Intelligent dynamic scoring multi-question-type question bank management system and method

Abstract

The invention relates to the technical field of education measurement and computer management, in particular to an intelligent dynamic scoring multi-question-type question bank management system and method, and the system comprises a multi-question-type support module, an intelligent difficulty adjustment module, an intelligent scoring module, an identity verification and examination monitoring module, a data analysis and feedback module and a system safety module. According to the system, a hierarchical architecture design is adopted, multiple question types including choice questions, gap filling questions, interpretation questions and combination questions are supported, an intelligent scoring technology and a comprehensive data analysis tool are introduced, the difficulty can be dynamically adjusted, intelligent scoring and real-time feedback are achieved, the examination requirements of different subjects and levels are met, and the examination efficiency is improved. And the examination efficiency and the scoring objectivity are improved.

Description

Technical Field

[0001] The present invention relates to the field of educational measurement and computer management technology, and in particular to an intelligent dynamic scoring multi-question type question bank management system and method. Background Art

[0002] With the continuous advancement of educational technology, the application of question bank management systems in the field of education is becoming increasingly widespread. However, existing question bank management systems still have many shortcomings in terms of functionality, intelligence and user experience. Traditional question bank management systems often only support a limited number of question types, such as multiple-choice questions and fill-in-the-blank questions, which cannot meet the needs of diversified examinations. At the same time, these systems also have obvious shortcomings in intelligent scoring and data analysis, resulting in low scoring efficiency, inaccurate analysis of examination results, and difficulty in providing strong data support for teaching. Especially in the current educational environment, teachers' needs for question bank management systems are becoming increasingly diversified. They not only require the system to support multiple question types, but also hope that the system has functions such as intelligent scoring, real-time feedback, and personalized difficulty adjustment to improve examination efficiency and scoring objectivity. In addition, with the development of big data and artificial intelligence technology, teachers have also put forward higher requirements for the data analysis capabilities of question bank management systems, expecting the system to provide comprehensive data analysis tools to support precise teaching and personalized learning.

[0003] Existing question bank management systems mostly adopt traditional architecture, which has problems such as poor scalability and difficult maintenance. At the same time, there are obvious deficiencies in functionality and intelligence, and they cannot meet the diverse needs of modern education examinations. Specifically, most of these systems only support a single question type, such as multiple-choice questions and fill-in-the-blank questions, and lack support for complex question types such as interpretation questions and combination questions. In terms of intelligent scoring, the existing system functions are limited and cannot provide accurate scoring and real-time feedback. In addition, the data analysis tools are simple and cannot provide teachers with comprehensive test analysis and student performance reports.

[0004] Furthermore, traditional question bank systems are mostly designed with a single functional module, lacking comprehensiveness and intelligence. They usually do not have the ability to dynamically adjust the difficulty and number of questions based on the candidates' answers, nor can they provide a personalized examination experience. In the design and scoring of complex questions, the existing systems have obvious shortcomings and cannot effectively simulate real scenarios or comprehensively examine students' knowledge application capabilities. For example, when handling interpretation questions, they cannot restore the real interpretation environment. In summary, the existing question bank management system needs to be improved in terms of question type support, intelligent scoring and data analysis.

[0005] Therefore, in order to solve the above problems, the present invention proposes an intelligent dynamic scoring multi-question type question bank management system and method. Summary of the Invention

[0006] In order to overcome the problems of single function and low intelligence in the prior art, the present invention proposes an intelligent dynamic scoring multi-question type question bank management system and method.

[0007] The technical solution of the present invention is: an intelligent dynamic scoring multi-question type question bank management system, including:

[0008] Multiple question type support module, used to manage multiple-choice questions, fill-in-the-blank questions, true-or-false questions, subjective questions, interpretation questions, combination questions and attachment questions;

[0009] Intelligent difficulty adjustment module, dynamically adjusting the difficulty of questions based on the test takers' performance;

[0010] Intelligent scoring module, combining speech recognition, semantic analysis and fuzzy matching technology to achieve automatic scoring of all question types;

[0011] Identity verification and exam monitoring module, integrating facial recognition and remote video monitoring technology;

[0012] Data analysis and feedback module, generating multi-dimensional visual reports and supporting export;

[0013] System security module, used to protect the security of data storage, transmission and access.

[0014] Preferably, the intelligent difficulty adjustment module dynamically evaluates the matching relationship between the examinee's ability level and the difficulty of the questions through the Bayesian statistical method, and collects the examinee's answer accuracy and answering time in real time as algorithm input. When the examinee's answer accuracy is higher than the preset threshold and the answering time is short, the difficulty level of subsequent questions is automatically increased, otherwise the difficulty is reduced. At the same time, the module supports teachers to customize adjustment parameters according to actual examination needs, including accuracy threshold, answering time threshold and difficulty grading rules, and dynamically screens questions in the question bank that meet the new difficulty level through algorithms and SQL query statements implemented in the back-end PHP language.

