Device and method for improving education quality through knowledge point intelligent short board analysis

Abstract

The invention relates to the technical field of education, and provides a method and device for improving education quality through knowledge point intelligent short board analysis. The system captures behavior data of a user in the answering process in real time by establishing a question bank containing parameters such as difficulty coefficients and knowledge point codes. The data update a knowledge point mastering model of the user through a real-time processing system, and an item response theory (IRT) model is adopted to calculate the knowledge point mastering degree according to the answer result of the user and the question difficulty. On the basis, the system updates the knowledge graph of the user in real time, and the overall mastering degree and difficulty level performance are covered. Afterwards, according to the updated knowledge graph, questions suitable for the current learning level of the user are recommended through an intelligent algorithm, and weak links are helped to be strengthened; the new behavior data generated in the answering process of the user is dynamically updated, and the recommendation effect is continuously optimized, so that personalized learning path planning is realized, and the learning efficiency and the education quality are improved.

Description

Technical Field

[0001] The present invention relates to the field of educational technology, and in particular to online education, personalized learning systems and intelligent recommendation algorithms, and specifically to a device and method for improving education quality through intelligent shortcoming analysis of knowledge points. Background Art

[0002] With the rapid development of information technology and the popularization of the Internet, online education platforms have become an important part of modern education. Students can access rich learning resources through the Internet and arrange their own learning time and progress. However, traditional online education systems usually use unified teaching content and exercises, lack in-depth analysis of each student's individual learning situation, and cannot meet students' personalized learning needs.

[0003] In the learning process, students often have shortcomings in certain knowledge points due to differences in the depth of knowledge understanding and acceptance ability. If these shortcomings are not discovered and remedied in time, it may lead to an incomplete knowledge structure and affect the subsequent learning effect. In addition, when students encounter difficulties in the learning process, if they do not receive timely and effective help, they may feel frustrated and reduce their interest and enthusiasm in learning.

[0004] The existing education system lacks real-time and accurate analysis of students' mastery of knowledge points, and is unable to provide students with personalized learning suggestions and exercise recommendations. In terms of question recommendations, many systems simply make recommendations based on students' scores or answer records, without fully considering the difficulty of the questions, the relevance of knowledge points, and the students' actual ability level. As a result, the difficulty of the recommended questions does not match the students' abilities. They are either too simple to consolidate and improve their knowledge, or too difficult to cause students to feel frustrated.

[0005] Therefore, there is an urgent need for a method and device that can analyze students' learning behavior in real time, accurately calculate their mastery of each knowledge point, and recommend the most suitable exercise questions to students based on this, so as to improve the quality of education. Summary of the invention

[0006] The purpose of the present invention is to provide a device and method for improving the quality of education through intelligent shortcoming analysis of knowledge points, which can record and analyze students' learning behavior in real time, accurately calculate their mastery of each knowledge point, and recommend questions suitable for students' current ability level based on this, thereby realizing personalized learning path planning, filling knowledge gaps, and improving learning efficiency and education quality.

[0007] To achieve the above object, the present invention provides a method for improving education quality through intelligent short board analysis of knowledge points, which specifically includes the following steps:

[0008] Create a question bank with rich questions. Each question contains the following parameters: Exam content: the main content of the exam, including the question stem, options, etc.

[0009] Question difficulty: The value range is 0 to 1, and the larger the value, the easier the question.

[0010] Knowledge point code (knowledge_point_code): The unique code of the knowledge point associated with the question.

[0011] Correct answer: The standard answer to the question.

[0012] Question type (category): for example, multiple-choice questions, fill-in-the-blank questions, true-or-false questions, etc.

[0013] Question tags: used to identify specific attributes of questions, such as "high-frequency test points", "easy-to-make mistakes", etc.

[0014] Applicable region: Indicates the region to which the question applies, such as "Beijing", "Shanghai", etc.

[0015] Is it a real exam question (is_real_exam): Indicates whether the question appears in a real exam.

