Exercise processing method and device, electronic equipment and storage medium

Abstract

The application provides an exercise processing method and device based on artificial intelligence, electronic equipment and computer readable storage medium; the method comprises the following steps: obtaining exercise data, wherein the exercise data comprises stem data and solution data; presenting the stem data in a man-machine interaction interface, and determining a plurality of solution knowledge points corresponding to the solution data; determining the relationship of the plurality of solution knowledge points in a knowledge graph corresponding to the subject of the exercise data, and determining at least one solution step included in the solution data according to the relationship; performing title classification processing on each solution step to determine the title of each solution step; and presenting at least one solution step and the corresponding title in the man-machine interaction interface. Through the application, the efficiency and accuracy of user understanding of the solution process of the exercise can be improved.

Description

Technical Field

[0001] This invention relates to artificial intelligence technology, and more particularly to an artificial intelligence-based problem-solving method, apparatus, electronic device, and computer-readable storage medium. Background Technology

[0002] Artificial intelligence (AI) is the theory, methods, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0003] Taking Computer-Aided Instruction (CAI) as an example, CAI refers to various teaching activities conducted with the assistance of computers. In related technologies, users can use traditional search engines to find questions or solutions of interest. However, users can only learn by simply viewing the questions or solutions, making it difficult to understand the solution process, thus reducing learning efficiency. Summary of the Invention

[0004] This invention provides an artificial intelligence-based problem-solving method, apparatus, electronic device, and computer-readable storage medium, which can improve the efficiency and accuracy of users' understanding of the problem-solving process.

[0005] The technical solution of this invention is implemented as follows:

[0006] This invention provides an artificial intelligence-based problem-solving method, comprising:

[0007] Acquire exercise data, wherein the exercise data includes question stem data and solution data;

[0008] The question data is presented in the human-computer interaction interface, and multiple problem-solving knowledge points corresponding to the solution data are identified;

[0009] In the knowledge graph of the subject corresponding to the exercise data, the relationship between the multiple problem-solving knowledge points is determined, and at least one problem-solving step included in the problem-solving data is determined based on the relationship;

[0010] Each of the problem-solving steps is categorized by title to determine the title of each problem-solving step;

[0011] At least one of the problem-solving steps and its corresponding title are presented in the human-computer interaction interface.

[0012] The above scheme also includes:

[0013] Obtain answer data;

[0014] Identify the solution data that matches the question stem data corresponding to the answer data;

[0015] Identify the problem-solving steps where there is a difference between the answer data and the corresponding solution data, and determine the problem-solving knowledge points and titles corresponding to the problem-solving steps with the difference;

[0016] The human-computer interaction interface presents the different problem-solving steps, corresponding problem-solving knowledge points, and titles.

[0017] The above scheme also includes:

[0018] The question stem data is encoded to determine the first feature set corresponding to the question stem data;

[0019] Identify at least one question stem data in the question bank that matches the question stem data, and use it as the recalled question stem data;

[0020] Based on the recalled question stem data, a second feature set corresponding to the question stem data is determined;

[0021] The features in the first feature set and the features in the second feature set are fused together to obtain a fused feature set;

[0022] Classification processing is performed based on multiple fusion features in the fusion feature set to determine at least one question stem knowledge point corresponding to the question stem data.

[0023] In the above scheme, the step of encoding the question stem data to determine the first feature set corresponding to the question stem data includes:

[0024] Determine the known content and the solution content included in the question data, and determine the subject concepts and formulas included in the known content and the solution content respectively;

[0025] Determine the subject-specific terminology corresponding to the formula;

[0026] The subject concepts and subject terms are encoded separately to obtain multiple first features, and

[0027] The plurality of first features are determined as the first feature set.

[0028] In the above scheme, determining the second feature set corresponding to the recalled question stem data based on the recalled question stem data includes:

[0029] Each knowledge point included in the recalled question stem data is used as a candidate knowledge point;

[0030] Determine the similarity between each recalled question stem and the question stem data;

[0031] Each candidate knowledge point is combined with its corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge point, and these multiple combined features are determined as the second feature set.

[0032] Wherein, the knowledge point similarity is the similarity between the recall question stem data to which the candidate knowledge point belongs and the question stem data.

[0033] In the above scheme, the step of fusing the features in the first feature set and the features in the second feature set to obtain a fused feature set includes:

[0034] Combine all features in the first feature set to obtain a first combined feature;

[0035] The first combined feature is fused with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points.

[0036] Multiple fusion features are determined as the fusion feature set.

[0037] In the above scheme, the classification process based on multiple fusion features in the fusion feature set to determine at least one question stem knowledge point corresponding to the question stem data includes:

[0038] Each fusion feature in the fusion feature set is mapped to the probability that the corresponding candidate knowledge point belongs to the knowledge point in the question stem.

[0039] The multiple fusion features are sorted in descending order according to the probabilities;

[0040] The candidate knowledge points corresponding to the fusion features that are sorted in descending order are determined as the knowledge points in the question stem.

[0041] This invention provides an artificial intelligence-based problem-solving method, comprising:

[0042] Acquire exercise data, wherein the exercise data includes solution data;

[0043] Identify the multiple problem-solving knowledge points corresponding to the problem-solving data;

[0044] In the knowledge graph of the subject corresponding to the exercise data, the relationship between the multiple problem-solving knowledge points is determined, and at least one problem-solving step included in the problem-solving data is determined based on the relationship;

[0045] Each of the problem-solving steps is categorized by title to determine the title of each problem-solving step.

[0046] In the above scheme, the exercise data includes question stem data, and the method further includes:

[0047] Identify at least one knowledge point corresponding to the question stem data.

[0048] The above scheme also includes:

[0049] The exercise data, the knowledge points in the question stem, the knowledge points for solving the problem, and the title of each of the problem-solving steps are stored in the database.

[0050] In the above scheme, determining the multiple problem-solving knowledge points corresponding to the problem-solving data includes:

[0051] The problem-solving data is divided into multiple candidate problem-solving steps;

[0052] For each candidate solution step, the following processing is performed to determine the solution knowledge points for each candidate solution step:

[0053] The candidate problem-solving steps are encoded to determine a first feature set corresponding to the candidate problem-solving steps;

[0054] Identify at least one solution step in the exercise bank that matches the candidate solution step, and use it as a recalled solution step;

[0055] Based on the aforementioned recall problem-solving steps, a second feature set corresponding to the candidate problem-solving steps is determined;

[0056] The features in the first feature set and the features in the second feature set are fused together to obtain a fused feature set;

[0057] Based on multiple fusion features in the fusion feature set, classification processing is performed to determine the problem-solving knowledge points of the candidate problem-solving steps.

[0058] In the above scheme, the step of encoding the candidate problem-solving steps to determine the first feature set corresponding to the candidate problem-solving steps includes:

[0059] Identify the subject concepts and formulas included in the candidate problem-solving steps;

[0060] Determine the subject-specific terminology corresponding to the formula;

[0061] The subject concepts and corresponding subject terms in the formulas are encoded separately to obtain multiple first features, and

[0062] The plurality of first features are determined as the first feature set.

[0063] In the above scheme, determining the second feature set corresponding to the candidate problem-solving steps based on the recall problem-solving steps includes:

[0064] Each knowledge point included in the recall and problem-solving steps is used as a candidate knowledge point;

[0065] Determine the similarity between each recall problem-solving step and the candidate problem-solving step;

[0066] Each candidate knowledge point is combined with its corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge point, and these multiple combined features are determined as the second feature set.

[0067] The knowledge point similarity is the similarity between the recall problem-solving step to which the candidate knowledge point belongs and the candidate problem-solving step.

[0068] In the above scheme, the step of fusing the features in the first feature set and the features in the second feature set to obtain a fused feature set includes:

[0069] Combine all features in the first feature set to obtain a first combined feature;

[0070] The first combined feature is fused with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points, and the multiple fused features are determined as the fused feature set.

[0071] In the above scheme, the classification process based on multiple fusion features in the fusion feature set to determine the problem-solving knowledge points of the candidate problem-solving steps includes:

[0072] Each fusion feature in the fusion feature set is mapped to the probability that the corresponding candidate knowledge point belongs to the problem-solving knowledge point;

[0073] The multiple fusion features are sorted in descending order according to the probabilities;

[0074] The candidate knowledge points corresponding to the fusion features that are sorted in descending order are determined as the problem-solving knowledge points of the candidate problem-solving steps.

[0075] In the above scheme, determining the relationship between the multiple problem-solving knowledge points in the knowledge graph of the subject corresponding to the exercise data, and determining at least one problem-solving step included in the problem-solving data based on the relationship, includes:

[0076] Determine the node in the knowledge graph corresponding to the problem-solving knowledge point of each candidate problem-solving step;

[0077] When there is a parent-child relationship between the nodes corresponding to the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data, the adjacent candidate problem-solving steps corresponding to the nodes with parent-child relationship are merged into a new problem-solving step, which is used as the target problem-solving step.

[0078] When there is no parent-child relationship between the nodes corresponding to the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data, the adjacent candidate problem-solving steps corresponding to the nodes that do not have a parent-child relationship are independently used as target problem-solving steps.

[0079] The target problem-solving steps are defined as the problem-solving steps included in the problem-solving data.

[0080] In the above scheme, the step of classifying the titles of each problem-solving step to determine the title of each problem-solving step includes:

[0081] For each of the problem-solving steps, the following processing is performed:

[0082] Identify the problem-solving knowledge points and formulas included in the problem-solving steps;

[0083] Determine the subject-specific terminology corresponding to the formula;

[0084] The problem-solving knowledge points included in the problem-solving steps and the corresponding subject terms of the formula are combined to obtain a combination feature;

[0085] The combined features are mapped to probabilities belonging to different titles, and the title corresponding to the highest probability is determined as the title of the problem-solving step.

