Learning assistance method, device and electronic device based on question bank knowledge graph

Abstract

The present application discloses a learning assistance method, device, storage medium and electronic device based on a question bank knowledge graph, which relates to the technical field of data processing, and includes: first, extracting knowledge points based on teaching material texts to obtain a basic knowledge point cluster; evaluating the importance of the knowledge points in the basic knowledge point cluster, and screening according to the evaluation results to obtain an important knowledge point cluster; determining the knowledge point distribution information of the basic knowledge point cluster and the important knowledge point cluster in the teaching material, generating a knowledge tree containing the knowledge point distribution information, and determining the association strength between each knowledge point according to the number of branches of the knowledge tree; constructing a question bank knowledge graph according to the association strength between each knowledge point; matching the question bank knowledge graph with the user's learning conversation record, and monitoring the user's question brushing situation according to the matching result. By extracting important knowledge point clusters from the teaching material and constructing a question bank knowledge graph, personalized learning assistance is realized.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and particularly to a learning assistance method, device, storage medium and electronic device based on a question bank knowledge graph. Background Art

[0002] During the process of students doing exercises, they usually encounter a large number of knowledge points. With the continuous development of educational technology, how to efficiently extract knowledge points from textbooks, construct a knowledge system, and assist teaching through question banks and visualization tools has become an urgent problem to be solved in the education field.

[0003] Currently, the difficulty level of knowledge points is usually determined by statistically analyzing the frequency of question appearance and the error rate of students. For example, questions that appear frequently and have a high error rate are often considered to correspond to more difficult knowledge points, while questions that appear infrequently and have a low error rate are considered to correspond to easier knowledge points.

[0004] However, important knowledge points and basic knowledge points are not effectively distinguished in the existing question bank knowledge graph, and it is difficult to accurately reflect the complex relationships between knowledge points, resulting in students being unable to accurately brush questions for key knowledge points. Summary of the Invention

[0005] In view of this, the present application provides a learning assistance method, device, storage medium and electronic device based on a question bank knowledge graph, mainly aiming to improve the technical problem that the existing question bank knowledge graph in the current technology is difficult to accurately reflect the complex relationships between knowledge points, resulting in students being unable to accurately brush questions for key knowledge points.

[0006] In a first aspect, the present application provides a learning assistance method based on a question bank knowledge graph, including:

[0007] Extracting knowledge points based on textbook texts to obtain a basic knowledge point cluster;

[0008] Evaluating the importance of the knowledge points in the basic knowledge point cluster, and screening according to the evaluation results to obtain an important knowledge point cluster;

[0009] Determining the knowledge point distribution information of the basic knowledge point cluster and the important knowledge point cluster in the textbook, generating a knowledge tree including the knowledge point distribution information, and determining the association strength between each knowledge point according to the number of branches of the knowledge tree;

[0010] Constructing a question bank knowledge graph according to the association strength between each knowledge point, where the question bank knowledge graph is used to display the questions corresponding to each knowledge point and the association relationships between different knowledge points;

[0011] Match the knowledge graph of the question bank with the user's learning conversation record, and monitor the user's question-solving situation according to the matching result.

[0012] In a second aspect, the present application provides a learning assistance device based on a knowledge graph of a question bank, including:

[0013] An extraction module, configured to extract knowledge points based on textbook texts to obtain a basic knowledge point cluster;

[0014] An evaluation module, configured to evaluate the importance of the knowledge points in the basic knowledge point cluster, and screen to obtain an important knowledge point cluster according to the evaluation result;

[0015] A generation module, configured to determine the knowledge point distribution information of the basic knowledge point cluster and the important knowledge point cluster in the textbook, generate a knowledge tree including the knowledge point distribution information, and determine the association strength between each knowledge point according to the number of branches of the knowledge tree;

[0016] A construction module, configured to construct a knowledge graph of a question bank according to the association strength between each knowledge point, where the knowledge graph of the question bank is used to display the questions corresponding to each knowledge point and the association relationship between different knowledge points;

[0017] A monitoring module, configured to match the knowledge graph of the question bank with the user's learning conversation record, and monitor the user's question-solving situation according to the matching result.

[0018] In a third aspect, the present application provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the learning assistance method based on the knowledge graph of the question bank described in the first aspect is implemented.

[0019] In a fourth aspect, the present application provides an electronic device, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, and when the processor executes the computer program, the learning assistance method based on the knowledge graph of the question bank described in the first aspect is implemented.

[0020] With the above technical solutions, a learning assistance method, device, storage medium, and electronic device based on a question bank knowledge graph provided by the present application, compared with the current existing technologies, the present application first extracts knowledge points based on textbook texts to obtain a basic knowledge point cluster; evaluates the importance of the knowledge points in the basic knowledge point cluster, and filters according to the evaluation results to obtain an important knowledge point cluster; determines the knowledge point distribution information of the basic knowledge point cluster and the important knowledge point cluster in the textbook, generates a knowledge tree including the knowledge point distribution information, and determines the association strength between each knowledge point according to the number of branches of the knowledge tree; constructs a question bank knowledge graph according to the association strength between each knowledge point, and the question bank knowledge graph is used to display the questions corresponding to each knowledge point and the association relationship between different knowledge points; matches the question bank knowledge graph with the user's learning conversation record, and monitors the user's question-solving situation according to the matching result. Through the analysis and processing of text information, a basic knowledge point cluster can be obtained from the textbook and further an important knowledge point cluster can be extracted, thereby effectively distinguishing important knowledge points and basic knowledge points, improving the accuracy of important knowledge point screening. By further constructing a question bank knowledge graph, associating knowledge points with questions, and using the graph to monitor and analyze the student's question-solving situation, students can accurately brush questions for key knowledge points, thereby realizing personalized learning assistance. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0022] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0023] Figure 1 Shows a schematic flowchart of a learning assistance method based on a question bank knowledge graph provided by an embodiment of the present application;

[0024] Figure 2 Shows a schematic flowchart of a learning assistance method based on a question bank knowledge graph provided by an embodiment of the present application;

[0025] Figure 3 Shows a schematic flowchart of a learning assistance method based on a question bank knowledge graph provided by an embodiment of the present application;

[0026] Figure 4 Shows a schematic flowchart of an example provided by an embodiment of the present application;

[0027] Figure 5 Shows a schematic flowchart of a learning assistance method based on a question bank knowledge graph provided by an embodiment of the present application;

[0028] Figure 6 Shows a schematic flowchart of an example provided by an embodiment of the present application;

[0029] Figure 7 Shows a schematic flowchart of an example provided by an embodiment of the present application;

[0030] Figure 8 Shows a schematic flowchart of a learning assistance method based on a question bank knowledge graph provided by an embodiment of the present application;

[0031] Figure 9 Shows a schematic flowchart of a learning assistance method based on a question bank knowledge graph provided by an embodiment of the present application;

[0032] Figure 10 Shows a schematic flowchart of a learning assistance method based on a question bank knowledge graph provided by an embodiment of the present application;

[0033] Figure 11 Shows a schematic flowchart of an example provided by an embodiment of the present application;

[0034] Figure 12 Shows a schematic structural diagram of a learning assistance device based on a question bank knowledge graph provided by an embodiment of the present application;

[0035] Figure 13 Shows a schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0036] The embodiments of the present application will be described in more detail with reference to the accompanying drawings. It should be noted that, without conflict, the embodiments and features in the embodiments of the present application may be combined with each other.

