Standard line digital review platform

Abstract

The invention relates to the field of online review, and discloses a standard online digital review platform which comprises a platform application and interaction module used for providing a unified user interface and processing operation instructions of different roles; the block chain and intelligent contract module is connected with the platform application and interaction module and is used for deploying a preset review process intelligent contract and storing a key review operation log in a non-tampering manner; the AI analysis engine module is connected with the block chain and intelligent contract module; and the database module is connected with the platform application and interaction module. The technical problems that process data of an existing review system is easy to tamper, objective quantitative analysis is difficult to carry out, and an analysis model cannot be self-optimized are solved, the effects that key vouchers in the review process cannot be tampered and the whole process is traceable are achieved, a self-optimization closed loop combining analysis and management and control is constructed, and the review efficiency is improved. Therefore, the fairness and reliability of review are improved.

Description

Technical Field

[0001] This invention relates to the field of online review technology, and in particular to a standard online digital review platform. Background Technology

[0002] The review process is crucial in the development of industry or national standards. Traditional paper-based reviews or early digital review systems were typically used to achieve basic online submission and comment aggregation functions. However, these systems have limited control over the review process. The recording of key operations during the review process (such as task allocation, review time, and comment submission) relies on centralized databases, resulting in insufficient data credibility and traceability. This makes it difficult to effectively guarantee the transparency and fairness of the review process.

[0003] To enhance data credibility, existing technologies attempt to utilize blockchain technology to store evaluation data, thereby ensuring the immutability and transparency of evaluation results.

[0004] However, existing technologies only solve the problem of solidifying and storing evaluation results data, but do not provide effective constraints and reliable records for the execution of the review process itself. The state transition logic of the review process is still controlled by traditional application servers, and the critical procedural timestamps lack the guarantee of immutability. Therefore, when objective analysis of review behavior is required, the system cannot provide high-confidence raw data.

[0005] Furthermore, existing digital review systems generally lack the ability to deeply analyze review process data, making it difficult to automatically identify abnormal review behaviors or predict potential review biases from reliable data. At the same time, the systems do not provide a mechanism to combine analysis results with process control; that is, anomalies identified in the analysis cannot trigger process intervention, and the conclusions of manual review cannot be used for iterative optimization of the analysis model, resulting in a lack of an adaptive quality monitoring closed loop. Summary of the Invention

[0006] The purpose of this invention is to provide a standard online digital review platform, which solves the problem that existing digital review systems lack a mechanism for reliably recording the review process, making it difficult to objectively quantify and analyze review behavior, and lack a self-optimizing closed loop that links analysis results with process intervention.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The standard online digital review platform includes: The platform application and interaction module is used to provide a unified user interface and handle operation commands from different roles; The blockchain and smart contract module is connected to the platform application and interaction module, and is used to deploy preset review process smart contracts and store key review operation logs in an immutable manner. The AI ​​analysis engine module is connected to the blockchain and smart contract module and is used to obtain trusted data from the blockchain and smart contract module to perform indicator calculation, behavior analysis and bias identification. The database module, connected to the platform application and interaction module, is used for non-credential auxiliary data, including: storing the original text of the standard draft and basic user information. Preferably, the platform application and interaction module includes: The user management unit is used to handle the accounts and permissions of administrators, reviewers, and those being reviewed. The task submission unit is used to send the original draft standard to the database module and the task metadata to the blockchain and smart contract module. The review execution unit is used to retrieve the original draft standard from the database module for review, collect the reviewer's operation data, and send the operation data to the blockchain and smart contract module. The analysis and alarm display unit is used to receive and display the analysis results data from the AI ​​analysis engine module, and to initiate process intervention operations and submit manual review conclusions.

[0008] Preferably, the manual review conclusion submitted by the analysis and alarm display unit is sent to the AI ​​analysis engine module as training data for model iteration.