[0086] In the above scheme, the method further includes:

[0087] The question stem data is encoded to determine the first feature set corresponding to the question stem data;

[0088] Identify at least one question stem data in the question bank that matches the question stem data, and use it as the recalled question stem data;

[0089] Based on the recalled question stem data, a second feature set corresponding to the question stem data is determined;

[0090] The features in the first feature set and the features in the second feature set are fused together to obtain a fused feature set;

[0091] Classification processing is performed based on multiple fusion features in the fusion feature set to determine at least one question stem knowledge point corresponding to the question stem data.

[0092] The above scheme also includes:

[0093] Obtain answer data;

[0094] Identify the solution data that matches the question stem data corresponding to the answer data;

[0095] Identify the problem-solving steps where there is a difference between the answer data and the corresponding solution data, and determine the problem-solving knowledge points and titles corresponding to the problem-solving steps with the difference;

[0096] The human-computer interaction interface presents the different problem-solving steps, corresponding problem-solving knowledge points, and titles.

[0097] In the above scheme, the step of encoding the question stem data to determine the first feature set corresponding to the question stem data includes:

[0098] Determine the known content and the solution content included in the question data, and determine the subject concepts and formulas included in the known content and the solution content respectively;

[0099] Determine the subject-specific terminology corresponding to the formula;

[0100] The subject concepts and subject terms are encoded separately to obtain multiple first features, and

[0101] The plurality of first features are determined as the first feature set.

[0102] In the above scheme, determining the second feature set corresponding to the recalled question stem data based on the recalled question stem data includes:

[0103] Each knowledge point included in the recalled question stem data is used as a candidate knowledge point;

[0104] Determine the similarity between each recalled question stem and the question stem data;

[0105] Each candidate knowledge point is combined with its corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge point, and these multiple combined features are determined as the second feature set.

[0106] Wherein, the knowledge point similarity is the similarity between the recall question stem data to which the candidate knowledge point belongs and the question stem data.

[0107] In the above scheme, the step of fusing the features in the first feature set and the features in the second feature set to obtain a fused feature set includes:

[0108] Combine all features in the first feature set to obtain a first combined feature;

[0109] The first combined feature is fused with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points.

[0110] Multiple fusion features are determined as the fusion feature set.

[0111] In the above scheme, the classification process based on multiple fusion features in the fusion feature set to determine at least one question stem knowledge point corresponding to the question stem data includes:

[0112] Each fusion feature in the fusion feature set is mapped to the probability that the corresponding candidate knowledge point belongs to the knowledge point in the question stem.

[0113] The multiple fusion features are sorted in descending order according to the probabilities;

[0114] The candidate knowledge points corresponding to the fusion features that are sorted in descending order are determined as the knowledge points in the question stem.

[0115] This invention provides an artificial intelligence-based exercise processing device, comprising:

[0116] The acquisition module is used to acquire exercise data, wherein the exercise data includes question stem data and solution data;

[0117] The exercise presentation module is used to present the question stem data in the exercise data in the human-computer interaction interface;

[0118] The knowledge point determination module is used to determine multiple problem-solving knowledge points corresponding to the problem-solving data;

[0119] The step determination module is used to determine the relationship between the multiple problem-solving knowledge points in the knowledge graph of the subject corresponding to the exercise data, and to determine at least one problem-solving step included in the problem-solving data based on the relationship.

[0120] The title classification module is used to perform title classification processing on each of the problem-solving steps to determine the title of each of the problem-solving steps;

[0121] The exercise presentation module is also used to present at least one of the problem-solving steps and the corresponding title in the human-computer interaction interface.

[0122] In the above scheme, the knowledge point determination module is further configured to segment the problem-solving data into multiple candidate problem-solving steps; perform the following processing on each candidate problem-solving step to determine the problem-solving knowledge points of each candidate problem-solving step: encode the candidate problem-solving step to determine a first feature set corresponding to the candidate problem-solving step; determine at least one problem-solving step in the exercise bank that matches the candidate problem-solving step as a recall problem-solving step; determine a second feature set corresponding to the candidate problem-solving step based on the recall problem-solving step; fuse the features in the first feature set and the features in the second feature set to obtain a fused feature set; and perform classification processing based on multiple fused features in the fused feature set to determine the problem-solving knowledge points of the candidate problem-solving step.

[0123] In the above scheme, the knowledge point determination module is further used to determine the subject concepts and formulas included in the candidate problem-solving steps; determine the subject terms corresponding to the formulas; encode the subject concepts and the subject terms corresponding to the formulas respectively to obtain multiple first features, and determine the multiple first features as the first feature set.

[0124] In the above scheme, the knowledge point determination module is further configured to: use the knowledge points included in each recall problem-solving step as candidate knowledge points; determine the similarity between each recall problem-solving step and the candidate problem-solving step; combine each candidate knowledge point and its corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge points, and determine the multiple combined features as the second feature set; wherein, the knowledge point similarity is the similarity between the recall problem-solving step to which the candidate knowledge point belongs and the candidate problem-solving step.

[0125] In the above scheme, the knowledge point determination module is further configured to combine all features in the first feature set to obtain a first combined feature; fuse the first combined feature with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points, and determine the multiple fused features as the fused feature set.

[0126] In the above scheme, the knowledge point determination module is further configured to map each fusion feature in the fusion feature set to the probability that the corresponding candidate knowledge point belongs to the problem-solving knowledge point; sort the multiple fusion features in descending order according to the probability; and determine the candidate knowledge points corresponding to the fusion features that are sorted first in the descending order as the problem-solving knowledge points of the candidate problem-solving steps.

[0127] In the above scheme, the step determination module is further configured to determine the node corresponding to the problem-solving knowledge point of each candidate problem-solving step in the knowledge graph; when there is a parent-child relationship between the nodes corresponding to the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data, the adjacent candidate problem-solving steps corresponding to the nodes with parent-child relationships are merged into a new problem-solving step as the target problem-solving step; when there is no parent-child relationship between the nodes corresponding to the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data, the adjacent candidate problem-solving steps corresponding to the nodes without parent-child relationships are each independently used as the target problem-solving step; the target problem-solving step is determined as the problem-solving steps included in the problem-solving data.

[0128] In the above scheme, the title classification module is further configured to perform the following processing on each problem-solving step: determine the problem-solving knowledge points and formulas included in the problem-solving step; determine the subject terms corresponding to the formulas; combine the problem-solving knowledge points included in the problem-solving step and the subject terms corresponding to the formulas to obtain combination features; map the combination features to probabilities of belonging to different titles, and determine the title corresponding to the highest probability as the title of the problem-solving step.

[0129] In the above scheme, the knowledge point determination module is further configured to encode the question stem data to determine a first feature set corresponding to the question stem data; determine at least one question stem data in the question bank that matches the question stem data as recall question stem data; determine a second feature set corresponding to the question stem data based on the recall question stem data; fuse the features in the first feature set and the features in the second feature set to obtain a fused feature set; and perform classification processing based on multiple fused features in the fused feature set to determine the question stem knowledge points.

[0130] In the above scheme, the knowledge point determination module is further used to determine the known content and the solution content included in the question stem data, and to determine the subject concepts and formulas included in the known content and the solution content respectively; to determine the subject terms corresponding to the formulas; to encode the subject concepts and the subject terms respectively to obtain multiple first features, and to determine the multiple first features as the first feature set.

[0131] In the above scheme, the knowledge point determination module is further configured to: use the knowledge points included in each recalled question stem data as candidate knowledge points; determine the similarity between each recalled question stem data and the question stem data; combine each candidate knowledge point and its corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge points, and determine the multiple combined features as the second feature set; wherein, the knowledge point similarity is the similarity between the recalled question stem data to which the candidate knowledge point belongs and the question stem data.

[0132] In the above scheme, the knowledge point determination module is further configured to combine all features in the first feature set to obtain a first combined feature; the first combined feature is fused with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points, and the multiple fused features are determined as the fused feature set.

[0133] In the above scheme, the knowledge point determination module is further configured to map each fusion feature in the fusion feature set to the probability that the corresponding candidate knowledge point belongs to the question stem knowledge point; sort the multiple fusion features in descending order according to the probability; and determine the candidate knowledge points corresponding to the fusion features that are sorted first in the descending order as the question stem knowledge points.

[0134] In the above scheme, the AI-based exercise processing device further includes: a judgment module, used to acquire answer data; determine solution data that is consistent with the question stem data corresponding to the answer data; determine that there are differences between the answer data and the corresponding solution data, and determine the solution knowledge points corresponding to the differences; and present the differences and the corresponding solution knowledge points in the human-computer interaction interface.

[0135] This invention provides an artificial intelligence-based exercise processing device, comprising:

[0136] The acquisition module is used to acquire exercise data, wherein the exercise data includes question stem data and solution data;

[0137] The knowledge point determination module is used to determine multiple problem-solving knowledge points corresponding to the problem-solving data;

[0138] The step determination module is used to determine the relationship between the multiple problem-solving knowledge points in the knowledge graph of the subject corresponding to the exercise data, and to determine at least one problem-solving step included in the problem-solving data based on the relationship.

[0139] The title classification module is used to perform title classification processing on each of the problem-solving steps to determine the title of each of the problem-solving steps.

[0140] This invention provides an electronic device, comprising:

[0141] Memory is used to store executable instructions for a computer;

[0142] The processor, when executing computer-executable instructions stored in the memory, implements the problem-solving method based on artificial intelligence provided in the embodiments of the present invention.

[0143] This invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the AI-based problem-solving method provided in this invention.