[0037] In order to improve the technical problem that in the current existing technology, the question bank knowledge graph is difficult to accurately reflect the complex relationships between knowledge points, and thus students cannot accurately brush questions for key knowledge points. This embodiment provides a learning assistance method based on a question bank knowledge graph, as Figure 1 shown, the method includes:

[0038] Step 101: Extract knowledge points based on textbook texts to obtain a basic knowledge point cluster.

[0039] Exemplarily, the basic knowledge point cluster is a set of knowledge units composed of all basic concepts, terms, facts, rules, etc. identified after text analysis and processing of textbooks or learning materials. First, the textbook content is cleaned and preprocessed, including removing irrelevant characters (such as punctuation marks), converting to lowercase, stemming or lemmatization, etc., to ensure the effectiveness of subsequent analysis. The text is segmented into words or phrases, and each vocabulary is annotated using text analysis techniques, such as text mining, natural language processing, semantic analysis, etc. Based on the extracted keywords and context information, named entity recognition or other machine learning models are applied to identify specific knowledge points. For each identified knowledge point, models such as Word2Vec, GloVe, or BERT can be used to convert it into a vector form, and these vectors can capture the semantic relationships between words, thus forming identification vectors.

[0040] In some examples, the generated identification vectors are grouped by clustering algorithms (such as K-means, hierarchical clustering, etc.), and the cluster composed of similar knowledge points is the basic knowledge point cluster, thereby clarifying the theme structure and association pattern in the textbook content.

[0041] Step 102: Evaluate the importance of the knowledge points in the basic knowledge point cluster, and screen according to the evaluation results to obtain the important knowledge point cluster.

[0042] Exemplarily, by classifying the knowledge points in the basic knowledge point cluster in detail and determining their difficulty levels according to the refinement degree of the knowledge points. For example, the refinement degree corresponding to the knowledge points can be determined according to the chapters, sections, and units to which the knowledge points belong in the textbook text; or the refinement degree corresponding to the knowledge points can also be determined according to the form of the identification vectors corresponding to the knowledge points (single unit or multi-unit combination form); and then, based on the refinement degree of the knowledge points, the difficulty level is determined. The higher the refinement degree of the knowledge points, the higher the corresponding difficulty level. By applying the method of this embodiment, the systematicness and hierarchy of the knowledge points are ensured, providing a basis for the generation of questions with different difficulty levels in the future.

[0043] In some examples, different difficulty levels correspond to different weight scores of the index importance of each knowledge point. Based on different difficulty levels and weight scores, each knowledge point can be evaluated to obtain an importance score. According to the importance score, the important knowledge point cluster can be further screened out from the basic knowledge point cluster, which is a set of core knowledge points crucial for understanding the target subject or achieving specific learning goals. By identifying and extracting important knowledge points, students can focus on learning key information and avoid wasting time on minor details. Important knowledge points are often the basis for constructing complex concepts. Mastering these core points helps to deeply understand the structure and logic of the subject, thus better solving complex problems.

[0044] Step 103: Determine the knowledge point distribution information of the basic knowledge point clusters and important knowledge point clusters in the teaching materials, generate a knowledge tree containing the knowledge point distribution information, and determine the association strength between each knowledge point according to the number of branches of the knowledge tree.

[0045] In some examples, by deeply analyzing the teaching material text, extract and classify the basic knowledge points, determine the difficulty level according to their refinement degree and complexity, and use a pre-trained evaluation model to evaluate the importance of each knowledge point, so as to distinguish the important knowledge point clusters. Then create nodes for each knowledge point and establish connections based on the logical relationships (such as causal and subordinate relationships) between knowledge points to construct a knowledge tree. At the same time, the difficulty level and importance score can be marked on the nodes.

[0046] Exemplarily, the association strength between knowledge points can be measured by analyzing the number of branches of each node. The more branches indicate that the knowledge point is more difficult. When necessary, adjust the association strength according to the importance score. Visualize the knowledge tree using a graphical tool, which can intuitively display the knowledge point distribution and interconnections, helping teachers and students better understand the course content and formulate effective learning plans. The knowledge tree not only clearly shows the hierarchical structure of knowledge points but also strengthens students' understanding of key knowledge points and their associations.

[0047] Step 104: Construct a knowledge graph of the question bank according to the association strength between each knowledge point. The knowledge graph of the question bank is used to display the questions corresponding to each knowledge point and the association relationships between different knowledge points.

[0048] Exemplarily, by classifying the knowledge points in the basic knowledge point clusters and important knowledge point clusters in detail and determining their difficulty levels according to the refinement degree of the knowledge points, the systematicness and hierarchy of the knowledge points are ensured, providing a basis for subsequent question generation. The question bank contains various types of questions such as multiple-choice questions, fill-in-the-blank questions, short-answer questions, and essay questions. The questions in the question bank correspond to the important knowledge points in the teaching material content. Using text analysis and data mining techniques, automatically extract the important knowledge point clusters, generate the corresponding question bank, and generate the corresponding knowledge graph of the question bank based on the question bank, avoiding the subjectivity and inefficiency of manual screening and improving the accuracy and efficiency of knowledge extraction.

[0049] For example, mark the important knowledge point clusters that are core concepts or principles with wide applications in the knowledge graph of the question bank. Different colors or symbols can be used for marking to distinguish ordinary knowledge points and important knowledge points. The relationships between knowledge points in the knowledge graph of the question bank are clearly visible, and the important knowledge point clusters are highlighted.

[0050] In some examples, knowledge points can be distributed on a two-dimensional grid, with each knowledge point cluster occupying a cell. The size of the cell can be adjusted according to the complexity and importance of the knowledge point. For the cell where the important knowledge point cluster is located, it is marked with a special identifier (such as color, etc.) to form a key learning grid. A navigation path can also be added to the learning map to indicate the learning order from basic knowledge points to advanced knowledge points, helping students better plan their learning paths. The setting of the key learning grid not only highlights the key knowledge points but also provides a clear learning path for students, thereby improving learning efficiency and effectiveness.

[0051] Step 105: Match the knowledge graph of the question bank with the user's learning conversation record, and monitor the user's question-solving situation according to the matching result.

[0052] Exemplarily, by constructing a detailed knowledge graph of the question bank, where each knowledge point node is marked with the corresponding difficulty level and importance degree and linked to relevant questions, and at the same time collecting and processing the user's conversation record, using natural language processing technology to identify the knowledge points mentioned by the user. Then match these knowledge points with the nodes in the knowledge graph, and help the guardian side timely monitor the student's question-solving situation according to the matching result. If the problem is not solved, it can timely help the student to answer and improve the question-solving efficiency.

[0053] For example, on this basis, the system can also record the user's question-solving data (such as the number of completed questions, correct rate, answering time, etc.), track their learning progress and mark the mastery situation on the knowledge graph. Provide personalized feedback and suggestions according to the user's question-solving performance, dynamically adjust the learning path to optimize the recommendation of learning resources, realize the effective monitoring and management of the user's question-solving situation, improve learning efficiency and provide a personalized learning experience.

[0054] Compared with the current existing technologies, in this embodiment, knowledge points are first extracted based on teaching materials texts to obtain a basic knowledge point cluster; the importance of the knowledge points in the basic knowledge point cluster is evaluated, and the important knowledge point cluster is obtained by screening according to the evaluation results; the knowledge point distribution information of the basic knowledge point cluster and the important knowledge point cluster in the teaching materials is determined, a knowledge tree containing the knowledge point distribution information is generated, and the association strength between each knowledge point is determined according to the number of branches of the knowledge tree; according to the association strength between each knowledge point, a question bank knowledge graph is constructed, and the question bank knowledge graph is used to display the questions corresponding to each knowledge point and the association relationship between different knowledge points; the question bank knowledge graph is matched with the user's learning conversation record, and the user's question brushing situation is monitored according to the matching result. Through the analysis and processing of text information, a basic knowledge point cluster can be obtained from the teaching materials and further important knowledge point clusters can be extracted, thus effectively distinguishing important knowledge points from basic knowledge points, improving the accuracy of important knowledge point screening. By further constructing a question bank knowledge graph, associating knowledge points with questions, and using the graph to monitor and analyze the student's question brushing situation, students can accurately brush questions for key knowledge points, thereby realizing personalized learning assistance.

