Cross-platform teaching tool data collaboration and personalized resource recommendation system based on MCP protocol

Abstract

The invention discloses a cross-platform teaching tool data collaboration and personalized resource recommendation system based on an MCP protocol, and belongs to the technical field of education. According to the system, a multi-platform teaching tool is in unified butt joint through the MCP protocol adaptation layer, and the problem of data splitting is solved; the data co-processing layer cleans and fuses multi-tool data, generates a teacher / student / parent data view and stores an evidence through a block chain; the personalized recommendation engine generates role-based resource recommendation based on a teaching knowledge graph and multi-dimensional behavior analysis; the user interaction layer supports multi-terminal access, and the data security module guarantees compliance. According to the system, the cross-platform data collaboration efficiency is improved by 70% or above, the resource recommendation matching degree reaches 92%, the system is suitable for K12 education, higher education and vocational education scenes, teaching tool data barriers are effectively broken through, and the teaching data utilization rate and the resource recommendation accuracy are improved.

Description

Technical Field

[0001] This invention relates to the field of educational technology and data collaboration, specifically to a cross-platform teaching tool data integration and resource recommendation system based on the MCP (Multi-Platform Collaboration) protocol. By adapting to multiple types of teaching tools through a unified protocol and constructing a multi-role data association model, it realizes cross-platform collaboration of teaching data and personalized resource push, and is applicable to various educational scenarios such as K-12, higher education, and vocational education. Background Technology

[0002] With the digital transformation of education, teachers need to use a variety of teaching tools in the teaching process (such as lesson preparation tools (Seewo Whiteboard), homework grading systems (Xiaoyuan Mental Arithmetic), learning analysis platforms (iFlytek Smart Learning Network), and classroom interaction tools (ClassIn)). However, existing technologies have significant pain points: Cross-platform data fragmentation: Different teaching tools belong to different vendors and have inconsistent data interfaces (e.g., homework data is in CSV format and student learning data is in JSON format). Teachers need to manually export / import data, and cross-tool data integration is time-consuming (average 1.5 hours per teacher per day), and data redundancy or loss is likely to occur. Low collaboration efficiency: The lack of a unified cross-platform collaboration protocol means that data cannot be synchronized between tools in real time (e.g., classroom interaction and answer data can only be synchronized to the learning platform after 24 hours), resulting in delayed teaching decisions; Low accuracy of resource recommendations: Existing recommendations are mostly based on data from a single tool (such as recommending exercises only based on homework errors), without integrating behavioral data from multiple tools (such as lesson preparation focus and classroom interaction feedback), and the matching degree between recommended resources and teaching needs is less than 60%. Data security risks: Cross-platform data transmission uses plaintext transmission or simple encryption, which makes it easy to leak students' privacy data (such as grades and wrong answers), and there is a lack of data operation traceability mechanism, making it impossible to locate the responsibility for data tampering; Insufficient adaptation to the needs of multiple roles: It fails to differentiate the needs of teachers (who need teaching resources), students (who need learning resources), and parents (who need learning progress reports), and the recommended content is homogeneous and cannot meet personalized needs.

[0003] While some existing educational data platforms attempt to integrate tool data, they mostly employ a "one-to-one interface adaptation" approach, resulting in high adaptation costs (developing a new tool requires 7-10 days) and a lack of real-time collaboration based on a unified protocol, thus failing to fundamentally solve the aforementioned problems. Therefore, there is an urgent need to design a cross-platform data collaboration and personalized recommendation system based on a unified protocol to break down data barriers between teaching tools. Summary of the Invention

[0004] This invention aims to address the pain points of data collaboration and resource recommendation in existing teaching tools, and achieves the following technical objectives: Supports cross-platform data collaboration for ≥15 mainstream teaching tools (Seewo Whiteboard, Xiaoyuan Math, Zhixue.com, etc.), with a development cycle of ≤2 days for adapting to new tools; Real-time data synchronization across tools (latency ≤500ms), data integration accuracy ≥99%, and 80% reduction in teacher data processing time; Personalized resource recommendations achieve a matching rate of ≥90% (Teacher's side: teaching resources match the focus of lesson preparation; Student's side: learning resources match the feedback from incorrect answers and classroom interactions). Data transmission is encrypted using the national standard SM4, and operation logs are stored on the blockchain, meeting the compliance requirements of the Personal Information Protection Law and the Guidelines for Educational Data Security. It supports differentiated recommendations for teachers, students, and parents, with a role switching response time of ≤100ms.

