Artificial intelligence based document compliance review data management system and method

Abstract

The application discloses a document compliance auditing data management system and method based on artificial intelligence, relates to the technical field of document auditing, and comprises the following steps: recording the compliance auditing process of any document, generating a corresponding auditing log; auditing and analyzing any auditing log, generating an auditing result and feeding back; extracting abnormal features of the corresponding document according to preset feature categories and dividing the corresponding document to obtain an abnormal feature set of the corresponding document; analyzing the association between different features; simulating the cause-effect relationship between the associated features to obtain effective cause-effect feature pairs; generating an actual cause-effect relationship based on the auditing result fed back by any effective cause-effect feature; identifying abnormal features existing in the actual document auditing process, and performing abnormal traceability on the abnormal features; effectively reducing the subjectivity and omission risk of manual auditing, and greatly improving the auditing efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of document review technology, specifically to an artificial intelligence-based document compliance review data management system and method. Background Technology

[0002] Currently, document compliance audits mainly rely on manual review, which is inefficient and prone to subjective bias and omissions; traditional risk assessment methods are mostly based on experience and rules, making it difficult to capture the non-linear causal relationships between complex risk factors.

[0003] Existing data analysis techniques are mostly limited to correlation analysis, making it difficult to accurately identify causal relationships. For example, simple statistical analysis can find a correlation between document type and violations, but it cannot clearly explain whether the document type is the cause of the violation or the influence of other factors. There is a lack of effective tools and methods to systematically identify and quantify the causal relationships between risk factors, thereby enabling precise risk control and prevention. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based document compliance review data management system and method to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a document compliance review data management method based on artificial intelligence, the management method comprising the following steps:

[0006] Step S100: Record the compliance review process of any document and generate the corresponding review log; analyze the document data in any review log, generate the review results of the corresponding document, and provide feedback.

[0007] Step S200: Extract the corresponding documents from any audit log, extract and classify the abnormal features of the corresponding documents according to the preset feature categories, and obtain the abnormal feature set of the corresponding documents.

[0008] Step S300: Based on the audit results of any audit log, analyze the correlation between different features; based on the correlation between various features in different audit logs, simulate the causal relationship between the related features;

[0009] Step S400: Analyze the causal simulation of any associated feature in any audit log to obtain effective causal feature pairs; based on the audit results fed back in different audit logs for any effective causal feature pairs, generate the actual causal relationship between each feature;

[0010] Step S500: Identify abnormal features in the actual document review process, and trace the source of abnormal features based on the actual causal relationship between them.

[0011] Furthermore, step S100 includes the following steps:

[0012] Step S101: Pre-configure an audit strategy database and a feature attribute database. The audit strategy database pre-stores several audit strategies, each of which matches a corresponding set of feature attributes. The feature attribute database pre-stores several feature attributes, each of which matches a feature dataset. The audit strategies included in the audit strategy database include regulatory compliance strategies, data accuracy strategies, and policy compliance strategies. For example, audit strategies in financial reporting may include financial statement completeness strategies, tax compliance strategies, and disclosure standardization strategies. The feature attributes are keywords for each item in the corresponding audit strategy, such as operating revenue and costs / expenses.

[0013] Step S102: Whenever a document to be reviewed is received, natural language processing technology is used to perform entity recognition and feature extraction on the document data in the document to be reviewed. The recognition region of each entity in the document to be reviewed is extracted, and the recognition region of any feature is extracted. If the recognition region of a certain feature is in the recognition region of a certain entity, then the entity is matched with the feature to generate entity datasets for each entity. The document to be reviewed is divided into several entity datasets, and a review log is generated.

[0014] Step S103: Randomly select an audit strategy for the generated audit log, extract the feature attribute set and feature dataset that match the selected audit strategy, compare any feature attribute with the entity of the document to be audited, and compare any entity data with the features of the document to be audited. If all feature attributes and corresponding feature data in the selected audit strategy can be extracted from the document to be audited, then set the selected audit strategy as the target audit strategy.