[0016] A question-answering interface is built on the user side, and a user behavior recording module is embedded to capture various behavior data of users during the question-answering process. This module can record the following information of users in real time: Time of entering the question (bhv_time): The timestamp when the user starts reading or answering the question.

[0017] Response time (response_time): The time the user stays on the question, in seconds.

[0018] Click behavior: user interaction on the interface, such as clicking options, switching pages, etc.

[0019] User's answer: The answer submitted by the user.

[0020] Answer result (answer_result): Indicates whether the user's answer is correct, 1 for correct and 0 for incorrect.

[0021] Hint_requested: Whether the user requested the hint information.

[0022] Explanation_viewed: Whether the user has viewed the explanation of the question.

[0023] Whether to collect or mark the question (question_flagged): whether the user has added the question to favorites or marked it.

[0024] Whether the user exited the system on this question (exited_on_question): Indicates whether the user exited the system during the question-answering process.

[0025] The collected user behavior data is stream processed and calculated through real-time data processing systems (such as Kafka and Flink). The system architecture includes: Data stream transmission: User behavior data is transmitted to the real-time computing engine in the form of streaming data through message queue systems such as Kafka.

[0026] Real-time computing: Use real-time computing frameworks such as Flink to calculate user behavior data and generate updates on the user's knowledge mastery.

[0027] Data storage: Use relational database or NoSQL database to store user basic information, question information, knowledge point information, user behavior data and user knowledge graph.

[0028] A model based on Item Response Theory (IRT) is used to accurately calculate the user's mastery of knowledge points by combining factors such as the user's answer results, question difficulty, expected correct probability and learning rate.

[0029] : The user's ability value for a certain knowledge point, ranging from 0 to 1.

[0030] : The difficulty parameter of the question, ranging from 0 to 1. Since the larger the difficulty value, the simpler the question, when converted to the difficulty parameter, .

[0031] : The user's answer result, 1 for a correct answer and 0 for an incorrect answer.

[0032] : Learning rate, the value range is 0 to 1, the recommended value is 0.1, and it can be adjusted according to actual conditions.

[0033] step:

[0034] Convert the user's mastery level into a competency value: Convert question difficulty to difficulty parameter: Calculate the expected correct probability: According to the IRT model, the expected probability of a user answering the question correctly is: in, is the base of natural logarithms.

[0035] Update ability value: Update the user's ability value based on their actual answer results: When the user answers the question correctly and the expected probability of correctness is low, the ability value increases.

[0036] When the user answers a question incorrectly and the expected probability of being correct is high, the ability value decreases.

[0037] Limit the ability value range and update the mastery level: Limit the updated ability value to the range of 0 to 1 and convert it to mastery level:

[0038] Practice correct answer count (right_count): If the answer is correct, add 1.

[0039] Wrong_count: If you answer wrongly, add 1.

[0040] Total number of practice times (practice_count): Added by 1.

[0041] Correct rate (correct_rate): Master the update time (last_updated): record the current time.

[0042] Classify the questions by difficulty level, count the user's performance at each difficulty level, and update the corresponding mastery level (mastery_sub_level).

[0043] (5) Update the user’s knowledge graph After the user answers each question, the user's knowledge graph is updated in real time based on the calculated mastery level. The knowledge graph includes: Overall mastery level (mastery_level): The user's overall mastery of each knowledge point, ranging from 0 to 100.

[0044] Mastery level (mastery_sub_level): The user's mastery level at different difficulty levels, ranging from 0 to 100.

[0045] Statistics: including right_count, wrong_count, practice_count, correct_rate, etc.

[0046] (6) Intelligent topic recommendation Based on the updated user knowledge graph, an intelligent recommendation algorithm is applied to recommend topics suitable for users' current learning level.

[0047] Initial candidate pool: select all questions from the question bank that are related to the knowledge points that the user needs to strengthen.

[0048] Application of deduplication rules: Remove questions on which the user has already taken action from the candidate pool, including questions that the user has answered, attempted, viewed hints for, or solved.

[0049] Result candidate pool: Get a set of related questions that the user has never seen.

[0050] Calculate the utility score of the candidate topic for each candidate topic , calculate the utility score , measure the potential learning benefits of the topic to the user.