[0005] The core of this invention is a four-layer architecture of "MCP protocol adaptation - data collaborative processing - personalized recommendation - multi-terminal interaction". Through unified protocol integration with multiple tools, multi-dimensional data fusion, and role-based recommendation models, it achieves cross-platform teaching data collaboration and accurate resource recommendation. The MCP protocol is the core of this system for enabling cross-tool collaboration. It defines the data transmission format, interface specifications, and synchronization mechanism. The specific design is as follows: 1.1 MCP Protocol Core Specifications Data transmission format: The structure is "header identifier + data body + checksum". The header identifier includes the tool ID (e.g., Seewo Whiteboard ID=001), data type (homework / learning progress / interaction), and timestamp; the data body adopts JSON-LD format, supporting both structured (grades) and unstructured (lesson preparation materials links) data; the checksum uses SHA-256 to ensure data integrity. Interface Specification: Defines 3 core interfaces to adapt to different tool functions: Data upload interface ( / mcp / upload): The tool uploads data (such as homework grading results) to the system, supporting batch (≤1000 records per batch) and incremental (upload only changed data) uploads; Data pull interface ( / mcp / pull): The system pulls data (such as classroom interaction and answer records) from the tool, and supports filtering by time range (such as the last hour) and data type; Synchronization notification interface ( / mcp / notify): After data synchronization is complete, the system sends a notification to the tool to ensure that the tool updates the data in real time; Synchronization mechanism: Supports "real-time synchronization" (classroom interaction data, latency ≤500ms) and "scheduled synchronization" (homework data, once every 30 minutes). The synchronization strategy can be configured by the teacher in the interface.

[0006] 1.2 Tool Interface Adaptation Process When adding new teaching tools, only a "tool interface adapter" (based on the MCP protocol SDK) needs to be developed; no modification to the tool's native code is required. The adapter reads native data from the tool (such as Excel-formatted homework data from Xiaoyuan Math). The data format converter transforms raw data into the MCP protocol standard JSON-LD format; The protocol parsing unit verifies the data header identifier and checksum, and then passes the data collaborative processing layer. Adaptation Case: Key lesson preparation data from Seewo Whiteboard (ID=001), converted using the adapter, in MCP protocol data format example: { "header": {"toolId": "001", "dataType": "lesson_prep", "timestamp":"2025-10-05 09:30:00"}, "dataBody": {"teacherId": "T001", "course": "mathematics", "keyPoints": ["Solving quadratic equations", "Application of Vieta's formulas"]}, "checkCode": "a3b7c9d2e8f1..."} The multi-tool data adapted to the MCP protocol is cleaned and merged to generate a multi-role data view: 2.1 Data Cleaning Unit Addressing three types of data issues: Redundant data: Delete duplicate records (such as two assignment submissions by the same student), and keep the latest one; Error data: Correct formatting errors (e.g., score "95.5" should be corrected to "95.5"), logical errors (e.g., scores > 100 should be marked as invalid and require teacher confirmation); Missing data: Use "fill with similar data" (e.g., if student age is missing, fill with the average age of the class) or "mark as to be filled" (e.g., if homework grading results are missing, mark as "not graded"). The cleaned data is stored in a distributed database (HBase) and partitioned by "Teacher ID - Tool ID - Time" for easy and fast querying.

[0007] 2.2 Multi-role data fusion module Based on the "User ID" and associated with data from multiple tools, three types of role data views are generated: Teacher View: Integrates key points of lesson preparation (Seewo Whiteboard), homework error rate (Xiaoyuan Math), and classroom interaction accuracy rate (ClassIn) to generate a "Teaching and Learning Overview" (e.g., the key points of lesson preparation for "Quadratic Equations" correspond to a homework error rate of 35%). Student View: Integrates homework errors (Xiaoyuan Math), classroom answer records (ClassIn), and learning analysis reports (Zhixue.com) to generate a "personal learning portfolio" (e.g., "60% of the questions are wrong in algebra, so we need to strengthen our equation solving skills"). Parent View: Integrates student performance trends (ZhiXue.com), homework completion status (XiaoYuan KouSuan), and classroom focus (ClassIn) to generate "Learning Reports" (such as "This week's math homework accuracy improved by 15%"). The data fusion uses a "weighted association algorithm". For example, the association weight between students' "equation mistakes" and teachers' "equation lesson preparation focus" is 0.8, and the association weight between them and "geometry lesson preparation focus" is 0.2.

