A Method and System for Mutual Exclusion Detection and Resolution of Document Rule Effects under Multi-Specification Parallelism

By analyzing the rule execution records of the document management system, the system automatically detects and resolves mutual exclusion of effects between rules, thus solving the performance anomaly problem caused by resource contention in the existing technology and realizing rapid fault location and system optimization.

CN123114633APending Publication Date: 2026-11-17SHANDONG XIEHE UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611402722.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-10
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

In existing document management systems, when multiple automation rules are executed in parallel, the mutual exclusion of effects caused by resource contention or sequential dependencies cannot be effectively detected and resolved by existing technologies, resulting in long performance anomaly localization times and low efficiency.

Method used

By extracting rule execution records from the database or logs of the document management system, statistical methods are used to detect mutual exclusion of effects between rule pairs, and the root cause type is determined based on conflict characteristics. A resolution draft is generated, and rule configuration update instructions are automatically generated to resolve mutual exclusion of effects.

Benefits of technology

It automatically identifies and resolves mutually exclusive effects, reducing fault location time from days to minutes, improving document review efficiency, reducing mis-reviews and omissions, and achieving adaptive optimization of the system.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123114633A_ABST
    Figure CN123114633A_ABST
Patent Text Reader

Abstract

This invention aims to solve the problem of mutual exclusion effects causing a sharp performance drop due to resource contention when multiple rules are executed in parallel in a document management system. It provides a method and system for detecting and resolving mutual exclusion effects of document rules under multi-specification parallel execution. The method includes extracting execution records of each document rule from the document management system's database or logs; combining document rules in pairs to form rule pairs; performing mutual exclusion detection on each rule pair using statistical methods to determine if mutual exclusion exists; identifying rule pairs determined to be mutually exclusive or suspected of being mutually exclusive; determining the conflict root cause type based on conflict characteristics; and matching and outputting corresponding resolution drafts based on the root cause type. This invention automatically identifies mutual exclusion effects and generates resolution schemes such as merging rules by analyzing historical rule execution data, aiming to reduce fault detection time from days to minutes, without manual intervention.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of document automation governance technology in software engineering, and in particular to a method and system for detecting and resolving mutual exclusion of document rule effects under multiple parallel specifications. Background Technology

[0002] In large software companies in industries such as finance, telecommunications, and manufacturing, the automated review of software development documents, such as API documentation and code comments, increasingly relies on rule-driven systems. A typical document management pipeline may simultaneously implement multiple sets of standards, including coding style checks, API integrity verification, security vulnerability scanning, and terminology consistency verification, corresponding to dozens or even hundreds of automated rules. These rules are executed in parallel during document submission, modification, or release to ensure document quality.

[0003] However, parallel execution of multiple rules introduces a hidden problem: mutually exclusive effects. For example, in an API documentation generation pipeline, the "API integrity check rule" needs to load approximately 50MB of OpenAPI schema into memory, while the "security scan rule" needs to read the metadata of each documentation file. Each rule takes approximately 0.4 seconds and 0.3 seconds to execute independently. However, when they execute simultaneously, the total execution time spikes to 1.6 seconds due to competition for memory bandwidth and disk I / O. Similar phenomena are common in scenarios such as code comment inspection and test case verification.

[0004] Existing rule engines (such as Drools and EasyRules) and document management systems can only detect "conditional exclusion"—that is, two rules whose triggering conditions overlap but whose execution results contradict each other. For effect exclusion, existing technologies cannot define or detect it. When encountering performance anomalies, operations personnel usually suspect the inefficiency of a single rule or insufficient system resources, rather than the interaction effects between rules, resulting in an average fault location time of several days. According to statistics from a certain enterprise's API documentation platform, approximately 60% of performance anomaly events were later confirmed to be caused by effect exclusion, but the average fault discovery time still reached 3 days.

[0005] Current document rule management systems typically rely on pre-set static rules (such as format validation, permission checks, and terminology consistency audits) for automated document processing. As the number of rules increases and their logic becomes more complex, concurrent or sequential execution of different rules on the same document often leads to mutual interference due to shared computing resources (CPU, memory, I / O) or hidden dependencies. This results in decreased overall rule execution efficiency, increased response latency, and even missed or incorrect reviews. Currently, there is a lack of systematic automatic detection methods to address the performance mutual exclusion problem between rules in document management systems. On the one hand, existing solutions often focus on logical contradictions within the rule conditions themselves (such as conflicts between "allow" and "prohibit"), neglecting the negative performance interaction effects caused by resource contention or sequential dependencies when two rules are executed simultaneously. On the other hand, existing performance interference detection technologies are mainly geared towards operating system-level tasks or application-level services, failing to penetrate to the document rule granularity. Furthermore, they largely rely on manually set fixed thresholds or complex machine learning models, making it difficult to dynamically adapt to changes in rule execution behavior and lacking the ability to automatically identify and resolve mutual exclusion root causes. Meanwhile, traditional methods lack robust estimation mechanisms when co-occurring samples are insufficient, and cannot continuously optimize detection parameters through feedback loops.

[0006] Therefore, there is an urgent need for a technology that can automatically detect mutually exclusive rule effects in a document management system, identify the root cause type, and generate a resolution solution. Summary of the Invention

[0007] To address the problems existing in the prior art, this paper provides a method and system for detecting and resolving mutual exclusion of document rule effects under multiple parallel specifications. By quantitatively analyzing the historical performance data of parallel rule execution, it automatically identifies mutual exclusion of effects and provides a feasible resolution solution. This aims to solve the problem of mutual exclusion of effects when multiple automated rules are executed in parallel in existing software development document management systems.

[0008] To achieve the above objectives, the technical solution adopted by the present invention is as follows: On the one hand, this invention provides a method for detecting and resolving mutual exclusion of document rule effects under multi-specification parallelism, comprising the following steps: Extract the execution record, document characteristics, and rule configuration snapshot for each document rule from the document management system's database or logs; The document rules are combined in pairs to form rule pairs. For each rule pair, a statistical method is used to detect mutual exclusion of effects to determine whether there is mutual exclusion of effects. For rule pairs that are determined to have mutually exclusive effects or are suspected to have mutually exclusive effects, the conflict root cause type is identified based on the conflict characteristics, and the corresponding resolution draft is matched and output based on the root cause type.