[0015] Preferably, the intelligent scoring module uses differentiated scoring technologies for different question types, and achieves fast and automatic scoring for multiple-choice questions, fill-in-the-blank questions and judgment questions by accurately matching the candidates' answers with preset correct answers. The fill-in-the-blank questions use a fuzzy matching algorithm based on the Python NLP library to support keyword matching and partial correct answer recognition. For interpretation questions, the voice recognition technology is used to transcribe the candidate's interpretation into text, and compare it with the reference translation. A comprehensive score is given based on the four dimensions of pronunciation clarity, intonation accuracy, language fluency and content completeness. For subjective questions, the semantic similarity and keyword coverage of the candidates' answers are analyzed through natural language processing technology, and intelligent scoring suggestions are generated for the teacher's reference. The teacher finally manually confirms the score based on the suggestions.

[0016] Preferably, the interpretation module supports two examination modes: consecutive interpretation and simultaneous interpretation. Candidates receive audio materials through the front-end interface and interpret in real time. The system uses deep learning models or third-party APIs to convert the candidate's voice into text, and then compares it with pre-stored reference translations in multiple dimensions. The scoring algorithm integrates voice feature analysis tools to evaluate pronunciation accuracy, naturalness of intonation, and fluency of speaking speed. At the same time, NLP technology is used to calculate the semantic matching degree between the content and the reference translation, and finally generates a scoring report containing detailed indicators. Candidates can view the scoring results and reference translations in real time and improve weak links in a targeted manner.

[0017] Preferably, the combination question module integrates various question types into a unified scenario case or problem scenario through a dynamic algorithm. The question bank administrator can flexibly design question combination strategies. The system dynamically adjusts the difficulty and question type ratio of subsequent combination questions according to the test settings and the candidates' real-time answering performance. In the scoring stage, the system comprehensively calculates the total score according to the preset weights based on the scores of each sub-question type, and supports teachers to manually adjust the weight distribution. The front end dynamically renders the scenario description and related question types through the Vue framework. Candidates complete answers to multiple question types in a coherent context. The system records the answer data in real time and feedbacks the periodic scores, which not only examines the single mastery of knowledge points, but also evaluates the comprehensive application ability.

[0018] Preferably, the identity authentication and examination monitoring module integrates multimodal identity authentication technology, including a face recognition system based on OpenCV and FaceNet models, which captures the facial features of the examinee through a camera and compares them with pre-stored data. Password verification uses the OAuth 2.0 protocol to support multi-factor authentication such as SMS / email verification codes. The remote proctoring function uses WebRTC technology to achieve real-time video streaming transmission on the examinee side. The proctoring interface built on the teacher side based on the Vue framework can monitor multiple examinee screens at the same time. The abnormal behavior detection algorithm analyzes the video stream in real time, identifies violations and automatically issues alarms.

[0019] Preferably, the data analysis and feedback module automatically aggregates test data through back-end services, uses Python's Pandas library for data cleaning and statistics, and uses the Matplotlib library to generate visual charts to intuitively display the class average score, highest / lowest score and question discrimination. For individual students, the system generates detailed ability maps, marks the strengths and weaknesses of knowledge points, and provides personalized learning suggestions. Teachers can customize report templates and export test analysis reports in PDF or Excel format. It supports multi-dimensional data screening by class, subject and time period, providing data support for teaching adjustments and question bank optimization.

[0020] Preferably, the system security module adopts a layered protection strategy. Sensitive data is encrypted and stored in the MySQL database using the AES-256 algorithm. The HTTPS protocol and SSL / TLS certificate are mandatory for encrypting the communication link during the data transmission phase. The system regularly calls vulnerability scanning tools to detect potential risks, and intercepts SQL injection and XSS attacks through WAF. The RBAC model is implemented for authentication data to limit the operating permissions of administrators, teachers and candidates. At the same time, data is regularly backed up through the Qiniu Cloud backup service, and one-click recovery is supported.

[0021] As a preferred embodiment, the intelligent dynamic scoring multi-question type question bank management method includes the following steps:

[0022] S1, design and implement question bank data structure and dynamic generation algorithm for various question types;

[0023] S2, adjusts the difficulty of questions in real time based on the test takers’ answer data;

[0024] S3, calls the intelligent scoring algorithm to automatically score the answers to all question types;

[0025] S4, ensuring exam fairness through identity verification and remote monitoring;

[0026] S5, analyzes the test data and generates visual reports.

[0027] Preferably, step S3 includes:

[0028] S1, uses fuzzy matching algorithm to support partially correct answers for fill-in-the-blank questions;

[0029] S2, for subjective questions, extracts keywords and semantic similarity using NLP technology to assist teachers in scoring;

[0030] S3, comprehensive scoring of interpretation questions through speech feature analysis and content similarity assessment.

[0031] Beneficial effects of the present invention:

[0032] 1. The multi-question support module enables systematic integrated management of multiple-choice questions, fill-in-the-blank questions, interpretation questions, and combination questions. Each question type is equipped with an independent data structure and scoring logic, solving the problem of traditional question bank systems with a single question type and inability to assess comprehensive application skills. In particular, the innovative interpretation question module supports both consecutive interpretation and simultaneous interpretation modes, and combines voice recognition technology to realistically reproduce interpretation test scenarios, significantly expanding the scope of application of the question bank system.