[0008] 2.3 Blockchain Evidence Storage Unit A consortium blockchain (Fabric) is used to store data operation logs (such as teachers exporting learning data and system-synchronized homework data). The logs include the operator, operation time, data type, and operation result. Each log block generates a unique hash value, which is associated with the hash of the previous block to ensure that the log is immutable; administrators can query the log through the blockchain explorer to locate the responsibility for data operations.

[0009] Based on multi-role data views and teaching knowledge graphs, differentiated resource recommendations are generated: 3.1 Construction of Instructional Knowledge Graph Using subject-specific knowledge points as the core (such as the mathematical concept of "quadratic equations in one variable"), a three-level graph structure is constructed: First-level node: Subject (Mathematics, Chinese); Second-level node: Chapter (Junior High School Mathematics Grade 8, First Semester); Level 3 nodes: Knowledge points (definition, solution, and application of quadratic equations in one variable); The graph links teaching resources (courseware, exercises, videos). For example, "Solving quadratic equations in one variable" links 3 courseware, 20 exercises, and 1 explanatory video. Resource tags include difficulty (easy / medium / difficult) and applicable scenarios (lesson preparation / practice / review).

[0010] 3.2 Multidimensional Behavioral Analysis Extract key behavioral features for different roles: Teachers should focus on the following aspects: lesson preparation (e.g., frequently reviewing the "Vieta's Formulas" slides), homework error rate (e.g., 40% error rate for "Applications of Quadratic Equations"), and classroom interaction topics (e.g., asking questions in class primarily about "solutions of equations"). Students: Types of incorrect questions (e.g., "completing the square" method accounts for 70% of incorrect questions), time spent answering questions (e.g., "word problems" take an average of more than 5 minutes), and resource click records (e.g., frequent clicks on "equation video explanations"). Parents: Check key learning points (e.g., check math scores 3 times a week) and pay attention to problems (e.g., repeatedly ask "how to improve the accuracy of equations"); The algorithm is used to calculate the weight of behavioral features. For example, the weight of "key points of lesson preparation - Vieta's formulas" is 0.7, and the weight of "homework error rate - application of equations" is 0.3.

[0011] 3.3 Recommended Strategy Scheduling Different recommendation strategies are deployed based on role and behavioral characteristics: For teachers: a dual approach of "lesson preparation focus + homework error rate". For example, if a teacher's lesson preparation focus is "Vieta's formulas" and the homework error rate for this knowledge point is 35%, it is recommended to use "Vieta's formulas extended courseware" + "collection of common mistakes". For students: Driven by "type of incorrect questions + time spent answering questions". For example, if a student has many incorrect questions on the "completing the square method" and spends a lot of time answering questions, we recommend "step-by-step explanation video of the completing the square method" + "basic practice questions". For parents: Driven by "learning focus + issues to be addressed", for example, if parents are concerned about math scores and inquire about equation learning, we recommend "Family Tutoring Equation Practice Plan" + "Grade Improvement Guide"; Recommendation results are sorted by "Matching Degree = Behavioral Feature Weight × Resource Tag Relevance" (Matching Degree = Behavioral Feature Weight × Resource Tag Relevance), with 10 resources displayed per page. It supports "Refresh" (recalculate recommendations) and "Favorite" (mark as frequently used resources).

[0012] 4.1 Data Security Module Data transmission: Encryption is performed using the national standard SM4 (key is updated every 24 hours) to prevent data leakage during transmission; Access permissions: Permissions are set based on the RBAC model (teachers can view data of students in their own class, parents can only view data of their own child, and administrators can view all data). Permission changes require approval (e.g., if a teacher requests to view data of other classes, it requires the principal's approval). Data anonymization: Student names, student IDs and other private data are anonymized (e.g., "Zhang San" is displayed as "Zhang*", "2023001" is displayed as "2023***"), and only administrators can view the complete data.

[0013] 4.2 User Interaction Layer Design Teacher side: Provides "Data Overview" (cross-tool data visualization charts), "Resource Recommendation" (sorted by lesson preparation progress), and "Data Export" (supports Excel / PDF formats); Student side: Provides functions such as "Personal Error Notebook" (integrating error questions from multiple tools), "Recommended Resources" (sorted by learning progress), and "Learning Report" (weekly / monthly learning summary); Parent App: Offers features such as "Learning Progress Briefing" (text and image version), "Resource Recommendation" (family tutoring resources), and "Teacher Communication" (one-click contact with the homeroom teacher); It supports access via Web, APP (iOS / Android), and mini-program, with real-time data synchronization across multiple platforms (e.g., resources saved in the APP can be viewed on the Web). Detailed Implementation

