A cross-disciplinary teaching motivation method and system

Abstract

The application discloses a kind of interdisciplinary teaching incentive method and system, solve the problem that existing teaching incentive system is difficult to objectively measure the interdisciplinary improvement of existing works, incentive distribution is not transparent and lacks credible evidence;When the original creator stores the work, the system extracts multi-modal features to generate multi-modal feature vectors, and establishes an unforgeable genesis relay record, the challenger needs to complete the pre-challenge learning to obtain the certificate, enter the double queue, after submitting the improved work, according to the improvement type, it is automatically routed to the verifier for verification, according to the improvement amplitude, the number of inheritance and multi-disciplinary calculation red, green and blue star level, the improvement contribution value, the work comprehensive star value and the inheritance points are distributed, all relay logs are stored through hash chain and three-level snapshot, and reverse recommendation and hot discipline push are supported;The application realizes the objective quantification, transparent incentive and whole-process traceability of interdisciplinary improvement, effectively stimulates the students' continuous innovation motivation for existing works.

Description

Technical Field

[0001] This invention relates to the field of educational informatization technology, specifically to an interdisciplinary teaching incentive method and system. Background Technology

[0002] In engineering education and interdisciplinary innovation practices in higher education institutions, encouraging students to continuously improve their existing works is an important way to cultivate their innovative abilities. Currently, there are several online learning incentive systems and experimental teaching platforms. For example, patent application CN119919256A discloses an online teaching incentive system that classifies and statistically analyzes learning record data to generate target learning data for incentives and formulates reward schemes; patent application CN112637271A discloses an open experimental teaching platform based on the Internet of Things, which achieves remote management and sharing of experimental resources by combining offline experimental environments with cloud servers; and patent application CN108876212A discloses an experimental instrument sharing service platform that provides sharing modules for experimental instruments, research software, and researcher information, as well as a research training platform.

[0003] However, the aforementioned existing technologies mainly incentivize the use of learning behavior data (such as course completion and test scores) or the reservation of experimental resources, and fail to effectively address the incentive problem in the interdisciplinary improvement process based on existing works.

[0004] Therefore, there is an urgent need for a method that can incentivize interdisciplinary improvements to existing works. Summary of the Invention

[0005] In view of the shortcomings of the prior art, the present invention aims to provide an interdisciplinary teaching incentive method and system to optimize the incentive of existing works.

[0006] To solve the above problems, the present invention adopts the following technical solution: On the one hand, this invention provides an interdisciplinary teaching incentive method, including: The creator or manager stores physical works in the smart cabinet and / or uploads digital works to the works library; Genesis relay records are generated by extracting multimodal features of physical and / or digital works and anchoring the hash values ​​to a blockchain or a third-party evidence storage platform. Challengers obtain relevant data on digital works by initiating challenges and / or obtain physical works according to the challenge order. Challengers optimize their works and store the physical works and improved physical works in the smart cabinet and / or upload the improved digital works and initiate work review. Review the improvements made to physical works and / or digital works, and calculate the challenger's contribution value and overall star rating for the work; The works are displayed and sorted based on their overall star rating. For each work, challenger information and contribution value for improvement are listed based on the challenge completion time, and a relay log is generated.

[0007] On the other hand, the present invention provides an interdisciplinary teaching incentive system, including a work storage module, a feature extraction and evidence storage module, a challenge module, an audit module, and a display module; The work storage module is used to provide smart cabinets for creators or managers to store physical works and / or upload digital works to the work library; The feature extraction and evidence storage module is used to extract multimodal features of physical and / or digital works to generate a genesis relay record, and anchor the hash value to a blockchain or a third-party evidence storage platform. The challenge module is used to obtain relevant data of digital works based on the challenger's initiation of the challenge and / or obtain physical works according to the challenge order. The challenger optimizes the works and stores the physical works and improved physical works in the smart cabinet and / or uploads the improved digital works and initiates the work review. The review module is used to review the improved physical works and / or improved digital works, and to calculate the challenger's improvement contribution value and the overall star value of the work; The display module is used to display and sort works based on their comprehensive star rating. For each work, challenger information and contribution value to the improvement of the work are listed based on the challenge completion time, and a relay log is generated.

[0008] The beneficial effects of this invention are as follows: Through an incentive model combining red, green, and blue three-dimensional star ratings, evolutionary contribution values, and inheritance points, each substantial improvement made by students receives quantitative recognition, forming a long-term, sustainable innovation relay mechanism. This greatly motivates students to repeatedly refine and iterate their work and solves the problem of technical works being forgotten within the university. It promotes interdisciplinary collaboration and integration by routing different improvement types (software, mechanics, design, law, etc.) to the corresponding discipline's faculty or student review panels, enabling students from engineering, humanities, and arts majors to contribute their expertise around the same physical work. This breaks down traditional disciplinary barriers, promotes the development of campus research works into mature products, and cultivates students' multifaceted thinking and teamwork abilities. Furthermore, it fosters a positive campus innovation culture, creating a visible role model effect and effectively encouraging more students to actively participate in interdisciplinary practical activities. Attached Figure Description

[0009] Figure 1 This is a flowchart of an interdisciplinary teaching incentive method.

[0010] Figure 2 A flowchart for calculating the challenger's legacy points.

[0011] Figure 3A flowchart for personalized recommendations and trending topic pushes.