[0009] Preferably, in the deployment of the blockchain and smart contract module, the review process smart contract includes: A finite state machine model is used to define the set of states of a review project, which includes at least the state to be assigned, the state under review, the state to be reviewed, and the state that has been archived. A state transition function is used to control the flow of the review item in the state set based on the received transaction and the preset permission verification logic; The trusted log generation unit is used to generate trusted log entries containing operator identity, state change details and operation data during state transition execution, and write them into the blockchain ledger. A status timestamp management mechanism is used to capture the start timestamp and end timestamp respectively when the review project enters and exits the review status, and write the start timestamp and end timestamp into the trusted log entry.

[0010] Preferably, the status timestamp management mechanism is specifically used for: When the review project enters the review process under the operation of a specific reviewer, the start timestamp is captured; When the reviewer subsequently initiates a transaction that causes the review process to exit, the end timestamp is captured.

[0011] Preferably, the AI ​​analysis engine module includes: A data acquisition unit is used to acquire trusted log entries from the blockchain and smart contract module; The indicator quantification calculation unit calculates the trusted review duration, review consistency, and reviewer historical baseline based on the trusted log entries. The review behavior pattern analysis unit uses the credible review duration and the reviewer's historical baseline to calculate the review duration Z-Score to identify abnormal behavior. The latent bias prediction unit is used to build feature vectors and run machine learning models to predict the probability of bias in a single review.

[0012] Preferably, the trusted review duration is calculated by extracting the start and end timestamps from the trusted log entries.

[0013] Preferably, the review consistency is calculated by extracting the review scores from the trusted log entries submitted by all reviewers corresponding to the same project.

[0014] Preferably, in the potential bias prediction unit, the feature vector includes at least: the credible review duration, the review score, and the review duration Z-Score.

[0015] Preferably, the AI ​​analysis engine module further includes: A composite alarm trigger is used to generate an alarm signal and send it to the platform application and interaction module when the review duration Z-Score or bias prediction probability meets the preset alarm rules. The platform application and interaction module can also initiate an intervention transaction to the blockchain and smart contract module based on the alarm signal, forcibly changing the status of the review project corresponding to the alarm signal from the archived status to the pending review status.

[0016] In summary, the present invention has at least one of the following beneficial technical effects: 1. This invention defines the review process state machine through smart contracts and encapsulates key operations in the review process as trusted log entries and writes them into the blockchain ledger; it solves the technical problems of key process data being easily tampered with, opaque, and difficult to trace in traditional review systems; it achieves the effect of key voucher data in the review process being immutable and traceable throughout the process, thereby improving the procedural fairness and transparency of the review process.

[0017] 2. This invention obtains trusted log entries containing immutable start and end timestamps provided by the blockchain and smart contract module, and calculates the trusted review duration and review duration Z-Score based on these entries; it solves the technical problem in traditional reviews where the evaluation of reviewer behavior relies on subjective judgment and lacks objective data support; it achieves the effect of objectively quantifying and analyzing review behavior, and improves the ability to analyze review quality and the reliability of the results.

[0018] 3. This invention generates alarms through the AI ​​analysis engine module, and the platform application and interaction module responds to the alarms to initiate process intervention, and feeds back the administrator's manual review conclusions as new training data to the AI ​​analysis engine module; it solves the technical problem that the analysis model is difficult to iterate based on the actual review results after deployment, which leads to model solidification or decreased accuracy; it achieves a closed-loop effect of self-optimization of the analysis model, enabling the platform to continuously learn and improve its analysis accuracy. Attached Figure Description

[0019] Figure 1 This is a platform module architecture diagram of the present invention; Figure 2 This is a diagram illustrating the platform application and interaction module architecture of the present invention. Figure 3 This is a diagram of the blockchain and smart contract module architecture of the present invention; Figure 4 This is a diagram of the AI ​​analysis engine module architecture of the present invention. Detailed Implementation

[0020] See attached document Figure 1 -Appendix Figure 4 This invention provides a standard online digital review platform, which includes: The platform application and interaction module 100 is built on cloud services, providing a unified user interface and handling operation instructions for different roles (such as administrators and reviewers).