[0144] The embodiments of the present invention have the following beneficial effects:

[0145] Based on the relationships between multiple problem-solving knowledge points corresponding to the problem-solving data in the knowledge graph, the problem-solving steps for the corresponding problem-solving data can be accurately determined. By determining the titles of the corresponding problem-solving steps and presenting the problem-solving steps and their corresponding titles to users in a convenient and intuitive way, compared with directly presenting the problem-solving data to users in related technologies, it is beneficial to improve the efficiency and accuracy of users' understanding of the problem-solving process, thereby improving the quality of learning and saving human and labor resources. Attached Figure Description

[0146] Figure 1A and Figure 1B This is a schematic diagram of the structure of the AI-based exercise processing system 100 provided in an embodiment of the present invention;

[0147] Figure 2A and Figure 2B This is a schematic diagram of the structure of the electronic device 500 provided in an embodiment of the present invention;

[0148] Figure 3 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention;

[0149] Figure 4 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention;

[0150] Figure 5 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention;

[0151] Figure 6 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention;

[0152] Figure 7A and Figure 7BThis is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention;

[0153] Figure 8A This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention;

[0154] Figure 8B This is a schematic diagram of an application scenario provided by an embodiment of the present invention;

[0155] Figure 9 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention. Detailed Implementation

[0156] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on the present invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0157] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0158] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein is for the purpose of describing embodiments of the invention only and is not intended to limit the invention.

[0159] Before providing a further detailed description of the embodiments of the present invention, the nouns and terms involved in the embodiments of the present invention will be explained, and the nouns and terms involved in the embodiments of the present invention shall be interpreted as follows.

[0160] 1) Computer-assisted instruction refers to a teaching method that uses computers to help or replace teachers in performing some teaching tasks, imparting knowledge to students and providing skills training.

[0161] 2) Knowledge graphs, known in library and information science as knowledge domain visualization or knowledge domain mapping maps, are a series of various graphics that display the development process and structural relationships of knowledge. They use visualization techniques to describe knowledge resources and their carriers, mining, analyzing, constructing, drawing, and displaying knowledge and their interrelationships. Knowledge graphs are a modern theory that combines theories and methods from applied mathematics, computer graphics, information visualization technology, and information science with methods such as bibliometric citation analysis and co-occurrence analysis. They utilize visualized graphs to vividly display the core structure, development history, cutting-edge fields, and overall knowledge architecture of a discipline, achieving the goal of multidisciplinary integration. They can provide practical and valuable references for disciplinary research.

[0162] 3) A knowledge point, or test point, is a relatively independent smallest unit of knowledge, theory, principle, or idea. In educational practice, a knowledge point is a general term for a particular piece of knowledge, specifically referring to knowledge in textbooks or exams.

[0163] The advantages of the internet have enabled the rapid development of computer-aided instruction, accelerating the digitization of content and online publishing. As a crucial component of science education, science exercises are not only experiencing a rapid increase in quantity but also facing significant changes in their application. Current technologies can only find similar questions based on the literal similarity of the standard answers, but they cannot understand the knowledge points involved or the corresponding solution process.

[0164] To address the aforementioned technical problems, embodiments of the present invention provide an artificial intelligence-based problem-solving method, apparatus, electronic device, and computer-readable storage medium. These embodiments combine knowledge graphs and natural language processing technologies, enabling the breakdown of the problem-solving process and providing technical support for automatic problem explanation. Furthermore, they allow for the analysis of students' answers, accurately identifying the knowledge point where the student who answered the problem incorrectly made the mistake.

[0165] The following describes exemplary applications of the electronic devices provided in the embodiments of the present invention. The electronic devices provided in the embodiments of the present invention can be implemented as various types of user terminals such as laptops, tablets, desktop computers, set-top boxes, and mobile devices (e.g., mobile phones, portable music players, personal digital assistants, dedicated messaging devices, portable gaming devices), or as servers.

[0166] The following describes an exemplary application when the device is implemented as a terminal.

[0167] See Figure 1A , Figure 1AThis is a schematic diagram of the structure of an AI-based exercise processing system 100 provided in an embodiment of the present invention. The AI-based exercise processing system 100 includes a server 200, a network 300, and a terminal 400, which will be described separately.

[0168] Server 200 is the backend server for client 410, used to respond to client 410's exercise retrieval request and send the corresponding exercise data to client 410.

[0169] Network 300 is used as a medium for communication between server 200 and terminal 400, and can be a wide area network, a local area network, or a combination of both.

[0170] Terminal 400 is used to run client 410, which is a client with auxiliary teaching functions. Client 410 is used to receive exercise data, including question stem data and solution data, sent by server 200, and to present the question stem data in the human-computer interaction interface; it is also used to determine the solution steps and corresponding titles included in the solution data (the process of determining the solution steps and corresponding titles will be described in detail below), and to present the solution steps and corresponding titles in the human-computer interaction interface.

[0171] Here, client 410 is a computer program or computer program product running on terminal 400. Specifically, client 410 can be an application (APP) with auxiliary education function, such as an auxiliary education APP or an education system APP; it can also be an auxiliary education mini-program that can be embedded in any APP; or it can be a browser with auxiliary education function.

[0172] The following will describe an exemplary application when an electronic device is implemented as a server.

[0173] See Figure 1B , Figure 1B This is a schematic diagram of the structure of an AI-based exercise processing system 100 provided in an embodiment of the present invention. The AI-based exercise processing system 100 includes a server 200, a network 300, and a terminal 400, which will be described separately.

[0174] Server 200 is the backend server for client 410. It is used to respond to client 410's exercise retrieval request, retrieve the corresponding exercise data including question stem data and solution data, determine the solution steps and corresponding titles included in the solution data (the process of determining the solution steps and corresponding titles will be explained in detail below), and send the question stem data, solution steps and corresponding titles to client 410.

[0175] Network 300 is used as a medium for communication between server 200 and terminal 400, and can be a wide area network, a local area network, or a combination of both.

[0176] Terminal 400 is used to run client 410, which is a client with auxiliary teaching functions. Client 410 is used to receive the question data, solution steps and corresponding title sent by server 200 and display them in the human-computer interaction interface.

[0177] The embodiments of the present invention can be implemented with the help of cloud technology, which refers to a hosting technology that unifies a series of resources such as hardware, software, and networks within a wide area network or local area network to realize the calculation, storage, processing, and sharing of data.

[0178] Cloud technology is a collective term for network technology, information technology, integration technology, management platform technology, and application technology applied to the cloud computing business model. It can form resource pools, allowing for on-demand, flexible, and convenient use. Cloud computing technology will become a crucial support. The backend services of technical network systems require substantial computing and storage resources; for example, the portal websites of educational systems.

[0179] As an example, server 200 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. Terminal 400 can be a smartphone, tablet, laptop, desktop computer, smart speaker, or smartwatch, but is not limited to these. Terminal 400 and server 200 can be directly or indirectly connected via wired or wireless communication, which is not limited in this embodiment of the invention.

[0180] The structure of the electronic device provided in the embodiments of the present invention will be described next. The electronic device can be a terminal 400 or a server 200. See also Figure 2A and Figure 2B , Figure 2A and Figure 2B This is a schematic diagram of the structure of the electronic device 500 provided in an embodiment of the present invention.

[0181] Below, taking electronic device 500 as an example, which is terminal 400 or server 200, we will combine... Figure 2A Explain the structure of electronic device 500. Figure 2AThe illustrated electronic device 500 includes at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together via a bus system 540. It is understood that the bus system 540 is used to implement communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2A The general labeled all buses as Bus System 540.

[0182] The processor 510 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0183] User interface 530 includes one or more output devices 531 that enable the presentation of media content, including one or more speakers and / or one or more visual displays. User interface 530 also includes one or more input devices 532, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

[0184] The memory 550 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 550 may optionally include one or more storage devices physically located away from the processor 510.

[0185] The memory 550 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 550 described in this embodiment is intended to include any suitable type of memory.

[0186] In some embodiments, memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0187] Operating system 551 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0188] The network communication module 552 is used to reach other computing devices via one or more (wired or wireless) network interfaces 520, exemplary network interfaces 520 including: Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0189] Presentation module 553 is configured to enable the presentation of information (e.g., a user interface for operating peripheral devices and displaying content and information) via one or more output devices 531 (e.g., a display screen, a speaker, etc.) associated with user interface 530;

[0190] The input processing module 554 is used to detect and translate one or more user inputs or interactions from one or more input devices 532.

[0191] It should be noted that the user interface 530, presentation module 553, and input processing module 554 mentioned above are optional. Specifically, when the electronic device 500 is a terminal 400, the electronic device 500 includes the user interface 530, presentation module 553, and input processing module 554; when the electronic device 500 is a server 200, the electronic device 500 does not include the user interface 530, presentation module 553, and input processing module 554.

[0192] In some embodiments, when the electronic device 500 is a terminal 400, the AI-based problem-solving device provided in this embodiment can be implemented by various forms of computer programs or computer program products running on the terminal 400, such as the operating system 551, client 410, software modules, and scripts described above. Thus, tasks such as determining the problem-solving steps and corresponding titles can be completed directly using the terminal 400's own computing resources.

[0193] In some embodiments, when the electronic device 500 is a server 200, the AI-based problem-solving device provided in this embodiment of the invention can be implemented by various forms of computer programs or computer program products running on the server 200. For example, the back-end service program of an auxiliary teaching system (stored in the storage medium of the server 200 in the form of computer-executable instructions and run by the processor of the server 200) integrates at least one of various raw data, intermediate data at all levels and final results received from other devices with some data or results already on the server 200 to complete tasks such as determining the problem-solving steps and the corresponding title.

[0194] As an example, Figure 2AAn AI-based exercise processing device 555, stored in memory 550, is shown. It includes the following software modules: an acquisition module 5551, a knowledge point determination module 5552, a step determination module 5553, and a title classification module 5554. These modules are logically linked and can therefore be arbitrarily combined or further divided according to their implemented functions. The functions of each module will be described below.