[0055] As a refinement and extension of the above embodiment, when determining the association strength between each knowledge point according to the number of branches of the knowledge tree and constructing a question bank knowledge graph, the following methods can be used but are not limited to, such as Figure 2 shown, the method includes:

[0056] Step 201: Divide the difficulty level of each knowledge point according to the number of branches of the knowledge tree.

[0057] In some examples, first create a node for each knowledge point and establish connections based on the logical relationship between knowledge points. Then calculate the number of direct dependent child nodes (i.e., the number of branches) of each knowledge point node, and divide the knowledge points into three difficulty levels: basic, medium, and high according to the number of branches: Knowledge points with basic difficulty usually have few or no dependent child nodes and are easy to understand and master; knowledge points with medium difficulty have a certain number of child nodes depending on it and require a certain level of understanding and application ability; knowledge points with high difficulty have a large number of child nodes depending on it, involve complex concepts and applications, and require in-depth understanding and flexible application.

[0058] Exemplarily, other factors such as the depth, breadth of knowledge points and student feedback can also be comprehensively considered and dynamically adjusted to ensure the accuracy and reasonableness of the difficulty level division, so as to help teachers and students better plan the learning path and improve learning efficiency.

[0059] Step 202: Determine the association strength between the first knowledge point and the second knowledge point according to the question overlap situation between the first knowledge point and the second knowledge point at adjacent difficulty levels.

[0060] Exemplarily, when constructing a knowledge point network, the existence form of the first-level classification An is an identification vector of a single unit, and the existence forms of the second-level and third-level classifications are forms combined by multiple units. By successively superimposing the identification vectors, multiple knowledge point networks form a knowledge point cluster, covering all the combination forms of knowledge points. The general expression of the knowledge point network is: U = (k, β), where k is the knowledge point set and β is the edge set. Based on the knowledge point set U = (k, β), a knowledge graph G = (k, β, r) in the question bank is constructed, where r is the association strength.

[0061] Specifically, based on the question set corresponding to the first knowledge point and the question set corresponding to the second knowledge point, the association strength between the first knowledge point and the second knowledge point is determined. The definition of the association strength is as follows:

[0062] (Formula 1)

[0063] In Formula 1, where, is the question set corresponding to the first knowledge point, is the question set corresponding to the second knowledge point, is the total number of questions marked with the first knowledge point, is the total number of questions marked with the second knowledge point, is the total number of questions marked with both the first knowledge point and the second knowledge point.

[0064] Step 203: Based on the association strength between the first knowledge point and the second knowledge point, update the questions corresponding to the difficulty levels of the first knowledge point and / or the second knowledge point in the knowledge graph of the question bank.

[0065] Exemplarily, if the association strength is relatively high (for example, exceeding a preset threshold), it indicates that these two knowledge points often appear together in the questions and are closely related. At this time, the question combination corresponding to these two knowledge points in the question bank can be adjusted: for the first knowledge point and the second knowledge point, add comprehensive questions that cover both, ensuring that students can master the related knowledge points synchronously; at the same time, dynamically adjust the question difficulty distribution according to the association strength. For example, reassign some of the questions originally belonging to high-difficulty knowledge points to low-difficulty knowledge points as advanced practice, or introduce more basic questions for high-difficulty knowledge points as a prelude. This can not only strengthen the logical connection between knowledge points but also improve the diversity and adaptability of the questions, increase comprehensive questions and optimize the connection between the question difficulty levels and knowledge points, helping students learn and master knowledge more systematically.

[0066] For example, questions can also be pushed according to the difficulty level of knowledge points. This push mechanism based on the knowledge point hierarchy and difficulty level can effectively help students gradually master knowledge. First, the difficulty of knowledge points can be classified and graded, which can be divided into different levels such as basic, intermediate, and advanced. The basic difficulty corresponds to the first-level knowledge points, which are used to help students consolidate basic knowledge and understand basic concepts; the intermediate difficulty corresponds to the second-level knowledge points, which are used to help students apply basic knowledge to more complex problem-solving; the advanced difficulty corresponds to the third-level knowledge points, which are used to help students improve their comprehensive application ability, requiring students to be able to integrate and flexibly use multiple knowledge points, etc. The mutual relationship and combination of knowledge points can also be considered to ensure that all combinations of knowledge points are covered. For example, if a certain knowledge point A is the basis of another knowledge point B, then it should be ensured that students have mastered A before pushing questions.

[0067] As a refinement and extension of the above embodiments, when updating the questions corresponding to the difficulty levels of the first knowledge point and / or the second knowledge point in the knowledge graph of the question bank based on the association strength between the first knowledge point and the second knowledge point, the following methods can be used but are not limited to, such as Figure 3 As shown, the method includes:

[0068] Step 301: Adjust the question update strategy corresponding to the first knowledge point and / or the second knowledge point according to the association strength and the preset association strength threshold.

[0069] Exemplarily, according to the association strength and the structure of the knowledge tree, construct a knowledge graph. Set the association strength threshold E. If > E, it is considered that the knowledge points are strongly associated, and then they are associated to under the knowledge points, otherwise they are weakly associated and then associated to under the knowledge points; according to the structure of the knowledge tree, finally obtain the knowledge graph G in the question bank.

[0070] Step 302: Update the questions corresponding to the difficulty levels of the first knowledge point and / or the second knowledge point in the knowledge graph of the question bank according to the question update strategy.

[0071] In some examples, according to the form of the identification vector corresponding to the knowledge point, the classification node level corresponding to the knowledge point can be obtained. Furthermore, the comprehensive difficulty level of the questions can be calculated in combination with the refinement degree of the knowledge point and other factors (such as application frequency, complexity, etc.). Furthermore, questions of different difficulty levels can be appropriately mixed in the same set of questions, and the question bank can be updated regularly with the changes of educational goals and the development of technology, adding new knowledge points and question types. The question bank includes questions corresponding to the first-level, second-level, and third-level knowledge points. The question bank is stored in the cloud server for use by the learning tutoring system, such as Figure 4 As shown, the learning tutoring system includes a cloud server and multiple user terminals.

[0072] Exemplarily, the question configuration in the question bank can be adjusted based on the association strength, by adding comprehensive questions that cover both of these knowledge points, and adjusting the difficulty distribution of the questions as needed. For example, more transitional questions from basic to advanced can be introduced in the low-difficulty level, while more complex comprehensive questions can be added in the high-difficulty level to ensure that students can gradually master the related knowledge points. In addition, dynamically updating the question tags and classifications in the question bank can reflect the latest knowledge point association situations, thereby optimizing the diversity and adaptability of the questions and enhancing the learning effect.

[0073] As a refinement and extension of the above embodiments, when evaluating the importance of the knowledge points in the basic knowledge point cluster and screening to obtain the important knowledge point cluster according to the evaluation results, the following methods can be used but are not limited to, such as Figure 5 As shown, this method includes:

[0074] Step 401: Use a pre-trained evaluation model to evaluate the importance corresponding to the knowledge points in the basic knowledge point cluster.