[0009] Furthermore, the above-mentioned method for detecting and resolving mutual exclusion of document rule effects under multiple parallel specifications also includes: The resolution draft is output to the application interface of the rule engine in the form of structured data. The rule engine automatically generates rule configuration update instructions and performs rule hot update according to the resolution strategy. The performance metrics of rule execution are collected before and after the resolution strategy takes effect. These performance metrics are used as feedback samples to optimize and adjust the detection parameters in the mutual exclusion detection of effects, forming a closed-loop feedback and adaptive optimization of the detection parameters.

[0010] On the other hand, a system for detecting and resolving mutual exclusion of document rule effects under multi-norm parallelism is provided, for implementing the aforementioned method for detecting and resolving mutual exclusion of document rule effects under multi-norm parallelism, including: The data acquisition module is used to extract the execution record, document characteristics, and rule configuration snapshot of each document rule from the database or logs of the document management system. The mutual exclusion detection module is used to combine document rules into rule pairs, and for each rule pair, perform mutual exclusion detection using statistical methods to determine whether there are mutual exclusions. The resolution draft generation module is used to identify the conflict root cause type based on the conflict characteristics of rule pairs that are judged to have mutually exclusive effects or suspected mutually exclusive effects, and to match and output the corresponding resolution draft based on the root cause type.

[0011] On the other hand, the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-mentioned method for detecting and resolving mutual exclusion of document rule effects under multiple parallel specifications.

[0012] On the other hand, the present invention provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method for detecting and resolving mutual exclusion of document rule effects under multiple parallel specifications.

[0013] On the other hand, the present invention provides a computer program product stored on a computer-readable storage medium and including computer instructions, which, when executed by a processor, cause a computer device to implement the steps of the above-described method for detecting and resolving mutual exclusion of document rule effects under multiple parallel specifications.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a method for detecting and resolving mutual exclusion of document rule effects under multi-specification parallelism. The method includes extracting execution records of each document rule from the database or logs of a document management system; combining document rules in pairs to form rule pairs; performing mutual exclusion detection on each rule pair using statistical methods to determine if mutual exclusion exists; identifying the conflict root cause type based on conflict characteristics for rule pairs determined to be mutually exclusive or suspected of being mutually exclusive, and matching and outputting the corresponding resolution draft based on the root cause type. This invention automatically identifies mutual exclusion and generates resolution schemes such as merging rules by analyzing historical rule execution data, aiming to shorten fault detection time from days to minutes without manual intervention. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the structures shown in these drawings without creative effort.

[0016] Figure 1 This is a flowchart of a method for detecting and resolving mutual exclusion of document rule effects under multi-specification parallelism in one embodiment; Detailed Implementation The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0017] Reference Figure 1 One embodiment provides a method for detecting and resolving mutual exclusion of document rule effects under multiple parallel specifications, including the following steps: Extract the execution record, document characteristics, and rule configuration snapshot for each document rule from the document management system's database or logs; The document rules are combined in pairs to form rule pairs. For each rule pair, a statistical method is used to detect mutual exclusion of effects to determine whether there is mutual exclusion of effects. For rule pairs that are determined to have mutually exclusive effects or are suspected to have mutually exclusive effects, the conflict root cause type is identified based on the conflict characteristics, and the corresponding resolution draft is matched and output based on the root cause type.

[0018] The document rules include, but are not limited to, API integrity checks, security scans, terminology consistency checks, coding standard checks, and test case standard checks. Document management systems include, but are not limited to, software development document management systems. The software development document management rules in a software development document management system refer to a set of judgment conditions and execution logic used to automatically process documents during the software development process. Examples include: API integrity rules (checking if each parameter in the OpenAPI specification is described), security scan rules (detecting whether the document contains hard-coded keys), and terminology consistency rules (verifying whether the terminology matches the project's terminology library). To ensure document quality, document management systems typically enable multiple automated rules simultaneously, such as API integrity checks, security scans, and terminology checks, and execute them in parallel to improve efficiency. However, when these rules run simultaneously, they easily compete for resources such as CPU, memory, and disk I / O, leading to a sharp drop in overall processing performance—for example, two rules might only take 0.4 seconds and 0.3 seconds to execute individually, but in parallel, the time spikes to 1.6 seconds. This phenomenon is called "mutually exclusive effects." Mutual exclusion of effects refers to a phenomenon where two or more rules perform normally when executed individually, but when executed simultaneously on the same document, the overall processing performance deteriorates significantly due to contention for system resources such as CPU, memory bandwidth, and disk I / O. Existing rule engines can only detect logical contradictions (such as mutually exclusive conditions) and are completely unable to detect this performance degradation caused by resource contention. Operations personnel often spend several days locating the root cause. When multiple rules are executed in parallel, the method for detecting and resolving mutual exclusion of document rule effects under multi-specification parallelism provided in the above embodiments detects performance mutual exclusion phenomena caused by resource contention and automatically generates resolution solutions. This invention is applicable to software development document management scenarios such as automatic API document review, code comment quality inspection, and test case specification verification in large software enterprises. The core positioning of this invention is a rule governance engine that periodically analyzes the historical execution data of all rule pairs within the document management system, automatically detects mutual exclusion of effects, and provides administrators with actionable resolution suggestions.

[0019] Specifically, the first step is to connect to the document management system and collect historical execution records of each rule from the system, providing a data foundation for subsequent statistical analysis. The technical approach involves connecting to the document management system's database using a "full initialization + incremental periodic" strategy to collect rule execution records, document characteristics, and rule configuration snapshots. Read-only connections are established via JDBC or RESTful API, supporting integration with Jira, Confluence, GitLab, and self-developed platforms. Only SELECT permissions are requested, and data transmission uses TLS encryption.

[0020] The rule execution record includes rule ID, rule version, execution time, trigger document ID, execution time (milliseconds), execution result (success / failure), and resource usage (including peak memory (megabytes) and disk IOPS). The document characteristics include document ID, document type (API document / code comment / design document), document size (bytes), and number of pages; The rule configuration snapshot includes the rule ID, rule version, rule content (such as regular expressions), effective time, and expiration time.

[0021] For collecting historical execution records of each rule in the document management system, the initial run collects all data from the last 180 days (if insufficient, it starts from the earliest record). Subsequent runs incrementally collect newly added records every 6 hours. Each record is accompanied by a timestamp and a unique identifier to avoid duplication. After collecting historical execution records of each rule in the document management system, records with abnormal execution times (e.g., less than or equal to 0 milliseconds or greater than 5 minutes) or interrupted due to system errors are removed. Individual execution records and co-occurring execution records are aligned by document ID and timestamp. All data is uniformly stored in a Parquet format data warehouse, partitioned by date.