[0012] Figure 4 This is a schematic diagram of an interdisciplinary teaching incentive system. Detailed Implementation

[0013] The present invention will be further described in detail below with reference to specific embodiments.

[0014] It should be noted that these embodiments are only used to illustrate the present invention and are not intended to limit the present invention. Simple improvements to the method under the premise of the present invention are all within the scope of protection claimed by the present invention.

[0015] See Figure 1 This is an interdisciplinary teaching motivation method, including: S100: The creator or manager stores physical works in the smart cabinet and uploads digital works to the works library.

[0016] In addition, the creator or administrator fills in the work category and baseline statement based on the physical and / or digital works.

[0017] The smart cabinets in this embodiment can be deployed in key areas of the campus, such as libraries, teaching buildings, laboratory buildings, engineering training centers, canteens, and dormitory lobbies. Submitted works can be graduation theses, journal articles, experimental reports, research projects, and / or related physical works.

[0018] Each compartment integrates the following components: Electronic lock: An electromagnetic lock controlled by an edge controller to ensure the secure access to items. It can only be opened when authorized by the system (such as when a student scans a code to verify their identity) or when forced by an administrator.

[0019] Camera module: It has a built-in high-definition wide-angle camera, which is used to take photos or short videos of the items in the cabinet at regular intervals during the storage, retrieval and challenge of items, as a status record and handover certificate.

[0020] Status sensor group: RFID reader: Used to read passive RFID tags affixed to physical artworks, uniquely identifying the artwork's identity.

[0021] Weight sensor: A high-precision strain gauge sensor used to detect changes in the weight of items inside the cabinet, helping to determine whether items have been taken or returned.

[0022] Programmable microcontroller: As the local execution unit of the cabinet, it receives instructions from the edge controller, controls the electronic lock, LED light strip and speaker, and collects sensor data. It can also execute programs for improved designs and drive actuators to produce physical actions (such as motor rotation, LED flashing, voice broadcasting), and display the execution results in real time on the display screen and / or the large screen of the cabinet cluster.

[0023] Actuator: LED light strip: surrounds the inside of the cabinet door, using different colors (such as green for vacancy, red for occupation, and blue for challenge) and flashing modes to indicate the status.

[0024] Display screen / touch screen: Installed above the cabinet door or integrated cabinet compartment, used to display QR codes, queue information, challenge progress, and work display content (such as images and videos).

[0025] Speaker: Used to emit prompts, such as "successful scan," "challenge started," and "timeout reminder."

[0026] Cabinet main control board: Each cabinet (containing multiple compartments) is equipped with a high-performance main control board running a lightweight Linux system. This main control board handles edge computing tasks, avoiding the transmission of all data to the cloud and reducing latency and bandwidth pressure.

[0027] The creator or manager places the physical artwork into a designated, empty smart cabinet. After the cabinet door is closed, the smart cabinet performs the following: Sensor confirmation: The weight sensor reading is stable within a certain range, and the RFID reader detects the RFID tag affixed by the user to the artwork (the tag ID was bound to the artwork ID when the user submitted the artwork).

[0028] Taking the initial photo: The camera inside the cabinet automatically takes a high-resolution photo as visual proof of the artwork's initial state. The photo is uploaded to object storage, and its hash value is also recorded in an extended field of the genesis record.

[0029] Generate and print the QR code: Generate a unique URL (pointing to the artwork details page) for this cabinet-artwork binding relationship and convert it into a QR code. The QR code will then be displayed on the cabinet door screen. Subsequent challengers can scan this QR code to obtain information about the artwork.

[0030] Display Status: The LED light strip in the cabinet switches to a solid green (or other configurable status indicating "idle - available for challenge"), indicating that the work is available for other students to challenge; the display status of online works is shown on the screen.

[0031] When an creator or administrator intends to store or upload physical and / or digital works, they can select or fill in the work category using the set category selector. Categories are predefined enumeration values ​​covering major subject areas, including at least: Code-based applications: Suitable for embedded programs, desktop applications, web services, algorithm implementations, etc.

[0032] Image Design Category: Suitable for graphic design, UI / UX design, 3D model renderings, photography, etc.

[0033] Chemical Experiment Reports: Suitable for experimental reports on organic / inorganic synthesis, analysis and detection, and materials preparation.

[0034] Mechanical structure category: applicable to mechanism design, component CAD models, assemblies, etc.

[0035] The creator selects one or more categories that best match the core attributes of the work.

[0036] The baseline statement includes: Performance metric name: Use a precise term to describe a core capability of the work. For example, for a line-following car, the metric could be "maximum line-following speed"; for a chemical synthesis experiment, the metric could be "target product yield"; and for a robotic arm, the metric could be "end-point positioning accuracy".

[0037] Performance baseline value: The actual measured value of this indicator on the current work, with standard units (such as m / s, %, MPa, ms).

[0038] Performance indicator measurement method: Describe in detail how this current value was obtained. It must be specific to the extent of reproducibility. For example, "On a standard black tape tracking path, use a laser speedometer to record the average speed of the car traveling 2 meters, repeat the measurement 10 times, remove the maximum and minimum values, and take the average."

[0039] Performance measurement conditions: Record all environmental and equipment conditions that may affect the measurement results. For example, "room temperature 25°C, humidity 50%, battery voltage stable at 7.4V, and track material is matte PVC".