[0015] Step S104: Retrieve each target audit strategy to conduct compliance audits on the documents to be audited. If an anomaly is found in the selected feature, mark the selected feature as an anomaly. Obtain the audit time point, audit strategy adopted, and anomaly mark for any selected feature to obtain the audit information group of the selected feature. Summarize the audit information groups of each feature in the document to be audited to obtain the audit result of the document to be audited and store it in the generated audit log.

[0016] Furthermore, step S200 includes the following steps:

[0017] Step S201: Randomly select an audit log and extract the audit information group of each feature in the selected audit log; randomly select a feature, and if there is an abnormal marker in the audit information group of the selected feature, then set the selected feature as an abnormal feature.

[0018] Step S202: Obtain the entities that match any abnormal feature in the selected audit log, and divide all abnormal features into several abnormal feature sets according to the matching entities; arbitrarily select an abnormal feature set, extract the feature attributes of the entity in the selected abnormal feature set, and set the extracted feature attributes as the abnormal features in the selected audit log.

[0019] Step S203: Extract the review time point and the review strategy adopted from the review information group of each feature, sort each feature according to the order of the review time points, and obtain the review time interval between any two adjacent features; if the review strategies adopted by two adjacent features are different, set the review time interval between the two adjacent features as the reference time interval, and calculate the average time interval by averaging all reference time intervals; by analyzing the conversion time between different review strategies, compare the conversion time with the review time interval in the same review strategy to determine whether features belong to the same batch under the same review strategy. Only features from different batches will have a causal relationship in the future.

[0020] Step S204: If several adjacent and consecutive features adopt the same review strategy, the features are initially divided into a set of similar features, and all features are divided into several sets of similar features; arbitrarily select a set of similar features, and obtain the review time interval between two adjacent features in the selected set of similar features. If the review time interval exceeds the average time interval, the selected set of similar features is divided into two feature sets through the two adjacent features, and several feature sets are regenerated.

[0021] Step S205: Randomly select an abnormal feature, determine the key set where the selected abnormal feature is located, and divide all abnormal features into several abnormal feature sets for the corresponding documents of the audit log according to the distribution of each abnormal feature in each feature set.

[0022] Furthermore, step S300 includes the following steps:

[0023] Step S301: Randomly select an audit log, extract several abnormal feature sets from the selected audit log, sort each abnormal feature set according to the chronological order of the audit time points, and arbitrarily select two abnormal feature sets, and arbitrarily select one abnormal feature from each of the two abnormal feature sets as two comparison features.

[0024] Step S302: Pre-build a semantic association database, match the semantic association between each feature attribute in the feature attribute library, obtain the association value between any two feature attributes, if the association value between two feature attributes exceeds the preset association threshold, then set the two feature attributes to have an association relationship; obtain the feature attributes corresponding to two comparison features, if there is an association relationship between two comparison features, then set the two selected abnormal features as an association feature group, and obtain several association feature groups for the selected audit log;

[0025] Step S303: Randomly select a related feature group, adjust the two abnormal features in the selected related feature group according to the order of the audit time points to generate a sequential related feature group; extract all sequential related feature groups in different audit logs, and arbitrarily select two sequential related feature groups. If the audit strategies adopted by the two sequential related feature groups are the same, then set the two sequential key feature groups as the same type of feature group. Summarize several similar feature groups to generate a set of similar feature groups. Summarize the previous and next abnormal features of each similar feature group in the set of similar feature groups to generate a first feature set and a second feature set, and set a causal relationship between the first feature set and the second feature set; to determine whether a causal relationship exists, it is necessary to determine whether the two abnormal features exist simultaneously in multiple audit logs.

[0026] Furthermore, step S400 includes the following steps:

[0027] Step S401: Randomly select a set of similar feature groups to obtain a first feature set and a second feature set of the selected set of similar feature groups. Randomly select an abnormal feature from the first feature set and the second feature set to generate a simulated feature group.