[0021] The blockchain and smart contract module 200 serves as the system's trust layer, used to deploy pre-defined review process smart contracts and to store key review operation logs in an immutable manner.

[0022] AI analytics engine module 300 is used to obtain trusted data from blockchain and smart contract module 200 and perform indicator calculations, behavior analysis and bias identification.

[0023] Database module 400 is used to store non-credential auxiliary data such as the original text of the standard draft and basic user information.

[0024] Its operation process is as follows: First, the administrator defines the review process rules through the platform application and interaction module 100, and deploys the review process smart contract on the blockchain and smart contract module 200. This contract predefines the state machine model of the review task, including states such as pending assignment, under review, and archived.

[0025] When the reviewee submits a draft standard through the platform application and interaction module 100, the module sends the original draft standard and other unstructured files to the database module 400 for storage, and at the same time sends the task metadata to the blockchain and smart contract module 200, triggering the smart contract to set the initial state of the task to pending assignment.

[0026] During the review process, the administrator performs the assignment operation through the platform application and interaction module 100. When the review begins, the reviewer retrieves the original draft standard from the database module 400 through the platform application and interaction module 100 for review. After the review is completed, the review comments are submitted to the blockchain and smart contract module 200.

[0027] After verifying the transaction permission, the blockchain and smart contract module 200 executes the state transition logic defined in the smart contract, such as transitioning from pending allocation to review.

[0028] While the state transition is being executed, the smart contract constructs a trusted log entry from the key information of this operation, including the operator's identity (Uid), the operation timestamp, the details of the state change, and the operation data, and writes it into the blockchain ledger for persistent storage.

[0029] The AI ​​analysis engine module 300 continuously listens to and obtains newly generated trusted log entry data from the blockchain and smart contract module 200.

[0030] The AI ​​analysis engine module 300 combines trusted log entry data obtained from the blockchain and smart contract module 200 with auxiliary data obtained from the database module 400 to perform quantitative calculations of indicators, such as calculating review time and review consistency, and running behavioral pattern analysis models and bias prediction models.

[0031] The AI ​​analysis engine module 300 sends its analysis results (such as abnormal behavior alerts or bias probabilities) to the platform application and interaction module 100 for display.

[0032] Based on the alarm information displayed by the platform application and interaction module 100, the administrator can decide whether to trigger the blockchain and smart contract module 200 to forcibly change the status of a specific task to pending review.

[0033] Meanwhile, the administrator's manual review conclusions (such as confirming bias) are fed back to the AI ​​analysis engine module 300 as new training data for iterative optimization of the bias prediction model. The platform application and interaction module 100 serves as the system's user interaction layer and is deployed based on cloud services. This module includes a built-in user management unit to handle accounts and permissions for at least three roles: administrators, reviewers, and those being reviewed. This user management unit can assign preset system function access permissions to different roles based on a Role-Based Access Control (RBAC) model. For example: The administrator role is granted permissions to create review projects, assign tasks, and view all analysis reports; The reviewer role is granted permission to view assigned tasks and submit review comments; The reviewee is granted the authority to submit draft standards and view the review status.

[0034] For the specific implementation of user authentication, well-known technologies in the field can be adopted, such as methods based on username / password, digital certificates, or third-party authentication services.

[0035] The platform application and interaction module 100 includes a task submission unit, a review and execution unit, and an analysis and alarm display unit.

[0036] In the task submission unit, the reviewer uploads the standard draft file (e.g., PDF or Word document) and related metadata (e.g., draft title, technical field category, abstract) through the interface provided by this unit. Upon receiving the submission request, this unit performs the following operations: Unstructured documents, such as the original draft of the standard, are sent to database module 400 for storage. At the same time, the task metadata (such as the generated project ID and submitter information) is encapsulated and sent to the blockchain and smart contract module 200 to trigger the initialization of the review process.