[0022] The mutual exclusion of effects refers to the situation where two rules perform normally when executed individually, but when executed simultaneously on the same document, the performance index value deteriorates due to system resource competition or other reasons, and the deterioration is statistically significant.

[0023] In a preferred embodiment, for each rule pair, mutual exclusion detection of effects is performed using statistical methods, including: For each rule pair, its individual execution records and co-occurrence execution records are merged into a record set corresponding to each rule pair. Each record in the record set carries a binary label. The binary label has a value of 0, which indicates that the record is an individual execution record, and a value of 1, which indicates that the record is a co-occurrence execution record. Individual execution records include execution records in which only the first rule in the rule pair is executed and execution records in which only the second rule in the rule pair is executed. Co-occurrence execution records are execution records in which the first rule and the second rule in the rule pair are executed simultaneously.

[0024] The sample size was calculated based on statistical power analysis (α=0.05, 1-β=0.8, moderate effect size d=0.5). If the rule version has changed, the interaction effect of different versions is calculated in segments to avoid cross-version data contamination. The rule for determining version change is: when the hash value of the rule code changes (the system can configure the change sensitivity, for example, only when the edit distance exceeds a certain proportion or the difference in structured content is greater than a preset threshold is it considered a significant change), the interaction effect is calculated in segments.

[0025] Count the number of records where each rule is currently executed by the CCP, and determine whether the preset sample size (e.g., no less than 30 records) has been reached. If the number of records where each rule is currently executed by the CCP has not reached the preset sample size, perform small sample approximation processing and output rule pairs with suspected mutual exclusion markers.

[0026] If the number of records executed by each rule against the CCP reaches a preset sample size threshold, then a mutual exclusion check will be performed on the following steps for the rule to execute: Calculate the performance metrics of all execution records in the record set where only the first rule is executed, the performance metrics of all execution records where only the second rule is executed, and the performance metrics of all execution records where both the first and second rules are executed simultaneously.

[0027] This invention extracts performance metrics (such as execution time, missed review rate, peak memory usage, etc.) for each rule executed individually and simultaneously with other rules from the database or logs of a document management system, and adaptively selects performance metrics based on the rule type. Specifically, for each rule's historical execution records, CPU utilization, memory utilization, disk I / O frequency, and deviation rate of the rule's output results are calculated; the dominant type of the rule is determined through a preset classification decision tree, and then the corresponding performance metrics are selected. For CPU-intensive or I / O-intensive rules (such as loading large schemas or scanning files), select execution time as the performance metric. For accuracy-sensitive rules (such as terminology validation), the missed review rate or the false review rate is selected as the performance indicator. For resource-intensive rules (such as memory loading), select peak memory (megabytes) or disk IOPS as performance metrics.

[0028] The interaction effect value Δ of the rule pair is calculated based on the performance metrics of all execution records in the record set where only the first rule is executed, all execution records where only the second rule is executed, and all execution records where both the first and second rules are executed simultaneously.

[0029] When calculating the interaction effect value Δ, only records within the time window with the same rule version are used; if the substantive content of the rule code changes, including but not limited to changes in rule logic, judgment conditions or execution actions, it is considered a rule version change, and the interaction effect value of different versions is calculated in segments; non-substantive changes such as code comments or formatting adjustments do not trigger version change determination.

[0030] The binary labels of all execution records in the record set are randomly shuffled while keeping the total number of labels with a value of 1 constant, resulting in a shuffled label sequence. Based on this shuffled label sequence, the records in the record set are regrouped, with records with a label of 0 grouped into a pseudo-isolated group and records with a label of 1 grouped into a pseudo-co-occurrence group. Then, the virtual interaction effect value is calculated. Repeat the random shuffling and the calculation process of the virtual interaction effect value after random shuffling M times (e.g., repeat 1000 times) to obtain the empirical distribution of the virtual interaction effect value composed of M virtual interaction effect values. The interaction effect values ​​of the rule pairs are compared with the empirical distribution of the dummy interaction effect values ​​to calculate the original p-values ​​of the rule pairs. , This represents the number of values ​​in the empirical distribution of the virtual interaction effect that are greater than or equal to the interaction effect value Δ. When detecting multiple rule pairs simultaneously, the Benjamini-Hochberg method is used to control the false detection rate (target FDR ≤ 5%). Specifically, the original p-values ​​of all rule pairs are obtained, and the Benjamini-Hochberg method is used to correct the original p-values ​​of each rule pair to obtain the corrected q-value for each rule pair. The q-value represents the adjusted false detection rate after correction by the Benjamini-Hochberg method. If q < the preset significance level, the effects of the corresponding rule pairs are considered mutually exclusive and significant in controlling for the multiple hypothesis testing error rate. Set dynamic threshold If the interaction effect value Δ of the rule pair is greater than the threshold If q < preset significance level, then the judgment rules are mutually exclusive in terms of their effects. The default value of the preset significance level is 0.05.

[0031] In this invention, mutual exclusion of effects is quantified as follows: the actual performance index value when two rules are executed simultaneously exhibits a statistically significant degradation compared to the maximum performance index value when each rule is executed individually. It is important to note that "mutual exclusion conflict" in existing technologies typically refers to logical contradictions. The "mutual exclusion of effects" defined in this invention is fundamentally different: the former is a static conflict at the logical level, which can be discovered through static analysis; the latter is a dynamic conflict at the performance level, which must be detected based on statistical analysis of historical performance data, and its root cause is resource competition rather than logical contradiction. The mutual exclusion of effects proposed in this invention cannot be discovered through static analysis and must be identified based on statistical analysis of historical execution data.

[0032] Specifically, suppose there are N historical records for the rule pair (R1, R2), where n1 and n2 represent the individual executions of the two rules, and n3 represents the co-executions (N = n1 + n2 + n3). Each record contains a performance metric value. and tags (0 represents execution alone, 1 represents execution in co-occurrence). Wherein Indicates a record index. The permutation test process is as follows: Randomly shuffle the label sequence of all records (keeping the total number of label 1 records constant at n³), regroup based on the shuffled labels, and calculate Δ_perm. Repeat this process M times (M≥1000) to obtain M Δ_perm values. Compare the originally calculated Δ with these M Δ_perm values, calculate the proportion of the original Δ that exceeds or equals the empirical distribution, and obtain the p-value.