[0075] Among them, the evaluation indicators include one or more of the number of identification vectors corresponding to the knowledge points at all levels in the knowledge point cluster, the weight scores corresponding to the importance of the knowledge points at all levels, the number of occurrences of the knowledge points corresponding to the examination points, and the number of error frequency occurrences of the knowledge points corresponding to the examination points.

[0076] In some examples, the teaching material content can be uploaded to the first knowledge point analysis module, and the knowledge points in the text can be extracted through text analysis technology and classified and identified hierarchically. Hierarchical classification and identification means assigning specific marks to the knowledge points or nodes at each level in a multi-level classification structure for easy distinction and management. Through hierarchical classification and identification, the hierarchical relationship of the knowledge points can be clearly displayed, which helps to construct the entire knowledge network.

[0077] For example, the first-level classification node can be used as the main node and classified by chapter; the second-level classification node can be used as the subordinate node of the first-level classification node and classified by section, representing each subsection in each chapter; the third-level classification node can be used as the subordinate node of the second-level classification node and classified by unit results to further subdivide the specific knowledge points or units under each section, etc.

[0078] Among them, the first-level classification node appears in the form of an identification vector of a single unit, and the multi-level classification node appears in the form of a combination of identification vectors of multiple units.

[0079] Exemplarily, extract the knowledge points in the teaching material and classify and identify them hierarchically. Divide the hierarchical classification and identification into n levels. For example, when n = 3, then U = [An, Bm, Ck], such as Figure 6As shown in the figure, where: the first-level classification An is used as the main node and classified by chapter to obtain An = [A1, A2... An]; the second-level classification Bm is used as the subordinate node of An and classified by section to obtain Bm = [A1b1, A1b2, A2b1... Anbm]; the third-level classification Ck is used as the subordinate node of Bm and classified by unit results to obtain Ck = [A1b1c1, A1b2c2, A1b2c3... Anbmck].

[0080] In some examples, the identification code is used to reflect the hierarchical structure of knowledge points (such as chapters, sections, units). Each knowledge point corresponds to a unique identification code. The specific form of the identification code value can include numerical numbers, letter numbers, mixed numbers, etc. These identification code values can be designed according to the hierarchical structure of knowledge points and converted into identification vectors through one-hot encoding or other methods for further processing and analysis.

[0081] In some examples, by encoding the hierarchical identification, a unique identification code is matched for each knowledge point, and a corresponding identification vector is generated according to the identification code. A knowledge point network is constructed based on the identification vector. The knowledge point network is used to represent the relationships between knowledge points, and then multiple knowledge point networks are combined into a comprehensive set of identification vectors Uori (i.e., the basic knowledge point cluster).

[0082] In some examples, the course video can also be recorded and stored through Internet of Things devices (such as cameras, smart blackboards, etc.); then, audio-to-text processing is performed through a pre-trained speech-to-text model. The speech-to-text model can adopt the OpenAI open-source model Whisper to construct a set of identification vectors corresponding to the knowledge point identification and generate a knowledge point network; based on the set of identification vectors (Dn, Em, Fk), the basic knowledge point cluster L = [Dn, Em, Fk] is obtained.

[0083] In some examples, the coverage of the knowledge point network means that the knowledge points are classified according to their importance and complexity, and all possible combinations of knowledge points are ensured to be covered. Through the construction of multi-level classification identification and identification vectors, the knowledge system of teaching materials can be systematically represented to form a complete knowledge framework.

[0084] Specifically, obtain the importance score of each knowledge point in the basic knowledge point cluster. Among them, the importance score G is obtained by using a pre-trained evaluation model (Formula Two); according to the importance score, determine the importance corresponding to each knowledge point. Among them, Formula Two is specifically as follows:[[]]

[0085] (Formula Two)

[0086] In Formula 2, Z, X, V, and Y are the numbers of identification vectors corresponding to the first-level, second-level, third-level, and n-level knowledge points in the basic knowledge point cluster, respectively. , , , are the weight coefficients corresponding to the importance levels of the first-level, second-level, third-level, and n-level identification vectors, respectively. is the weighted score corresponding to the importance level of the n-level identification vector, and < < because as the refinement degree of the knowledge points gets deeper, the proportion of importance is more.

[0087] In some examples, the weight coefficients should reflect the importance and complexity of the knowledge points at each level. For example: The knowledge points corresponding to the first-level classification are usually relatively broad concepts with a low weight. The knowledge points corresponding to the second-level classification further refine the content with a moderate weight. The knowledge points corresponding to the third-level classification are relatively detailed knowledge points with the highest weight. Adjust the weight coefficients according to the actual situation to ensure that the scoring model can accurately reflect the actual importance of the knowledge points. Over time and with the introduction of new knowledge, it may be necessary to re-evaluate and adjust the weight coefficients. In addition to the refinement degree of the knowledge points, other factors (such as the application frequency and difficulty of the knowledge points) can also be considered to adjust the weight coefficients.

[0088] Exemplarily, it is also possible to generate the basic knowledge point cluster Zori = [Wn, Rm, Hk] by obtaining the historical question bank and based on the set of identification vectors (Wn, Rm, Hk); then screen the basic knowledge point cluster through a pre-trained key knowledge point evaluation model (Formula 3), where Formula 3 is specifically as follows:

[0089] (Formula 3)

[0090] Among them, is the number of times the test point knowledge point appears; is the number of times the error frequency of the test point knowledge point appears; Gr is the number of branches of the test point knowledge point; α is the weighted score of the corresponding index. Set the scoring threshold T2. When > T2, screen out the combination of key knowledge points to form the second most important knowledge point cluster Znew.

[0091] Step 402: Screen in the basic knowledge point cluster based on the importance level corresponding to each knowledge point and a preset screening threshold, and determine the set of screened knowledge points as the important knowledge point cluster.

[0092] Among them, the preset screening threshold is the initial value set for the importance level of the identification vector.

[0093] Exemplarily, a screening threshold T1 is set. When G > T1, the identified vector of knowledge points obtained by screening is used as the important knowledge point cluster Unew. As Figure 7 shown, the initial threshold can be set by using methods such as customization, average method, median, etc., and this embodiment does not limit this.

[0094] By applying the method of this embodiment, the important knowledge point cluster can be effectively screened out. Select a suitable method for setting the screening threshold according to the actual situation to ensure that the screening result can accurately reflect the importance of the knowledge points.

[0095] Optionally, when adjusting the difficulty level of the knowledge points in the question bank knowledge graph, the following methods can be used but are not limited to, such as Figure 8 shown, and this method includes:

[0096] Step 501: Based on the learning data characteristics of the user in the question bank knowledge graph, construct a knowledge base classified according to knowledge points of corresponding difficulty levels. The learning data characteristics are used to analyze the user's learning process and learning effect.

[0097] For example, the learning data characteristics may include one or more of the online learning duration, the number of help requests, the number of effective help times, the response efficiency, the accuracy rate, the learning progress, the answering duration, and the proficiency of a single knowledge point. The answering duration starts timing by the background after the user enters the answering interface. At this time, there may be situations such as hanging up, not answering questions for a long time, waiting for a long time to start answering, or re - entering the current question after exiting the answering.

[0098] In some examples, the knowledge points can be classified according to their difficulty levels, similar - difficulty knowledge points are grouped together, and their classification is dynamically adjusted to reflect the user's latest learning progress. For knowledge points of different difficulty levels, corresponding question sets are generated or updated. Through this method of classifying knowledge points based on the user's learning data characteristics, the constructed knowledge base can not only accurately match the user's personalized needs but also help the user gradually improve the mastery of knowledge points at each difficulty level, thereby realizing more efficient learning path planning and resource allocation.