[0037] Within the review execution unit, the role is geared towards the reviewer. After a reviewer logs in, this unit retrieves a list of tasks assigned to that reviewer from the system. Once the reviewer selects a specific task, this unit is responsible for retrieving the original draft standard from database module 400 for review. Simultaneously, this unit presents a structured review comment form.

[0038] This form is used to collect operational data from reviewers. The operational data includes at least one or more quantitative review scores and qualitative review comments. When the reviewer completes the form and confirms submission, the review execution unit encapsulates the collected operational data, along with the reviewer's identity and project ID, into a transaction payload and sends it to the blockchain and smart contract module 200 to trigger a state transition in the smart contract.

[0039] The Analysis and Alerts Display Unit is primarily designed for administrators. This unit receives analysis results data from the AI ​​Analysis Engine Module 300. It then visualizes the received data, for example, displaying overall review statistics in a dashboard or showing flagged abnormal reviews and tasks with high bias probability in a list format.

[0040] This analysis and alarm display unit also provides interactive functions for process intervention. When an administrator determines that manual review is required based on the displayed alarm information, they can initiate a specific intervention transaction to the blockchain and smart contract module 200 through the unit's interface to call the smart contract to forcibly set the task status to pending review.

[0041] In addition, this unit includes an input interface for manual review of conclusions. After completing the manual review, the administrator submits the conclusion through this interface. This conclusion data will be sent to the AI ​​analysis engine module 300 as training data for model iteration.

[0042] The blockchain and smart contract module 200, acting as the system's trust layer or trust foundation, is responsible for executing predefined review process rules and solidifying key credential data during the review process in an immutable manner. In one embodiment, this module can be deployed in a consortium blockchain network environment to ensure controllability of node access, transaction performance, and data privacy. The core of this module is a pre-deployed smart contract for the review process.

[0043] The smart contract for this review process internally defines a finite state machine model to programmatically manage and constrain the lifecycle of each review item. This state machine defines a set of states S, which includes at least: (Pending Assignment Status) indicates that the draft standard has been submitted by the reviewer and is awaiting further action from the administrator (such as assigning a reviewer).

[0044] (Assigned status) indicates that the administrator has assigned one or more reviewers to the project, but the reviewers have not yet confirmed or started their review work.

[0045] (Under Review Status) indicates that the designated reviewer has received the task and is currently reviewing and writing comments.

[0046] (Pending review status) indicates that the review results of this project (e.g., by AI analysis engine module 300 or the administrator) have been marked as having potential anomalies or biases and require manual review.

[0047] (Archived status) indicates that the review process for this project has been completed normally and the review comments have been summarized.

[0048] The smart contract further defines state transition functions to control the flow of states from the current state to the target state. These state transition functions are tightly coupled with the permission verification logic. When the smart contract receives a transaction request from the platform application and interaction module 100, it first verifies whether the identity (Uid) of the operator initiating the transaction meets the preset permission conditions.

[0049] For example, used to trigger For transactions involving state transitions, the initiator's Uid must have an administrator role identifier.

[0050] For example, used to trigger For a state transition transaction, the initiator's Uid must be consistent with the project's Uid. The status must match the reviewer's Uid. For a state transition to occur, the initiator's Uid must have an administrator role identifier. If permission verification fails, the transaction will be rejected, and the state will not change.

[0051] The smart contract also includes a trusted log generation unit. This unit is triggered after the state transition function is verified and successfully executed. Its function is to generate a trusted log entry and emit it as an event or write it directly into the state tree of the blockchain ledger.

[0052] The LogEntry is defined as a structured data body, which contains at least the following fields: Project ID, Operator ID (Uid), and State before the transfer. The state after the transfer and operation data The operational data here refers to the business data carried during the transaction and collected by the platform application and interaction module 100, such as the quantitative review scores submitted by the reviewers. Or the hash value of the review comments text.

[0053] To support the AI ​​analysis engine module 300 in accurately analyzing review behavior (especially review duration), the smart contract also implements a state timestamp management mechanism. In this mechanism, when a project (Pid) enters a state that requires timing for the first time under the operation of a specific reviewer (Uid), the smart contract captures the timestamp of the current block and uses this timestamp as the starting timestamp, associating it with Pid, ​​Uid, and the current state, and storing it in the contract's internal storage.