[0033] 2. Intelligent scoring and real-time feedback functions significantly improve exam efficiency and reduce the time and cost of manual scoring. At the same time, through identity authentication and intelligent scoring technology, the fairness of the exam and the objectivity of the scoring are ensured.

[0034] 3. The present invention can dynamically adjust the difficulty and number of questions to meet the ability levels of different candidates and provide a personalized examination experience. In addition, the design of combination questions and interpretation questions can comprehensively examine students' comprehensive application ability and practical application ability. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 Shown is a schematic diagram of the system structure of the present invention;

[0036] Figure 2 Shown is a schematic diagram of the steps of the intelligent difficulty adjustment algorithm of the present invention. DETAILED DESCRIPTION

[0037] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.

[0038] The present invention provides an embodiment: an intelligent dynamic scoring multi-question type question bank management system and method, including:

[0039] Multiple question type support module, used to manage multiple-choice questions, fill-in-the-blank questions, true-or-false questions, subjective questions, interpretation questions, combination questions and attachment questions;

[0040] Intelligent difficulty adjustment module, dynamically adjusting the difficulty of questions based on the test takers' performance;

[0041] Intelligent scoring module, combining speech recognition, semantic analysis and fuzzy matching technology to achieve automatic scoring of all question types;

[0042] Identity verification and exam monitoring module, integrating facial recognition, fingerprint recognition and remote video monitoring technologies;

[0043] Data analysis and feedback module, generating multi-dimensional visual reports and supporting export;

[0044] System security module, used to protect the security of data storage, transmission and access.

[0045] Further, the specific technical solutions of the present invention are described:

[0046] See also Figure 1The present invention specifically includes the design and implementation of multiple question types, supporting multiple-choice questions, fill-in-the-blank questions, true-or-false questions, etc.; an intelligent difficulty adjustment algorithm, which dynamically adjusts the difficulty of questions in real time to achieve a personalized test experience; an intelligent scoring and feedback mechanism, which combines voice recognition, semantic analysis and fuzzy matching technology to achieve automatic scoring and real-time feedback for all question types, assisting teachers in manual scoring; identity authentication and test monitoring, which integrates face recognition and remote video monitoring technology, and detects abnormal behavior in real time through deep learning to ensure test security and fairness; a data analysis and feedback tool, which conducts multi-dimensional data analysis, generates visual reports and teaching suggestions, and supports PDF / Excel export; and a system security design to ensure the security of data storage, transmission and access.

[0047] The design and implementation of the various question types mentioned above support multiple-choice questions, fill-in-the-blank questions, true-or-false questions, subjective questions, interpretation questions, combination questions and attachment questions. Each question type has its own specific implementation process and technical architecture, including the design of the question bank data structure, algorithms for random selection and dynamic generation of questions, intelligent scoring and auxiliary scoring technology, as well as front-end display and user interaction.

[0048] The multiple-choice questions are the most common question types. The system supports single-choice and multiple-choice questions. The question design includes the question stem, options, and correct answers. The system uses a random algorithm to extract questions from the question bank to ensure that the question combination is different for each exam. The specific implementation steps are as follows:

[0049] In the first step, the question bank administrator enters the multiple-choice question data in the background, including the question stem, options and correct answers. The question bank data structure: a relational database table structure is used to store multiple-choice question data, and the fields include the question stem, options, correct answer, difficulty level, subject, etc.

[0050] In the second step, the system randomly selects questions based on the exam settings and generates the test paper. The random selection algorithm uses a SQL query statement combined with a random function (ORDER BY RAND()) to achieve random selection of questions. At the same time, the system also selects questions that meet the requirements based on the exam settings (such as difficulty level and subject).

[0051] In the third step, candidates select answers on the front-end interface, and the system records the answers in real time. The front-end display uses the Vue framework to build a multiple-choice question component, which supports dynamic rendering of questions and options. After the candidates select the answers, the answers are submitted to the back-end server through AJAX technology.

[0052] Step 4. After submission, the system will automatically score and display the results.

[0053] The fill-in-the-blank questions are used to test students' memory and comprehension of knowledge points. They support single-blank and multiple-blank filling and provide intelligent prompts. The specific implementation steps are as follows:

[0054] The first step is for the question bank administrator to enter the fill-in-the-blank question data, including the question stem and correct answer. The question bank data structure: the fill-in-the-blank question data is stored in a MySQL table, and the fields include the question stem, correct answer (supports multiple or blank answers), difficulty level, subject, etc.

[0055] In the second step, the system dynamically generates fill-in-the-blank questions based on the exam settings. Intelligent prompt algorithm: This intelligent prompt function is implemented based on keyword matching and fuzzy query. When the examinee enters a partial answer, the system matches the keywords in the question bank through SQL fuzzy query (LIKE statement) and returns prompt information.

[0056] In the third step, after the candidates fill in the answers, the system will perform fuzzy matching scoring on the answers (supporting keyword matching). Scoring algorithm: uses a fuzzy matching algorithm, supports keyword matching and partially correct answers, and the scoring results are calculated based on the degree of matching (such as full marks for complete matching, and proportional scores for partial matching).

