019] The knowledge base receives a solution execution instruction from a user, matches the content of the search instruction with the attributes of the solution template, and retrieves the best solution template;

[0020] When the inference engine receives the user's solution execution instruction, it retrieves and calls the corresponding solution template from the knowledge base, starts to execute the solution, gradually infers the next step node, simulates the expert's intervention, and if the user's participation is required, it sends it to the user in the form of a task for execution, and receives the interactive data fed back by the user; if the user's participation is not required, the task is automatically executed, and so on, until the solution execution is completed;

[0021] The global database is used to temporarily suspend and store the scheme in execution when the scheme requires user feedback; and to store the values ​​of global variables modified during the execution process;

[0022] During the execution of the scheme by the inference engine, the interpreter provides the user with explanations of the execution of each step of the inference engine to assist the user in the learning and understanding process.

[0023] Furthermore, the content of the solution template includes: a flowchart corresponding to a knowledge point, the direction of an arrow, the execution result of the previous step, expert rules and global data conditions;

[0024] The properties of the solution template include: whether to enable, the discipline to which it belongs, and the application scenario.

[0025] Furthermore, the flowchart includes: general nodes, task orientations and business nodes. The general nodes include nodes of various types during the execution of the solution. The task orientations indicate the direction of node execution and guide the next execution node. The business nodes include user interaction actions during the execution of the solution.

[0026] Further, for a branch node, the inference engine executes a conditional expression on the branch node, and if the conditional expression is judged to be true, the next node of the outgoing edge that is true is selected as the next node to be executed;

[0027] For a business node, the inference engine sends instructions to the user interaction module according to the type of the node, and then stores the intermediate result data of the solution execution in the global database, and suspends the solution;

[0028] After executing the current node, the next node to be executed is obtained according to the direction of the outgoing arrow, and so on until the end node.

[0029] Furthermore, each time the inference engine executes a node, a change in the variable in the global database will occur. The interpreter extracts the value before the change and the value after the change from the global database, and then generates initial log data. At the same time, a timestamp label, node type, node identifier and node execution time are added to the initial log data, stored in the global database, and associated with the execution plan in the context to obtain an interpretation log of the execution plan.

[0030] Furthermore, the inference engine retrieves a solution template to be executed from the knowledge base, replaces the placeholders in the solution template with real values, and obtains a real executable solution after duplication, which is stored in the global database.

[0031] Furthermore, the instance data of each completed execution plan includes: an identification of the completed execution plan, execution process data and execution result data.

[0032] Furthermore, the inductive learning module classifies the node execution inputs and outputs in all instance data according to the node type and places them into the corresponding data training set; marks the instance data with positive output, and clusters the instance input data of the node with positive output through a clustering algorithm, and obtains the optimal data parameters according to the clustering results as the knowledge points obtained by inductive learning.

[0033] Furthermore, according to the optimal data parameters obtained by the inductive learning module, the default parameters of the corresponding node types in the solution in the solution generator are adjusted.

[0034] The AI-based learning training program generation device of the present invention has the following beneficial effects:

[0035] (1) AI-based student intervention task generation technology and model-based expert knowledge optimization technology use big data technology to build models and accurately push the most suitable resources to students, making it easier for students to choose resources. Each student will get the most suitable resources without wasting time.

[0036] (2) Using AIGC technology, historical actions are used to accurately generate personalized next-step actions that are in line with the individual. This has a good learning effect on all students, does not waste time, can adapt to the individual situations of all students, is transparent to differences and has a good intervention effect, has a good problem-solving effect and a controllable problem-solving process, so that students can better learn and master knowledge, which is conducive to improving their academic performance.

[0037] (3) Adopting digital twin technology that processes expert knowledge, using standard intervention actions, and digital twin teaching and research experts’ knowledge makes the learning process more scientific, makes learning simpler and more efficient, and the learning effect more significant. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings are only used to illustrate specific embodiments and are not considered to limit the present invention. In the entire drawings, the same reference symbols represent the same components. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0039] Figure 1 This is a structural block diagram of an AI-based learning and training program generation device according to an embodiment of the present invention;

[0040] Figure 2 A schematic diagram of an AI-based learning and training program generation device according to an embodiment of the present invention;

[0041] Figure 3 A schematic diagram of knowledge points of an embodiment of the present invention;

[0042] Figure 4 The present invention is a flowchart of an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in the field without making creative work should fall within the scope of protection of the present invention.