[0054] When the reviewer (Uid) subsequently initiates a transaction that causes the state to exit (e.g.) While executing state transitions, smart contracts perform the following operations: Get the timestamp of the current block and define it as the end timestamp.

[0055] Retrieve the previously stored start timestamp from the contract's internal storage based on Pid and Uid.

[0056] The start timestamp and the end timestamp Together with the key field, they are written into the aforementioned trusted log entries.

[0057] Ultimately, including The complete LogEntry is permanently stored on the blockchain ledger. This mechanism ensures that the raw time data used to calculate the review duration is generated synchronously with the review operation, strongly correlated with it, and immutable, providing a high-confidence data source for subsequent AI analysis.

[0058] The AI ​​analytics engine module 300, serving as the system's analytics service layer, is responsible for in-depth processing and modeling of the trusted data embedded in the blockchain and smart contract module 200. This module includes a data acquisition unit, an indicator quantification and calculation unit, a review behavior pattern analysis unit, and a potential bias prediction unit.

[0059] The data acquisition unit is responsible for communicating with the blockchain and smart contract module 200. In one embodiment, this unit acquires newly generated trusted log entries on the blockchain by listening to predefined events in the smart contract or by periodically polling via an API interface (e.g., JSON-RPC). This unit parses the acquired LogEntry data and extracts key fields (such as...) In (etc.), and load them into an analytical database to support efficient aggregate queries and statistical analysis.

[0060] The indicator quantification calculation unit is responsible for calculating the core indicators required by the analysis model based on the acquired LogEntry data.

[0061] One function of this quantitative calculation unit is to calculate the trusted review duration. This unit, for a specific reviewer (Ui) and project, retrieves the corresponding data from the analytical database, based on the state transitions (e.g., by smart contracts) of the smart contract. A LogEntry generated and recorded at a specific time. The start timestamp is extracted from this LogEntry. and end timestamp Trustworthy review duration Calculated using the following formula: ; In the formula, For reviewers For the project Trustworthy review duration; The end timestamp extracted from this LogEntry; This is the start timestamp extracted from this LogEntry.

[0062] Another function of this quantitative calculation unit is to calculate review consistency. This unit queries the same project. All corresponding Each reviewer submits a LogEntry, from which their respective review scores are extracted. First, the average score for the project is calculated. : ; In the formula, For the project The average score; The number of reviewers who participated in the review of this project; For reviewers For the project The scores were then calculated. Subsequently, the variance of review consistency was calculated. : ; In the formula, For the project The variance of reviewer consistency. The larger this value, the greater the disagreement among reviewers in their scoring, and the lower the consistency.

[0063] The quantitative calculation unit for this indicator also includes a function for building a reviewer's historical baseline. This function is specific to each reviewer. The system queries the reliable review duration records of all historical review tasks in the analytical database, and uses standard statistical methods to dynamically calculate the reviewer's historical average review duration and historical duration standard deviation based on this historical data.

[0064] The review behavior pattern analysis unit uses the calculated indicators to conduct statistical analysis of reviewers' behavior. This unit identifies statistically abnormal behavior by calculating the review duration Z-score. The review duration Z-score is calculated using the following formula: ; In the formula, For reviewers For the project The review time Z-Score; This is the credible review duration calculated for this review. This refers to the reviewer's historical average review time. This represents the standard deviation of the reviewer's historical duration. The value objectively quantifies the degree to which the duration of the current review deviates from the individual's historical average. A significantly negative value indicates... Values ​​(e.g., less than -2) indicate hasty review behavior; significantly positive values ​​indicate... Values ​​(e.g., greater than 2) indicate delayed review behavior. The objectivity of this model's analysis stems from its input. It is based on the blockchain and smart contract module 200, ensuring immutability. and It was calculated.