[0057] In the fourth step, if the candidate has difficulty answering a question, the system will provide keyword prompts. The front-end display uses the Vue framework to build a fill-in-the-blank component, which supports dynamic rendering of the question stem and fill-in-the-blank box. After the candidate enters the answer, the system will provide real-time feedback on the scoring result.

[0058] The judgment questions are used to test students' ability to judge whether knowledge points are correct or not. They support simple judgment questions and complex judgment questions such as logical judgment questions. The specific implementation steps are as follows:

[0059] The first step is for the question bank administrator to enter the true or false question data, including the question stem and correct answer. The question bank data structure: the true or false question data is stored in a MySQL table, and the fields include the question stem, correct answer ("true" or "false"), difficulty level, subject, etc.

[0060] In the second step, the system randomly selects questions and generates a test paper. The random selection algorithm is similar to the multiple-choice questions. Random selection of questions is achieved through SQL query statements combined with random functions.

[0061] In the third step, the candidate selects "correct" or "wrong", and the system automatically scores and displays the results. The scoring algorithm is as follows: simple judgment questions directly match the candidate's answer and the correct answer; complex judgment questions (such as logical judgment questions) use logical operators (AND, OR) to implement scoring logic.

[0062] The subjective questions include short-answer questions and essay questions, which are used to test students' comprehensive analysis ability. The system supports manual scoring by teachers and intelligent scoring assistance. The specific implementation steps are as follows:

[0063] In the first step, the question bank administrator enters the subjective question data, including the question stem and scoring criteria. The subjective question data is stored in a MySQL table, and the fields include question stem, reference answer, scoring criteria, difficulty level, subject, etc.

[0064] In the second step, the candidates enter the answers on the front-end interface.

[0065] In the third step, the system submits the answers to the teacher for manual grading, while also providing intelligent grading assistance (such as keyword matching and semantic analysis). This intelligent grading assistance utilizes natural language processing (NLP) technology, combined with keyword matching and semantic analysis. It also uses Python's NLTK or Spacy libraries to perform text analysis on the test takers' answers, extracting keywords and semantic similarity.

[0066] In the fourth step, the teacher assigns a final grade based on the scoring criteria and intelligent scoring suggestions. The scoring algorithm calculates intelligent scoring suggestions based on keyword matching and semantic similarity. The teacher can manually grade based on the intelligent scoring suggestions, and the final grade is confirmed by the teacher.

[0067] The interpretation questions mentioned above are special questions of this system, used to test the examinee's interpretation ability. The system supports consecutive interpretation and simultaneous interpretation modes, and provides voice recognition and instant scoring functions. The specific implementation steps are as follows:

[0068] In the first step, the question bank administrator enters the interpretation question data, including audio materials and reference translations. The question bank data structure: the interpretation question data is stored in MongoDB, and the fields include audio materials (binary files), reference translations, difficulty level, subject, etc.

[0069] In the second step, the candidate listens to the audio material on the front-end interface and performs interpretation.

[0070] In the third step, the system transcribes the candidate's interpretation into text using speech recognition technology. This technology uses a deep learning model (such as the Google Speech-to-Text API or a custom TensorFlow model) to transcribe the candidate's interpretation into text. The system supports real-time speech recognition, allowing candidates to view the transcript while interpreting.

[0071] In the fourth step, the system instantly scores the candidate's interpretation, including pronunciation, intonation, fluency, and content accuracy. The scoring algorithm combines the speech recognition results with the reference translation to score based on four dimensions: pronunciation, intonation, fluency, and content accuracy. It uses Python speech processing libraries (such as librosa) to analyze speech features and combines natural language processing (NLP) techniques to assess content similarity.

[0072] In the fifth step, candidates can review the scoring results and reference translations and make targeted improvements. The interpretation component is built using the Vue framework, supporting audio playback and voice input. After the candidate completes the interpretation, the system displays the scoring results and reference translations in real time.

[0073] The combination questions mentioned above integrate multiple question types to test students' comprehensive application ability. The system supports multiple combination strategies such as scenario simulation and case analysis. The specific implementation steps are as follows:

[0074] In the first step, the question bank administrator designs combination questions, integrating multiple-choice questions, fill-in-the-blank questions, true-or-false questions, etc. into a scenario or case. The question bank data structure: the combination question data is stored in a MySQL table, and the fields include scenario description, related question types (such as multiple-choice questions, fill-in-the-blank questions, true-or-false questions), difficulty level, subject, etc.

[0075] In the second step, the system dynamically generates combination questions based on the exam settings. The dynamic combination algorithm dynamically adjusts the type and difficulty of the combination questions based on the exam settings and the candidates' responses. The system generates the combination questions in real time through back-end services and pushes them to the candidates.

[0076] In the third step, candidates complete the combination test on the front-end interface. The system dynamically adjusts the difficulty of subsequent questions based on their responses. The scoring algorithm combines the scoring results of each question type to calculate the total score for the combination test. The system supports dynamic adjustment of the difficulty of subsequent questions based on the response.