[0028] Step S402: Randomly select an audit log, obtain the audit information group of each feature in the selected audit log, compare the two abnormal features in the simulation adjustment group with each feature, if there are two features in the selected audit log that are the same as the two abnormal features, then obtain the audit information group of the two features. If there are abnormal markers in the audit information groups of the two features, and the order of the audit time points and the audit strategy adopted are the same, then set the simulation feature group as a valid causal feature pair.

[0029] Step S403: Extract all valid causal feature pairs from the audit logs, and arbitrarily select a set of valid causal feature pairs. Compare the selected valid causal feature pairs with each feature in the remaining audit logs. If the selected valid causal feature pairs also exist in the remaining audit logs, then the selected valid causal feature pairs are set as actual causal feature pairs. The actual causal feature pairs require that the valid causal feature pairs show anomalies in all audit logs, which indicates that a causal relationship actually exists.

[0030] Step S404: Randomly select two actual causal feature pairs. If the latter feature in one actual causal feature pair and the former feature in the other actual causal feature pair generate a new actual causal feature pair, then connect the two selected actual causal feature pairs to obtain a causal feature tracing chain. Connect each actual causal feature pair to generate a causal feature tracing network.

[0031] Furthermore, step S500 includes the following steps:

[0032] Step S501: Monitor the document review process in real time, perform entity recognition and feature extraction on the document data in the real-time document, retrieve the review strategy database and feature attribute database for comparison, generate review information groups for each feature in the real-time document; obtain the abnormal markers of the review information groups for each feature, and obtain several abnormal features of the real-time document.

[0033] Step S502: Randomly select two abnormal features from the aforementioned abnormal features, extract the review time points from the review information group containing the two selected abnormal features, and generate expected causal feature groups according to the chronological order of the review time points to obtain several expected causal feature groups for the real-time document.

[0034] Step S503: Retrieve the causal feature tracing network and compare it with the several expected causal feature groups. If the several expected causal feature groups are continuously connected in the causal feature tracing network, generate the expected tracing chain, obtain the source features of each abnormal feature, and provide an anomaly repair reminder for each source feature. Re-identify the abnormal features of the real-time document after repair until the real-time document is reviewed.

[0035] To better implement the above methods, a document compliance audit data management system is also proposed. The management system includes a document data acquisition module, a document feature recognition module, a causal relationship simulation module, a causal relationship generation module, and an audit anomaly capture module.

[0036] The document data acquisition module is used to record the compliance review process of any document and generate corresponding review logs; it also performs review analysis on the document data in any review logs, generates the review results of the corresponding documents, and provides feedback.

[0037] The document feature recognition module is used to extract the corresponding documents from any audit log, extract and classify the abnormal features of the corresponding documents according to the preset feature categories, and obtain the abnormal feature set of the corresponding documents.

[0038] The causal relationship simulation module is used to analyze the correlation between different features based on the audit results of any audit log; and to simulate the causal relationship between related features based on the correlation between various features in different audit logs.

[0039] The causal relationship generation module is used to analyze the causal simulation of any associated feature in any audit log to obtain effective causal feature pairs; based on the audit results fed back in different audit logs for any effective causal feature pairs, the actual causal relationship between each feature is generated.

[0040] The anomaly detection module is used to identify abnormal features in the actual document review process and to trace the source of the anomalies based on the actual causal relationship between the abnormal features.

[0041] Furthermore, the document data acquisition module includes a data acquisition and processing unit and an audit result feedback unit;

[0042] The data acquisition and processing unit is used to record the compliance review process of any document and generate corresponding review logs; the review result feedback unit is used to review and analyze the document data in any review log, generate the review results of the corresponding document, and provide feedback.

[0043] Furthermore, the causal association simulation module includes a feature association analysis unit and a causal relationship simulation unit;

[0044] The feature association analysis unit is used to analyze the association between different features based on the review results of any review log; the causal relationship simulation unit is used to simulate the causal relationship between associated features based on the association between various features in different review logs.