[0044] Furthermore, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessarily obscuring the concepts disclosed in the present invention.

[0045] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the orientation or position relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. is based on the orientation or position relationship shown in the drawings, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. The terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0046] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples of methods and systems consistent with some aspects of the present invention as detailed in the appended claims.

[0047] like Figure 1 As shown, the present invention proposes an AI-based learning training program generation device, including: a program generator 100, an instance database 200, an inductive learning module 300 and a user interaction module 400.

[0048] Specifically, the solution generator 100 receives the digitally processed knowledge points and stores them in the internal knowledge base in the form of corresponding matching solution templates according to the content of the knowledge points. When the solution generator 100 receives the user's solution execution instruction, it calls the corresponding solution template and infers the next node, simulating the expert's intervention until the solution execution is completed. At the same time, during the solution execution process, the solution generator 100 provides the user with explanation content for the execution of each step node to assist the user's learning and understanding process.

[0049] It should be noted that the “user” in the relevant embodiments of the present invention may be a student.

[0050] refer to Figure 2 , the solution generator 100 includes: a knowledge entry module, a knowledge base, an inference engine, a global database and an interpreter.

[0051] The knowledge entry module receives knowledge points from external knowledge entry providers, organizes and analyzes the abstract knowledge in the knowledge points, and generates corresponding solution templates.

[0052] That is, knowledge points are abstract knowledge in the mind of the knowledge inputter, which is transformed into concrete knowledge through the knowledge input module and expressed as solution templates. Knowledge points are input into the knowledge input module, which organizes them into solution templates, and then stores the solution templates in the knowledge base.

[0053] When the system is cold started, the teaching and research experts will digitize the solution, process, and steps of a certain knowledge point into a solution template. The solution template is entered into the system through the knowledge entry module and stored in the knowledge base to prepare for the next step of solution execution. In this way, the knowledge of the teaching and research experts is digitally twinned into the computer system, so that student problems can be accurately solved.

[0054] The knowledge points entered into the knowledge entry module are expressed as flow charts and global variables.

[0055] like Figure 3 As shown, the flowchart includes three types of nodes: general nodes, task-oriented nodes, and business nodes.

[0056] Common nodes include various types of nodes in the solution execution process. Common nodes include: start node (i.e., the solution starts here), end node (i.e., the solution ends here), branch node (i.e., the solution starts to fork from here). Among them, the representation of branch conditions is mainly expressed by conditional expressions. The expression is a j2ee boolean statement, and global variables can be used in the statement. Conditional expressions can only be attached to branch nodes.

[0057] Task orientation indicates the direction of node execution and guides which node needs to be executed next.

[0058] The business node includes user interaction actions during the execution of the solution. User interaction actions include, for example, action selector, watching videos, watching PPT, watching examples, doing exercises, and doing assessments. User interaction actions are not limited to the above examples and can be expanded as needed.

[0059] Global variables are used by all nodes in the flowchart. They can use variables and change their values. Mapping is done by the type of the variable.

[0060] Variables have types, there are three categories with 20 types in total, you can only choose one. The three types are basic type, list type, and character type.

[0061] The basic types include TypeString, TypeNumber, and TypeBoolean, whose default values ​​are true or false. The default value of TypeDate is a string in the format of yyyy-MM-dd HH:mm:ss.

[0062] List types include TypeListBoolean, TypeListDate, TypeListMapBoolean, TypeListMapDate, TypeListMapNumber, TypeListMapString, TypeListNumber, and TypeListString.

[0063] Typical types include TypeMapBoolean, TypeMapDate, TypeMapListBoolean, TypeMapListDate, TypeMapListNumber, TypeMapListString, TypeMapNumber, TypeMapString

[0064] The knowledge base stores the knowledge points from the knowledge entry module in the form of solution templates. The content of the solution template includes: the flowchart corresponding to the knowledge point, the direction of the arrow, the execution result of the previous step, expert rules and global data. The properties of the solution template include: whether it is enabled, the subject it belongs to and the application scenario.