[0195] Next, taking electronic device 500 as an example of terminal 400, combined with... Figure 2B Explain the structure of electronic device 500. Figure 2B The illustrated electronic device 500 includes at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together via a bus system 540. It is understood that the bus system 540 is used to implement communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2B The general labeled all buses as Bus System 540.

[0196] It should be noted that the processor 510, memory 550, network interface 520, user interface 530, and bus system 540 have similar functions to those described above, and will not be repeated here.

[0197] As an example, Figure 2B An AI-based problem-solving device 555, stored in memory 550, is shown. Figure 2A Building upon this, the AI-based exercise processing device 555 may further include an exercise presentation module 5555. These modules are logically connected and can therefore be arbitrarily combined or further divided according to their implemented functions. The functions of each module will be described below.

[0198] See Figure 3 , Figure 3 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention, which will be combined with... Figure 3 The steps shown are explained.

[0199] It should be noted that, Figure 3 The method shown can be executed by the terminal or the server alone, or by the terminal and the server working together. Figure 3 The execution subject of the steps of the method shown can be either server 200 or terminal 400. For example, for terminal 400, it can be executed by various forms of computer programs or computer program products running on terminal 400, such as operating system 551, client 410, software modules and scripts as described above.

[0200] In step S101, the exercise data is obtained.

[0201] See here. Figure 8B The exercise data includes question stem data 801 and solution data 802.

[0202] In step S102, multiple (i.e. at least two) problem-solving knowledge points corresponding to the problem-solving data are determined.

[0203] See Figure 4 , Figure 4 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention. Figure 3 Step S102 may specifically include steps S1021 to S1026.

[0204] In step S1021, the problem-solving data is divided into multiple candidate problem-solving steps.

[0205] In some embodiments, key characters in the problem-solving data are identified, and the problem-solving data is segmented into multiple candidate problem-solving steps based on keywords and / or key characters.

[0206] Here, since semicolons and periods typically indicate the end of a sentence (or statement) in problem-solving, the key characters can be semicolons and / or periods. In science exam questions, a sentence in the problem-solving process usually includes "because..." and "therefore...", so the keywords can be "because" and / or "therefore".

[0207] As an example, the statement between two key characters is identified as a candidate solution step; the statement containing the keyword is identified as a candidate solution step.

[0208] Next, steps S1022 to S1026 are performed on each candidate solution step obtained in step S1021 to determine the solution knowledge points of each candidate solution step.

[0209] In step S1022, the candidate problem-solving steps are encoded to determine the first feature set of the corresponding candidate problem-solving steps.

[0210] In some embodiments, the subject concepts and formulas included in the candidate problem-solving steps are determined; the subject terms corresponding to the formulas are determined; the subject concepts and the subject terms corresponding to the formulas are encoded respectively to obtain multiple first features, and the multiple first features are determined as a set of first features.

[0211] Here, the encoding can be one-hot encoding. Specifically, it uses an N (N is a positive integer) bit state register to encode N states. Each state has its own independent register bit, and at any given time, only one bit is valid.

[0212] For example, the process of determining the subject-specific terminology for a corresponding formula can be: when formula (x-1) 2 +(y-2) 2 When =9, it can be determined that the professional term corresponding to this formula is the standard equation of a circle. In this way, the textual information of the corresponding formula can be obtained, so as to encode the formula.

[0213] In step S1023, at least one solution step in the problem set that matches the candidate solution step is determined as the recall solution step.

[0214] In some embodiments, the similarity between all problem-solving steps in the exercise bank and candidate problem-solving steps is determined; based on the similarity, all problem-solving steps in the exercise bank are sorted in descending order, and at least one problem-solving step that appears first in the descending order is selected as the recalled problem-solving step. Thus, since the candidate problem-solving steps and the recalled problem-solving steps have a high degree of similarity and overlap in the knowledge points they encompass, selecting a recalled problem-solving step that is similar to the candidate problem-solving steps is beneficial for subsequently determining the knowledge points of the candidate problem-solving steps.

[0215] In step S1024, based on the recall problem-solving steps, a second feature set corresponding to the candidate problem-solving steps is determined.

[0216] In some embodiments, the knowledge points included in each recall problem-solving step are used as candidate knowledge points; the similarity between each recall problem-solving step and the candidate problem-solving step is determined; each candidate knowledge point and its corresponding knowledge point similarity are combined to obtain multiple combined features that correspond one-to-one with the candidate knowledge points, and the multiple combined features are determined as a second feature set.

[0217] Here, knowledge point similarity is the similarity between the recall problem-solving steps to which the candidate knowledge point belongs and the candidate problem-solving steps.

[0218] For example, when the similarity between the recalled solution step A and the candidate solution step is 0.3, solution step A includes knowledge points 1 and 2, the similarity between the recalled solution step B and the candidate solution step is 0.4, and solution step B includes knowledge points 2, 3, and 4, the knowledge point similarity of knowledge point 1 is 0.3 (the similarity between the recalled solution step A and the candidate solution step), the knowledge point similarity of knowledge point 1 is 0.3 + 0.4 = 0.7 (the sum of the similarity between the recalled solution step A and the candidate solution step, and the similarity between the recalled solution step B and the candidate solution step), the knowledge point similarity of knowledge point 3 is 0.4 (the similarity between the recalled solution step B and the candidate solution step), and the knowledge point similarity of knowledge point 4 is 0.4 (the similarity between the recalled solution step B and the candidate solution step). In this way, each candidate knowledge point can be weighted based on the knowledge point similarity to improve the accuracy of subsequently determining the solution knowledge points based on the candidate knowledge points.

[0219] In step S1025, the features in the first feature set and the features in the second feature set are fused to obtain a fused feature set.

[0220] Here, the features included in the second feature set correspond one-to-one with the candidate knowledge points. The features included in the fused feature set also correspond one-to-one with the candidate knowledge points.

[0221] In some embodiments, all features in the first feature set are combined to obtain a first combined feature; the first combined feature is fused with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with multiple candidate knowledge points, and the multiple fused features are determined as a fused feature set.

[0222] Taking a first feature set including first feature A and first feature B, and a second feature set including second feature C (corresponding to candidate knowledge point 1), second feature D (corresponding to candidate knowledge point 2), and second feature E (corresponding to candidate knowledge point 3) as an example, firstly, first feature A and second feature B in the first feature set are combined to obtain combined feature 1; then, combined feature 1 is fused with second feature C, second feature D, and second feature E respectively to obtain fused feature 1-C, fused feature 1-D, and fused feature 1-E, where fused feature 1-C corresponds to candidate knowledge point 1; fused feature 1-D corresponds to candidate knowledge point 2; and fused feature 1-E corresponds to candidate knowledge point 3. In this way, features from the first feature set and features from the second feature set are fused to improve the comprehensiveness and accuracy of subsequently determining the problem-solving knowledge points based on the fused features.

[0223] In step S1026, classification processing is performed based on multiple fusion features in the fusion feature set to determine the problem-solving knowledge points of the candidate problem-solving steps.

[0224] In some embodiments, each fusion feature in the fusion feature set is mapped to the probability that the corresponding candidate knowledge point belongs to the problem-solving knowledge point; the multiple fusion features are sorted in descending order according to the probability; and the candidate knowledge points corresponding to the first (or more) fusion features in the descending order are determined as the problem-solving knowledge points of the candidate problem-solving steps.

[0225] Here, the selection of the first fusion features in the descending sort can be achieved by choosing the fusion features that are first in the descending sort and have a probability greater than a probability threshold; or by choosing the fusion features that are first in the descending sort and whose quantity is equal to a quantity threshold. The probability threshold can be a default value, a user-defined value, or a value determined based on the probabilities of all fusion features, for example, the average probability of all fusion features. The quantity threshold can also be a default value, a user-defined value, or a value determined based on the number of fusion features included in the fusion feature set, for example, half the number of fusion features included in the fusion feature set.

[0226] The embodiments of the present invention are based on the high degree of overlap of knowledge points included in candidate problem-solving steps and similar recall problem-solving steps. By determining the knowledge points of candidate problem-solving steps based on the knowledge points of similar recall problem-solving steps, not only is it possible to eliminate the need to redetermine the knowledge points of candidate problem-solving steps based on semantic analysis, thus saving computational resources, but it also improves the accuracy of determining problem-solving knowledge points.

[0227] In step S103, the relationship between multiple problem-solving knowledge points is determined in the knowledge graph of the subject corresponding to the exercise data, and at least one problem-solving step is determined in the problem-solving data based on the relationship.

[0228] In some embodiments, the node corresponding to the problem-solving knowledge point of each candidate problem-solving step in the knowledge graph is determined; when there is a parent-child relationship between the nodes corresponding to the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data, the adjacent candidate problem-solving steps corresponding to the nodes with parent-child relationship are merged into a new problem-solving step as the target problem-solving step; when there is no parent-child relationship between the nodes corresponding to the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data, the adjacent candidate problem-solving steps corresponding to the nodes without parent-child relationship are each independently used as the target problem-solving step; the target problem-solving step is determined as the problem-solving steps included in the problem-solving data.

[0229] For example, Figure 8BIn the problem-solving data, the nodes corresponding to the knowledge points "definition of vertical angles" and "properties of vertical angles" in the knowledge graph have a parent-child relationship. Therefore, the candidate solution steps for the corresponding knowledge point "definition of vertical angles" and the corresponding knowledge point "properties of vertical angles" can be combined into one solution step (i.e., "step one" in step 806). The nodes corresponding to the knowledge points "definition of adjacent supplementary angles" and "properties of angle bisectors" in the knowledge graph do not have a parent-child relationship. Therefore, the candidate solution steps for the corresponding knowledge points "definition of adjacent supplementary angles" and the corresponding knowledge point "properties of angle bisectors" can be treated as two independent solution steps (i.e., "step three" and "step four" in step 806).