[0040] Core Knowledge Point Tags: Set up a subject-specific knowledge base (e.g., mechanical principles, automatic control, organic chemistry, computer graphics). The creator or administrator needs to select 3 to 5 tags most relevant to the work from the knowledge base, such as "PID control algorithm," "infrared sensor array," "esterification reaction mechanism," and "finite element static analysis." These tags will be used for subsequent knowledge point testing and reverse recommendation.

[0041] The materials for digital works may include: Code files: Accepts source code files such as .c, .py, .ino, .js, etc. The system will automatically analyze their structure.

[0042] Drawing files: Accepts CAD or engineering drawing files such as .stp, .igs, .pdf, and .dwg.

[0043] Media files: Accepts demo videos in .mp4 format and key photos in .jpg / .png format.

[0044] Data Recording: We accept experimental data tables in .xlsx or .csv format, or original experimental records scanned into PDF format on this page.

[0045] S200: Extract multimodal features of physical and / or digital works to generate a genesis relay record, and anchor the hash value to a blockchain or a third-party evidence storage platform.

[0046] Based on the category of the work, corresponding multimodal features are extracted, and a unified feature vector is generated.

[0047] This embodiment provides a feasible method for multimodal feature extraction, including: Feature extraction for code-based works: Basic characteristics: code file size (KB), number of valid lines of code (excluding blank lines and comments), total number of functions or methods, maximum nesting depth of conditional branches (if-else / switch), whether it contains blocking delay functions (such as delay()), whether it contains floating-point arithmetic, and whether it contains serial port or network communication instructions.

[0048] Structural features: A language-independent parser is invoked to parse the code into an abstract syntax tree. Five structural features are extracted from the tree structure: the total number of conditional branch nodes in the abstract syntax tree, the maximum nesting depth of branch nodes, the total number of loop nodes (for / while), the number of different nodes in the function call graph, and the total number of arithmetic operator nodes.

[0049] A polynomial extension is performed to compute all second-order interaction terms of the basic and structural features (e.g., feature A multiplied by feature B), resulting in a 52-dimensional feature vector. A pre-trained principal component analysis projection matrix on code samples is then used to reduce this 52-dimensional vector to 64 dimensions. This projection matrix is ​​computed offline and stored in the system.

[0050] Feature extraction for graphic design works: Load a streamlined convolutional neural network model (such as the first three layers of MobileNetV2) to extract global features from the image. Perform global average pooling on the feature map output by the network to obtain a 32-dimensional vector that encodes the image's color distribution, texture pattern, and edge density.

[0051] Calculate handcrafted design features, including image information entropy, the ratio of compressed file size to original file size, edge density detected using the Canny operator, number of corners detected using the Shi-Tomasi algorithm, and horizontal and vertical symmetry indices of the image.

[0052] The ORB algorithm is used to detect keypoints in an image. The number of keypoints, the distribution density of keypoints on the image plane (measured by the variance of nearest neighbor distance), and K-Means clustering (e.g., K=8) are extracted from the ORB descriptors of all keypoints. The distances from each sample to its cluster center are summed to form a 16-dimensional local feature vector.

[0053] The above features are directly concatenated to form the final multimodal feature vector of the work, which is unified into 64-dimensional features.

[0054] Feature extraction for works involving chemical experiments: From the uploaded experimental protocol PDF or text, natural language processing technology is used to automatically identify and count text features: the number of experimental steps (defined by numerical codes or sequential conjunctions), the total number of reagents, the number of reaction condition parameters (such as temperature, pressure, pH, and light), the number of special equipment (such as rotary evaporators and gas chromatographs), and the number of safety protection measures mentioned in the document (such as goggles and fume hoods).

[0055] Analyze the results section of the experimental report to check if it contains quantifiable output data. If so, calculate the statistical characteristics of this data: mean, standard deviation, range, and data completeness (whether parallel samples and error bars are included), and give a score.

[0056] A safety keyword dictionary is constructed, mapping keywords such as "flammable," "explosive," "highly toxic," and "corrosive" to risk levels 1-5. Experimental protocols are scanned, keywords are matched, and risk scores are accumulated. The final risk level (1-5) is represented as a 5-dimensional one-hot encoding. There are four such level dimensions (corresponding to different risk types), resulting in a final 20-dimensional safety risk feature.

[0057] The above features are combined and unified to 64 dimensions.

[0058] Feature extraction for mechanical structure works: Use a CAD file parsing library (such as Open CASCADE) to read the 3D model file. Extract the analytical features: the total number of individual parts, the hierarchy depth of the assembly (number of recursive nesting levels), the ratio of the total area of ​​all surfaces to the volume of the part's bounding box, the total number of holes, the total number of chamfers, and the total number of arrays. Encode this information into a vector.

[0059] Analyze the assembly constraint characteristics in the work, whether there are kinematic pairs such as hinges, slide rails, gears, and cams, as well as the total number and type of kinematic pairs (rotary pairs, prismatic pairs, screw pairs), and the degree of freedom of the mechanism.

[0060] Based on the characteristics of the part, such as whether it has overhangs, thin walls, thin rods, and material properties, determine the most suitable manufacturing process (such as FDM 3D printing, SLA photopolymerization, CNC milling, laser cutting, manual assembly), process complexity, estimated time, and material loss coefficient.

[0061] Material properties (if any), or to be supplemented by the original author. Record the type of material (plastic, metal, wood), the estimated total weight, and a material cost index (linked to the average market price).

[0062] The above features are concatenated and unified into a 64-dimensional vector.