[0065] The latent bias prediction unit is used to build a machine learning model to predict the probability of bias in a single review. This unit first constructs a feature vector X for each review. This feature vector X includes at least: credible review duration, review score, review duration Z-score, score bias, and other auxiliary features obtainable from database module 400, such as the technical field classification of the draft standard and the reviewer's professional field profile.

[0066] This unit employs one or more machine learning algorithms (such as logistic regression, random forest, or gradient boosting tree) as the prediction model. The training process of this model relies on a labeled dataset. The labels of this dataset... (Where 1 indicates bias and 0 indicates no bias) is derived from the conclusions provided by administrators after manual review during the pending review process. This unit will analyze and calculate the... value, value and The probability value is sent to the platform application and interaction module 100 for display and alarm.

[0067] Embodiments of the present invention also provide a closed-loop mechanism for alarm, intervention and feedback, which coordinates the platform application and interaction module 100, the blockchain and smart contract module 200 and the AI ​​analysis engine module 300.

[0068] The AI ​​analytics engine module 300 integrates a composite alarm trigger. This trigger continuously monitors the outputs of the metric quantification calculation unit and the potential bias prediction unit. The trigger is pre-configured with a set of alarm thresholds, such as a behavioral threshold for the review duration Z-score, a threshold for review consistency variance, and a threshold for bias prediction probability. These thresholds can be fixed values ​​or percentiles dynamically calculated based on historical data distribution.

[0069] The trigger is activated when the metric value calculated by the AI ​​analysis engine module 300 meets the preset alarm rules. The alarm rules can be set to trigger a single metric (e.g., ) or a combination of indicators triggering (e.g., and ).

[0070] Once the trigger is activated, the AI ​​analysis engine module 300 generates an alarm signal. This alarm signal includes the project ID (Pid), the ID of the reviewer involved, the specific alarm type, and relevant metric values. The alarm signal is then sent to the platform application and interaction module 100, where its analysis and alarm display unit presents it.

[0071] After receiving alarm information through the platform application and interaction module 100 interface, the administrator can manually analyze the alarm. If the administrator believes that the review project needs to be manually reviewed, the administrator can initiate a process intervention operation through a specific function interface on the platform application and interaction module 100 (such as the "Initiate Review" button).

[0072] The intervention operation is encapsulated as a specific type of intervention transaction by the platform application and interaction module 100 and sent to the blockchain and smart contract module 200.

[0073] Upon receiving the intervention transaction, the blockchain and smart contract module 200's built-in state transition function first verifies whether the transaction initiator's Uid has administrator privileges. If the verification is successful, the smart contract executes a pre-defined intervention logic, forcibly changing the current state of the project ID (Pid) to a pending review state. This state change is recorded on the blockchain ledger, enabling programmatic intervention in the trusted review process driven by AI analysis results.

[0074] Once a project enters the pending review stage, the platform application and interaction module 100 will provide a manual review interface for the administrator (or the designated review committee). On this interface, reviewers can view the original draft standard, marked review comments, and the alert basis (such as Z-Score value) provided by the AI ​​analysis engine module 300.

[0075] After completing their manual assessment, reviewers submit their final review conclusions through this interface. This conclusion is a high-quality labeled data point, for example, marked as confirming bias or confirming no bias / false positive.

[0076] After obtaining the review conclusion, the platform application and interaction module 100 associates it with the original feature vector X of the review (which was generated and saved by the AI ​​analysis engine module 300 when making bias predictions).

[0077] The associated data (i.e.) The data pairs are fed back to the AI ​​analysis engine module 300 and added to the training dataset of the potential bias prediction model.

[0078] The AI ​​analysis engine module 300 can retrain or incrementally train its potential bias prediction model using an enhanced training set containing this new feedback data after a periodic (e.g., daily or weekly) or cumulative period of a certain amount of new labeled data.

[0079] Through the closed-loop mechanism of alarm, intervention, and feedback described above, this invention achieves a self-optimizing and adaptive system. The analysis results of the AI ​​model are used to drive interventions in trusted processes, while the human review conclusions of the interventions, in turn, serve as high-quality training data to iteratively improve the accuracy of subsequent predictions by the AI ​​model.