[0077] Step 4: After submission, the system will comprehensively score and display the results. Front-end display: The combination question component is built using the Vue framework, which supports dynamic rendering of scenario descriptions and related question types. After the candidate completes the combination question, the system will provide real-time feedback on the total score and the score of each question type.

[0078] The attachment question allows teachers to upload documents, pictures, audio or video related to the exam content to enhance the interactivity and practicality of the exam. The specific implementation steps are as follows:

[0079] In the first step, the question bank administrator designs a combination of questions, integrating multiple-choice questions, fill-in-the-blank questions, and true-or-false questions into a scenario or case. Attachment storage: Attachment files are stored in Qiniu Cloud Storage Service, which supports multiple file formats (such as PDF, Word, images, audio, and video).

[0080] In the second step, the system dynamically generates combination questions based on the exam settings. Attachment management: Teachers upload attachments through the backend management system, and the system automatically records the attachment metadata (such as file name, file type, and topic). Attachments and topics are associated with each other and stored in the MySQL database through foreign keys.

[0081] In the third step, candidates complete the combination test on the front-end interface. The system dynamically adjusts the difficulty of subsequent questions based on their responses. Online preview and annotation: The front-end uses the Vue framework combined with third-party libraries (such as PDF.js and Video.js) to implement online preview of attachments. Candidates can annotate on attachments, and the system uses WebRTC technology to achieve real-time annotation.

[0082] Step 4. After submission, the system will comprehensively score and display the results. Security control: All attachments are virus scanned and content reviewed before uploading to prevent the spread of malicious files. Access rights to attachments are strictly controlled, and candidates can only access designated attachments during the examination time.

[0083] See also Figure 2 Furthermore, the intelligent difficulty adjustment algorithm is a dynamic adjustment algorithm based on the test-taker's performance. The algorithm adjusts the difficulty of subsequent questions in real time according to factors such as the test-taker's accuracy rate and answering time. The implementation steps are as follows:

[0084] The first step is to implement the difficulty adjustment algorithm in the back-end server and store the test takers' answer data.

[0085] In the second step, after each question is answered, the algorithm calculates the new difficulty level based on the current answer situation.

[0086] In the third step, the system extracts questions from the question bank according to the new difficulty level and pushes them to the candidates.

[0087] Fourth, the algorithm supports custom parameters (such as accuracy threshold and answering time threshold), and teachers can adjust them according to examination requirements.

[0088] The algorithm logic is: initialize the difficulty level (such as elementary, intermediate, and advanced); calculate the accuracy rate and answering time based on the candidates' answers; if the accuracy rate is higher than the threshold and the answering time is short, increase the difficulty level; if the accuracy rate is lower than the threshold or the answering time is too long, reduce the difficulty level; extract questions from the question bank according to the new difficulty level.

[0089] The technical implementation of the intelligent difficulty adjustment algorithm is as follows:

[0090] Data collection: The system collects the examinee's answer data (such as answering time, whether the answer is correct or not) in real time through the back-end service.

[0091] Algorithm implementation: The dynamic difficulty adjustment algorithm is implemented in the back-end server using PHP. The algorithm is based on Bayesian statistical methods and dynamically calculates the ability level of the examinee and the difficulty of the questions.

[0092] Question bank query: Based on the new difficulty level, eligible questions are extracted from the question bank through SQL query statements. The system supports filtering questions based on multiple dimensions such as difficulty level, subject, and question type.

[0093] Furthermore, the intelligent scoring and feedback mechanism uses an intelligent scoring algorithm to automatically score different question types. For multiple-choice questions, fill-in-the-blank questions, and true-or-false questions, the system scores by matching answers; for interpretation questions and subjective questions, the system scores by combining speech recognition and semantic analysis. The implementation steps are as follows:

[0094] The first step is to implement an intelligent scoring algorithm in the backend server to support scoring logic for various question types.

[0095] In the second step, after the candidates submit their answers, the system calls the scoring algorithm to score.

[0096] In the third step, the scoring results are fed back to the candidates in real time, including scores and scoring details (such as pronunciation scores, content scores, etc.).

[0097] Step 4: For subjective questions, the system provides intelligent scoring suggestions, and teachers can manually score based on the suggestions.

[0098] The algorithm logic is: for multiple-choice questions and true-or-false questions, directly match the candidates' answers with the correct answers; for fill-in-the-blank questions, use a fuzzy matching algorithm to support keyword matching and partially correct answers; for interpretation questions, score based on the speech recognition results and reference translations, taking into account pronunciation, intonation, fluency and content accuracy; for subjective questions, combine intelligent scoring assistance (such as keyword matching, semantic analysis) and manual scoring by teachers.

[0099] Technical implementation method of intelligent scoring and feedback mechanism:

[0100] Scoring of multiple-choice and true-or-false questions: SQL query statements are used to match the candidate's answers with the correct answers and calculate the scores.

[0101] Scoring of fill-in-the-blank questions: Use Python's NLP library (such as NLTK) to implement a fuzzy matching algorithm to calculate the degree of match between the candidate's answer and the correct answer.