[0099] Step 502: Use the knowledge base to evaluate the user's mastery of each knowledge point.

[0100] In some examples, according to the information identification and preset relationships of the basic knowledge point cluster and the important knowledge point cluster, the learning grids corresponding to all the knowledge point clusters included in a chapter of the teaching material content are arranged to form a learning area. The preset relationships between knowledge points can include hierarchical relationships, causal relationships, etc. Each learning grid contains all the relevant information of the knowledge point cluster, such as the knowledge point name, description, difficulty level, etc.

[0101] In some examples, a learning map is generated based on the learning areas, and each learning grid in the learning map corresponds to all the questions corresponding to a knowledge point cluster in the question bank. Each learning grid in the learning map corresponds to all the questions corresponding to a knowledge point cluster in the question bank. The learning map can be a graphical representation or a structured list, and can provide personalized learning paths according to the student's learning progress and knowledge mastery, improving learning efficiency.

[0102] For example, the learning map can provide personalized learning paths according to the student's learning progress and knowledge mastery, improving learning efficiency. Specifically, the completion status of the student in each learning grid can be recorded through the learning map, the mastery of each knowledge point can be evaluated through tests or exercises, and then according to the student's evaluation results, suitable learning paths and questions can be recommended. Suppose a student has mastered the knowledge point of "variables and constants", but encounters difficulties in "expressions and equations". The learning map can suggest that the student first review the relevant basic knowledge and then practice more complex questions.

[0103] Exemplarily, the area corresponding to the important knowledge point cluster on the learning map is determined as the key monitoring area. The dynamic update mechanism of the key monitoring area realizes intelligent management through multi-source data fusion and machine learning technology. Its core process includes: constructing a multi-dimensional association network of operation behavior, knowledge graph, learning path, and spatial topology, and establishing a dual-mode user portrait database containing static attributes and dynamic behavior characteristics based on the question bank knowledge graph or using the improved K-means++ clustering algorithm (Improved K-means++ Clustering Algorithm, K-means++) combined with the long short-term memory neural network (Long Short-Term Memory Neural Network, LSTM) neural network; identifying the operation-intensive area through spatial clustering analysis of the dual-mode user portrait database (Dual-Mode User Profile Database with Static Attributes and Dynamic Behavioral Features, DBSCAN), innovatively introducing a non-linear evaluation model of the number of terminals and the degree of knowledge mastery, dynamically adjusting the monitoring boundary through an adaptive grid division algorithm, forming an intelligent closed-loop system including behavior heat map analysis, resource intelligent allocation, and multi-dimensional evaluation of learning efficiency, realizing early warning of knowledge blind spots, path optimization, and dynamic matching of teacher allocation, and promoting the upgrade of the monitoring of students' learning conditions to cognitive perception-based intelligent monitoring.

[0104] In some examples, by performing clustering analysis on user operation behavior data, information on key monitoring areas with multiple different characteristics is formed. The clustering center points in the required areas are used as key points or key location points, different area information is set, and the number of user terminals within the scope of this area (the number of terminals can reflect the user's mastery of the knowledge points at this location. If the number of people who have mastered it is large, the number of user terminals in this area is small, indicating that most users can master it) is obtained as the basis for setting key points, so that new key monitoring area information can be constructed or incremented.

[0105] Step 503: Dynamically adjust the difficulty level of knowledge points in the knowledge graph of the question bank according to the user's mastery of each knowledge point.

[0106] Exemplarily, based on the user's mastery of knowledge points, dynamically adjusting the difficulty level and question configuration of knowledge points can enable the knowledge graph of the question bank to better meet the user's personalized learning needs.

[0107] Optionally, when matching the knowledge graph of the question bank with the user's learning conversation record and monitoring the user's question-solving situation according to the matching result, the following methods can be used but are not limited to, such as Figure 9 As shown, this method includes:

[0108] Step 601: Identify the management detection feature data corresponding to the learning conversation record through a deep learning model.

[0109] For example, a model combining a pre-trained language model based on Bidirectional Encoder Representations from Transformers (BERT) and a Recurrent Neural Network (RNN) model can be used to identify the management detection feature data of the chat content, and a content comparison model can be used to determine whether the chat content belongs to the knowledge base content.

[0110] Step 602: Based on the management detection feature data, use a learning quality scoring model to score the learning conversation record.

[0111] In some examples, user A initiates a question / request for help to user B and sends information related to the question to the chat box; during the chat, the system monitors the chat box conversation content (learning conversation record) in real time; finally, user A gives feedback that the problem has been solved / unsolved / ends the conversation as a termination monitoring instruction; assume that a single conversation contains N messages, and K of them are identified as knowledge base content, and the following core indicators are defined: Dependency Index (DI) and Cognitive Leap Span (CLS).

[0112] Exemplarily, the dependence index between the chat content within the time window T and the knowledge base content is judged, and multiple time windows T are statistically analyzed to obtain whether the chat content of the user at this stage is related to learning. The calculation formula of the dependence index is as shown in Formula Four:

[0113] (Formula Four)

[0114] wherein, is the similarity between the i-th chat content and the knowledge base content, is the enhancement term. The higher the similarity, the stronger the dependence.

[0115] Exemplarily, the thinking jump degree is used to evaluate the reasoning logic of the chat content context. A semantic analysis model and a word vector embedding model are used to extract the semantic feature vectors of the chat content , and based on the DIKWP model, the semantic jump distance between adjacent messages is calculated , and after normalization, the weighted sum is obtained to get CLS, as shown in Formula Five:

[0116] (Formula Five)

[0117] wherein, , is the mean and standard deviation of in the global conversation data; is the time decay weight, ( , recency effect reinforcement).

[0118] Exemplarily, the learning conversation records are scored, and the chat content quality scoring model is as shown in Formula Six:

[0119] (Formula Six)

[0120] wherein, is the dynamic penalty factor, is the proportional penalty for the part exceeding the threshold, is the indicator function, , wherein does not belong to the knowledge base content, is the message number threshold.

[0121] Step 603: Determine the matching degree between the knowledge graph of the question bank and the learning conversation records according to the learning quality scores corresponding to the learning conversation records.

[0122] Exemplarily, after calculating the learning quality score Q, these knowledge points are matched with the knowledge points in the question bank knowledge graph to evaluate their consistency. Among them, the matching degree can be determined by comparing the coincidence degree of the knowledge points mentioned in the conversation with the corresponding knowledge points in the question bank and the user's performance on these knowledge points. A high matching degree means that the question bank knowledge graph can well reflect the user's learning content and needs, and the user's learning process is relatively smooth, while a low matching degree may indicate that the user has unresolved problems or low question-solving efficiency.

[0123] Step 604: If the matching degree is lower than the first warning threshold, trigger the warning mechanism and generate a warning message based on the user's question-solving situation.

[0124] Exemplarily, a first warning threshold G1 can be set. When Q < G1, a warning is triggered, and the warning message is reported to the guardian terminal app. The warning message includes: question / help-seeking information, chat start / end time, dialogue feedback results, etc.; helping the guardian terminal to monitor the student's question-solving situation in a timely manner, and if the problem is not solved, it can help the student to answer in a timely manner to improve the question-solving efficiency.

[0125] Optionally, when matching the question bank knowledge graph with the user's learning conversation record and monitoring the user's question-solving situation according to the matching result, the following methods can also be used but are not limited to, such as Figure 10 As shown, this method includes:

[0126] Step 701: Map the textbook and the learning conversation record to a unified multi-modal space through feature space projection, and calculate the geometric distance and semantic similarity of cross-modal features.