[0230] The embodiments of the present invention can merge steps that have a coherent solution and separate steps that do not. This not only makes it easier for users to understand the solution process, but also reduces the number of titles generated subsequently, saving computing resources.

[0231] In step S104, a title classification process is performed on each problem-solving step to determine the title of each problem-solving step.

[0232] In some embodiments, the title classification model is invoked to perform the following processing on each problem-solving step: determine the problem-solving knowledge points and formulas included in the problem-solving step; determine the subject terms of the corresponding formulas; combine the problem-solving knowledge points included in the problem-solving step and the subject terms of the corresponding formulas to obtain combined features; map the combined features to probabilities of belonging to different titles, and determine the title corresponding to the highest probability as the title of the problem-solving step.

[0233] Here, the problem-solving steps can include one or more problem-solving knowledge points. The title classification model is trained based on sample problem-solving steps and the titles labeled for the sample problem-solving steps.

[0234] This invention determines the title of a problem-solving step based on the knowledge points and formulas included in the problem-solving process. This precise determination of the title enables users to understand the problem-solving process based on the title, thereby improving the efficiency and accuracy of users' understanding of the problem-solving process, thus enhancing the quality of learning and saving human resources.

[0235] See Figure 5 , Figure 5 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention. Figure 3 Step S105 may be included after step S101. Step S105 and steps S102 to S104 are not in any order and can be executed in parallel or sequentially.

[0236] In step S105, at least one knowledge point corresponding to the question stem data is determined.

[0237] See Figure 6 , Figure 6 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention. Figure 5 Step S105 may specifically include steps S1051 to S1055.

[0238] In step S1051, the question stem data is encoded to determine the first feature set corresponding to the question stem data.

[0239] In some embodiments, the known content and the solution content included in the question stem data are determined, and the subject concepts and formulas included in the known content and the solution content are determined respectively; the subject terms corresponding to the formulas are determined; the subject concepts and subject terms are encoded respectively to obtain multiple first features, and the multiple first features are determined as a set of first features.

[0240] Here, the encoding can be one-hot encoding.

[0241] For example, the process of determining the subject terminology corresponding to a formula can be as follows: when the formula (x-1)2+(y-2)2=9, it can be determined that the professional term corresponding to this formula is the standard equation of a circle. In this way, the textual information of the corresponding formula can be obtained, so as to encode the formula.

[0242] In step S1052, at least one question stem data in the question bank that matches the question stem data is determined as the recalled question stem data.

[0243] In some embodiments, the similarity between all question stem data in the question bank and the question stem data to be matched is determined; based on the similarity, all question stem data in the question bank are sorted in descending order, and at least one question stem data that appears first in the descending order is selected as the recall question stem data. Thus, since the question stem data to be matched and the recall question stem data have a high degree of similarity and overlap in the knowledge points they encompass, selecting recall question stem data that is similar to the question stem data is beneficial for subsequently determining the knowledge points of the question stem data.

[0244] In step S1053, based on the recalled question stem data, a second feature set corresponding to the question stem data is determined.

[0245] In some embodiments, the knowledge points included in each recalled question stem data are used as candidate knowledge points; the similarity between each recalled question stem data and the question stem data is determined; each candidate knowledge point and its corresponding knowledge point similarity are combined to obtain multiple combined features that correspond one-to-one with the candidate knowledge points, and the multiple combined features are determined as a second feature set.

[0246] Here, knowledge point similarity is the similarity between the recall question data to which the candidate knowledge point belongs and the question data.

[0247] For example, when the similarity between the recalled question data A and the question data is 0.3, question data A includes knowledge points 1 and 2, the similarity between the recalled question data B and the question data is 0.4, and question data B includes knowledge points 2, 3, and 4, the knowledge point similarity of knowledge point 1 is 0.3 (the similarity between the recalled question data A and the question data), the knowledge point similarity of knowledge point 1 is 0.3 + 0.4 = 0.7 (the sum of the similarity between the recalled question data A and the question data, and the similarity between the recalled question data B and the question data), the knowledge point similarity of knowledge point 3 is 0.4 (the similarity between the recalled question data B and the question data), and the knowledge point similarity of knowledge point 4 is 0.4 (the similarity between the recalled question data B and the question data). In this way, each candidate knowledge point can be weighted based on the knowledge point similarity to improve the accuracy of subsequently determining the question knowledge points based on the candidate knowledge points.

[0248] In step S1054, the features in the first feature set and the features in the second feature set are fused to obtain a fused feature set.

[0249] Here, the features included in the second feature set correspond one-to-one with the candidate knowledge points. The features included in the fused feature set also correspond one-to-one with the candidate knowledge points.

[0250] In some embodiments, all features in the first feature set are combined to obtain a first combined feature; the first combined feature is fused with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with multiple candidate knowledge points, and the multiple fused features are determined as a fused feature set.

[0251] Taking a first feature set including first feature A and first feature B, and a second feature set including second feature C (corresponding to candidate knowledge point 1), second feature D (corresponding to candidate knowledge point 2), and second feature E (corresponding to candidate knowledge point 3) as an example, firstly, first feature A and second feature B in the first feature set are combined to obtain combined feature 1; then, combined feature 1 is fused with second feature C, second feature D, and second feature E respectively to obtain fused feature 1-C, fused feature 1-D, and fused feature 1-E, where fused feature 1-C corresponds to candidate knowledge point 1; fused feature 1-D corresponds to candidate knowledge point 2; and fused feature 1-E corresponds to candidate knowledge point 3. In this way, features from the first feature set and features from the second feature set are fused to improve the comprehensiveness and accuracy of subsequently determining the knowledge points in the question stem based on the fused features.

[0252] In step S1055, classification processing is performed based on multiple fusion features in the fusion feature set to determine at least one question stem knowledge point corresponding to the question stem data.

[0253] In some embodiments, each fusion feature in the fusion feature set is mapped to the probability that the corresponding candidate knowledge point belongs to the question stem knowledge point; the multiple fusion features are sorted in descending order according to the probability; and the candidate knowledge points corresponding to the first (or more) fusion features in the descending order are determined as the question stem knowledge points.

[0254] Here, the selection of the first fusion features in the descending sort can be achieved by choosing the fusion features that are first in the descending sort and have a probability greater than a probability threshold; or by choosing the fusion features that are first in the descending sort and whose quantity is equal to a quantity threshold. The probability threshold can be a default value, a user-defined value, or a value determined based on the probabilities of all fusion features, for example, the average probability of all fusion features. The quantity threshold can also be a default value, a user-defined value, or a value determined based on the number of fusion features included in the fusion feature set, for example, half the number of fusion features included in the fusion feature set.

[0255] The embodiments of the present invention are based on the fact that the knowledge points included in the question stem data and similar recall question stem data have a high degree of overlap. The knowledge points of the question stem data are determined based on the knowledge points of similar recall question stem data. This not only eliminates the need to redetermine the knowledge points of the question stem data based on semantic analysis, thus saving computing resources, but also improves the accuracy of determining the knowledge points of the question stem data.

[0256] In some embodiments, after step S105, the method may further include: storing the exercise data, the knowledge points in the question stem, the knowledge points for solving the problem, and the title of each solution step into a database; wherein the exercise data, the knowledge points in the question stem, the knowledge points for solving the problem, and the title of each solution step stored in the database can be obtained by indexing the knowledge points in the question stem. This provides data support for subsequent application scenarios such as question recommendation and automatic question explanation.

[0257] Below, we will take the application of the AI-based problem-solving method provided in this embodiment of the invention to a scenario where a client with auxiliary teaching functions presents the problem-solving process as an example. See below. Figure 7A and Figure 7B , Figure 7A and Figure 7B This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention. Step S104 may be followed by step S106.

[0258] In step S106, the client presents the question stem data, at least one solution step, and the corresponding title in the human-computer interaction interface.

[0259] As a successor Figure 1A See examples. Figure 7A The execution entity for steps S101 to S104 is the client 410 in the terminal 400. Step S101 may include: the client obtaining the exercise data sent by the server. In this way, the client completes the process of determining the solution steps and the corresponding title by itself or by calling the operating system 551, and presents the question data, at least one solution step, and the corresponding title in the human-computer interaction interface.

[0260] As a successor Figure 1B See examples. Figure 7B The execution entity for steps S101 to S104 is the server 200. Therefore, the alternative step for step S106 can be: the client obtains the question data, at least one solution step, and the corresponding title sent by the server, and presents the question data, at least one solution step, and the corresponding title in the human-computer interaction interface.

[0261] In some embodiments, in response to a presentation operation on the exercise data, the client presents the question stem data, at least one solution step, and the corresponding title in the human-computer interaction interface in the order of solving the problem.

[0262] Here, the presented operation can be any form of operation pre-set by the operating system and does not conflict with the registered operation; or it can be any form of operation defined by the user and does not conflict with the registered operation. The presented operation includes at least one of the following: click operation (e.g., single-finger click operation, multi-finger click operation, or multiple click operation); swipe operation according to a specific trajectory or direction; voice operation; motion operation (e.g., up and down shaking operation or curved motion operation).

[0263] In other embodiments, the client obtains the user's answer data; when the answer data and the corresponding solution data are consistent, a sign indicating that the answer is correct is displayed on the human-computer interaction interface; when the answer data and the corresponding solution data are inconsistent, a sign indicating that the answer is incorrect is displayed on the human-computer interaction interface, and at least one solution step of the corresponding solution data and the corresponding title are displayed.

[0264] Here, presenting at least one solution step and the corresponding title of the corresponding solution data may include at least one of the following: automatically presenting at least one solution step and the corresponding title of the corresponding solution data; presenting at least one solution step and the corresponding title of the corresponding solution data when the number of times the client presents the identifier representing the incorrect answer on the human-computer interaction interface reaches a threshold; or presenting at least one solution step and the corresponding title of the corresponding solution data in response to the presentation operation of the exercise data.