[0015] Application Background: A middle school math teacher (ID=T001) needs to use four tools: Seewo Whiteboard (lesson preparation), Xiaoyuan Mental Arithmetic (homework correction), ClassIn (classroom interaction), and Zhixue.com (learning analysis). The teacher needs to address the issues of "time-consuming cross-tool data integration" and "resource recommendation." MCP Protocol Adaptation: Adapted to the above 4 tools. Teachers can enable "Real-time Synchronization" (classroom interaction data) and "Scheduled Synchronization" (homework data, once every 30 minutes) in the system settings. Knowledge Graph: A knowledge graph for eighth-grade mathematics (including 12 chapters, 48 ​​knowledge points, and 200+ related resources); Recommended strategy: Teachers should be driven by "lesson preparation focus + homework error rate", while students should be driven by "type of wrong questions + time spent answering questions".

[0016] Implementation process 3.1 Data Collaboration Process Teachers use Seewo whiteboard to prepare lessons, highlighting "solutions to quadratic equations in one variable," and the data is synchronized to the system after being converted by the MCP protocol adaptation layer. Students completed the equations assignment on Xiaoyuan Mental Arithmetic (50 students submitted, error rate 36%). The assignment data was synchronized to the system on a regular basis. The data cleaning unit deleted 2 duplicate submission records and corrected 1 grade format error ("98 points" was corrected to "98"). The system integrates key points of lesson preparation (Seewo Whiteboard) with homework error rates (Xiaoyuan Math) to generate a teacher view: The key point of lesson preparation, "Solving Quadratic Equations," corresponds to a homework error rate of 36%, indicating that teaching needs to be strengthened. In class, teachers initiated interactive equation-solving activities through ClassIn (with a 64% accuracy rate), and the data was synchronized to the system in real time, updating the teacher's view and students' personal profiles.

[0017] 3.2 Personalized Recommendation Process Teacher recommendations: Based on "Lesson preparation focus - equation solving" (weight 0.6) and "homework error rate -36%" (weight 0.4), the system recommends "step-by-step courseware on equation solving" (matching degree 94%) and "10 common mistakes" (matching degree 92%). Student Recommendations: In Student A's (ID=S001) homework, 80% of the questions on "Complete the Formula" were incorrect, and the time spent answering the questions exceeded 6 minutes. The system recommends "Complete the Formula Explanation Video" (93% match) and "5 Basic Completion Practice Questions" (90% match). Parent Recommendation: When Student A's parent views the learning progress report, the system recommends a "Family Equation Practice Plan" (88% match) based on "Parents' Concern for Math Grades" (weight 0.7) and "Consultation on Equation Learning" (weight 0.3). Beneficial effects

[0018] Compared with the prior art, the present invention has the following significant advantages: Cross-platform collaboration efficiency is greatly improved: Based on the MCP protocol, it achieves unified adaptation of ≥15 teaching tools, with the adaptation cycle of new tools ≤2 days, cross-tool data synchronization latency ≤500ms, and teacher data processing time reduced by 80%; High accuracy of resource recommendations: Integrating data from multiple tools and knowledge graphs, role-based recommendations have a matching rate of ≥90%, and resources on the teacher's end are highly consistent with teaching needs, while resources on the student's end are highly consistent with learning weaknesses; Data security compliance: It adopts the national cryptographic standard SM4 encryption and blockchain notarization to meet the requirements of education data security regulations. The privacy data anonymization rate is 100%, and data tampering is traceable. Highly adaptable to multiple roles: It differentiates the needs of teachers, students, and parents, provides personalized data views and resource recommendations, and offers real-time synchronization across multiple platforms (Web / APP / Mini Program), making it highly user-friendly; It has good scalability: it supports the addition of new teaching tools (quickly adapted through the MCP protocol SDK) and new educational scenarios (such as adding "practical training tools" for vocational education) without reconstructing the core system architecture.