[0065] The knowledge base has a retrieval function, and can retrieve suitable solution templates according to the above attributes. Specifically, the knowledge base receives the user's solution execution instruction, matches the content of the retrieval instruction with the attributes of the solution template, and retrieves the best solution template.

[0066] The following is an explanation of the solution execution process in combination with the functions of the inference engine.

[0067] When the inference engine receives the user's solution execution instruction, it retrieves and calls the corresponding solution template from the knowledge base, starts to execute the solution, gradually infers the next step node, simulates expert intervention, and if user participation is required, it will be sent to the user in the form of a task and receive interactive data from the user; if user participation is not required, the task will be executed automatically, and so on, until the solution is completed.

[0068] That is, when a user executes a plan, the inference engine dynamically generates the next step based on the plan template flow chart, the direction of the arrow, the execution result of the previous step, the expert rules, and the global data situation, thereby simulating the expert to intervene. If user participation is required, it will be sent to the user in the form of a task for execution, so that the user does not have to choose what to do by himself, but just follows the task. If user participation is not required, the system will automatically execute the task next, and so on until the end node, and the plan ends.

[0069] 1) Generation of plans to be executed

[0070] First, the inference engine retrieves the solution template to be executed from the knowledge base, using the application scenario and the subject to which it belongs as the main search criteria, and filtering out solution templates that have not been enabled. Then, the placeholders in the solution template are replaced with real values, such as the student ID, subject ID, assessment paper ID, etc. Finally, a copy of the replaced solution template is made into a real executable solution, which is stored in the global database.

[0071] 2) Execution of pending plans

[0072] a. Start and suspend the program

[0073] First, the execution of the plan is indicated by the start node, and the next node to be executed is obtained according to the unique outgoing arrow of the start node.

[0074] If it is a branch node, the inference engine executes the conditional expression on the branch node. If the conditional expression is judged to be true, the next node of the true outgoing edge is selected as the next node to be executed.

[0075] If it is a business node, the inference engine sends an instruction to the user interaction module 400 according to the type of the node, and then stores the intermediate result data of the solution execution in the global database, and the solution is suspended.

[0076] Finally, after executing the current node, the next node to be executed is obtained according to the direction of the outgoing arrow, and so on until the end node.

[0077] b. Awakening of the solution

[0078] After receiving feedback from the user interaction module 400, the inference engine retrieves the suspended solution from the global database, and then continues to execute the subsequent process (refer to the logical process of a above).

[0079] The global database is used to temporarily suspend and store the executing plan when the plan requires user feedback. When there is feedback data, the system will retrieve the plan to be executed from this module and then execute it.

[0080] Each node in the execution process may modify the global variables, and the modified global variables will be stored in the global database. Therefore, when the next node is executed, the latest value of the global variable will be seen.

[0081] During the execution of the inference engine, the interpreter provides the user with explanations of each step of the execution of the inference engine to assist the user's learning and understanding process.

[0082] When the inference engine executes the plan, the interpreter will explain each step of its execution so that humans can understand the process and steps of its reasoning. Whenever the inference engine executes a node, a change of variables in the global database will occur. The interpreter extracts the value before and after the change from the global database, and then generates the initial log data. At the same time, the timestamp label, node type, node identifier and node execution time are added to the initial log data and stored in the global database. The log data generated by the execution is then stored in the global database and associated with the execution plan in the context. In this way, the explanation log of the plan can be obtained through the plan.

[0083] The example database 200 receives and stores the example data of each executed solution from the solution generator 100, and generates a set of examples to be learned. The example data of each executed solution includes: an identifier of the executed solution, execution process data, and execution result data.

[0084] Specifically, after the execution of the scheme is completed, the reasoning machine will send out the identification of the executed scheme, the execution process data, and the execution result data in the form of a message, so that the instance database 200 can receive these instance data through the message. After receiving the process data and result data of the instance, the instance database 200 stores the input parameters and output parameters of each step of the execution in the instance data in the database to prepare data for the next step of the inductive learning module 300.