[0063] The multimodal feature vectors of all works (whether original or improved versions) are stored in a high-dimensional vector database of the feature library. To support efficient similarity search, a KD-Tree spatial index is established. Whenever the number of newly added works reaches 100, the system automatically triggers an index reconstruction, integrating the feature vectors of the new works into the KD-Tree.

[0064] Once the multimodal features of the work are extracted, this embodiment provides a feasible way to generate a genealogy relay record: Genesis relay records include: Work ID: A globally unique work ID. It follows the format of "subject code (e.g., MECH, SOFT, CHEM) + storage date (YYYYMMDD) + serial number of the day".

[0065] Original creator ID: The depositor's student ID number or system user ID.

[0066] Storage and retrieval timestamps and cabinet numbers: System server time, accurate to milliseconds; cabinet number is the cabinet number where the physical artwork is stored.

[0067] Work categories: code / image / chemistry / mechanics.

[0068] Baseline Declaration: The form content of the baseline declaration is serialized and stored as a JSON string.

[0069] Knowledge point tag list: An array of tags selected by the user.

[0070] Artwork hash: Each artwork has an independent MD5 hash value, and the timestamp of the upload time is recorded.

[0071] Feature vector: Stores a foreign key ID that points to a multimodal feature vector in the feature library.

[0072] Preorder hash: This field is empty for the initial genesis relay record.

[0073] Initial vitality index: initialized to 100, or it can be the evaluation score of the work by experts / teachers.

[0074] Red star: Initialized to 0.

[0075] Blue star: Initialized to 0.

[0076] Green star: Initialized to 0.

[0077] Overall star rating: Initialized to 0.

[0078] The hash value of the current genesis relay record: Calculate the SHA256 (a string concatenated from all fields except the current genesis relay record hash in a fixed order).

[0079] In this embodiment, after generating the Genesis Relay record, pre-challenge learning tasks are configured according to different works. When the challenger completes the pre-challenge learning task, reward points are issued and a challenger's challenge certificate is generated.

[0080] Based on core knowledge point tags, the creator or administrator configures a set of structured pre-challenge learning tasks, which include video viewing and interactive verification tasks and knowledge point adaptive testing tasks.

[0081] The task of watching and interacting with the tutorial video requires administrators or creators to upload a tutorial video. The video content must focus on the core principles of the work, the correct usage method, the innovative points, and possible directions for improvement.

[0082] The system incorporates an interactive verification mechanism within the video playback module to prevent students from idling and racking up scores. The specific mechanism is as follows: Random pop-ups: During video playback, the system will pop up two verification questions at random time points (e.g., at the 30th and 150th seconds). The questions will be either multiple choice or click-based. For example, "Please click on the location of the sensor currently being shown in the video" or "What are the abbreviations for the key algorithms mentioned in the video?"

[0083] Countdown and Retry: A 15-second countdown begins on the screen after each verification question. The challenger must respond correctly before the countdown ends. If the answer is incorrect or the timeout occurs, video playback pauses and a message appears: "Verification failed, please watch again." The challenger can only replay the entire video from the beginning.

[0084] Progress tracking: The system uses the HTML5 video API to accurately track playback progress. A task is only marked "completed" when playback reaches 100% and both verification questions are answered correctly on their first appearance. Replays due to any verification failure will not earn points.

[0085] The adaptive knowledge point test task automatically generates a set of adaptive test questions based on 3-5 core knowledge point tags associated with the work. The question bank is pre-entered, and each knowledge point tag contains at least 5 different multiple-choice questions.

[0086] The testing process is as follows: After the challenger enters the testing module, the system randomly selects three tags from all the knowledge point tags of the work. For each selected tag, the system randomly selects one multiple-choice question from the question bank corresponding to that tag.

[0087] The order of the four options in each question is randomly shuffled to prevent challengers from mechanically memorizing the positions of the options.

[0088] Challengers answer questions one by one. The system immediately determines the correctness of each question. If the answer is correct, the challenger receives instant points, which are recorded by the system. If the answer is incorrect, no points are deducted, but a modal window pops up displaying the correct answer and a detailed explanation of why that answer was chosen, along with the relevant underlying principles. The challenger must close the window to proceed to the next question.

[0089] To pass the knowledge point test, challengers must answer at least 2 out of the 3 randomly selected questions correctly (with an accuracy rate of ≥66.7%).

[0090] If the challenger fails the test (answers fewer than 2 questions correctly), they can retry immediately. However, during the retry, the system will re-select questions from the corresponding knowledge point tag (different questions under the same tag). If the challenger fails three consecutive retryes, the system will determine "Pre-challenge Verification Failure," exit the entire pre-challenge process, and deduct all points from the attempt. The challenger will need to exit the mini-program, scan the QR code again, and restart the entire process, including watching the video. This is to prevent blind guessing and encourage challengers to genuinely learn.

[0091] When a challenger completes the pre-challenge learning task, a challenge token is generated. This token has an expiration date (starting from the completion time). Within the validity period, the student challenger can use this token to enter the challenge waiting area at any time. If no challenge is initiated within the validity period, the token automatically expires, and the challenger needs to complete the entire pre-challenge process again (all points will not be deducted, but the challenge qualification needs to be re-obtained; a re-challenge after the challenge token expires will not earn points for that work). This effectively prevents challengers from completing the pre-challenge early but not making any physical improvements, thus occupying a queue spot.