[0051] Knowledge point priority score : The lower the mastery level of a knowledge point, the higher its priority.

[0052] Difficulty Level Priority Score : , is the difficulty level weight.

[0053] Matching function : Use Gaussian function to measure the degree of match between question difficulty and user ability , Controls the width of the matching degree. The recommended value is 0.2.

[0054] Utility score :

[0055] Sort by: Based on utility score Sort the candidate questions from high to low.

[0056] Sorting rules: If the utility scores are the same, they can be further sorted based on factors such as title quality and content diversity.

[0057] Recommended list: Select the first N questions after sorting and recommend them to the user, 10 questions each time.

[0058] User behavior monitoring: During the user's answering process, the behavior data of each question is collected in real time.

[0059] Next batch trigger timing: When the user answers the ninth question, the system starts preparing the next batch of recommended questions.

[0060] Update user profile: Use the collected behavioral data to update the user's understanding of the user and .

[0061] Generate the next batch of recommendations: Based on the updated user knowledge graph, repeat the above steps to generate a new recommendation list.

[0062] Deduplication rules: Purpose: Avoid recommending questions that users have already done and improve user experience.

[0063] Implementation: Remove topics on which the user has already performed actions from the candidate pool.

[0064] Behavior definition: includes questions that have been answered, attempted, or have had hints or solutions viewed.

[0065] Sorting rules: Basis: Based on utility score Sort the candidate topics.

[0066] Sorting strategy: When the utility scores are the same, factors such as title quality, content diversity, and title novelty are considered.

[0067] Presentation method: The sorted question IDs are given to the front end and presented to the user one by one. DETAILED DESCRIPTION

[0068] Implementation Method 1: System Overall Architecture The system architecture of the present invention mainly includes the following modules: 1. User layer: User terminal devices: including PCs, tablets, smart phones and other devices. Users answer questions through the question-answering interface.

[0069] Question-answering interface and user behavior recording module: Provides a friendly user interaction interface and embeds a behavior recording function to collect user behavior data in real time.

[0070] 2. Data transmission layer: Message queue system (such as Kafka): transmits user behavior data to the data processing layer in the form of streaming data.

[0071] 3. Data processing layer: Real-time computing engine (such as Flink): performs real-time computing on streaming data to update the user's knowledge level.

[0072] 4. Data storage layer: Database: used to store question banks, user basic information, knowledge point information, user behavior data, and user knowledge graphs, etc.

[0073] 5. Business logic layer: Mastery degree calculation module: Calculates the user's mastery of knowledge points based on user behavior data.

[0074] Intelligent recommendation algorithm module: Generates a personalized topic recommendation list based on the user's knowledge graph.

[0075] 6. Front-end presentation layer: Recommended question display: present recommended questions to users one by one.

[0076] System process overview: Users answer questions through the question-answering interface of the terminal device, and their behavioral data is collected in real time.

[0077] The behavior data is transmitted to Flink via Kafka for real-time calculation and processing.

[0078] Flink calculates the user's mastery level, updates the knowledge graph, and stores the results in the database.

[0079] The intelligent recommendation algorithm generates a list of recommended topics based on the latest knowledge graph.

[0080] The front end receives a list of recommended topic IDs and presents them to the user in order.

[0081] The above process is repeated to achieve real-time updates and recommendations.

[0082] Implementation Method 2: Data Model Design (1) User basic information table Stores basic user information, including user ID, user name, gender, age, region, membership level, registration time, last login time, etc.

[0083] (2) Knowledge point material table Stores relevant information of knowledge points, including knowledge point ID, knowledge point code, knowledge point name, parent knowledge point code, related knowledge point list, associated test ID, associated textbook ID, etc.

[0084] (3) List of materials Stores question information, including question ID, question code, question content, question difficulty, knowledge point code, question type, question label, applicable region, whether it is a real exam question, etc.

[0085] (4) User behavior table Stores the user's behavior data for a single question, including user ID, question code, behavior time, answer result, answer time, user answer, whether to view hints, whether to view analysis, whether to collect or mark the question, etc.