[0033] In one embodiment, the interaction effect value Δ of the rule pair is calculated as follows: The median or mean of the performance metrics of all execution records in the record set where only the first rule is executed, the performance metrics of all execution records where only the second rule is executed, and the performance metrics of all execution records where both the first and second rules are executed are taken as the performance metric values ​​when only the first rule is executed, the performance metric values ​​when only the second rule is executed, and the performance metric values ​​when both the first and second rules are executed, respectively. The interaction effect value Δ of a rule pair is the difference between the actual performance of the two rules when executed simultaneously and the maximum performance of each rule when executed individually. The formula is as follows:

[0034] Where T_together is the performance metric value when the first rule and the second rule are executed simultaneously, and T_alone1 and T_alone2 are the performance metric values ​​when only the first rule is executed and when only the second rule is executed, respectively. Positive interaction values ​​indicate performance degradation (mutually exclusive effects), while negative values ​​indicate synergistic effects.

[0035] The interaction effect value of the rule pairs recalculated after each random shuffling of labels is defined as the virtual interaction effect value. The calculation formula is as follows:

[0036] in This represents the median or mean of the performance metrics for all execution records in the pseudo-co-occurrence group. , These are the median or mean of the performance metrics for all execution records in the pseudo-isolated group that belong to the execution of only the first rule, and the median or mean of the performance metrics for all execution records in the pseudo-isolated group that belong to the execution of only the second rule.

[0037] In one embodiment, a dynamic threshold is set based on the following method. : For each rule pair, a sliding window is maintained (the window width is a preset value, defaulting to 30 days; this preset value can be adjusted according to system load characteristics and rule execution frequency, typically ranging from 10 to 90 days; windows shorter than 10 days are more susceptible to random fluctuations, while windows longer than 90 days introduce excessive historical noise). The mean μ_history and standard deviation σ_history are obtained based on the interaction effect value sequence of all co-occurring execution records within the sliding window. The above ranges are determined based on statistical power analysis: a sample size shorter than 10 days is insufficient to meet the minimum sample requirement, and a sample size longer than 90 days introduces noise exceeding the signal gain.

[0038] Dynamic threshold , where k is a configurable coefficient. The initial value of k is determined based on the quantiles of the normal interaction effects of historically mutually exclusive rule pairs with no effects (normal rule pairs can be manually labeled or pre-selected based on business experience). Specifically, it is determined as follows: obtain the normal interaction effect sequence of historically mutually exclusive rule pairs with no effects and calculate the 95th percentile of the sequence, denoted as P_percentile, and simultaneously calculate the standard deviation of the sequence, denoted as σ_history_normal. If a pair of mutually exclusive rules with no historical effect cannot be obtained, then k=3 is used as the initial value.

[0039] In another embodiment, to enhance adaptability to non-normally distributed data, a dynamic threshold is set based on the following method. : The Shapiro-Wilk test is used to determine whether the interaction effect value sequence follows a normal distribution; When the original p-value of a rule pair is greater than 0.05, the interaction effect value sequence is determined to follow a normal distribution. For each rule pair, a sliding window is maintained. Based on the interaction effect value sequence of all co-occurring execution records within the sliding window, its mean μ_history and standard deviation σ_history are obtained, along with a dynamic threshold. , where k is a configurable coefficient; When the original P-value of the rule pair is less than or equal to 0.05, it is determined that the interaction effect value sequence does not follow a normal distribution. The dynamic threshold threshold = percentile(Δ_history, 95%) is calculated using the nonparametric quantile method, where percentile(Δ_history, 95%) represents the 95th percentile of the historical interaction effect sequence.

[0040] Furthermore, severity levels can be categorized, and specific recommended actions can be given based on the severity level. In one embodiment, the severity levels, judgment criteria, and corresponding recommended actions are shown in Table 1: Table 1

[0041] Note: The same significance threshold (q<0.05) is used for both moderate and severe cases to ensure consistency in judgment. The preset significance level can be configured to other values ​​such as 0.01 according to actual business needs.

[0042] In one embodiment, if the number of currently executed records for a rule does not reach a preset sample size, a small sample approximation process is performed, outputting rule pairs that are suspected to have mutually exclusive effects, including the following steps: When the number of co-occurring execution records is lower than a preset sample size threshold, the interaction effect value is estimated using the sample mean.

[0043] in The average performance metric for all co-occurring execution records. , These are the average performance metrics for executing only the first rule and executing only the second rule, respectively. The posterior mean was calculated using the empirical Bayesian method. ,in The total number of co-occurring execution records, The prior mean, This is the prior strength parameter. The default value of λ is 10, and this value is determined through cross-validation (selecting the value that minimizes the mean square error of the posterior estimate within the range of λ∈[5,20]) or the BIC criterion. The value of λ can also be pre-configured by the system administrator according to the business scenario, with a configuration range of 5~20.

[0044] like Divide by If the ratio exceeds a preset relative threshold (typically 20%), it is marked as a rule pair with suspected mutually exclusive effects. .

[0045] When the co-occurrence sample size is lower than a preset sample size threshold (typically 30 samples), the sample median fluctuates significantly. Therefore, the sample mean is used to estimate the interaction effect, i.e., let... The interaction effect is the arithmetic mean of the differences between the performance metrics of co-occurrence records and the maximum values ​​of individual executions. Based on this, an empirical Bayesian method is used to robustly estimate the interaction effect. It is assumed that globally mutually exclusive rules with no effect follow a prior distribution for the interaction effect. The estimation method for prior parameters is as follows: If there are no pre-labeled normal rule pairs, a stratified random sampling method is used. First, stratify according to rule type, then randomly sample 100 pairs within each stratum, and remove outlier samples whose interaction effect values ​​exceed three standard deviations or the 99th percentile of the global mean. Calculate the mean and variance of the interaction effect as the prior. If there are pre-labeled normal rule pairs, their mean and variance are used directly. The reason for choosing the normal distribution as the prior distribution is that, according to the central limit theorem, with a sufficient sample size, the distribution of the sample mean approaches a normal distribution; and according to the goodness-of-fit test of historical data, the normal distribution has the best fit.

[0046] After obtaining the prior distribution parameters, the posterior mean is calculated using Bayes' theorem, combined with small sample observations of the current rule pairs. The posterior mean is a contracted estimator of the interaction effect; it integrates global prior information (reflecting typical interaction effects of normal rule pairs) with the current small sample observations, offering a more comprehensive estimate compared to simply using the sample mean. More stable.