[0092] S300: Challengers obtain relevant data on digital works based on the challenge they initiate and / or obtain physical works according to the challenge order. Challengers optimize their works and store the physical works and improved physical works in the smart cabinet and / or upload the improved digital works and initiate a review of the works.

[0093] Before challengers obtain digital artwork data based on the challenge and / or physical artwork according to the challenge order, the process also includes: The system includes a challenge waiting area and a learning area. The challenge waiting area stores the order and information of students who already hold valid challenge credentials, while the learning area stores the information of students who have not yet completed the pre-challenge learning task. For tasks that cannot be initiated by multiple challengers simultaneously (such as tasks that require improvement based on physical works but have already been taken, or tasks that require program burning and queuing), the challengers are initiated according to their order in the challenge waiting area, and the estimated waiting time is calculated.

[0094] In this embodiment, a challenge waiting area and a learning area are set up, specifically including: The challenge waiting area is an ordered queue used to store information about students who already possess valid challenge credentials. Students are queued according to the time they obtained their credentials (i.e., the timestamp of completing the pre-challenge).

[0095] The learning area is used to store information about students who have not yet completed the pre-challenge.

[0096] Once a challenger enters the queue, their initial position is in the waiting area. They must complete a pre-challenge task within the queue. Once completed, the challenger's information is removed from the waiting area and added to the end of the challenge waiting area.

[0097] After students scan the QR code, a unified queuing interface will appear on the screen or mobile app, containing the following information: Current Challenger Information: Displays the challenger's nickname who is currently using the cabinet, along with their current task and progress (e.g., "Li Hua is currently using a physical artwork, estimated to have 2 days remaining").

[0098] Challenger's queue position: Displays the challenger's number in the challenge waiting area (if they already have a pass), or displays the learning area (requires completion of the learning process first).

[0099] Estimated wait time: Calculate the sum of the predicted equivalent times of all tasks preceding the challenger in the queue, and add it to the remaining estimated time of the currently executing task, displayed as an estimated wait of X minutes. This estimated time is dynamically updated as tasks are executed.

[0100] The setting of challenge waiting areas and learning areas can reasonably prioritize tasks that cannot be performed by multiple challengers together and estimate the expected waiting time.

[0101] When a challenger's turn comes in the queue, a notification is sent via the mini-program and / or the cabinet door screen. The challenger clicks "Start Challenge" within a certain time; otherwise, it is considered a forfeit, the student's information is placed back at the head of the queue, and one instance of timeout is recorded. Two consecutive timeouts will remove the challenger from the queue, requiring them to rejoin. After successfully starting the challenge, the challenger unlocks the cabinet's electronic lock using the QR code obtained through the mini-program, allowing them to retrieve their physical artwork (if physical modifications are needed) and / or all materials related to their digital artwork.

[0102] This embodiment also provides a feasible method for initiating a review of works, including: Submit a challenge statement, which includes the improvement goals, improvement values, improvement types, verification criteria, or a custom verification method.

[0103] Specifically, the improvement targets are: select one or more metrics from the list to improve (e.g., select "maximum tracking speed" and "steering response delay").

[0104] Improvement Value: For each selected metric, challengers should enter an improved expected value or a measured value. If the improvement can be achieved in software simulation, enter the expected value; if physical testing has already been conducted, enter the measured value.

[0105] Improvement Types: A selection box is provided for multidisciplinary improvement types, including but not limited to: software algorithm optimization, mechanical structure optimization, electrical / circuit improvement, appearance design, legal compliance (such as open source license compatibility, patent search), media packaging (such as product introduction videos, promotional copy), multilingual translation, business model analysis, and user experience research. Students can select multiple types simultaneously.

[0106] Verification criteria: If the work has a preset standard testing process (such as "Line-following car speed test standard V1.0"), the challenger can directly select the template, and the system will automatically fill in the testing methods, tools, number of times, etc.

[0107] Custom validation methods: Self-designed indicator validation methods, including method steps, measurement tools, number of measurements, data processing methods, etc.

[0108] After the challengers initiate the review of the works, they also conduct multimodal feature extraction and similarity matching to improve the works.

[0109] The same feature extraction process is applied to the uploaded code, drawings, or descriptive text of the improved work to generate a multimodal feature vector.

[0110] Using the KD-Tree index, search the feature library for the top 3 historical works with the smallest cosine distance to the new feature vector. The search scope includes the original work as well as all previously successfully challenged improved works.

[0111] Read the actual equivalent duration of these 3 similar works when they performed the challenge in the past. Calculate the arithmetic mean of these 3 durations as the predicted equivalent duration for this improved task. If there are fewer than 3 similar works in the feature library (such as early works), use the default duration for this category.

[0112] S400: Review the improvements made to physical works and / or digital works, and calculate the challenger's contribution to improvement and the overall star rating of the work.

[0113] Review improved physical works and / or improved digital works, and route them to the most appropriate verifier for review.

[0114] Maintain a dynamic routing table, which administrators can add, delete, modify, and query through the backend interface. The columns of the routing table include: Improvement Type, Work Category, Verifier Role, Verification Method, and Pass Conditions. See Table 1 for an example.

[0115] Table 1 Routing Table

[0116] After logging into the system, verifiers can see their assigned verification tasks in the pending verification task workbench. Clicking to enter the interface includes the following areas: Challenge Information Summary: Shows the baseline statement of the original work and the improvement statement filled out by the challenger.