Claims

1. A standard online digital review platform, characterized in that, include: The platform application and interaction module is used to provide a unified user interface and handle operation commands from different roles; The blockchain and smart contract module is connected to the platform application and interaction module, and is used to deploy preset review process smart contracts and store key review operation logs in an immutable manner. The AI ​​analysis engine module is connected to the blockchain and smart contract module and is used to obtain trusted data from the blockchain and smart contract module to perform indicator calculation, behavior analysis and bias identification. The database module, connected to the platform application and interaction module, is used for non-credential auxiliary data, which includes: storing the original text of the standard draft and basic user information.

2. The online digital review platform for standards according to claim 1, characterized in that, The platform application and interaction modules include: The user management unit is used to handle the accounts and permissions of administrators, reviewers, and those being reviewed. The task submission unit is used to send the original draft standard to the database module and the task metadata to the blockchain and smart contract module. The review execution unit is used to retrieve the original draft standard from the database module for review, collect the reviewer's operation data, and send the operation data to the blockchain and smart contract module. The analysis and alarm display unit is used to receive and display the analysis results data from the AI ​​analysis engine module, and to initiate process intervention operations and submit manual review conclusions.

3. The online digital review platform for standards according to claim 2, characterized in that, The manual review conclusions submitted by the analysis and alarm display unit are sent to the AI ​​analysis engine module as training data for model iteration.

4. The online digital review platform for standards according to claim 1, characterized in that, In the deployment of the blockchain and smart contract module, the review process smart contract includes: A finite state machine model is used to define the set of states of a review project, which includes at least the state to be assigned, the state under review, the state to be reviewed, and the state that has been archived. A state transition function is used to control the flow of the review item in the state set based on the received transaction and the preset permission verification logic; The trusted log generation unit is used to generate trusted log entries containing operator identity, state change details and operation data during state transition execution, and write them into the blockchain ledger. A status timestamp management mechanism is used to capture the start timestamp and end timestamp respectively when the review project enters and exits the review status, and write the start timestamp and end timestamp into the trusted log entry.

5. The online digital review platform for standards according to claim 4, characterized in that, The aforementioned status timestamp management mechanism is specifically used for: When the review project enters the review process under the operation of a specific reviewer, the start timestamp is captured; When the reviewer subsequently initiates a transaction that causes the review process to exit, the end timestamp is captured.

6. The online digital review platform for standards according to claim 1, characterized in that, The AI ​​analysis engine module includes: A data acquisition unit is used to acquire trusted log entries from the blockchain and smart contract module; The indicator quantification calculation unit calculates the trusted review duration, review consistency, and reviewer historical baseline based on the trusted log entries; The review behavior pattern analysis unit uses the credible review duration and the reviewer's historical baseline to calculate the review duration Z-Score to identify abnormal behavior. The latent bias prediction unit is used to build feature vectors and run machine learning models to predict the probability of bias in a single review.

7. The online digital review platform for standards according to claim 6, characterized in that, The trusted review duration is calculated by extracting the start and end timestamps from the trusted log entries.

8. The online digital review platform for standards according to claim 6, characterized in that, The consistency of the review is calculated by extracting the review scores from the trusted log entries submitted by all reviewers for the same project.

9. The online digital review platform for standards according to claim 6, characterized in that, In the potential bias prediction unit, the feature vector includes at least: the credible review duration, the review score, and the review duration Z-Score.

10. The online digital review platform for standards according to claim 6, characterized in that, The AI ​​analysis engine module also includes: A composite alarm trigger is used to generate an alarm signal and send it to the platform application and interaction module when the review duration Z-Score or bias prediction probability meets the preset alarm rules. The platform application and interaction module can also initiate an intervention transaction to the blockchain and smart contract module based on the alarm signal, forcibly changing the status of the review project corresponding to the alarm signal from the archived status to the pending review status.

Images