[0045] Furthermore, the causal association generation module includes a relationship simulation verification unit and a feature causal generation unit;

[0046] The relationship simulation and verification unit is used to analyze the causal simulation of any associated feature in any audit log to obtain effective causal feature pairs; the feature causal generation unit is used to generate the actual causal relationship between each feature based on the audit results fed back in different audit logs for any effective causal feature pairs.

[0047] Compared with the prior art, the beneficial effects of the present invention are:

[0048] 1. This application achieves automatic recording, feature extraction, and abnormal feature identification in the document review process, which can effectively reduce the subjectivity and omission risk of manual review, greatly improve review efficiency and accuracy, and also improve the level of intelligence in document compliance review;

[0049] 2. This invention, through a causal reasoning mechanism, can mine causal relationships between features from multi-dimensional and multi-log data, thereby constructing a causal feature tracing network. This enables the systematic identification and tracing of the root causes of anomalies, providing deeper and more scientific decision support for risk prevention and control.

[0050] 3. This invention helps staff control the accuracy of document compliance review and improves review efficiency by monitoring abnormal features in real time during the review process and combining historical causal networks to provide intelligent reminders and repair suggestions. Attached Figure Description

[0051] Figure 1 A schematic diagram illustrating the steps of an AI-based document compliance review data management method;

[0052] Figure 2 This is a schematic diagram of the structure of an AI-based document compliance review data management system. Detailed Implementation

[0053] 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 some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Example: Figures 1 to 2 As shown, this invention provides an artificial intelligence-based method for managing document compliance review data. The management method includes the following steps:

[0055] Step S100: Record the compliance review process of any document and generate the corresponding review log; analyze the document data in any review log, generate the review results of the corresponding document, and provide feedback.

[0056] Step S100 includes the following steps:

[0057] Step S101: Pre-configure an audit strategy database and a feature attribute database. The audit strategy database pre-stores several audit strategies, wherein any audit strategy matches a corresponding feature attribute set. The feature attribute database pre-stores several feature attributes, wherein any feature attribute matches a feature dataset.

[0058] Step S102: Whenever a document to be reviewed is received, natural language processing technology is used to perform entity recognition and feature extraction on the document data in the document to be reviewed. The recognition region of each entity in the document to be reviewed is extracted, and the recognition region of any feature is extracted. If the recognition region of a certain feature is in the recognition region of a certain entity, then the entity is matched with the feature to generate entity datasets for each entity. The document to be reviewed is divided into several entity datasets, and a review log is generated.

[0059] Step S103: Randomly select an audit strategy for the generated audit log, extract the feature attribute set and feature dataset that match the selected audit strategy, compare any feature attribute with the entity of the document to be audited, and compare any entity data with the features of the document to be audited. If all feature attributes and corresponding feature data in the selected audit strategy can be extracted from the document to be audited, then set the selected audit strategy as the target audit strategy.

[0060] Step S104: Retrieve each target audit strategy to conduct compliance audits on the documents to be audited. If an anomaly is found in the selected feature, mark the selected feature as an anomaly. Obtain the audit time point, audit strategy adopted, and anomaly mark for any selected feature to obtain the audit information group of the selected feature. Summarize the audit information groups of each feature in the document to be audited to obtain the audit result of the document to be audited and store it in the generated audit log.

[0061] Example 1: A compliance audit of a company's annual financial report is set up. A pre-configured audit strategy database includes financial statement integrity strategy, tax compliance strategy, and disclosure standardization strategy. The financial statement integrity strategy matches a set of characteristic attributes, including operating revenue and net profit. The system uses NLP technology to identify entities such as "operating revenue" and "net profit" in the report, which match the characteristic attribute set of the financial statement integrity strategy. If the financial statement integrity strategy is invoked to check the completeness of the financial statements, and if data is missing in the net profit section, an audit log is generated, marking this characteristic as an anomaly.