[0102] Scoring of interpretation questions: Speech recognition results (using the Google Speech-to-Text API or a custom TensorFlow model) and reference translations are combined to score based on pronunciation, intonation, fluency, and content accuracy.

[0103] Scoring of subjective questions: Use Python's NLTK library or Spacy library to perform text analysis on candidates' answers, extract keywords and semantic similarity, and generate intelligent scoring suggestions.

[0104] Preferably, the identity authentication and examination monitoring adopt multiple identity authentication technologies to ensure the authenticity of the examinee's identity and the legitimacy of the examination, support face recognition, fingerprint recognition and password verification, support remote proctoring function, and teachers can monitor the examinee's examination environment through video to ensure the fairness of the examination.

[0105] The authentication technology implementation method includes:

[0106] Face recognition: Use the OpenCV library combined with deep learning models (such as FaceNet) to implement face recognition. Candidates take facial photos with a camera before the exam, and the system calls the pre-trained deep learning model to extract and compare facial features.

[0107] Fingerprint recognition: Integrates with the hardware fingerprint recognition device, connects to the test terminal via USB interface or Bluetooth, and the system calls the fingerprint recognition SDK of the device to complete fingerprint collection and comparison.

[0108] Password verification: OAuth2.0 protocol is used to implement password verification, and multi-factor authentication is supported (such as SMS verification code, email verification code).

[0109] The remote proctoring technology implementation method includes:

[0110] Video monitoring: The candidate side installs the WebRTC plug-in to support real-time video transmission, and the system realizes real-time communication between the teacher side and the candidate side through the RTP / RTCP protocol.

[0111] Proctoring interface: The teacher uses the Vue framework to build a proctoring interface that supports simultaneous monitoring of multiple candidates. The teacher can view the candidate's video feed, issue warnings, or terminate the exam.

[0112] Anomaly detection: The system uses deep learning algorithms (such as YOLOv5) to analyze video footage in real time and detect abnormal behaviors of candidates (such as leaving their seats, using mobile phones, etc.). When abnormal behavior is detected, the system automatically sends a warning to the teacher.

[0113] Furthermore, the data analysis and feedback tool provides comprehensive data analysis tools, providing teachers with accurate exam analysis and student performance reports. This includes analysis of exam data, student data, and question bank data. Exam data analysis primarily collects statistics on exam score distribution, average scores, highest scores, and lowest scores. Student performance analysis generates detailed reports for each student, including answer status, scores, and knowledge mastery. Question bank data analysis analyzes the difficulty, discrimination, and reliability of questions, providing a basis for question bank optimization.

[0114] The implementation steps are:

[0115] The first step is to implement the data analysis module in the back-end server to support multiple data analysis functions.

[0116] In the second step, the system automatically collects test data after the test and generates an analysis report.

[0117] In the third step, teachers can view and download analysis reports in the background, and support exporting them to PDF or Excel format.

[0118] In the fourth step, the system provides teaching suggestions based on the analysis results to help teachers adjust their teaching methods.

[0119] Technical implementation of data analysis and feedback tools:

[0120] First, the system collects test data in real time through back-end services and stores it in a MySQL database. The data includes candidates' answers, scores, answering time, test progress, etc.

[0121] Second, Python's Pandas and Matplotlib libraries are used to implement data analysis and visualization functions. The system supports the generation of bar charts, line charts, and pie charts to intuitively display the results of test data analysis.

[0122] Third, the system generates reports in PDF or Excel format based on the analysis results, and supports teachers to customize the report content and format.

[0123] Furthermore, the security design includes data encryption and security protection, and identity authentication data management.

[0124] The data encryption and security protection mentioned above uses advanced encryption technology and security measures to ensure the security of candidates' personal information and test data. The technical implementation methods include:

[0125] First, sensitive data (such as candidate information and test records) is encrypted and stored in the database, using the AES-256 encryption algorithm to ensure data confidentiality.

[0126] Second, the system uses the HTTPS protocol for data transmission to ensure the security of the data transmission process. The communication link is encrypted using SSL / TLS certificates.

[0127] Third, the system regularly scans for security vulnerabilities and uses WAF (Web Application Firewall) to prevent common security threats such as SQL injection and XSS attacks.

[0128] The authentication data management mentioned above strictly manages the authentication data to ensure the integrity and availability of the data. The technical implementation methods include:

[0129] First, the system regularly backs up authentication data, uses Qiniu Cloud's backup service to store backup data, and supports data recovery functions to ensure rapid recovery in the event of data loss or damage.

[0130] Second, the access rights to authentication data are strictly controlled and only authorized personnel can access it. The system adopts the RBAC (role-based access control) model to assign different access rights to different roles (such as administrators, teachers, and candidates).