[0127] In some examples, the knowledge point learning data in the question bank knowledge graph can be obtained to construct a knowledge base. The knowledge point learning data includes data such as textbooks, question banks and answers, and courseware. After clustering analysis using a pre-trained classification model, the knowledge base classification result is obtained, and the knowledge base classification result is classified according to the corresponding knowledge points. The classification model is a model that combines a language model pre-trained by BERT and an RNN model, and can divide the knowledge base and chat content into three categories: text, formula, and chart.

[0128] Exemplarily, the textbook text, formula, chart and learning conversation record are projected into a unified high-dimensional space through a heterogeneous encoder to achieve double alignment of semantics and structure, fuse the geometric distance (cosine similarity) and semantic similarity (concept association strength), establish an interpretable measurement system for cross-modal features, dynamically adjust the decision threshold according to the historical data distribution, balance the precision rate and recall rate, and solve the distribution shift problem caused by knowledge base update.

[0129] In some examples, the training method of the model includes: inputting textbook PDF / text + formula LaTeX + chart images at the textbook end, and inputting chat text + screenshots (including formulas / charts) at the user end. The feature encoder is configured conditionally (the core inference functional module), and the text encoding uses an improved Curriculum-BERT model (knowledge distillation). On the basis of the standard BERT, a textbook domain thesaurus (such as mathematical symbols, professional terms) is added, and chapter-level position encoding is added as shown in Formula Seven:

[0130] (Formula Seven)

[0131] Among them, c is the chapter number, l is the hierarchical depth (for example, the chapter is 1, the section is 2, and the subsection is 3), pos is the position of the character in the paragraph, is the hidden layer dimension of the BERT model, and i is the dimension index, The term realizes the attenuation attention to the deep chapter structure. The deeper the hierarchy (such as the subsection), the smaller the fluctuation amplitude of the position encoding.

[0132] In some examples, the process of formula graph convolutional encoding includes: Step 1: Parse LaTeX into a syntax tree → convert it into a graph structure G=(V,E), where the node set V is operators, operands, special symbols, and the edge set E is the syntax dependency relationship; Step 2: Iteratively update the graph convolution, as shown in Formula Eight:

[0133] (Formula Eight)

[0134] Among them, is the feature vector of node v at the kth layer (the layer number k usually takes 3 to balance local and global features), is the neighbor set of node v, is the normalization coefficient of edge (v,u) (usually the node degree), is the learnable weight matrix of the kth layer, is the GELU activation function

[0135] In some examples, the dynamic threshold decision mechanism first needs to calculate the basic similarity:

[0136] (Formula Nine)

[0137] Among them, is the feature vector of the student chat content, is the feature vector of the knowledge base, is the standard deviation of the feature space

[0138] Secondly, calculate the adaptive threshold:

[0139] (Formula Ten)

[0140] Among them, is the initial threshold (default 0.85), η is the maximum adjustment amplitude, k is the adjustment rate, t is the current training round, is the starting round for triggering adjustment

[0141] Then calculate the decision function:

[0142] (Formula Eleven)

[0143] In some examples, the multi-modal pre-training steps include: establishing a cross-modal unified representation space. Step 1: Knowledge base parsing: Convert text into a Token sequence with hierarchical tags, convert formulas into syntactic graph structures, and convert charts into coordinate annotation images; Step 2: Contrastive learning: Positive sample pairs: (knowledge base text segment, corresponding formula / chart), Negative sample pairs: Type 1: Randomly replace formulas in textbooks, Type 2: Shuffle the coordinate annotations of charts.

[0144] Calculate the loss function:

[0145] (Formula Twelve)

[0146] Among them, is the margin parameter (default 0.2), (a, p, n) is the anchor, positive example, negative example triple.

[0147] In some examples, discriminative fine-tuning can also be performed on the model to enhance the ability to distinguish textbook features from non-textbook features. Specifically, through mixed augmentation data generation (improving the accuracy of model recognition through complex samples), semantic-preserving perturbation, and synonym replacement in the following ways:

[0148] (Formula Thirteen)

[0149] Among them, x is the vocabulary, w is the synonym of x, Cos() is the similarity, combined with sentence pattern conversion, active-passive voice conversion, structure destruction operation, and random dropping of formula nodes as shown in the following formula:

[0150] (Formula Fourteen)

[0151] In the DropNode random deletion method, the chart offset is as shown in the following formula:

[0152] (Formula Fifteen)

[0153] Among them, is the offset.

[0154] Exemplarily, the training process can be carried out in three stages: primary, intermediate, and advanced. The sample complexity can gradually increase by stage to improve the accuracy of model recognition.

[0155] Step 702: Based on the geometric distance and semantic similarity of cross-modal features, determine the matching degree between the knowledge graph of the question bank and the learning conversation record through a decision function.

[0156] Exemplarily, as Figure 11 shown, first, the user input is processed to extract multi-modal features, which are then fused with the knowledge base features in the pre-computed cache. Through knowledge distillation technology, the cross-modally fused features are further optimized, and finally, a dynamic decision result is generated. This series of steps ensures that the system can efficiently process complex information and make real-time decisions based on comprehensive analysis.

[0157] In some embodiments, in the first step, the chat content "The Pythagorean theorem formula is ; in the second step, the processing flow includes a text encoder: extract the concept features of "Pythagorean theorem" , a formula encoder: parse the formula structure to generate , and the cross-modal fusion is as shown in the following formula:

[0158] (Formula XVI)

[0159] Based on the steps of the above embodiments, in the third step, retrieve the textbook library: find all entries related to the Pythagorean theorem in the mathematics textbook , and in the fourth step, perform decision calculation:

[0160] (Formula XVII)

[0161] For example, if the current threshold θ = 0.88, it can be determined that the matching degree of the learning conversation record meets the standard and belongs to the content of the knowledge graph of the question bank.

[0162] Step 703: If the matching degree is lower than the second warning threshold, trigger the warning mechanism and generate a warning message based on the user's question-solving situation.

[0163] Among them, the second warning threshold can be adaptively adjusted based on the historical data distribution.

[0164] Exemplarily, the second warning threshold G2 can be set. When < G2, trigger a warning, report the warning message to the guardian-end app. The warning message includes: question / help request information, chat start / end time, dialogue feedback result, etc.; help the guardian-end to monitor the student's question-solving situation in a timely manner, and if the problem is not solved, be able to help the student answer in a timely manner to improve the question-solving efficiency.

[0165] Compared with the current existing technologies, in this embodiment, through the analysis and processing of text information, the basic knowledge point clusters can be obtained from teaching materials and further the important knowledge point clusters can be extracted, thus effectively distinguishing the important knowledge points from the basic knowledge points, improving the accuracy of screening important knowledge points. By further constructing the knowledge graph of the question bank, the knowledge points are associated with the questions, and the learning situation of students is monitored and analyzed by using this graph, so that students can accurately brush questions for key knowledge points, thereby realizing personalized learning assistance.

[0166] Further, as Figures 1 to 10 a specific implementation of the method shown, this embodiment provides a learning assistance device based on the knowledge graph of the question bank, as Figure 12 shown. The device includes: an extraction module 81, an evaluation module 82, a generation module 83, a construction module 84, and a monitoring module 85.