[0062] Step S200: Extract the corresponding documents from any audit log, extract and classify the abnormal features of the corresponding documents according to the preset feature categories, and obtain the abnormal feature set of the corresponding documents.

[0063] Step S200 includes the following steps:

[0064] Step S201: Randomly select an audit log and extract the audit information group of each feature in the selected audit log; randomly select a feature, and if there is an abnormal marker in the audit information group of the selected feature, then set the selected feature as an abnormal feature.

[0065] Step S202: Obtain the entities that match any abnormal feature in the selected audit log, and divide all abnormal features into several abnormal feature sets according to the matching entities; arbitrarily select an abnormal feature set, extract the feature attributes of the entity in the selected abnormal feature set, and set the extracted feature attributes as the abnormal features in the selected audit log.

[0066] Step S203: Extract the review time point and the review strategy adopted from the review information group of each feature, sort each feature according to the order of the review time points, and obtain the review time interval between any two adjacent features; if the review strategies adopted by two adjacent features are different, set the review time interval between the two adjacent features as the reference time interval, and calculate the average time interval by averaging all the reference time intervals.

[0067] Step S204: If several adjacent and consecutive features adopt the same review strategy, the features are initially divided into a set of similar features, and all features are divided into several sets of similar features; arbitrarily select a set of similar features, and obtain the review time interval between two adjacent features in the selected set of similar features. If the review time interval exceeds the average time interval, the selected set of similar features is divided into two feature sets through the two adjacent features, and several feature sets are regenerated.

[0068] Step S205: Randomly select an abnormal feature, determine the key set where the selected abnormal feature is located, and divide all abnormal features into several abnormal feature sets for the corresponding documents of the audit log according to the distribution of each abnormal feature in each feature set.

[0069] Step S300: Based on the audit results of any audit log, analyze the correlation between different features; based on the correlation between various features in different audit logs, simulate the causal relationship between the related features;

[0070] Step S300 includes the following steps:

[0071] Step S301: Randomly select an audit log, extract several abnormal feature sets from the selected audit log, sort each abnormal feature set according to the chronological order of the audit time points, and arbitrarily select two abnormal feature sets, and arbitrarily select one abnormal feature from each of the two abnormal feature sets as two comparison features.

[0072] Step S302: Pre-build a semantic association database, match the semantic association between each feature attribute in the feature attribute library, obtain the association value between any two feature attributes, if the association value between two feature attributes exceeds the preset association threshold, then set the two feature attributes to have an association relationship; obtain the feature attributes corresponding to two comparison features, if there is an association relationship between two comparison features, then set the two selected abnormal features as an association feature group, and obtain several association feature groups for the selected audit log;

[0073] Step S303: Randomly select a related feature group, adjust the two abnormal features in the selected related feature group according to the order of the audit time points to generate a sequential related feature group; extract all sequential related feature groups in different audit logs, and arbitrarily select two sequential related feature groups. If the audit strategies adopted by the two sequential related feature groups are the same, then set the two sequential key feature groups as the same type of feature group. Summarize several similar feature groups to generate a set of similar feature groups. Summarize the previous and next abnormal features of each similar feature group in the set of similar feature groups to generate a first feature set and a second feature set, and set a causal relationship between the first feature set and the second feature set.

[0074] Example 2: Analyzing multiple financial report audit logs reveals that "abnormal operating revenue" and "abnormal income tax" occur simultaneously. Semantic association analysis shows a correlation between operating revenue and income tax, thus designating these two features as a related feature group. Furthermore, in different audit logs, "abnormal operating revenue" is audited first, followed by "abnormal income tax," and the same audit strategy is employed. This preliminarily simulates a causal relationship where "abnormal revenue may lead to abnormal tax."