[0131] Furthermore, the question bank management system of the present invention adopts a layered architecture design, including a front-end user interface, a back-end server, a database and a cloud service platform. The front-end user interface provides an interactive interface for candidates and teachers, supports access from multiple devices (such as computers, tablets, and mobile phones), and adopts a responsive design to ensure compatibility on different devices. The back-end server is responsible for processing business logic, including question bank management, examination process control, intelligent scoring, etc., and adopts a microservice architecture to split different functional modules into independent services to improve the scalability and maintainability of the system. The database mainly stores question bank data, candidate information, examination records, scoring results, etc., and adopts a hybrid architecture of relational database (MySQL) and non-relational database (MongoDB) to meet the storage requirements of different data types. The cloud service platform provides data storage, computing resources, identity authentication services, etc. to ensure the high availability and scalability of the system, and adopts AWS (Amazon Web Services) as the cloud service provider, utilizing its Elastic Compute Service (ECS), Simple Storage Service (S3) and Relational Database Service (RDS).

[0132] The technology selection of this invention involves the front-end technology stack, back-end technology stack, database, cloud service platform, and identity authentication technology, as follows:

[0133] Front-end technology stack: HTML5, CSS3, JavaScript (Vue framework). The Vue framework is used to build dynamic user interfaces, supports component-based development, and improves development efficiency and user experience. Back-end technology stack: PHP, with its efficient data processing capabilities and rich ecosystem, ensures application stability and scalability.

[0134] Database: MySQL (relational database), MongoDB (non-relational database). MySQL is used to store structured data, such as candidate information and test records; MongoDB is used to store unstructured data, such as audio files of interpretation questions and scoring records.

[0135] Cloud service platform: Alibaba Cloud ECS provides users with a stable and secure application environment, improves operation and maintenance efficiency, and reduces IT costs. Whether it is building a website, running an application, or conducting big data analysis, ECS can meet your needs.

[0136] Authentication technology: face recognition (based on OpenCV and deep learning model), fingerprint recognition (integrated with hardware devices), password verification (based on OAuth2.0 protocol).

[0137] The present invention provides an embodiment of a multi-question intelligent test paper, specifically:

[0138] This embodiment demonstrates the application of the system in a CET-4 simulation test. Teachers import a multi-question question bank containing listening multiple-choice questions, reading fill-in-the-blank questions, translation subjective questions, and interpretation questions through the backend. The system automatically generates test papers according to the test syllabus weights (listening 30% / reading 40% / translation 20% / interpretation 10%). For interpretation questions, simultaneous interpretation is used. Candidates interpret immediately after hearing the audio. The system converts the speech into text and compares it with the reference translation for semantic similarity. It automatically scores based on pronunciation accuracy (weighted 40%) and content completeness (weighted 60%). At the same time, the dynamic difficulty adjustment module adjusts the difficulty level based on the correct answer rate of the candidate's first 20 questions (if the correct rate is greater than 80%, the difficulty level of the reading questions will be increased). Finally, an analysis report is generated that includes the score rate of each question type and a radar chart of knowledge points and weaknesses. The teacher can export the PDF report for class review.

[0139] The present invention provides an embodiment of a dynamic medical qualification certification examination, specifically:

[0140] In the clinical medicine practice examination scenario, the system combines case analysis questions (attachment questions), drug compatibility multiple-choice questions, and treatment plan judgment questions. Candidates first view the uploaded CT image attachments, answer relevant multiple-choice questions, and then enter the dynamic difficulty adjustment stage. If they answer three pharmacology knowledge questions correctly in a row, the system will automatically push more difficult rare disease treatment plan judgment questions. The intelligent scoring module uses NLTK keyword extraction (such as the matching degree of terms such as "malignant tumor" and "lymph node metastasis") to assist teachers in scoring case analysis questions. After the exam, the system automatically counts the knowledge point mastery rate of each department (such as 72% accuracy rate for internal medicine and 85% accuracy rate for surgery), and generates a physician ability assessment matrix for reference by the review committee.

Claims

1. Intelligent dynamic scoring multi-question type question bank management system, characterized by: Includes: Multiple question type support module, used to manage multiple-choice questions, fill-in-the-blank questions, true-or-false questions, subjective questions, interpretation questions, combination questions and attachment questions; Intelligent difficulty adjustment module, dynamically adjusting the difficulty of questions based on the test takers' performance; Intelligent scoring module, combining speech recognition, semantic analysis and fuzzy matching technology to achieve automatic scoring of all question types; Identity verification and exam monitoring module, integrating facial recognition and remote video monitoring technology; Data analysis and feedback module, generating multi-dimensional visual reports and supporting export; System security module, used to protect the security of data storage, transmission and access.

2. The intelligent dynamic scoring multi-question type question bank management system according to claim 1, characterized in that: The intelligent difficulty adjustment module dynamically evaluates the matching relationship between the examinee's ability level and the difficulty of the questions through the Bayesian statistical method, and collects the examinee's answer accuracy and answering time in real time as algorithm input. When the examinee's answer accuracy is higher than the preset threshold and the answering time is short, the difficulty level of subsequent questions is automatically increased, and vice versa. At the same time, the module supports teachers to customize adjustment parameters according to actual examination needs, including the accuracy threshold, answering time threshold and difficulty grading rules, and dynamically screens questions in the question bank that meet the new difficulty level through algorithms implemented in the back-end PHP language and SQL query statements.