[0265] In some embodiments, the client may also present at least one question stem knowledge point and multiple problem-solving knowledge points in the human-computer interaction interface, while simultaneously presenting the question stem data. Thus, step S106 may be: the client presents at least one question stem knowledge point, multiple problem-solving knowledge points, question stem data, at least one problem-solving step, and a corresponding title in the human-computer interaction interface.

[0266] The embodiments of the present invention present the knowledge points of the question stem, the knowledge points of the solution, the solution steps and the corresponding title to the user in a convenient and intuitive way. Compared with the related technologies that directly present the solution data to the user, it is beneficial to improve the efficiency and accuracy of the user's understanding of the solution process, thereby improving the quality of learning and saving human resources.

[0267] The following description uses the example of applying the AI-based exercise processing method provided in this embodiment of the invention to a teaching client to review the answer data of users (e.g., students). Step S104 may be followed by steps S107 and S108.

[0268] In step S107, the client obtains the answer data.

[0269] Here, the answer data can be the answers to exercises submitted by users (e.g., students).

[0270] In some embodiments, the client responds to an audit operation by retrieving the exercise data submitted by the user.

[0271] Here, the review operation can be any form of operation pre-set by the operating system that does not conflict with the registered operation; or it can be any form of operation defined by the user that does not conflict with the registered operation. The review operation includes at least one of the following: click operation (e.g., single-finger click operation, multi-finger click operation, or multiple click operation); swipe operation according to a specific trajectory or direction; voice operation; motion control operation (e.g., up and down shaking operation or curved motion operation).

[0272] In step S108, the client determines the review result of the answer data based on the exercise data of the answer data and presents the review result on the human-computer interaction interface.

[0273] Here, the review results include at least one of the following: review indicators; discrepancies in the solution steps; the relevant knowledge points corresponding to the discrepancies in the solution steps; and the titles corresponding to the discrepancies in the solution steps. The review indicators include indicators indicating correct answers and indicators indicating incorrect answers; the review indicators can be graphic or text.

[0274] In some embodiments, the client may invoke its own or the corresponding service of the operating system 551 (e.g., the review service) to complete the review process of the answer data. The client may also invoke the corresponding service of the server (e.g., the review service) to complete the review process of the answer data through the server.

[0275] Thus, the alternative step S108 can be: the client sends the answer data to the server; the server determines the review result of the answer data based on the exercise data corresponding to the answer data, and sends the review result to the client; the client presents the review result in the human-computer interaction interface.

[0276] The following example illustrates the specific implementation of answer verification, using the client-side process of verifying answer data as an example.

[0277] In some embodiments, the client determines the solution data that is consistent with the question stem data of the answer data; determines the solution steps that differ between the answer data and the corresponding solution data, and determines the solution knowledge points and titles corresponding to the solution steps that differ; and presents the solution steps that differ, the corresponding solution knowledge points and titles in the human-computer interaction interface.

[0278] Here, the client can present different problem-solving steps, corresponding problem-solving knowledge points, and titles in the human-computer interaction interface, while also displaying an audit mark (a mark indicating a correct answer or a mark indicating an incorrect answer).

[0279] It should be noted that the process of server-side verification of answer data is similar to that of client-side verification of answer data, and will not be described in detail here.

[0280] This invention can accurately analyze the location of errors made by users during problem-solving, precisely identifying which knowledge point the problem occurred in the student's solution. This allows for the construction of a framework to assess each user's mastery of each knowledge point within the subject, thereby enabling the precise recommendation of relevant questions to reinforce and improve the student's understanding.

[0281] The following example, using science exam questions as an example, illustrates the artificial intelligence-based question processing method provided by the embodiments of the present invention.

[0282] This invention, based on a subject-specific knowledge graph, first analyzes the known and required content of the problem stem (i.e., the aforementioned problem stem data) using techniques such as semantic similarity matching, text classification, and search. Then, it analyzes the knowledge points applied in the problem-solving process (i.e., the aforementioned problem-solving data) using the same techniques. Next, based on the paths of the knowledge points moving across the knowledge graph during the problem-solving process, it determines whether the problem-solving process should be divided into steps. Finally, based on the path transfers of the knowledge points used in the problem-solving process, it determines the titles of each step. This invention systematically analyzes the problem-solving process and usage methods, assisting students in adaptive learning.

[0283] See Figure 8A and Figure 8B , Figure 8A This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention. Figure 8B This is a schematic diagram of an application scenario provided by an embodiment of the present invention, which will be combined with Figure 8A and Figure 8B Please provide an explanation.

[0284] In step S801, the exercise data is obtained.

[0285] See here. Figure 8B The exercise data includes question stem data 801 and solution data 802.

[0286] In step S802, the question stem data is analyzed.

[0287] In some embodiments, see Figure 8B Analyze the data in the question stem to determine the known content 803 and the solution content 804 in the data.

[0288] In some embodiments, the question stem data is analyzed to identify the knowledge points in the question stem data (i.e., the question stem knowledge points mentioned above), and the knowledge points in the question stem data can be used to associate exercises with the corresponding knowledge points for students to practice.

[0289] In step S803, the problem-solving data is analyzed.

[0290] In some embodiments, see Figure 8B Analyze the problem-solving data to determine the knowledge points 805 (i.e., the problem-solving knowledge points mentioned above) used in each step of the problem-solving data.

[0291] In step S804, the problem-solving data is divided into multiple steps (i.e., the problem-solving steps described above).

[0292] In some embodiments, see Figure 8B The problem-solving data is divided into multiple steps (806).

[0293] In step S805, a title is generated for each step.

[0294] In some embodiments, see Figure 8B Generate a corresponding title 807 for each step 806.

[0295] In step S806, the structured analysis results of the exercise data are stored in the database.

[0296] In some embodiments, the structured analysis of the exercise data includes at least one of the following: question stem data; solution data; knowledge points corresponding to the question stem data; knowledge points corresponding to the solution data; steps included in the solution data; and a title corresponding to each step.

[0297] See Figure 9 , Figure 9 This is a flowchart illustrating the problem-solving method based on artificial intelligence provided in an embodiment of the present invention.

[0298] Figure 9 In this system, when a user inputs exercise data, the system retrieves the corresponding question stem data and solution data. These two data are then processed separately, and the results are finally merged and stored in the database. The following sections will explain the analysis and processing procedures for the question stem data and solution data.

[0299] (I) Analysis and processing of the question data

[0300] First, the rule-matching analyzer analyzes the known content and the solution content of the question stem data to match the concepts (i.e., the subject concepts mentioned above) and formulas, and abstracts the concepts and formulas, such as formula (x-1). 2 +(y-2) 2 =9 can be abstracted into the standard equation of a circle, and the abstracted concepts and formulas are used as the feature set F1.

[0301] Secondly, using a search engine that stores subject-specific question banks (i.e., the aforementioned question bank), the top 10 question stem data (i.e., the aforementioned recalled question stem data) with the highest similarity to the input question stem data are selected from the search engine. The semantic similarity between the selected question stem data and the input question stem data is calculated, and the knowledge points of each selected question stem data are taken as candidate knowledge points. Based on the candidate knowledge points and the similarity between the question stem data to which each candidate knowledge point belongs and the input question stem data, the feature set F2 is obtained.

[0302] The semantic similarity between the selected question stem data and the input question stem data can be calculated by building a model based on Bidirectional Encoder Representations from Transformers (BERT), and the similarity can be calculated through the established model.

[0303] Finally, the feature sets F1 and F2 obtained above are used together as features input into the classification model to determine the knowledge points of the question data.

[0304] In this embodiment of the invention, similar question stem data to the input question stem data is searched through a search engine, and the knowledge points of the original question stem data are determined based on the knowledge points of the similar question stem data.

[0305] (II) Analysis and Processing of Problem-Solving Data

[0306] The analysis and processing of problem-solving data involves analyzing the knowledge points of each sentence, dividing the problem-solving data into steps based on the knowledge points, and generating a title for each step.

[0307] First, a similar analysis method to that used for the question stem data is employed to obtain the knowledge points for each sentence in the solution data. The difference lies in the fact that the question stem data analysis selects sentences with high similarity to the input question stem data, while the solution data analysis selects sentences with high similarity to the input solution data. In this way, the knowledge points for each sentence in the solution data can be obtained, and all the obtained knowledge points are combined into a knowledge point sequence.

[0308] Secondly, the knowledge point sequence is input into the knowledge graph of the corresponding subject. A tree traversal method is used to determine the relationship between the knowledge points of adjacent statements in the knowledge graph. When a parent-child relationship exists between the knowledge points of two adjacent statements in the knowledge graph, these two adjacent statements are merged into the same step; when no parent-child relationship exists between the knowledge points of two adjacent statements in the knowledge graph, these two adjacent statements are separated into two different steps.

[0309] Finally, the sequence of knowledge points for each step and the formulas included in that step are combined as features and input into a classification model used to generate a title for that step, thus selecting a title for that step.

[0310] After completing the above processing, the exercise data is stored in a structured database to provide data support for subsequent application scenarios such as question recommendation and automatic question explanation.

[0311] In this embodiment of the invention, not only can the known content and the solution content of a problem be accurately determined, but the knowledge points involved in each step of the answer can also be analyzed. When a student makes a mistake, this embodiment of the invention can accurately analyze the location of the student's error in the problem-solving process, so as to construct each student's mastery of each knowledge point in the subject, thereby accurately recommending relevant questions to consolidate and improve the student's level.