[0117] Verification Operation Area: For automated testing, a "Run Test" button is displayed. After clicking, the system executes the script in the background and automatically fills in the results (actual values, whether it passed or failed).

[0118] For teacher / student reviews, provide "pass / fail" radio buttons and text boxes to ask for scores.

[0119] Supporting information: If the teacher feels the evidence is insufficient, they can click "Request Supplementary Evidence." The system will send a notification to the challenger, who has a certain period of time to supplement the evidence. After the supplementary evidence is provided, the same verifier will receive a notification again.

[0120] This embodiment provides a feasible method for calculating the challenger's improvement contribution value and the overall star rating of the work: A red star represents the ratio of the improvement magnitude to 5% multiplied by an improvement type coefficient, or based on teacher scores, with a maximum of 10. The improvement type coefficient is set according to the specific improvement type, reflecting the difficulty of improvement in different technical fields. For example, one red star is awarded for every 5% improvement achieved and the improvement type coefficient is 1. If a challenger improves multiple indicators, the system will calculate the number of red stars for each indicator separately and then accumulate them.

[0121] The blue star represents the total number of times the work has been successfully improved.

[0122] Green stars indicate the number of improved types.

[0123] The improvement contribution value is the product of the weighted sum of red and green stars and the subject popularity. Subject popularity reflects whether the subject corresponding to the challenger's improved work is currently a hot topic, in order to guide more students to engage in hot fields.

[0124] The overall star rating of a work is the weighted sum of the total number of red stars, the total number of valid blue stars, and the total number of green stars.

[0125] The total number of red stars and green stars are historical cumulative values. The effective number of blue stars is the product of the number of blue stars and a time decay coefficient, which is set as follows: If the date of the most recent successful challenge is less than or equal to 100 days from today: the decay factor is 1.0; If the number of days since the most recent successful challenge is less than or equal to 300 days, the decay factor is 0.7. If the number of days since the most recent successful challenge is less than or equal to 600 days, the decay factor is 0.4. The most recent successful challenge was more than 600 days ago: the decay factor is 0.2.

[0126] join Figure 2 This embodiment also provides a feasible method for calculating the challenger's legacy points, including: Based on the relationship between the challenger's improved work and the original work, obtain the improvement route (i.e., A / original work-B2 / second improved work of the second improvement type-C4-D2, etc.). Based on the improved works involved in the improvement route, calculate the cumulative challenger contribution value of the improved works involved in the improvement route, multiply the cumulative challenger contribution value by the inheritance coefficient and add it to the initial vitality index to obtain the total inheritance value. The ratio of a challenger's contribution value multiplied by the inheritance coefficient to the total inheritance value is calculated as the challenger's inheritance score.

[0127] The challenger's legacy points will be used as a factor in revenue distribution, and the accumulated number of the challenger's legacy points will be used as a factor in evaluating the challenger's performance.

[0128] S500 sorts the entries based on their overall star rating. For each entry, it lists the challenger's information and contribution value for improvement based on the challenge completion time, and generates a relay log. The hash value of the generated relay log is anchored to a blockchain or a third-party evidence storage platform. The fields in the relay log include: Log ID, Parent Work ID, Improved Work ID, Challenger ID, Challenge Completion Timestamp, Improvement Type, Actual Improvement Amount, Red Star, Green Star, Valid Blue Star, Storage Cabinet Number, Verifier ID, Multimodal Feature Vector Hash, Actual Waiting Time, Overall Work Star Value, Improvement Contribution Value, Inheritance Points, Hash Value Pointing to the Previous Log, and Current Hash (the hash value obtained by concatenating the aforementioned fields).

[0129] Each new relay log entry has its preceding hash field valued as the current hash of the previous log entry. The preceding hash of the first log entry (genesis relay record / genesis log) is 256 bits of all zeros. All log entries form a tightly linked hash chain from beginning to end. Anyone can trace back from the latest log entry to the genesis log. If the content of any log entry in the middle is modified, its hash value will change, causing the hash values ​​of all subsequent log entries to be calculated incorrectly, thus breaking the entire chain and facilitating verification and tampering queries.

[0130] When the overall star rating reaches a certain value, the work is displayed on the screen. Works are sorted based on their overall star rating. For each work, challenger information and contribution value for improvement are listed based on the challenge completion time, and a relay log is generated.

[0131] To encourage students to discover works that may interest them and to guide interdisciplinary participation, this implementation provides personalized recommendations and trending topics. Figure 3 ,include: (1) Construct an interaction matrix between students and works, such as an m×n sparse matrix M. Where m is the number of students and n is the number of works. The values ​​of the matrix elements are obtained by weighted summation of the time spent browsing works, the number of times the works were challenged, and the number of times the challenges were successfully completed. The number of times the works were challenged represents the students' willingness to actively try to improve the works, and the number of times the challenges were successfully completed represents the actual results achieved by the students in the works.

[0132] (2) The weighted matrix decomposition method is used to learn the latent vectors of students and works from the interaction matrix.

[0133] The original M matrix is ​​approximately decomposed into the product of two low-rank matrices: the student factor matrix P and the work factor matrix Q.

[0134] The algorithm uses the alternating least squares method to solve the problem. Specific steps: With Q fixed, solve for the latent vector p of each student u. u (Minimize the weighted squared error).

[0135] With P fixed, solve for the latent vector q of each work i. i .

[0136] Alternate iterations until convergence (or the preset maximum number of iterations, such as 30).