Claims

1. A cross-platform teaching tool data collaboration and personalized resource recommendation system based on the MCP protocol, characterized in that, include: The MCP protocol adaptation layer is used to uniformly connect to teaching tools on multiple platforms. It reads the native data of the tool through the tool interface adapter, verifies the legality of the data through the protocol parsing unit, converts the data format to the MCP protocol standard format (header identifier + data body + SHA-256 checksum) through the data format converter, and realizes cross-tool data transmission based on the "real-time synchronization / timed synchronization" mechanism. The data collaboration processing layer receives standardized data output from the MCP protocol adaptation layer and processes redundant / erroneous / missing data through the data cleaning unit. The multi-role data fusion module generates teacher / student / parent data views based on "user ID" and associates data from multiple tools. The blockchain notarization unit uses a consortium blockchain to store data operation logs. The personalized recommendation engine generates a role-based resource recommendation list based on the teaching knowledge graph (three-level nodes: subject-chapter-knowledge point) and multi-dimensional behavioral analysis (teacher's lesson preparation focus / homework error rate, student's wrong question type / answer time, and parents' focus). The user interaction layer provides a visual interface for teachers (data overview, resource recommendations, data export), students (error notebook, learning report), and parents (learning progress report), supporting access from multiple platforms including Web / APP / Mini Program; The data security module is used to encrypt data transmission using the national cryptographic standard SM4, control access permissions based on the RBAC model, and perform anonymization processing on privacy data.

2. The system according to claim 1, characterized in that, The protocol specifications of the MCP protocol adaptation layer include: In the data transmission format, the header identifier includes a tool ID (2-digit code), data type (2-digit code: 01 = homework, 02 = learning situation, 03 = interaction, 04 = lesson preparation), and timestamp (accurate to milliseconds); the data body adopts JSON-LD format, supporting structured and unstructured data; the checksum is generated using the SHA-256 algorithm; the core interfaces include a data upload interface ( / mcp / upload, supporting batch / incremental upload), a data pull interface ( / mcp / pull, supporting time range / data type filtering), and a synchronization notification interface ( / mcp / notify, sending HTTP / HTTPS notifications after synchronization is completed); the synchronization mechanism supports "real-time synchronization" (delay ≤500ms, suitable for classroom interaction data) and "scheduled synchronization" (period configurable from 30 minutes to 24 hours, suitable for homework / learning situation data).

3. The system according to claim 1, characterized in that, The multi-role data fusion module of the data collaborative processing layer includes: a teacher view that integrates lesson preparation tool data (such as Seewo Whiteboard), homework tool data (such as Xiaoyuan Mental Arithmetic), and interactive tool data (such as ClassIn) to generate an overview of teaching and learning; a student view that integrates homework errors, classroom answer records, and learning analysis reports to generate a personal learning profile; and a parent view that integrates student performance trends, homework completion status, and classroom focus to generate a learning summary. Data fusion employs a weighted association algorithm, where the association weight of relevant data is ≥0.8, and the association weight of irrelevant data is ≤0.

2.

4. The system according to claim 1, characterized in that, The personalized recommendation engine employs the following recommendation strategies: For teachers, a dual-drive approach is used, prioritizing "lesson preparation focus (weight 0.6) + homework error rate (weight 0.4)" to recommend teaching resources such as courseware, exercises, and interactive question banks. For students, a dual-drive approach is used, prioritizing "error type (weight 0.7) + answering time (weight 0.3)" to recommend learning resources such as explanation videos, practice questions, and error correction exercises. For parents, a dual-drive approach is used, prioritizing "learning progress (weight 0.7) + concerns (weight 0.3)" to recommend resources such as family tutoring plans and tutoring guides. Recommendation results are sorted in descending order based on "matching degree = behavioral feature weight × resource tag relevance," and support "refresh" and "favorite" operations.

5. The system according to claim 1, characterized in that, The data security module employs the national standard SM4 symmetric encryption algorithm for data transmission, with the key automatically updated every 24 hours. Access permissions are set based on the RBAC model, allowing teachers to view only the data of their own students, and parents to view only their own child's data. Changes to permissions require approval and record-keeping. Privacy data (name, student ID, contact information) is anonymized using "surname + *", "partially hidden code", and "middle four digits hidden," ensuring that only administrators can view the complete data.

6. The system according to claim 1, characterized in that, The system supports compatibility with ≥15 mainstream teaching tools (Seewo Whiteboard, Xiaoyuan Mental Arithmetic, ClassIn, Zhixue.com, Xuexitong, etc.). New tools can be developed through the MCP protocol SDK development tool interface adapter, with a development cycle of ≤2 days and no need to modify the tool's native code.

7. The system according to claim 1, characterized in that, The accuracy of cross-tool data integration is ≥99%, the matching degree of personalized resource recommendation is ≥90%, the data processing latency is ≤500ms, and the role switching response time is ≤100ms, meeting the compliance requirements of the "Personal Information Protection Law of the People's Republic of China" and the "Education Data Security Guidelines".