[0312] The following will continue to combine Figure 2A The embodiment of the AI-based exercise processing device 555 provided in this invention is an exemplary structure of a software module. In some embodiments, such as... Figure 2A As shown, the software modules stored in the AI-based problem-solving device 555 in the memory 550 may include:

[0313] The acquisition module 5551 is used to acquire exercise data, wherein the exercise data includes question stem data and solution data;

[0314] The knowledge point determination module 5552 is used to determine multiple problem-solving knowledge points corresponding to the problem-solving data;

[0315] The step determination module 5553 is used to determine the relationship between the multiple problem-solving knowledge points in the knowledge graph of the subject corresponding to the exercise data, and to determine at least one problem-solving step included in the problem-solving data based on the relationship.

[0316] The title classification module 5554 is used to perform title classification processing on each of the problem-solving steps to determine the title of each of the problem-solving steps.

[0317] The following is combined with Figure 2B The embodiment of the AI-based exercise processing device 555 provided in this invention is an exemplary structure of a software module. In some embodiments, such as... Figure 2B As shown, the software modules stored in the AI-based problem-solving device 555 in the memory 550 may include:

[0318] The acquisition module 5551 is used to acquire exercise data, wherein the exercise data includes question stem data and solution data;

[0319] The exercise presentation module 5555 is used to present the question stem data in the exercise data in the human-computer interaction interface;

[0320] The knowledge point determination module 5552 is used to determine multiple problem-solving knowledge points corresponding to the problem-solving data;

[0321] The step determination module 5553 is used to determine the relationship between the multiple problem-solving knowledge points in the knowledge graph of the subject corresponding to the exercise data, and to determine at least one problem-solving step included in the problem-solving data based on the relationship.

[0322] The title classification module 5554 is used to perform title classification processing on each of the problem-solving steps to determine the title of each of the problem-solving steps;

[0323] The exercise presentation module 5555 is also used to present at least one of the problem-solving steps and the corresponding title in the human-computer interaction interface.

[0324] In the above scheme, the knowledge point determination module 5552 is further configured to segment the problem-solving data into multiple candidate problem-solving steps; perform the following processing on each candidate problem-solving step to determine the problem-solving knowledge points of each candidate problem-solving step: encode the candidate problem-solving step to determine a first feature set corresponding to the candidate problem-solving step; determine at least one problem-solving step in the exercise bank that matches the candidate problem-solving step as a recall problem-solving step; determine a second feature set corresponding to the candidate problem-solving step based on the recall problem-solving step; fuse the features in the first feature set and the features in the second feature set to obtain a fused feature set; and perform classification processing based on multiple fused features in the fused feature set to determine the problem-solving knowledge points of the candidate problem-solving step.

[0325] In the above scheme, the knowledge point determination module 5552 is further used to determine the subject concepts and formulas included in the candidate problem-solving steps; determine the subject terms corresponding to the formulas; encode the subject concepts and the subject terms corresponding to the formulas respectively to obtain multiple first features, and determine the multiple first features as the first feature set.

[0326] In the above scheme, the knowledge point determination module 5552 is further configured to: use the knowledge points included in each recall problem-solving step as candidate knowledge points; determine the similarity between each recall problem-solving step and the candidate problem-solving step; combine each candidate knowledge point and its corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge points, and determine the multiple combined features as the second feature set; wherein, the knowledge point similarity is the similarity between the recall problem-solving step to which the candidate knowledge point belongs and the candidate problem-solving step.

[0327] In the above scheme, the knowledge point determination module 5552 is further configured to combine all features in the first feature set to obtain a first combined feature; fuse the first combined feature with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points, and determine the multiple fused features as the fused feature set.

[0328] In the above scheme, the knowledge point determination module 5552 is further configured to map each fusion feature in the fusion feature set to the probability that the corresponding candidate knowledge point belongs to the problem-solving knowledge point; sort the multiple fusion features in descending order according to the probability; and determine the candidate knowledge points corresponding to the fusion features that are sorted first in descending order as the problem-solving knowledge points of the candidate problem-solving steps.

[0329] In the above scheme, the step determination module 5553 is further used to determine the node corresponding to the problem-solving knowledge point of each candidate problem-solving step in the knowledge graph; when there is a parent-child relationship between the nodes corresponding to the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data, the adjacent candidate problem-solving steps corresponding to the nodes with parent-child relationship are merged into a new problem-solving step as the target problem-solving step; when there is no parent-child relationship between the nodes corresponding to the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data, the adjacent candidate problem-solving steps corresponding to the nodes without parent-child relationship are each independently used as the target problem-solving step; the target problem-solving step is determined as the problem-solving steps included in the problem-solving data.

[0330] In the above scheme, the title classification module 5554 is further configured to perform the following processing on each problem-solving step: determine the problem-solving knowledge points and formulas included in the problem-solving step; determine the subject terms corresponding to the formulas; combine the problem-solving knowledge points included in the problem-solving step and the subject terms corresponding to the formulas to obtain combination features; map the combination features to probabilities of belonging to different titles, and determine the title corresponding to the highest probability as the title of the problem-solving step.

[0331] In the above scheme, the knowledge point determination module 5552 is further configured to encode the question stem data to determine a first feature set corresponding to the question stem data; determine at least one question stem data in the question bank that matches the question stem data as recall question stem data; determine a second feature set corresponding to the question stem data based on the recall question stem data; fuse the features in the first feature set and the features in the second feature set to obtain a fused feature set; and perform classification processing based on multiple fused features in the fused feature set to determine the question stem knowledge points.

[0332] In the above scheme, the knowledge point determination module 5552 is further used to determine the known content and the solution content included in the question stem data, and to determine the subject concepts and formulas included in the known content and the solution content respectively; to determine the subject terms corresponding to the formulas; to encode the subject concepts and the subject terms respectively to obtain multiple first features, and to determine the multiple first features as the first feature set.

[0333] In the above scheme, the knowledge point determination module 5552 is further configured to: take the knowledge points included in each recalled question stem data as candidate knowledge points; determine the similarity between each recalled question stem data and the question stem data; combine each candidate knowledge point and the corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge points, and determine the multiple combined features as the second feature set; wherein, the knowledge point similarity is the similarity between the recalled question stem data to which the candidate knowledge point belongs and the question stem data.

[0334] In the above scheme, the knowledge point determination module 5552 is further configured to combine all features in the first feature set to obtain a first combined feature; the first combined feature is fused with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points, and the multiple fused features are determined as the fused feature set.

[0335] In the above scheme, the knowledge point determination module 5552 is further configured to map each fusion feature in the fusion feature set to the probability that the corresponding candidate knowledge point belongs to the question stem knowledge point; sort the multiple fusion features in descending order according to the probability; and determine the candidate knowledge points corresponding to the fusion features that are sorted first in descending order as the question stem knowledge points.

[0336] In the above scheme, the AI-based exercise processing device 555 includes: a judgment module, used to acquire answer data; determine solution data that is consistent with the question stem data corresponding to the answer data; determine that there are differences between the answer data and the corresponding solution data, and determine the solution knowledge points corresponding to the differences; and present the differences and the corresponding solution knowledge points in the human-computer interaction interface.

[0337] This invention provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the artificial intelligence-based problem-solving method described above in this invention.

[0338] This invention provides a computer-readable storage medium storing computer-executable instructions. When these computer-executable instructions are executed by a processor, they cause the processor to execute the AI-based problem-solving method provided in this invention. For example... Figure 3 , Figure 4 , Figure 5 , Figure 6 , Figure 7A , Figure 7B , Figure 8A or Figure 9 The illustrated problem-solving method is based on artificial intelligence, and the computer includes various computing devices such as smart terminals and servers.

[0339] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EP ROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a variety of devices including one or any combination of the above-mentioned memories.

[0340] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0341] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hypertext Markup Language document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0342] As an example, computer-executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0343] In summary, the embodiments of the present invention have the following beneficial effects:

[0344] (1) Based on the high overlap of knowledge points included in candidate problem-solving steps and similar recall problem-solving steps, the knowledge points of candidate problem-solving steps are determined according to the knowledge points of similar recall problem-solving steps. This not only eliminates the need to redetermine the knowledge points of candidate problem-solving steps based on semantic analysis, saving computational resources, but also improves the accuracy of knowledge point determination.

[0345] (2) It can merge steps that have a coherent solution and separate steps that do not have a coherent solution. This not only makes it easier for users to understand the solution process, but also reduces the number of titles generated later and saves computing resources.

[0346] (3) Based on the problem-solving knowledge points and formulas included in the problem-solving steps, determine the title of the problem-solving steps and accurately determine the title of the corresponding problem-solving steps so that users can understand the problem-solving process according to the title, thereby improving the efficiency and accuracy of users' understanding of the problem-solving process, thereby improving the quality of learning and saving human resources.

[0347] (4) Presenting the knowledge points of the question stem, the knowledge points of the solution, the solution steps and the corresponding title to users in a convenient and intuitive way is more effective than presenting the solution data directly to users in related technologies. This helps improve the efficiency and accuracy of users’ understanding of the solution process, thereby improving the quality of learning and saving human resources.

[0348] (5) It can accurately analyze the location of errors made by users in the problem-solving process, so as to construct the mastery of each user's knowledge points in the subject, thereby accurately recommending relevant questions to consolidate and improve students' level.

[0349] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of the present invention are included within the scope of protection of the present invention.