[0086] (5) User knowledge graph table Stores the user's mastery of each knowledge point, including user ID, knowledge point code, mastery level, mastery update time, number of correct answers in practice, number of incorrect answers in practice, total number of practice times, accuracy rate, difficulty level subset, etc.

[0087] Implementation method 3: Method for calculating the degree of user knowledge mastery This embodiment describes in detail how to calculate the user's mastery of knowledge points.

[0088] 1. Data collection: Obtain information such as the user's answer results, answering time, question difficulty, etc. from user behavior data.

[0089] 2. Ability value initialization: For new users or users who have not been exposed to a certain knowledge point, initialize their ability value A, which can be set to a default value, such as 0.5.

[0090] 3. Ability value update: According to the user's answer results, the calculation method based on the IRT model is used to update the user's ability value A for the knowledge point.

[0091] 4. Mastery level conversion: Convert the updated ability value A to mastery level for easy display and analysis.

[0092] 5. Difficulty level mastery update: Statistics on user performance at different difficulty levels and update the corresponding mastery levels.

[0093] 6. Statistical data update: Update the number of correct answers, wrong answers, total number of exercises, accuracy rate, etc. of the user on this knowledge point.

[0094] 7. Data storage: Store the updated data back into the user knowledge graph table for subsequent recommendation algorithms and data analysis.

[0095] Implementation method 4: Implementation of intelligent recommendation algorithm This embodiment describes in detail the specific implementation process of the intelligent recommendation algorithm.

[0096] 1. Construction of candidate question pool: Based on the user's knowledge graph, select the knowledge points that the user has a low level of mastery, extract relevant questions from the question bank, and build an initial candidate question pool.

[0097] 2. Deduplication: Apply deduplication rules to remove topics that the user has already acted on, ensuring that the recommended topics are new to the user.

[0098] 3. Utility score calculation: For each candidate question, calculate its utility score. The utility score is determined by factors such as knowledge point priority, difficulty level priority, and the degree of match between the question and the user's ability.

[0099] 4. Sorting of candidate questions: Sort the candidate questions in descending order according to their utility scores.

[0100] 5. Recommended question selection: Select the first N questions after sorting, usually 10 questions, and generate a recommended list.

[0101] 6. Real-time update mechanism: During the user's answering process, the user's knowledge graph is updated in real time, and the recommendation strategy is dynamically adjusted to ensure that the subsequent recommended questions are always adapted to the user's current level.

[0102] Implementation 5: Deduplication and Sorting Rules This implementation method describes in detail the specific application of deduplication and sorting rules.

[0103] Deduplication rules: Purpose: Avoid recommending duplicate questions and improve the user's learning experience.

[0104] Implementation method: Behavior record verification: Before recommending, check the user's behavior record to determine the topics on which the user has engaged in behavior.

[0105] Candidate pool filtering: remove these topics from the candidate topic pool to ensure the novelty of the recommendations.

[0106] Sorting rules: Sorting factors: Utility score: The candidate questions are sorted mainly based on their utility scores.

[0107] Question quality: When the utility scores are the same, questions with higher quality will be recommended first.

[0108] Diversity considerations: In order to prevent the recommended content from being too monotonous, diversity weights can be introduced during sorting to ensure the richness of topic types and content.

[0109] Sorting process: a. Calculate utility score: Calculate the utility score for each candidate question.

[0110] b. Preliminary sorting: sort in descending order based on utility scores.

[0111] c. Quality and diversity adjustment: Based on the initial ranking, the ranking results are adjusted considering the quality and diversity of the questions.

[0112] d. Generate the final recommendation list: Select the first N questions based on the adjusted ranking results.

[0113] in conclusion The present invention collects and analyzes the user's learning behavior in real time, adopts a scientific and effective calculation method, accurately calculates the user's mastery of each knowledge point, and establishes the user's knowledge map. Based on this, the intelligent recommendation algorithm is applied to recommend the most suitable questions for the user's current learning level, fill in the knowledge gap, realize personalized learning, and improve learning efficiency and education quality.