[0075] Step S400: Analyze the causal simulation of any associated feature in any audit log to obtain effective causal feature pairs; based on the audit results fed back in different audit logs for any effective causal feature pairs, generate the actual causal relationship between each feature;

[0076] Step S400 includes the following steps:

[0077] Step S401: Randomly select a set of similar feature groups to obtain a first feature set and a second feature set of the selected set of similar feature groups. Randomly select an abnormal feature from the first feature set and the second feature set to generate a simulated feature group.

[0078] Step S402: Randomly select an audit log, obtain the audit information group of each feature in the selected audit log, compare the two abnormal features in the simulation adjustment group with each feature, if there are two features in the selected audit log that are the same as the two abnormal features, then obtain the audit information group of the two features. If there are abnormal markers in the audit information groups of the two features, and the order of the audit time points and the audit strategy adopted are the same, then set the simulation feature group as a valid causal feature pair.

[0079] Step S403: Extract all valid causal feature pairs from the audit logs, and arbitrarily select a set of valid causal feature pairs. Compare the selected valid causal feature pairs with each feature in the remaining audit logs. If the selected valid causal feature pairs also exist in the remaining audit logs, then set the selected valid causal feature pairs as actual causal feature pairs.

[0080] Step S404: Randomly select two actual causal feature pairs. If the latter feature in one actual causal feature pair and the former feature in the other actual causal feature pair generate a new actual causal feature pair, then connect the two selected actual causal feature pairs to obtain a causal feature tracing chain. Connect each actual causal feature pair to generate a causal feature tracing network.

[0081] Example 3: Select "Revenue Anomaly → Tax Anomaly" from the simulated causal pair. Search the historical logs and find that these two anomalies appear simultaneously in multiple logs with the same time sequence. Mark this pair as "actual causal feature pair". At the same time, find "Cost Anomaly → Revenue Anomaly" and "Revenue Anomaly → Tax Anomaly" and connect them into a causal chain: "Cost Anomaly → Revenue Anomaly → Tax Anomaly", and construct a causal feature tracing network.

[0082] Step S500: Identify the abnormal features present in the actual document review process, and trace the source of the abnormal features based on the actual causal relationship between them;

[0083] Step S500 includes the following steps:

[0084] Step S501: Monitor the document review process in real time, perform entity recognition and feature extraction on the document data in the real-time document, retrieve the review strategy database and feature attribute database for comparison, generate review information groups for each feature in the real-time document; obtain the abnormal markers of the review information groups for each feature, and obtain several abnormal features of the real-time document.

[0085] Step S502: Randomly select two abnormal features from the aforementioned abnormal features, extract the review time points from the review information group containing the two selected abnormal features, and generate expected causal feature groups according to the chronological order of the review time points to obtain several expected causal feature groups for the real-time document.

[0086] Step S503: Retrieve the causal feature tracing network and compare it with the several expected causal feature groups. If the several expected causal feature groups are continuously connected in the causal feature tracing network, generate the expected tracing chain, obtain the source features of each abnormal feature, and provide an anomaly repair reminder for each source feature. Re-identify the abnormal features of the real-time document after repair until the real-time document is reviewed.

[0087] A document compliance audit data management system, comprising a document data acquisition module, a document feature recognition module, a causal relationship simulation module, a causal relationship generation module, and an audit anomaly capture module;

[0088] The document data acquisition module is used to record the compliance review process of any document and generate corresponding review logs; it also performs review analysis on the document data in any review logs, generates the review results of the corresponding documents, and provides feedback.

[0089] The document feature recognition module is used to extract the corresponding documents from any audit log, extract and classify the abnormal features of the corresponding documents according to the preset feature categories, and obtain the abnormal feature set of the corresponding documents.

[0090] The causal relationship simulation module is used to analyze the correlation between different features based on the audit results of any audit log; and to simulate the causal relationship between related features based on the correlation between various features in different audit logs.

[0091] The causal relationship generation module is used to analyze the causal simulation of any associated feature in any audit log to obtain effective causal feature pairs; based on the audit results fed back in different audit logs for any effective causal feature pairs, the actual causal relationship between each feature is generated.