3. The intelligent dynamic scoring multi-question type question bank management system according to claim 1, characterized in that: The intelligent scoring module uses differentiated scoring technologies for different question types. It achieves fast and automatic scoring for multiple-choice questions, fill-in-the-blank questions, and judgment questions by precisely matching the examinee's answers with preset correct answers. For fill-in-the-blank questions, a fuzzy matching algorithm based on the Python NLP library is used to support keyword matching and partial correct answer recognition. For interpretation questions, speech recognition technology is used to transcribe the examinee's interpretation into text, and the text is compared with the reference translation. A comprehensive score is given based on four dimensions: pronunciation clarity, intonation accuracy, language fluency, and content completeness. For subjective questions, natural language processing technology is used to analyze the semantic similarity and keyword coverage of the examinee's answers, and intelligent scoring suggestions are generated for the teacher's reference. The teacher will finally manually confirm the score based on the suggestions.

4. The intelligent dynamic scoring multi-question type question bank management system according to claim 1, characterized in that: The interpretation module supports two examination modes: consecutive interpretation and simultaneous interpretation. Candidates receive audio materials through the front-end interface and interpret in real time. The system uses deep learning models or third-party APIs to convert the candidate's voice into text, and then compares it with pre-stored reference translations in multiple dimensions. The scoring algorithm integrates voice feature analysis tools to evaluate pronunciation accuracy, naturalness of intonation, and fluency of speaking speed. At the same time, NLP technology is used to calculate the semantic match between the content and the reference translation, and finally generates a scoring report containing detailed indicators. Candidates can view the scoring results and reference translations in real time and improve weak links in a targeted manner.

5. The intelligent dynamic scoring multi-question type question bank management system according to claim 1, characterized in that: The combination question module integrates various question types into a unified situational case or problem scenario through a dynamic algorithm. The question bank administrator can flexibly design question combination strategies. The system dynamically adjusts the difficulty and question type ratio of subsequent combination questions based on the test settings and the candidates' real-time answering performance. In the scoring stage, the system calculates the total score according to the preset weights based on the scores of each sub-question type, and supports teachers to manually adjust the weight distribution. The front-end dynamically renders the situation description and related question types through the Vue framework. Candidates complete answers to multiple question types in a coherent context. The system records the answer data in real time and feedbacks the periodic scores, which not only examines the single mastery of knowledge points, but also evaluates the comprehensive application ability.

6. The intelligent dynamic scoring multi-question type question bank management system according to claim 1, characterized in that: The identity authentication and examination monitoring module integrates multimodal identity authentication technology, including a face recognition system based on OpenCV and FaceNet models, which captures the facial features of the examinee through the camera and compares them with pre-stored data. Password verification uses the OAuth 2.0 protocol to support multi-factor authentication such as SMS / email verification codes. The remote proctoring function uses WebRTC technology to realize real-time video streaming transmission on the examinee side. The teacher's proctoring interface built on the Vue framework can monitor multiple examinee screens at the same time. The abnormal behavior detection algorithm analyzes the video stream in real time, identifies violations and automatically issues alarms.

7. The intelligent dynamic scoring multi-question type question bank management system according to claim 1, characterized in that: The data analysis and feedback module automatically aggregates test data through back-end services, uses Python's Pandas library for data cleaning and statistics, and uses the Matplotlib library to generate visual charts to intuitively display the class average score, highest / lowest score, and question discrimination. For individual students, the system generates detailed ability maps, marks the strengths and weaknesses of knowledge points, and provides personalized learning suggestions. Teachers can customize report templates and export test analysis reports in PDF or Excel format. It supports multi-dimensional data filtering by class, subject, and time period, providing data support for teaching adjustments and question bank optimization.

8. The intelligent dynamic scoring multi-question type question bank management system according to claim 1, characterized in that: The system security module adopts a layered protection strategy. Sensitive data is encrypted and stored in the MySQL database using the AES-256 algorithm. The HTTPS protocol and SSL / TLS certificates are mandatory for encrypting the communication link during the data transmission phase. The system regularly calls vulnerability scanning tools to detect potential risks, and intercepts SQL injection and XSS attacks through WAF. The RBAC model is implemented for authentication data to limit the operating permissions of administrators, teachers and candidates. At the same time, data is regularly backed up through the Qiniu Cloud backup service, which supports one-click recovery.

9. An intelligent dynamic scoring multi-question type question bank management method, using the intelligent dynamic scoring multi-question type question bank management system according to claims 1-9, characterized in that: The following steps are included: S1, design and implement question bank data structure and dynamic generation algorithm for various question types; S2, adjusts the difficulty of questions in real time based on the test takers’ answer data; S3, calls the intelligent scoring algorithm to automatically score the answers to all question types; S4, ensuring exam fairness through identity verification and remote monitoring; S5, analyzes the test data and generates visual reports.

10. The intelligent dynamic scoring multi-question type question bank management method according to claim 9, characterized in that: The step S3 comprises: S1, uses fuzzy matching algorithm to support partially correct answers for fill-in-the-blank questions; S2, for subjective questions, extracts keywords and semantic similarity using NLP technology to assist teachers in scoring; S3, comprehensive scoring of interpretation questions through speech feature analysis and content similarity assessment.

Images