[0047] Next, for rule pairs determined to be mutually exclusive or potentially mutually exclusive, their root causes (resource competition, sequence dependency, logical contradiction) are analyzed, and corresponding resolution strategies are matched. In one embodiment, for rule pairs determined to be mutually exclusive or potentially mutually exclusive, the conflict root cause type is determined based on conflict characteristics. The conflict root cause types include at least resource competition, sequence dependency, and logical contradiction, wherein the methods for determining resource competition, sequence dependency, and logical contradiction include: If both rules in a rule pair involve a large amount of I / O or memory allocation, and the number of processor cores is limited, then it is determined to be a resource contention type. Check whether there is a non-empty intersection between the input parameter set of the second rule and the output result set of the first rule in the rule pair. If the second rule must wait for the intermediate state data of the first rule before execution, and the intermediate state data includes shared memory objects or calculation results, then it is determined to be a sequential dependency. If the conditions of the first and second rules in a rule pair are logically mutually exclusive, then the pair is classified as logically contradictory. This determination can be achieved by statically analyzing the conditional expressions of the rules, for example, by using symbolic execution or the conflict detection function built into the rule engine.

[0048] In one embodiment, for rule pairs determined to be non-resource contention-type, a corresponding resolution strategy is directly matched from a pre-set strategy library based on the root cause type as a resolution draft. There are no restrictions on the corresponding resolution strategies matched from the pre-set strategy library; those skilled in the art can make reasonable designs and selections based on existing technology.

[0049] Without loss of generality, the pre-built strategy library in one embodiment is shown in Table 2: Table 2

[0050] For rule pairs determined to be resource-competitive, the corresponding resolution strategy is directly matched from the pre-built strategy library as the resolution draft, or the resolution draft of the resource-competitive rule pair is generated according to the following methods, including: Lexical and syntactic analysis (e.g., based on ANTLR or a self-developed parser) is performed on the source code of each rule in the rule pair to extract the core elements of the rule, including the input data source type, the output field set, the execution prerequisites, and resource dependencies. The input data source type includes file path, API endpoint, and database table name, and the resource dependencies include file handle, database connection, and API call.

[0051] The source code of the rules is represented using a standardized DSL (Domain-Specific Language), such as a JSON Schema-based format. The input and output of each rule are annotated with standardized metadata fields. Specifically, the parser uses regular expressions to match input data source patterns: for example, matching `file\.open\(['"]([^'"]+)['"]\)` extracts the filename, matching `api\.call\(['"]([^'"]+)['"]\)` extracts the API endpoint, and matching `SELECT.*FROM` extracts the database table name. Output fields are identified similarly, matching patterns such as `return\s+[{\[]`. Execution preconditions are extracted by matching conditional statements such as `if\s*\(.*\)`; resource dependencies (such as file handles, database connections, and API calls) are extracted by matching function calls such as `fopen\|connect\|open`.

[0052] Compare the input data sources extracted by two rules to identify shareable resource types: If the input data sources of the two rules in a rule pair contain regular expression matching the same file path, then the file metadata cache is determined to be shareable; if the two rules call the same API endpoint (by comparing the endpoint URL string), then the API response cache is determined to be shareable; if the two rules access the same database table, then the database query result cache is determined to be shareable. For example: if the code of rule R1 contains `file.getCreationTime()` and `file.getSize()`, and the code of rule R2 contains `file.getMetaData()`, then both require file metadata, therefore the file metadata cache can be shared.

[0053] Based on the extracted execution preconditions and output field set, a data dependency graph G=(V,E) is constructed, where node V represents the execution steps of each rule and potential cache nodes, and directed edge E represents data dependency relationship. The dependency relationship determination rule is as follows: if the intersection of the output field set O(R1) of the first rule R1 and the input field set I(R2) of the second rule R2 in the rule pair is not empty, then add a directed edge R1→R2; if the first rule R1 and the second rule R2 in the rule pair share cacheable resources, then add an undirected shared edge between them and introduce a cache node.

[0054] Perform topological sorting on the data dependency graph to generate an execution order sequence. If there is a circular dependency (such as R1 depending on R2 and R2 depending on R1), stop automatic merging and output a prompt for manual intervention. Automatic merging is not recommended. Based on the above analysis results, a composite rule configuration in JSON / YAML format is generated as a resolution draft. This composite rule configuration includes at least a shared cache definition, an execution step sequence, and dependencies between each execution step. The shared cache definition includes at least the cache name, type, and lifecycle.

[0055] The following is an example configuration: { "shared_cache": { "name": "file_metadata_cache", "type": "memory", "ttl_seconds": 300}, "execution_steps": [ { "step_id": "step1", "source_rule": "API integrity check", "input": "cache", "output": "intermediate_result"}, { "step_id": "step2", "source_rule": "security scan", "input": "cache", "depends_on": ["step1"]} ], "final_output": { "combine": ["step1_output", "step2_output"]} } The "input": "cache" option indicates that data from a shared cache is being used.

[0056] For complex dependency scenarios that cannot be automatically merged, the system will downgrade to pushing manual intervention suggestions instead of forcibly generating a draft.

[0057] Thus, this invention automatically identifies rule pairs with mutually exclusive resource-competitive effects by analyzing historical rule execution data and using a combination of permutation testing and dynamic thresholding, without manual intervention. For detected mutually exclusive effects, especially resource-competitive ones, it automatically generates candidate configurations for merging rules (including shared cache definitions, execution step sequences, and dependencies between execution steps), allowing administrators to apply them with a single click. This invention significantly reduces fault location time: reducing the time to discover performance anomalies from several days of traditional manual troubleshooting to minutes. By resolving mutually exclusive effects, it significantly reduces the average document review time while ensuring that detection accuracy and error detection rate meet practical engineering requirements.

[0058] Furthermore, the detection results and resolution drafts can be pushed to administrators in a visual format, supporting one-click application. A mutual exclusion detection report is generated every 24 hours, including: a list of mutually exclusive rule pairs, an interaction effect trend graph, severity, and recommended resolution strategies. A plugin is embedded in the document management system's management backend to display mutually exclusive rule pairs in tabular form, highlighting their severity. Administrators can select "Adopt Merge Draft," "Adjust Order," or "Ignore" for each rule. After adoption, the system dynamically updates the rule configuration via the rule engine's API without requiring a service restart. For merged rules, the system marks the original two rules as "deprecated" and deploys the composite rule.