[0137] The weighting is reflected in the loss function: observed interactions (matrix element values ​​are greater than 0) are given higher weights (e.g., 1.0), and unobserved interactions (matrix element values ​​are equal to 0) are given lower weights (e.g., 0.01) to avoid overfitting.

[0138] After the solution is obtained, each student is represented as a multidimensional vector p. u Each work is also a multidimensional vector q i The dot product (or cosine similarity) of two vectors reflects the student's potential preference for the work.

[0139] (3) When a challenger enters the homepage of the mini program, the latent vector of the challenger is read from the database; all works that have not yet been challenged by the student are traversed, and the cosine similarity between the latent vector of each work and the latent vector of the challenger is calculated; the 5 works with the highest similarity are selected as recommended works, and an explanation is generated for each recommended work.

[0140] The reasons for the explanation include: If the similarity primarily comes from browsing duration features (which can be determined by analyzing the contribution of each part of the interaction matrix), display "Because you have viewed similar works".

[0141] If it comes from the number of challenges, it will display "Because you have challenged similar works before".

[0142] If the challenge is successful, it will display "Because your area of ​​expertise is [Improvement Type Tag extracted from historical successful challenges], this work needs this type of improvement".

[0143] In addition, a subject popularity score is calculated based on the performance of each subject / improvement type within a time period. The score is the product of the challenge success rate, the lifespan value-added rate of the work, and the proportion of participating students. The higher the score, the higher the quality of output, the faster the growth, and the wider the participation in that subject during that time period.

[0144] The challenge success rate is the ratio of the number of successful challenges to the total number of challenges for all works in that subject within a given time period.

[0145] The vitality of a work is the product of its improvement contribution value and its vitality coefficient. The vitality value-added rate of a work is the ratio of the sum of the vitality increments within a time period to the total vitality at the beginning of the time period.

[0146] The percentage of participating students is the ratio of the number of challengers who attempted works in that subject within the specified time period to the total number of challengers.

[0147] Based on the popularity of academic disciplines, popular disciplines are identified and recommended. Disciplines whose popularity exceeds a certain threshold within a given time period are considered popular and will be recommended. For example: On the display screen of the smart cabinet, the color blocks of popular subjects will be highlighted (such as orange frames with flame icons).

[0148] The system pushes a mini-program message to students who have not participated in any challenges in this subject: "[Hot News] XX subject is undergoing a hot innovation! Click to view recommended works and participate in the challenge to earn points." The message includes a link to the most popular work in this subject.

[0149] See Figure 4 Based on the aforementioned method, an interdisciplinary teaching incentive system is designed, including a work storage module 100, a feature extraction and evidence storage module 200, a challenge module 300, an audit module 400, and a display module 500. The artwork storage module 100 is used to provide smart cabinets for creators or managers to store physical artworks and / or upload digital artworks to the artwork library; The Feature Extraction 200 and Evidence Preservation Module is used to extract multimodal features of physical and / or digital works to generate a genesis relay record and anchor the hash value to a blockchain or a third-party evidence preservation platform. The Challenge Module 300 is used to obtain relevant data of digital works based on the challenger's initiation of the challenge and / or obtain physical works according to the challenge order. The challenger optimizes the works and stores the physical works and improved physical works in the smart cabinet and / or uploads the improved digital works and initiates the work review. The review module 400 is used to review the improvements made to physical works and / or digital works, and to calculate the challenger's contribution to improvement and the overall star rating of the work; The display module 500 is used to display and sort works based on their overall star rating. For each work, challenger information and contribution value for improvement are listed based on the challenge completion time, and a relay log is generated.

[0150] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described with reference to preferred embodiments, those skilled in the art should understand that various changes in form and detail can be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A cross-disciplinary teaching incentive method, characterized in that, include: The creator or manager stores physical works in the smart cabinet and / or uploads digital works to the works library; Genesis relay records are generated by extracting multimodal features of physical and / or digital works and anchoring the hash values ​​to a blockchain or a third-party evidence storage platform. Challengers obtain relevant data on digital works by initiating challenges and / or obtain physical works according to the challenge order. Challengers optimize their works and store the physical works and improved physical works in the smart cabinet and / or upload the improved digital works and initiate work review. Review the improvements made to physical works and / or digital works, and calculate the challenger's contribution value and overall star rating for the work; The works are displayed and sorted based on their overall star rating. For each work, challenger information and contribution value for improvement are listed based on the challenge completion time, and a relay log is generated.

2. The interdisciplinary teaching incentive method according to claim 1, characterized in that, It also includes the creation of works and the baseline declaration filled in by the creator or manager based on physical works and / or digital works; the baseline declaration includes the performance indicator name, the performance indicator baseline value, the performance indicator measurement method, the performance indicator measurement conditions and the core knowledge point tags.