[0092] The anomaly detection module is used to identify abnormal features in the actual document review process and to trace the source of the anomalies based on the actual causal relationship between the abnormal features.

[0093] The document data acquisition module includes a data acquisition and processing unit and an audit result feedback unit.

[0094] The data acquisition and processing unit is used to record the compliance review process of any document and generate corresponding review logs; the review result feedback unit is used to review and analyze the document data in any review log, generate the review results of the corresponding document, and provide feedback.

[0095] The causal association simulation module includes a feature association analysis unit and a causal relationship simulation unit.

[0096] The feature association analysis unit is used to analyze the association between different features based on the review results of any review log; the causal relationship simulation unit is used to simulate the causal relationship between associated features based on the association between various features in different review logs.

[0097] The causal association generation module includes a relationship simulation and verification unit and a feature causal generation unit.

[0098] The relationship simulation and verification unit is used to analyze the causal simulation of any associated feature in any audit log to obtain effective causal feature pairs; the feature causal generation unit is used to generate the actual causal relationship between each feature based on the audit results fed back in different audit logs for any effective causal feature pairs.

[0099] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. An artificial intelligence based document compliance audit data management method characterized by: The management method comprises the following steps: Step S100: record the compliance audit process of any document, generate the corresponding audit log, and perform audit analysis on the document data in any audit log to generate the audit result of the corresponding document and provide feedback; Step S200: extract the corresponding document of any audit log, extract the abnormal features of the corresponding document according to the preset feature categories, and divide them to obtain the abnormal feature set of the corresponding document; Step S300: analyze the association between different features based on the audit result of any audit log; simulate the causal relationship between the associated features based on the association between the features in different audit logs; Step S400: analyze the causal simulation of any associated feature in any audit log to obtain the effective causal feature pair; generate the actual causal relationship between the features based on the audit result fed back by any effective causal feature pair in different audit logs; Step S500: identify the abnormal features existing in the actual document audit process, and perform abnormal traceability on the abnormal features based on the actual causal relationship between the abnormal features; The step S300 comprises the following steps: Step S301: randomly select an audit log, extract a plurality of abnormal feature sets in the selected audit log, sort the abnormal feature sets according to the chronological order of the audit time points, and randomly select two abnormal feature sets, and randomly select an abnormal feature from each of the two abnormal feature sets, and set the two selected abnormal features as two comparison features; Step S302: pre-construct a semantic association database, match the semantic association between the feature attributes in the feature attribute library, obtain the association value between any two feature attributes, and if the association value between the two feature attributes exceeds the preset association threshold, set that there is an association relationship between the two feature attributes; obtain the feature attributes corresponding to the two comparison features, and if there is an association relationship between the two comparison features, set the two selected abnormal features as an associated feature group, and obtain a plurality of associated feature groups of the selected audit log; Step S303: randomly select an associated feature group, adjust the two abnormal features in the selected associated feature group according to the chronological order of the audit time points to generate a sequential associated feature group; extract all sequential associated feature groups in different audit logs, and randomly select two sequential associated feature groups, if the two sequential associated feature groups adopt the same audit strategy, set the two sequential key feature groups as the same type feature group, aggregate a plurality of same type feature groups to generate a same type feature group set, aggregate and generate a first feature set and a second feature set for the previous abnormal feature and the subsequent abnormal feature of each same type feature group in the same type feature group set, and set the first feature set and the second feature set as the causal relationship; The step S400 comprises the following steps: Step S401: randomly select one same-class feature group set, obtain a first feature set and a second feature set of the selected same-class feature group set, randomly select one abnormal feature from the first feature set and the second feature set respectively, and generate one simulation feature group; Step S402: randomly select one audit log, obtain the audit information set of each feature in the selected audit log, compare the two abnormal features in the simulation feature group with each feature, if the two features in the selected audit log are the same as the two abnormal features respectively, obtain the audit information set of the two features, if the audit information set of the two features both have abnormal marks and the order of the audit time points and the adopted audit strategy are the same, set the simulation feature group as an effective cause-effect feature pair; Step S403: extract all the effective cause-effect feature pairs in the selected audit log, and randomly select one group of effective cause-effect feature pairs, compare the selected effective cause-effect feature pair with each feature in the remaining audit logs, if the selected effective cause-effect feature pair also has an effective cause-effect feature pair in the remaining audit logs, set the selected effective cause-effect feature pair as an actual cause-effect feature pair; Step S404: randomly select two actual cause-effect feature pairs, if the latter feature in one of the actual cause-effect feature pairs is the same as the former feature in the other actual cause-effect feature pair, generate a new actual cause-effect feature pair, connect the selected two actual cause-effect feature pairs to obtain a cause-effect feature trace chain, and connect each actual cause-effect feature pair with each other to generate a cause-effect feature trace network.