[0059] Furthermore, in one embodiment of a method for detecting and resolving mutual exclusion of document rule effects under multi-specification parallelism, after obtaining the resolution draft, the method further includes: The resolution draft is output to the application interface of the rule engine in the form of structured data. The rule engine automatically generates rule configuration update instructions and performs rule hot update according to the resolution strategy. The system collects performance metrics changes in rule execution before and after the resolution strategy takes effect. These performance metrics changes are used as feedback samples to optimize and adjust the detection parameters in the mutual exclusion detection, forming a closed-loop feedback and adaptive optimization of the detection parameters. This feedback optimization loop allows for dynamic optimization of the detection threshold parameter (the k-value in the dynamic threshold) based on real-time collection of performance metrics changes in rule execution before and after the resolution strategy takes effect, or based on the administrator's adoption of resolution suggestions and actual performance improvements, thus achieving adaptive evolution of the system. The k-value is dynamically optimized as follows: a validation set is set in the historical feedback samples, the candidate k-value set is traversed, and the k-value that maximizes the accuracy of mutual exclusion detection is selected and updated.

[0060] Furthermore, in one embodiment, a method for detecting and resolving mutual exclusion of document rule effects under multi-specification parallelism includes recording the administrator's adoption of resolution suggestions and dynamically optimizing the detection threshold and strategy matching, including: Feedback collection: Record the rule pair ID, predicted severity, recommended strategy, administrator actions (accept / reject / modify), and actual performance improvement for each alert.

[0061] Threshold parameter update: Whenever the accumulated number of feedback samples reaches a preset threshold (default is 100), the k value in the dynamic threshold is recalculated. The specific optimization method is as follows: A validation set is created from the historical feedback samples; the set of candidate k values ​​(k from 0.5 to 4.0, with a step size of 0.1) is traversed; the k value that maximizes the mutual exclusion detection accuracy (or F1 score) is selected, and the deployment is updated. This optimization process uses a deterministic search algorithm and does not involve training a machine learning model.

[0062] Policy library optimization (optional enhancement): Reduce the priority of recommendation policies that administrators frequently "reject"; increase the recommendation weight of policies that are "adopted" and show significant performance improvements.

[0063] In another embodiment, a system for detecting and resolving mutual exclusion of document rule effects under multi-specification parallelism is provided. This system implements the method for detecting and resolving mutual exclusion of document rule effects under multi-specification parallelism described in any of the above embodiments. The system can be implemented using a computer program, runs on at least one processor and memory, and interfaces with a document management system through a data acquisition interface. Specifically, the system includes: The data acquisition module is used to extract the execution record, document characteristics, and rule configuration snapshot of each document rule from the database or logs of the document management system. The mutual exclusion detection module is used to combine document rules into rule pairs, and for each rule pair, perform mutual exclusion detection using statistical methods to determine whether there are mutual exclusions. The resolution draft generation module is used to identify the conflict root cause type based on the conflict characteristics of rule pairs that are judged to have mutually exclusive effects or suspected mutually exclusive effects, and to match and output the corresponding resolution draft based on the root cause type.

[0064] Furthermore, the aforementioned system also includes: The application module is used to output the resolution draft in the form of structured data to the application interface of the rule engine. The rule engine automatically generates rule configuration update instructions and performs rule hot updates according to the resolution strategy. The feedback self-optimization module is used to collect the performance index changes of rule execution before and after the resolution strategy takes effect. The performance index changes are used as feedback samples to optimize and adjust the detection parameters in the effect mutual exclusion detection, forming a closed-loop feedback and adaptive optimization of the detection parameters.

[0065] The modules described above work together to automatically detect and adaptively resolve mutually exclusive document rule effects. Specifically, the implementation of each component module is the same as that described in the corresponding method embodiments above, and will not be repeated here.

[0066] This invention enables automatic detection and adaptive resolution of mutual exclusion of effects in a document management system under a multi-specification parallel execution environment. This significantly reduces pipeline latency caused by performance competition between rules and avoids the inefficiency of manual review. The system can be integrated into API documentation generation pipelines (such as SwaggerHub and GitLab CI), code comment checking tools, and enterprise document compliance review systems as an intelligent module for rule governance. It is particularly suitable for medium to large-sized software enterprises with numerous document specifications, rapidly increasing rule numbers, and increasingly prominent issues of mutual exclusion of effects. This solution is applicable to scenarios with high co-occurrence frequency of rule pairs (a minimum of 30 co-occurrences is recommended). For low-frequency co-occurrence rule pairs, manual review or empirical Bayesian approximation processing can be used.

[0067] Below is an example of a specific application scenario: Scenario setting: An API documentation auto-review pipeline runs three rules simultaneously, as follows: Rule R1: API integrity check. Requires loading the OpenAPI schema (approximately 50MB) into memory and parsing the parameters and return values ​​of each interface. Execution alone takes approximately 0.4 seconds.

[0068] Rule R2: Security Scan Rule. Reads the metadata (creation time, modifier, file size) of each document file and matches it against sensitive information patterns. Execution takes approximately 0.3 seconds.

[0069] Rule R3: Terminology Consistency Rule. Loads the project terminology library (approximately 200MB) and validates the terminology in the documentation. Execution alone takes approximately 0.5 seconds.

[0070] The pipeline processes approximately 200 API documents per day. Historical data records show that 210 documents were executed simultaneously by R1 and R2, 180 documents by R1 and R3, and 50 documents by R2 and R3.

[0071] Implementation process: Data collection: All rule execution records for the past 90 days were collected from the system's log database, totaling approximately 11,700 records. Individual execution and co-occurrence execution records were aligned by document ID and timestamp.

[0072] Effect mutual exclusion detection: For the rule pair (R1, R2): T_alone1 = 0.4s, T_alone2 = 0.3s, T_together median = 1.6s. We use Δ = 1.6 - max(0.4, 0.3) = 1.2s.

[0073] Permutation test: The labels were randomly shuffled 1000 times, resulting in a Δ_perm distribution with p=0.003. After BH correction, q=0.008 (assuming FDR=5%).

[0074] Dynamic threshold: Sliding window (last 30 days) μ_history=0.9s, σ_history=0.2s, initial k=1.2, threshold=0.9+1.2×0.2=1.14s. Δ=1.2s>threshold, and q=0.008<0.05, indicating mutually exclusive severe effects.