[0167] The extraction module 81 is configured to extract knowledge points based on teaching material texts to obtain basic knowledge point clusters;

[0168] The evaluation module 82 is configured to evaluate the importance of the knowledge points in the basic knowledge point clusters, and screen to obtain important knowledge point clusters according to the evaluation results;

[0169] The generation module 83 is configured to determine the knowledge point distribution information of the basic knowledge point clusters and the important knowledge point clusters in the teaching materials, generate a knowledge tree including the knowledge point distribution information, and determine the association strength between each knowledge point according to the number of branches of the knowledge tree;

[0170] The construction module 84 is configured to construct a knowledge graph of the question bank according to the association strength between each knowledge point. The knowledge graph of the question bank is used to display the questions corresponding to each knowledge point and the association relationships between different knowledge points;

[0171] The monitoring module 85 is configured to match the knowledge graph of the question bank with the learning conversation records of the user, and monitor the question brushing situation of the user according to the matching results.

[0172] In some examples of this embodiment, the generation module 83 is specifically configured to divide the difficulty levels of each knowledge point according to the number of branches of the knowledge tree; determine the association strength between the first knowledge point and the second knowledge point according to the question overlap situation between the first knowledge point and the second knowledge point at adjacent difficulty levels; correspondingly, the construction module 84 is specifically configured to update the questions at the corresponding difficulty levels of the first knowledge point and / or the second knowledge point in the knowledge graph of the question bank based on the association strength between the first knowledge point and the second knowledge point.

[0173] In some examples of this embodiment, the generation module 83 is further specifically configured to determine the association strength between the first knowledge point and the second knowledge point based on the question set corresponding to the first knowledge point and the question set corresponding to the second knowledge point, where the association strength is defined as:

[0174]

[0175] wherein, is the question set corresponding to the first knowledge point, is the question set corresponding to the second knowledge point, is the total number of questions annotating the first knowledge point, is the total number of questions annotating the second knowledge point, is the total number of questions annotating both the first knowledge point and the second knowledge point.

[0176] In some examples of this embodiment, the generation module 83 is further specifically configured to adjust the question update strategy corresponding to the first knowledge point and / or the second knowledge point according to the association strength and a preset association strength threshold; and update the questions corresponding to the difficulty levels of the first knowledge point and / or the second knowledge point in the question bank knowledge graph according to the question update strategy.

[0177] In some examples of this embodiment, the evaluation module 82 is specifically configured to evaluate the importance of the knowledge points corresponding to the basic knowledge point cluster by using a pre-trained evaluation model, and the evaluation indicators include one or more of the number of identification vectors corresponding to the knowledge points at all levels in the knowledge point cluster, the weight scores corresponding to the importance of the knowledge points at all levels, the number of occurrences of the knowledge points corresponding to the examination points, and the number of error frequency occurrences of the knowledge points corresponding to the examination points; perform screening in the basic knowledge point cluster based on the importance of each knowledge point and a preset screening threshold, and determine the screened knowledge point set as the important knowledge point cluster, where the preset screening threshold is an initial value set for the importance of each knowledge point.

[0178] In some examples of this embodiment, the construction module 84 is further specifically configured to construct a knowledge base classified according to knowledge points of corresponding difficulty levels based on the learning data characteristics of the user in the question bank knowledge graph, where the learning data characteristics are used to analyze the learning process and learning effect of the user; evaluate the user's mastery of each knowledge point by using the knowledge base; and dynamically adjust the difficulty levels of the knowledge points in the question bank knowledge graph according to the user's mastery of each knowledge point.

[0179] In some examples of this embodiment, the monitoring module 85 is specifically configured to identify the management detection feature data corresponding to the learning conversation record through a deep learning model; based on the management detection feature data, use a learning quality scoring model to score the learning conversation record; according to the learning quality score corresponding to the learning conversation record, determine the matching degree between the question bank knowledge graph and the learning conversation record; if the matching degree is lower than the first warning threshold, trigger a warning mechanism and generate a warning message based on the user's question-solving situation.

[0180] In some examples of this embodiment, the monitoring module 85 is further specifically configured to map the teaching material and the learning conversation record to a unified multimodal space through feature space projection, and calculate the geometric distance and semantic similarity of cross-modal features; based on the geometric distance and semantic similarity of the cross-modal features, determine the matching degree between the question bank knowledge graph and the learning conversation record through a decision function; if the matching degree is lower than the second warning threshold, trigger a warning mechanism and generate a warning message based on the user's question-solving situation, and the second warning threshold is adaptively adjusted based on historical data distribution.

[0181] It should be noted that for other corresponding descriptions of each functional unit involved in the learning assistance device based on the question bank knowledge graph provided in this embodiment, reference can be made to Figures 1 to 10 the corresponding description in, which will not be elaborated here.

[0182] Based on the method as shown in Figures 1 to 10 above, correspondingly, this embodiment further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the method as shown in Figures 1 to 10 above is implemented.

[0183] Based on such an understanding, the technical solution of this application can be embodied in the form of a software product, and the software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of various implementation scenarios of this application.

[0184] As shown in Figure 13 is a schematic hardware structure diagram of an electronic device of the present invention, including:

[0185] at least one processor 901; and,

[0186] a memory 902 communicatively connected to at least one of the processors 901; wherein,

[0187] The memory 902 stores instructions that can be executed by at least one of the processors. The instructions are executed by at least one of the processors so that at least one of the processors can execute the learning assistance method based on the question bank knowledge graph as described above.

[0188] Figure 13 Taking one processor 901 as an example.

[0189] The electronic device may further include: an input device 903 and a display device 904.

[0190] The processor 901, the memory 902, the input device 903, and the display device 904 may be connected through a bus or other means. Figure 13 Taking connection through a bus as an example.

[0191] As a non-volatile computer-readable storage medium, the memory 902 can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the learning assistance method based on the question bank knowledge graph in the embodiments of the present application. For example, Figures 1 to 8 The method flow shown. The processor 901 executes various functional applications and data processing by running the non-volatile software programs, instructions, and modules stored in the memory 902, that is, implements the learning assistance method based on the question bank knowledge graph in the above embodiments.

[0192] The memory 902 may include a program storage area and a data storage area. Among them, the program storage area can store an operating system and application programs required for at least one function; the data storage area can store data created according to the use of the learning assistance method based on the question bank knowledge graph, etc. In addition, the memory 902 may include high-speed random access memory, and may also include non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices. In some embodiments, the memory 902 may optionally include a memory remotely set relative to the processor 901, and these remote memories can be connected to the device for executing the learning assistance method based on the question bank knowledge graph through a network. Examples of the above network include but are not limited to the Internet, enterprise intranets, local area networks, mobile communication networks, and their combinations.

[0193] The input device 903 can receive input user clicks and generate signal inputs related to user settings and function controls of the learning assistance method based on the question bank knowledge graph. The display device 904 may include a display screen and other display devices.

[0194] When the one or more modules are stored in the memory 902 and run by the one or more processors 901, they execute the learning assistance method based on the question bank knowledge graph in any of the above method embodiments.

[0195] Optionally, the above-mentioned physical device may further include a user interface, a network interface, a camera, a Radio Frequency (RF) circuit, sensors, an audio circuit, a WI-FI module, etc. The user interface may include a display, an input unit such as a keyboard, etc. Optionally, the user interface may further include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), etc.

[0196] Those skilled in the art can understand that the above-mentioned physical device structure provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine some components, or have different component arrangements.

[0197] The storage medium may further include an operating system and a network communication module. The operating system is a program for managing the hardware and software resources of the above-mentioned physical device, and supports the operation of information processing programs and other software and / or programs. The network communication module is used to implement communication between components inside the storage medium, as well as communication between other hardware and software in the information processing physical device.