[0085] Among them, input parameters: configuration information of each node, such as: ID of the test paper of the test paper node, etc. Output parameters: result data after each node is executed, such as: score of the test paper node.

[0086] After the system has been running for a period of time, a large amount of data will be accumulated, and it will enter the hot start stage. Based on these big data, the system uses machine statistical learning related technologies to learn the best parameters through the inductive learning module 300. The teaching and research staff verifies these parameters and data, and then modifies and optimizes the default parameters of the node type in the corresponding solution template, making the solution more accurate and easier to use.

[0087] The inductive learning module 300 selects the positive output instance data from the set of instances to be learned, labels and clusters them, obtains the knowledge points learned by induction, and provides them to the knowledge inputter for query, so that the knowledge inputter can further input the knowledge points learned by induction into the solution generator 100. The positive output instance data is the instance data corresponding to the input node score higher than the output node score.

[0088] The inductive learning module 300 first classifies the node execution input and output in all instance data into the corresponding data training set according to the node type. Then the instance data with positive output is marked, and the marking rule is: the scheme execution effect is good, specifically judged as the input node score is higher than the output node score (the score of the entrance exam is higher than the entrance exam). Finally, the instance input data of the node with positive output is clustered using the clustering algorithm in machine learning, and the optimal data parameters are obtained according to the clustering results as the knowledge points obtained by inductive learning.

[0089] It should be noted that the instance input data is a group of discrete data. After the clustering algorithm, the optimal data parameters representing the group can be obtained.

[0090] For example, when making a test paper node, the input parameter is the test paper ID, which was previously filled in by the data entry personnel based on their own experience. After learning, an optimal test paper ID is obtained, and this parameter will automatically be displayed in the text input box as a default value.

[0091] Perform template optimization adjustment (knowledge entry): adjust the default parameters of the corresponding node types in the solution in the solution generator 100 according to the optimal data parameters obtained by the inductive learning module 300. This step can be completed by teaching and research personnel or automatically by the system.

[0092] The user interaction module 400 displays an interactive interface to the user so that the user can interact through the interactive interface, and sends data during the interaction process to the solution generator 100 to promote the execution of the solution.

[0093] Specifically, some tasks need to be performed by users and interact with users in order for the inference engine to continue to execute, such as doing evaluations and watching videos. At this time, a user interaction module 400 is needed to undertake these functions.

[0094] First, the user interaction module 400 displays different interactive interfaces according to different instruction types. For example, the instruction to take an assessment will display an interface for a user to take an assessment paper. Then, the user interacts with this interactive interface, such as taking an assessment, watching a video, watching a PPT, etc. Figure 4Finally, the user interaction module 400 transmits the data of the interaction process back to the inference engine, which then performs the next step, such as which buttons to click, how long the task takes to execute, and the user's answer input.

[0095] The AI-based learning training program generation device proposed in the present invention adopts the process-based digital twin technology of expert knowledge, AI-based student intervention task generation technology and model-based expert knowledge optimization technology, uses AIGC technology, combines digital twins with big data, uses historical records, and comprehensively derives the next intervention action. Through step-by-step execution, the corresponding learning problems are finally solved. It is transparent to student differences and has good intervention effects. It has good problem-solving effects and a controllable solution process.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the above embodiments, or replace some of the technical features therein by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention. Any changes or replacements that can be easily thought of by those skilled in the art within the technical scope disclosed in the present invention should be covered within the protection scope of the present invention.