[0114] The method and device of the present invention have the following characteristics and advantages: Wide applicability: can be applied to various online education platforms, learning applications and education management systems.

[0115] Good scalability: The system architecture is flexible and new modules and functions can be added according to needs.

[0116] Real-time and accuracy: Ensure the accuracy and timeliness of the user's knowledge graph through real-time calculation and updating.

[0117] Improve learning experience: Improve users' learning experience and results through personalized topic recommendations and learning path planning.

[0118] The present invention provides an effective technical means for the education industry, helps to realize the personalization and intelligence of education, and has broad application prospects and market value.

Claims

1. A method for improving education quality through intelligent short board analysis of knowledge points, characterized in that: The method comprises the following steps: Build a question bank. Each question in the question bank contains parameters such as the question itself, difficulty coefficient, knowledge point code, and correct answer; Establish a question-answering interface and embed a user behavior recording module to capture the user's behavior data during the question-answering process; Collect and record the user's test-taking behavior data, including the time of entering the question, the time of answering the question, click behavior, user answer, answer result, whether to view the prompt, whether to view the analysis, etc.; Through the real-time data processing system, the user behavior data is processed and calculated to generate a user's knowledge point mastery model, and the user's mastery of knowledge points is accurately calculated based on the preset calculation formula; Update the user's knowledge graph in real time. According to the changes in the user's mastery of knowledge points after each question answering, update the knowledge graph including the overall mastery level and the mastery level at different difficulty levels. Based on the updated knowledge graph, intelligent recommendation algorithms are applied to customize and recommend questions suitable for users' current learning level, so as to fill knowledge gaps, intelligently recommend questions, and learn by analogy.

2. The method according to claim 1, wherein in step 4, the step of calculating the user's mastery of the knowledge point comprises: Using a model based on Item Response Theory (IRT), combined with the user's answer results, question difficulty, expected correct probability and learning rate, the user's ability value for relevant knowledge points is calculated and updated; The updated ability value is converted into a mastery degree, with a value range of 0 to 100, which is used to indicate the user's mastery level of the knowledge point.

3. The method according to claim 1 or 2, wherein in step 6, the intelligent recommendation algorithm comprises: Calculate the priority score of each knowledge point. The priority is determined by the user's mastery of the knowledge point. The lower the mastery of the knowledge point, the higher the priority. Calculate the utility score of each candidate question based on the knowledge point priority, difficulty level priority, and the degree of match between the question difficulty and the user's ability; Filter out the questions that the user has already done through the deduplication rules to form a candidate question pool; Sort the candidate questions by utility score from high to low, and select the top N questions for recommendation.

4. The method according to claim 3, wherein the calculation formula of the utility score is: U q = S k q × S k q , d q × f θ k q , d q β q in: For the topic Utility score of For knowledge points Priority score; For knowledge points At the difficulty level Priority score under It is a matching function that measures the degree of match between the difficulty of the question and the user's ability.

5. The method according to any one of claims 1 to 4, wherein the user behavior data includes the user's answering behavior, clicking behavior, viewing of prompts and analysis for a single question, and whether the user exits the system on the question, etc.

6. A device for improving the quality of education through intelligent short board analysis of knowledge points, characterized in that: include: The question bank construction module is used to create a question bank containing parameters such as the question itself, difficulty coefficient, knowledge point code, and correct answer; The question-answering interface and user behavior recording module are used to provide the user with a question-answering interface and record the user's question-answering behavior data; The data processing module is used to process the user's behavior data in real time, calculate the user's mastery of knowledge points, and update the user's knowledge graph; The intelligent recommendation module is used to recommend suitable topics to users based on the updated knowledge graph and intelligent recommendation algorithm.

7. The device according to claim 6, wherein the data processing module adopts a real-time data processing system composed of Kafka and Flink to perform streaming calculation and processing on the user's behavior data.

8. The device according to claim 6 or 7, wherein the intelligent recommendation module generates a personalized topic recommendation list based on the user's knowledge mastery and performance at each difficulty level, applying utility functions and sorting rules.