[0075] (R1,R3): T_together=1.2s, Δ=1.2-0.5=0.7s, p=0.12, after BH correction q=0.15>0.05, no mutual exclusion effect.

[0076] (R2,R3): T_together=1.4s, Δ=1.4-0.5=0.9s, p=0.08, q=0.10>0.05, no mutual exclusion.

[0077] Root cause analysis and strategy recommendation: The system determines that (R1,R2) is a resource contention-type mutually exclusive effect, recommends the "merge rule" strategy, generates a composite rule draft, and shares the file metadata cache.

[0078] Warning and Adoption: The administrator receives the warning, reviews the draft, and confirms the application. After merging, the composite rule first loads the file metadata into the cache once, and then performs integrity checks and security scans sequentially. The actual average execution time is reduced to 0.58 seconds, a 64% performance improvement compared to the original parallel execution (1.6s). Based on a survey of 50 developers, the adoption rate of this suggestion is approximately 82%.

[0079] Feedback loop: Adoption events are recorded and used to optimize subsequent k values.

[0080] Mutually exclusive recognition rate: In this example, it was correctly recognized (R1 and R2 were detected, while R1, R3 and R2, R3 were not detected).

[0081] Conflict detection time: reduced from approximately 3 days (manual investigation) to 0.5 hours.

[0082] The average time for document review has decreased by 64% (from 1.6 seconds to 0.58 seconds).

[0083] On the other hand, the present invention provides a computer device including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the document rule effect mutual exclusion detection and resolution method provided in any of the above embodiments under multi-specification parallelism. The computer device can be a server. The computer device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device stores sample data. The network interface of the computer device is used for communication with external terminals via a network connection.

[0084] On the other hand, the present invention provides a computer-readable storage medium storing a computer program thereon, wherein when the computer program is executed by a processor, it implements the steps of the method for mutual exclusion detection and resolution of document rule effects under multi-specification parallelism provided in any of the above embodiments.

[0085] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0086] Matters not covered in this invention are common knowledge.

[0087] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0088] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism, characterized in that, Includes the following steps: Extract the execution record, document characteristics, and rule configuration snapshot for each document rule from the document management system's database or logs; The document rules are combined in pairs to form rule pairs. For each rule pair, a statistical method is used to detect mutual exclusion of effects to determine whether there is mutual exclusion of effects. For rule pairs that are determined to have mutually exclusive effects or are suspected to have mutually exclusive effects, the conflict root cause type is identified based on the conflict characteristics, and the corresponding resolution draft is matched and output based on the root cause type.

2. The method for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism as described in claim 1, characterized in that, Also includes: The resolution draft is output to the application interface of the rule engine in the form of structured data. The rule engine automatically generates rule configuration update instructions and performs rule hot update according to the resolution strategy. The performance metrics of rule execution are collected before and after the resolution strategy takes effect. These performance metrics are used as feedback samples to optimize and adjust the detection parameters in the mutual exclusion detection of effects, forming a closed-loop feedback and adaptive optimization of the detection parameters.

3. The method for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism according to claim 1 or 2, characterized in that, The rule execution record includes rule ID, rule version, execution time, trigger document ID, execution time, execution result, and resource consumption; the document characteristics include document ID, document type, document size, and number of pages; the rule configuration snapshot includes rule ID, rule version, rule content, effective time, and expiration time. For each rule pair, a statistical method is used to perform mutual exclusion detection of effects, including: For each rule pair, its individual execution records and co-occurrence execution records are merged into a record set corresponding to each rule pair. Each record in the record set carries a binary tag. The value of the binary tag is 0, which indicates that the record is an individual execution record, and a value of 1 indicates that the record is a co-occurrence execution record. Individual execution records include execution records in which only the first rule in the rule pair is executed and execution records in which only the second rule in the rule pair is executed. Co-occurrence execution records are execution records in which the first rule and the second rule in the rule pair are executed simultaneously. The number of records of each rule being executed against the CCP is counted to determine if the preset sample size has been reached. If the preset sample size threshold is reached, the following steps are performed to check the mutual exclusion of the effects of each rule's execution: Based on the rule type in the rule pair, select the performance index, calculate the performance index of all execution records in the record set where only the first rule is executed, the performance index of all execution records where only the second rule is executed, and the performance index of all execution records where the first rule and the second rule are executed simultaneously, and then calculate the interaction effect value Δ of the rule pair; The binary labels of all execution records in the record set are randomly shuffled while keeping the total number of labels with a value of 1 constant, resulting in a shuffled label sequence. Based on this shuffled label sequence, the records in the record set are regrouped, with records with a label of 0 grouped into a pseudo-isolated group and records with a label of 1 grouped into a pseudo-co-occurrence group. Then, the virtual interaction effect value is calculated. Repeat the process of randomly shuffling and calculating the virtual interaction effect value M times to obtain the empirical distribution of the virtual interaction effect value consisting of M virtual interaction effect values. The interaction effect values ​​of the rule pairs are compared with the empirical distribution of the dummy interaction effect values ​​to calculate the original p-values ​​of the rule pairs. , This represents the number of values ​​in the empirical distribution of the virtual interaction effect value that are greater than or equal to the interaction effect value Δ. Obtain the original p-values ​​for all rule pairs and correct them using the Benjamini-Hochberg method to obtain the corrected q-values ​​for each rule pair. The q-value represents the adjusted false detection rate after correction using the Benjamini-Hochberg method. If q < the preset significance level, the effects of the corresponding rule pairs are considered to be mutually exclusive and significant in terms of controlling for the error rate of multiple hypothesis testing. Set dynamic threshold If the interaction effect value Δ of the rule pair is greater than the threshold If q < preset significance level, then the judgment rules are mutually exclusive in terms of effect. The default value of the preset significance level is 0.

05. If the number of records currently executed by the rule against the CCP does not reach the preset sample size, then small sample approximation processing is performed, and rule pairs with suspected mutually exclusive effects are output.