[0198] Through the description of the above embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus a necessary general hardware platform, or can also be implemented by hardware. By applying the solution of this embodiment, compared with the current existing technologies, this embodiment first extracts knowledge points based on teaching material texts to obtain a basic knowledge point cluster; evaluates the importance of the knowledge points in the basic knowledge point cluster, and filters according to the evaluation results to obtain an important knowledge point cluster; determines the knowledge point distribution information of the basic knowledge point cluster and the important knowledge point cluster in the teaching material, generates a knowledge tree containing the knowledge point distribution information, and determines the association strength between each knowledge point according to the number of branches of the knowledge tree; constructs a question bank knowledge graph according to the association strength between each knowledge point, and the question bank knowledge graph is used to display the questions corresponding to each knowledge point and the association relationship between different knowledge points; matches the question bank knowledge graph with the user's learning conversation record, and monitors the user's question-solving situation according to the matching result. Through the analysis and processing of text information, a basic knowledge point cluster can be obtained from the teaching material and further an important knowledge point cluster can be extracted, thus effectively distinguishing important knowledge points and basic knowledge points, improving the accuracy of screening important knowledge points. By further constructing a question bank knowledge graph, associating knowledge points with questions, and using this graph to monitor and analyze the student's question-solving situation, students can accurately brush questions for key knowledge points, thereby realizing personalized learning assistance.

[0199] It should be noted that in this document, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising said element.

[0200] The above are only specific embodiments of the present application, enabling those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments described herein, but rather will conform to the broadest scope consistent with the principles and novel features claimed herein.

Claims

1. A learning assistance method based on question bank knowledge graph, characterized in that: include: Extract knowledge points based on textbook texts to obtain basic knowledge point clusters; Evaluate the importance of the knowledge points in the basic knowledge point cluster, and select important knowledge point clusters according to the evaluation results; Determine the knowledge point distribution information of the basic knowledge point cluster and the important knowledge point cluster in the textbook, generate a knowledge tree containing the knowledge point distribution information, and determine the association strength between each knowledge point according to the number of branches of the knowledge tree, including: dividing the difficulty level of each knowledge point according to the number of branches of the knowledge tree; determining the association strength between the first knowledge point and the second knowledge point according to the overlap of questions between the first knowledge point and the second knowledge point of adjacent difficulty levels, and the association strength is defined as: in, is a set of questions corresponding to the first knowledge point, is the set of questions corresponding to the second knowledge point, is the total number of questions marking the first knowledge point, The total number of questions marking the second knowledge point, is the total number of questions marked with both the first knowledge point and the second knowledge point; According to the association strength between the various knowledge points, a question bank knowledge graph is constructed, wherein the question bank knowledge graph is used to display the questions corresponding to each knowledge point and the association relationship between different knowledge points, including: based on the association strength between the first knowledge point and the second knowledge point, updating the questions of the difficulty level corresponding to the first knowledge point and / or the second knowledge point in the question bank knowledge graph; The question bank knowledge graph is matched with the user's learning conversation record, and the user's question-answering status is monitored based on the matching result.

2. The method according to claim 1, characterized in that The updating of questions of difficulty levels corresponding to the first knowledge point and / or the second knowledge point in the question bank knowledge graph based on the association strength between the first knowledge point and the second knowledge point includes: According to the association strength and a preset association strength threshold, adjusting a question update strategy corresponding to the first knowledge point and / or the second knowledge point; According to the question updating strategy, questions of difficulty levels corresponding to the first knowledge point and / or the second knowledge point are updated in the question bank knowledge graph.

3. The method according to claim 1, characterized in that The step of evaluating the importance of the knowledge points in the basic knowledge point cluster and screening the important knowledge point cluster according to the evaluation result includes: Using a pre-trained evaluation model to evaluate the importance of the knowledge points in the basic knowledge point cluster, the evaluation indicators include one or more of the number of identification vectors corresponding to the knowledge points at each level in the knowledge point cluster, the weight scores corresponding to the importance of the knowledge points at each level, the number of occurrences of the knowledge points corresponding to the test points, and the number of error frequencies of the knowledge points corresponding to the test points; The basic knowledge point cluster is screened based on the importance of each knowledge point and a preset screening threshold, and the screened knowledge point set is determined as the important knowledge point cluster, wherein the preset screening threshold is an initial value set for the importance of each knowledge point.

4. The method according to claim 1, characterized in that The method further comprises: Based on the learning data features of the user in the question bank knowledge graph, a knowledge base is constructed that is classified according to knowledge points of corresponding difficulty levels, and the learning data features are used to analyze the learning process and learning effect of the user; Using the knowledge base to evaluate the user's mastery of each knowledge point; The difficulty level of the knowledge points in the question bank knowledge graph is dynamically adjusted according to the user's mastery of each knowledge point.

5. The method according to claim 1, characterized in that The matching of the question bank knowledge graph with the user's learning conversation record and monitoring the user's question-answering situation according to the matching result includes: Identify management detection feature data corresponding to the learning conversation record through a deep learning model; Based on the management detection feature data, scoring the learning conversation record using a learning quality scoring model; Determining the matching degree between the question bank knowledge graph and the learning conversation record according to the learning quality score corresponding to the learning conversation record; If the matching degree is lower than the first warning threshold, the warning mechanism is triggered and warning information is generated based on the user's answering situation.

6. The method according to claim 1, characterized in that The matching of the question bank knowledge graph with the user's learning conversation record and monitoring the user's question-answering situation according to the matching result includes: Mapping the teaching material and the learning dialogue record to a unified multimodal space through feature space projection, and calculating the geometric distance and semantic similarity of cross-modal features; Based on the geometric distance and semantic similarity of the cross-modal features, determining the matching degree between the question bank knowledge graph and the learning dialogue record through a decision function; If the matching degree is lower than a second warning threshold, the warning mechanism is triggered and warning information is generated based on the user's question-solving situation. The second warning threshold is adaptively adjusted based on historical data distribution.

7. A learning assistance device based on a question bank knowledge graph, characterized in that: include: An extraction module is configured to extract knowledge points based on the textbook text to obtain basic knowledge point clusters; An evaluation module is configured to evaluate the importance of the knowledge points in the basic knowledge point cluster, and screen the important knowledge point cluster according to the evaluation result; The generation module is configured to determine the knowledge point distribution information of the basic knowledge point cluster and the important knowledge point cluster in the textbook, generate a knowledge tree containing the knowledge point distribution information, and determine the association strength between each knowledge point according to the number of branches of the knowledge tree, including: dividing the difficulty level of each knowledge point according to the number of branches of the knowledge tree; determining the association strength between the first knowledge point and the second knowledge point according to the overlap of questions between the first knowledge point and the second knowledge point of adjacent difficulty levels, and the association strength is defined as: in, is a set of questions corresponding to the first knowledge point, is the set of questions corresponding to the second knowledge point, is the total number of questions marking the first knowledge point, The total number of questions marking the second knowledge point, is the total number of questions marked with both the first knowledge point and the second knowledge point; A construction module is configured to construct a question bank knowledge graph according to the association strength between the various knowledge points, wherein the question bank knowledge graph is used to display the questions corresponding to each knowledge point and the association relationship between different knowledge points, including: updating the questions of the difficulty level corresponding to the first knowledge point and / or the second knowledge point in the question bank knowledge graph based on the association strength between the first knowledge point and the second knowledge point; The monitoring module is configured to match the question bank knowledge graph with the user's learning conversation record, and monitor the user's question-answering situation based on the matching result.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

9. An electronic device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

Images