Claims

1. A device for generating a learning training program based on AI, characterized in that: include: Solution generator, instance database, inductive learning module and user interaction module, among which, The solution generator is used to receive digitally processed knowledge points, and store them in the internal knowledge base in the form of corresponding matching solution templates according to the content of the knowledge points; and when receiving the user's solution execution instruction, the content of the retrieval instruction is matched with the attributes of the solution template, the best solution template is retrieved, the best solution template is called and the next step node is inferred, and the expert intervention is simulated until the solution execution is completed; at the same time, during the solution execution process, the user is provided with explanation content for the execution of each step node to assist the user's learning and understanding process; wherein, the content of the solution template includes: the flowchart corresponding to the knowledge point, the direction of the arrow, the execution result of the previous step, the expert rules and the global data situation; the attributes of the solution template include: whether it is enabled, the subject to which it belongs and the application scenario; The instance database is used to receive and store instance data of each executed scheme from the scheme generator to generate a set of instances to be learned; wherein the instance data of each executed scheme includes: an identifier of the executed scheme, execution process data and execution result data; The inductive learning module is used to filter out the instance data of positive output from the set of instances to be learned, mark and cluster them, obtain the knowledge points learned by induction, and provide them to the knowledge inputter for query, so that the knowledge inputter can further input the knowledge points learned by induction into the solution generator; wherein the instance data of the positive output is the instance data corresponding to the input node score higher than the output node score; The user interaction module is used to display an interactive interface to the user so that the user can interact through the interactive interface, and send data during the interaction process to the solution generator to promote the execution of the solution.

2. The AI-based learning training program generation device according to claim 1, characterized in that: The solution generator includes: a knowledge entry module, a knowledge base, an inference engine, a global database and an interpreter, wherein: The knowledge entry module receives knowledge points from external knowledge entry users, and organizes and analyzes the abstract knowledge in the knowledge points to generate corresponding solution templates; The knowledge base stores the knowledge points from the knowledge entry module in the form of solution templates; The knowledge base receives a solution execution instruction from a user, matches the content of the search instruction with the attributes of the solution template, and retrieves the best solution template; When the inference engine receives the user's solution execution instruction, it retrieves and calls the corresponding solution template from the knowledge base, starts to execute the solution, gradually infers the next step node, simulates the expert's intervention, and if the user's participation is required, it sends it to the user in the form of a task for execution, and receives the interactive data fed back by the user; if the user's participation is not required, the task is automatically executed, and so on, until the solution execution is completed; The global database is used to temporarily suspend and store the scheme in execution when the scheme requires user feedback; and to store the values ​​of global variables modified during the execution process; During the execution of the scheme by the inference engine, the interpreter provides the user with explanations of the execution of each step of the inference engine to assist the user in the learning and understanding process.

3. The AI-based learning training program generation device according to claim 2, characterized in that: The flowchart includes: general nodes, task orientation and business nodes. The general nodes include nodes of various types in the process of program execution. The task orientation indicates the direction of node execution and guides the next execution node. The business nodes include user interaction actions in the process of program execution.

4. The AI-based learning training program generation device according to claim 3, characterized in that: For a branch node, the inference engine executes the conditional expression on the branch node, and if the conditional expression is judged to be true, the next node of the outgoing edge that is true is selected as the next node to be executed; For a business node, the inference engine sends instructions to the user interaction module according to the type of the node, and then stores the intermediate result data of the solution execution in the global database, and suspends the solution; After executing the current node, the next node to be executed is obtained according to the direction of the outgoing arrow, and so on until the end node.

5. The AI-based learning training program generation device according to claim 2, characterized in that: Whenever the inference engine executes a node, a change will occur in the variables in the global database. The interpreter extracts the value before and after the change from the global database, and then generates initial log data. At the same time, a timestamp label, node type, node identifier and node execution time are added to the initial log data, stored in the global database, and associated with the execution plan in the context to obtain an interpretation log of the execution plan.

6. The AI-based learning training program generation device according to claim 2, characterized in that: The inference engine retrieves the solution template to be executed from the knowledge base, replaces the placeholders in the solution template with real values, and obtains a real executable solution after copying, and stores it in the global database.

7. The AI-based learning training program generation device according to claim 1, characterized in that: The inductive learning module classifies the node execution inputs and outputs in all instance data into corresponding data training sets according to the node types; The instance data with positive output is marked, and the instance input data of the node with positive output is clustered by a clustering algorithm. The optimal data parameters are obtained according to the clustering results as the knowledge points obtained by inductive learning.

8. The AI-based learning training program generation device according to claim 7, characterized in that: According to the optimal data parameters obtained by the inductive learning module, the default parameters of the corresponding node types in the solution in the solution generator are adjusted.

Images