4. The method for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism as described in claim 3, characterized in that, The interaction effect value Δ of the rule pair is calculated as follows: The median or mean of the performance metrics of all execution records in the record set where only the first rule is executed, the performance metrics of all execution records where only the second rule is executed, and the performance metrics of all execution records where both the first and second rules are executed are taken as the performance metric values ​​when only the first rule is executed, the performance metric values ​​when only the second rule is executed, and the performance metric values ​​when both the first and second rules are executed, respectively. The interaction effect value Δ of the rule pair is calculated using the following formula: Where T_together is the performance index value when the first rule and the second rule are executed simultaneously, and T_alone1 and T_alone2 are the performance index values ​​when only the first rule is executed and when only the second rule is executed, respectively; Virtual interaction effect value The calculation formula is as follows: in This represents the median or mean of the performance metrics for all execution records in the pseudo-co-occurrence group. , These are the median or mean of the performance metrics for all execution records in the pseudo-isolated group that belong to the execution of only the first rule, and the median or mean of the performance metrics for all execution records in the pseudo-isolated group that belong to the execution of only the second rule.

5. The method for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism as described in claim 4, characterized in that, Set dynamic thresholds using any of the following methods : For each rule pair, maintain a sliding window. Based on the interaction effect value sequence of all co-occurring execution records within the sliding window, obtain its mean μ_history and standard deviation σ_history, and a dynamic threshold. , where k is a configurable coefficient. Alternatively, the Shapiro-Wilk test can be used to determine whether the interaction effect value sequence follows a normal distribution. When the original p-value of the rule pair is greater than 0.05, the interaction effect value sequence is considered to follow a normal distribution. For each rule pair, a sliding window is maintained, and based on the interaction effect value sequence of all co-occurring execution records within the sliding window, its mean μ_history and standard deviation σ_history are obtained, along with a dynamic threshold. , where k is a configurable coefficient; when the original P-value of the rule pair is less than or equal to 0.05, it is determined that the interaction effect value sequence does not follow a normal distribution, and the dynamic threshold threshold=percentile(Δ_history, 95%) is calculated using the nonparametric quantile method, where percentile(Δ_history, 95%) represents the 95th percentile of the historical interaction effect sequence.

6. The method for mutual exclusion detection and resolution of document rule effects under multi-standard parallelism as described in claim 5, characterized in that, The configurable coefficient k is determined in the following way: Obtain the normal interaction effect sequence of mutually exclusive rule pairs with no historical effect and calculate the 95th percentile of the sequence, denoted as P_percentile. Simultaneously, calculate the standard deviation of the sequence, denoted as σ_history_normal. ; If a pair of mutually exclusive rules with no historical effect cannot be obtained, then k=3 is used as the initial value.

7. The method for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism as described in claim 3, characterized in that, If the number of records of the rule being enforced by the CCP does not reach the preset sample size, then a small sample approximation process is performed, outputting rule pairs that are suspected to have mutually exclusive effects, including the following steps: When the number of co-occurring execution records is lower than a preset sample size threshold, the interaction effect value is estimated using the sample mean. in The average performance metric for all co-occurring execution records. , These are the average performance metrics for executing only the first rule and executing only the second rule, respectively. The posterior mean was calculated using the empirical Bayesian method. ,in The total number of co-occurring execution records, The prior mean, These are prior strength parameters; like Divide by If the ratio exceeds a preset relative threshold, it is marked as a rule pair with suspected mutually exclusive effects. .

8. The method for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism as described in claim 3, characterized in that, For rule pairs judged to have mutually exclusive effects or suspected mutually exclusive effects, the conflict root cause type is determined based on conflict characteristics. Conflict root cause types include at least resource competition, sequential dependency, and logical contradiction. The methods for determining resource competition, sequential dependency, and logical contradiction include: If both rules in a rule pair involve a large amount of I / O or memory allocation, and the number of processor cores is limited, then it is determined to be a resource contention type. Check whether there is a non-empty intersection between the input parameter set of the second rule and the output result set of the first rule in the rule pair. If the second rule must wait for the intermediate state data of the first rule before execution, and the intermediate state data includes shared memory objects or calculation results, then it is determined to be a sequential dependency. If the judgment conditions of the first rule and the second rule in a rule pair are logically mutually exclusive, then it is judged as a logical contradiction.

9. The method for mutual exclusion detection and resolution of document rule effects under multi-standard parallelism as described in claim 8, characterized in that, For rule pairs determined to be non-resource-competitive, the corresponding resolution strategy is directly matched from the pre-defined strategy library based on the root cause type as the resolution draft. For rule pairs determined to be resource-competitive, the corresponding resolution strategy is directly matched from the pre-built strategy library as the resolution draft, or the resolution draft of the resource-competitive rule pair is generated according to the following methods, including: Lexical and syntactic analysis is performed on the source code of each rule in the rule pair to extract the core elements of the rule, including the input data source type, output field set, execution preconditions and resource dependencies. The input data source type includes file path, API endpoint and database table name, and the resource dependencies include file handle, database connection and API call. The comparison rule pairs extract input data sources from two rules. If the input data sources of the two rules contain regular expression matching the same file path, it is determined that the file metadata cache can be shared; if the two rules call the same API endpoint, it is determined that the API response cache can be shared; if the two rules access the same database table, it is determined that the database query result cache can be shared. Based on the extracted execution preconditions and output field set, a data dependency graph G=(V,E) is constructed, where node V represents the execution steps of each rule and potential cache nodes, and directed edge E represents data dependency relationship. The dependency relationship determination rule is as follows: if the intersection of the output field set O(R1) of the first rule R1 and the input field set I(R2) of the second rule R2 in the rule pair is not empty, then add a directed edge R1→R2; if the first rule R1 and the second rule R2 in the rule pair share cacheable resources, then add an undirected shared edge between them and introduce a cache node. Perform topological sorting on the data dependency graph to generate an execution order sequence. If circular dependencies exist, stop automatic merging and output a prompt for manual intervention. Based on the above analysis results, a composite rule configuration in JSON / YAML format is generated as a resolution draft. The composite rule configuration includes at least a shared cache definition, an execution step sequence, and dependencies between each execution step.

10. A system for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism, used to implement the method for detecting and resolving mutual exclusion of document rule effects under multi-standard parallelism as described in claim 1, characterized in that, include: The data acquisition module is used to extract the execution record, document characteristics, and rule configuration snapshot of each document rule from the database or logs of the document management system. The mutual exclusion detection module is used to combine document rules into rule pairs, and for each rule pair, perform mutual exclusion detection using statistical methods to determine whether there are mutual exclusions. The resolution draft generation module is used to identify the conflict root cause type based on the conflict characteristics of rule pairs that are judged to have mutually exclusive effects or suspected mutually exclusive effects, and to match and output the corresponding resolution draft based on the root cause type.