Claims

1. A problem-solving method based on artificial intelligence, characterized in that, The method includes: Acquire exercise data, wherein the exercise data includes question stem data and solution data; The question data is presented in the human-computer interaction interface; The problem-solving data is segmented into multiple candidate problem-solving steps. The candidate problem-solving steps are encoded to determine a first feature set corresponding to each candidate problem-solving step. At least one problem-solving step in the problem bank that matches the candidate problem-solving steps is identified as a recalled problem-solving step. Based on the recalled problem-solving steps, a second feature set corresponding to the candidate problem-solving steps is determined. Features in the first feature set and features in the second feature set are fused to obtain a fused feature set. Multiple fused features in the fused feature set are classified to determine the problem-solving knowledge points of the candidate problem-solving steps, and all the obtained problem-solving knowledge points are combined into a knowledge point sequence. Determine the node in the knowledge graph corresponding to the problem-solving knowledge point of each candidate problem-solving step, and input the knowledge point sequence into the knowledge graph; When a tree traversal method is used to determine that there is a parent-child relationship between the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data and the corresponding nodes in the knowledge graph, the adjacent candidate problem-solving steps corresponding to the nodes with parent-child relationships are merged into a new problem-solving step as the target problem-solving step. When there is no parent-child relationship between the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data and the corresponding nodes in the knowledge graph, the adjacent candidate problem-solving steps corresponding to the nodes without parent-child relationship are independently used as target problem-solving steps. The target problem-solving steps are defined as the problem-solving steps included in the problem-solving data; The title classification model is invoked to perform the following processing for each problem-solving step: determine the problem-solving knowledge points, formulas, and corresponding subject terms included in the problem-solving step; combine the problem-solving knowledge points and corresponding subject terms to obtain combined features; map the combined features to probabilities of belonging to different titles, and determine the title corresponding to the highest probability as the title of the problem-solving step; At least one of the problem-solving steps and a corresponding title are presented in the human-computer interaction interface; The system acquires answer data, determines the review result of the answer data based on the corresponding exercise data, and presents the review result on the human-computer interaction interface. The review result includes the problem-solving steps that differ, the problem-solving knowledge points corresponding to the problem-solving steps, and the title corresponding to the problem-solving steps.

2. The method according to claim 1, characterized in that, The encoding process for the candidate problem-solving steps to determine the first feature set corresponding to the candidate problem-solving steps includes: Identify the subject concepts and formulas included in the candidate problem-solving steps; Determine the subject-specific terminology corresponding to the formula; The subject concepts and corresponding subject terms in the formulas are encoded separately to obtain multiple first features, and The plurality of first features are determined as the first feature set.

3. The method according to claim 1, characterized in that, The step of determining the second feature set corresponding to the candidate problem-solving steps based on the recall problem-solving steps includes: Each knowledge point included in the recall and problem-solving steps is used as a candidate knowledge point; Determine the similarity between each recall problem-solving step and the candidate problem-solving step; Each candidate knowledge point is combined with its corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge point, and these multiple combined features are determined as the second feature set. The knowledge point similarity is the similarity between the recall problem-solving step to which the candidate knowledge point belongs and the candidate problem-solving step.

4. The method according to claim 3, characterized in that, The step of fusing the features in the first feature set and the features in the second feature set to obtain a fused feature set includes: Combine all features in the first feature set to obtain a first combined feature; The first combined feature is fused with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points, and the multiple fused features are determined as the fused feature set. The classification process based on multiple fusion features in the fusion feature set to determine the problem-solving knowledge points of the candidate problem-solving steps includes: Each fusion feature in the fusion feature set is mapped to the probability that the corresponding candidate knowledge point belongs to the problem-solving knowledge point; The multiple fusion features are sorted in descending order according to the probabilities; The candidate knowledge points corresponding to the fusion features that are sorted in descending order are determined as the problem-solving knowledge points of the candidate problem-solving steps.

5. The method according to claim 1, characterized in that, The method further includes: The question stem data is encoded to determine the first feature set corresponding to the question stem data; Identify at least one question stem data in the question bank that matches the question stem data, and use it as the recalled question stem data; Based on the recalled question stem data, a second feature set corresponding to the question stem data is determined; The features in the first feature set and the features in the second feature set are fused together to obtain a fused feature set; Classification processing is performed based on multiple fusion features in the fusion feature set to determine at least one question stem knowledge point corresponding to the question stem data.

6. A problem-solving method based on artificial intelligence, characterized in that, The method includes: Acquire exercise data, wherein the exercise data includes solution data; The problem-solving data is segmented into multiple candidate problem-solving steps. The candidate problem-solving steps are encoded to determine a first feature set corresponding to each candidate problem-solving step. At least one problem-solving step in the problem bank that matches the candidate problem-solving steps is identified as a recalled problem-solving step. Based on the recalled problem-solving steps, a second feature set corresponding to the candidate problem-solving steps is determined. Features in the first feature set and features in the second feature set are fused to obtain a fused feature set. Multiple fused features in the fused feature set are classified to determine the problem-solving knowledge points of the candidate problem-solving steps, and all the obtained problem-solving knowledge points are combined into a knowledge point sequence. Determine the node in the knowledge graph corresponding to the problem-solving knowledge point of each candidate problem-solving step, and input the knowledge point sequence into the knowledge graph; When a tree traversal method is used to determine that there is a parent-child relationship between the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data and the corresponding nodes in the knowledge graph, the adjacent candidate problem-solving steps corresponding to the nodes with parent-child relationships are merged into a new problem-solving step as the target problem-solving step. When there is no parent-child relationship between the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data and the corresponding nodes in the knowledge graph, the adjacent candidate problem-solving steps corresponding to the nodes without parent-child relationship are independently used as target problem-solving steps. The target problem-solving steps are defined as the problem-solving steps included in the problem-solving data; The title classification model is invoked to perform the following processing for each problem-solving step: determine the problem-solving knowledge points, formulas, and corresponding subject terms included in the problem-solving step; combine the problem-solving knowledge points and corresponding subject terms to obtain combined features; map the combined features to probabilities of belonging to different titles, and determine the title corresponding to the highest probability as the title of the problem-solving step; The system acquires answer data, determines the review result of the answer data based on the corresponding exercise data, and presents the review result on the human-computer interaction interface. The review result includes the problem-solving steps that differ, the problem-solving knowledge points corresponding to the problem-solving steps, and the title corresponding to the problem-solving steps.

7. A problem-solving device based on artificial intelligence, characterized in that, The device includes: The acquisition module is used to acquire exercise data, wherein the exercise data includes question stem data and solution data; The exercise presentation module is used to present the question data in the human-computer interaction interface; The knowledge point determination module is used to segment the problem-solving data into multiple candidate problem-solving steps, wherein the candidate problem-solving steps are encoded to determine a first feature set corresponding to the candidate problem-solving steps; at least one problem-solving step in the exercise bank that matches the candidate problem-solving steps is determined as a recalled problem-solving step; based on the recalled problem-solving steps, a second feature set corresponding to the candidate problem-solving steps is determined; the features in the first feature set and the features in the second feature set are fused to obtain a fused feature set; multiple fused features in the fused feature set are classified to determine the problem-solving knowledge points of the candidate problem-solving steps, and all the obtained problem-solving knowledge points are combined into a knowledge point sequence. The step determination module is used to determine the node corresponding to the solution knowledge point of each candidate solution step in the knowledge graph, and input the knowledge point sequence into the knowledge graph; when the tree traversal method is used to determine that there is a parent-child relationship between the solution knowledge points of adjacent candidate solution steps in the solution data and the corresponding nodes in the knowledge graph, the adjacent candidate solution steps corresponding to the nodes with parent-child relationship are merged into a new solution step as the target solution step; When there is no parent-child relationship between the problem-solving knowledge points of adjacent candidate problem-solving steps in the problem-solving data and the corresponding nodes in the knowledge graph, the adjacent candidate problem-solving steps corresponding to the nodes without parent-child relationship are independently taken as target problem-solving steps; the target problem-solving steps are determined as the problem-solving steps included in the problem-solving data; The title classification module is used to call the title classification model to perform the following processing for each problem-solving step: determine the problem-solving knowledge points, formulas, and corresponding subject terms included in the problem-solving step; combine the problem-solving knowledge points and corresponding subject terms to obtain combination features; map the combination features to probabilities of belonging to different titles, and determine the title corresponding to the highest probability as the title of the problem-solving step; The exercise presentation module is further configured to present at least one of the problem-solving steps and a corresponding title in the human-computer interaction interface; acquire answer data, and determine the review result of the answer data based on the exercise data corresponding to the answer data, and present the review result in the human-computer interaction interface. The review result includes the problem-solving steps that differ, the problem-solving knowledge points corresponding to the problem-solving steps, and the title corresponding to the problem-solving steps.

8. The apparatus according to claim 7, characterized in that, The knowledge point determination module is further configured to determine the subject concepts and formulas included in the candidate problem-solving steps; determine the subject terms corresponding to the formulas; encode the subject concepts and the subject terms corresponding to the formulas respectively to obtain multiple first features, and determine the multiple first features as the first feature set.

9. The apparatus according to claim 7, characterized in that, The knowledge point determination module is further configured to: use the knowledge points included in each recall problem-solving step as candidate knowledge points; determine the similarity between each recall problem-solving step and the candidate problem-solving step; combine each candidate knowledge point and its corresponding knowledge point similarity to obtain multiple combined features that correspond one-to-one with the candidate knowledge points, and determine the multiple combined features as the second feature set; wherein, the knowledge point similarity is the similarity between the recall problem-solving step to which the candidate knowledge point belongs and the candidate problem-solving step.

10. The apparatus according to claim 7, characterized in that, The knowledge point determination module is further configured to combine all features in the first feature set to obtain a first combined feature; fuse the first combined feature with each combined feature in the second feature set to obtain a fused feature that corresponds one-to-one with the multiple candidate knowledge points, and determine the multiple fused features as the fused feature set.

11. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; A processor, when executing executable instructions stored in the memory, implements the problem-solving method based on artificial intelligence as described in any one of claims 1 to 6.

12. A computer-readable storage medium storing executable instructions, characterized in that, When the executable instructions are executed by the processor, they implement the problem-solving method based on artificial intelligence as described in any one of claims 1 to 6.

13. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the processor, they implement the problem-solving method based on artificial intelligence as described in any one of claims 1 to 6.

Images