3. The interdisciplinary teaching incentive method according to claim 2, characterized in that, The extracted multimodal features of the work include: For code-based works, basic features are extracted, including code file size, number of valid lines of code, total number of functions or methods, maximum nesting depth of conditional branches, whether blocking delay functions are included, whether floating-point arithmetic is included, and whether serial port or network communication instructions are included. Structural features are extracted, including the total number of conditional branch nodes in the syntax tree, maximum nesting depth of branch nodes, total number of loop nodes, number of different nodes in the function call graph, and total number of arithmetic operator nodes. The basic and structural features are then multinomial-expanded to form a multimodal feature vector for code-based works. For image design works, global features including color distribution, texture pattern, and edge density are extracted. Handcrafted design features including information entropy, the ratio of compressed file size to original file size, edge density detected by the Canny operator, number of corner points detected by the Shi-Tomasi algorithm, and horizontal and vertical symmetry indices of the image are extracted. Local feature vectors including the number of keypoints, the distribution density of keypoints on the image plane, and K-Means clustering of ORB descriptors of all keypoints are extracted. The global features, handcrafted design features, and local features are concatenated to obtain the multimodal feature vector of the image design works. For works involving chemical experiments, textual features are extracted, including the number of experimental steps, the total number of reagents, the number of reaction condition parameters, the number of special equipment, and the number of safety protection measures mentioned in the document. Statistical features, including the mean, standard deviation, range, and data integrity of the output data, are also extracted. Safety risk features of the experimental scheme are extracted. The textual features, statistical features, and safety risk features are then concatenated to obtain a multimodal feature vector for works involving chemical experiments. For mechanical structure works, the following analytical features are extracted: total number of independent parts, assembly level depth, ratio of total area of ​​all curved surfaces to volume of part bounding box, total number of holes, total number of chamfers, and total number of arrays. The following assembly constraint features are extracted: whether there are hinges, slide rails, gears, cam kinematic pairs and the total number and type of kinematic pairs, as well as the degree of freedom of the mechanism. The following part features are extracted: whether there are overhangs, thin walls, thin rods, material properties, most suitable manufacturing process, process complexity, estimated time, and material loss coefficient.

4. The interdisciplinary teaching incentive method according to claim 3, characterized in that, Also includes: Pre-challenge learning tasks are configured according to different works. When the challenger completes the pre-challenge learning tasks, reward points are awarded and a challenge certificate is generated for the challenger.

5. The interdisciplinary teaching incentive method according to claim 4, characterized in that, It also includes setting up a challenge waiting area and a learning area. The challenge waiting area is used to store the order and information of students who have valid challenge credentials, and the learning area is used to store the information of students who have not yet completed the pre-challenge learning task. For tasks that cannot be initiated by multiple challengers at the same time, the challenge is initiated according to the order of challengers in the challenge waiting area, as well as the calculation of the estimated waiting time.

6. The interdisciplinary teaching incentive method according to claim 5, characterized in that, The submission of a challenge statement includes the following: the goal of improvement, the value of improvement, the type of improvement, the verification criteria or a custom verification method. The challenger's contribution to improvement is the product of the weighted sum of red and green stars and the subject's popularity. The overall star rating of the work is a weighted sum of the total number of red stars, the total number of valid blue stars, and the total number of green stars.

7. The interdisciplinary teaching incentive method according to claim 6, characterized in that, This also includes personalized recommendations and trending topics for challengers, including: Construct an interaction matrix between students and their work; We use weighted matrix factorization to learn the latent vectors of students and works from the interaction matrix; When a challenger enters the mini-program homepage, the latent vector of the challenger is read from the database; all works that have not yet been challenged by the student are traversed, and the cosine similarity between the latent vector of each work and the latent vector of the challenger is calculated; the 5 works with the highest similarity are selected as recommended works, and an explanation is generated for each recommended work. Subject popularity is scored based on the performance of each type of improvement within a time period. Based on the subject popularity, popular subjects are set and recommended.

8. The interdisciplinary teaching incentive method according to claim 7, characterized in that, The relay log includes: log ID, parent work ID, improved work ID, challenger ID, challenge completion timestamp, improvement type, measured improvement range, red star, green star, valid blue star, storage cabinet number, verifier ID, multimodal feature vector hash, actual waiting time, work's overall star value, improvement contribution value, inheritance points, hash value of the previous log, and current hash.

9. The interdisciplinary teaching incentive method according to claim 8, characterized in that, It also includes calculating the challenger's legacy points, including: Based on the relationship between the challenger's improved work and the original work, obtain the improvement path; Based on the improved works involved in the improvement route, calculate the cumulative challenger contribution value of the improved works involved in the improvement route, multiply the cumulative challenger contribution value by the inheritance coefficient and add it to the initial vitality index to obtain the total inheritance value. The ratio of a challenger's contribution value multiplied by the inheritance coefficient to the total inheritance value is calculated as the challenger's inheritance score. The challenger's legacy points will be used as a factor in revenue distribution, and the accumulated number of the challenger's legacy points will be used as a factor in evaluating the challenger's performance.

10. A cross-disciplinary teaching incentive system, characterized in that, It includes modules for storing works, extracting and storing features, challenges, review, and display. The work storage module is used to provide smart cabinets for creators or managers to store physical works and / or upload digital works to the work library; The feature extraction and evidence storage module is used to extract multimodal features of physical and / or digital works to generate a genesis relay record, and anchor the hash value to a blockchain or a third-party evidence storage platform. The challenge module is used to obtain relevant data of digital works based on the challenger's initiation of the challenge and / or obtain physical works according to the challenge order. The challenger optimizes the works and stores the physical works and improved physical works in the smart cabinet and / or uploads the improved digital works and initiates the work review. The review module is used to review the improved physical works and / or improved digital works, and to calculate the challenger's improvement contribution value and the overall star value of the work; The display module is used to display and sort works based on their comprehensive star rating. For each work, challenger information and contribution value to the improvement of the work are listed based on the challenge completion time, and a relay log is generated.

Images