2. The artificial intelligence based document compliance audit data management method as claimed in claim 1 wherein: The step S100 includes the following steps: Step S101: pre-configure an audit strategy database and a feature attribute library, a plurality of audit strategies are pre-stored in the audit strategy database, wherein any audit strategy is matched with a corresponding feature attribute set, and a plurality of feature attributes are pre-stored in the feature attribute library, wherein any feature attribute is matched with a feature data set; Step S102: whenever a document to be audited is received, use natural language technology to perform entity recognition and feature extraction on the document data in the document to be audited, extract the recognition area of each entity in the document to be audited, and extract the recognition area of any feature, if the recognition area of a certain feature is in the recognition area of a certain entity, match the certain entity with the certain feature to generate an entity data set of each entity, divide the document to be audited into a plurality of entity data sets, and generate an audit log; Step S103: randomly select an audit strategy for the generated audit log, extract the feature attribute set and the feature data set matched with the selected audit strategy, compare any feature attribute with the entity of the document to be audited, and compare any entity data with the feature of the document to be audited, if all the feature attributes and the corresponding feature data in the selected audit strategy can be extracted from the document to be audited, set the selected audit strategy as a target audit strategy; Step S104: retrieve each target audit policy to perform compliance audit on the to-be-audited document respectively, and if the selected feature has an anomaly, mark the selected feature as abnormal; obtain the audit time point, the adopted audit policy and the abnormal mark of any selected feature to obtain the audit information set of the selected feature, aggregate the audit information set of each feature in the to-be-audited document, obtain the audit result of the to-be-audited document and store it in the generated audit log. 3.The artificial intelligence-based document compliance auditing data management method of claim 2, wherein: The step S200 includes the following steps: Step S201: randomly select an audit log, extract the audit information set of each feature in the selected audit log; randomly select a feature, and if the audit information set of the selected feature has an abnormal mark, set the selected feature as an abnormal feature; Step S202: obtain the matched entity of any abnormal feature in the selected audit log, divide all abnormal features into a plurality of abnormal feature sets according to the matched entity; randomly select an abnormal feature set, extract the feature attribute of the entity where the selected abnormal feature set is located, and set the extracted feature attribute as the abnormal feature in the selected audit log; Step S203: extract the audit time point and the adopted audit policy from the audit information set of each feature, sort the features according to the chronological order of the audit time point to obtain the audit time interval of any adjacent two features; if the adopted audit policies of the adjacent two features are different, set the audit time interval of the adjacent two features as a reference time interval, and calculate the average value of all reference time intervals to obtain an average time interval; Step S204: if the adopted audit policies of a plurality of adjacent and continuous features are the same, preliminarily divide the plurality of features into a same-type feature set, and divide all features into a plurality of same-type feature sets; randomly select a same-type feature set, obtain the audit time interval of two adjacent features in the selected same-type feature set, and if the audit time interval exceeds the average time interval, divide the selected same-type feature set into two feature sets through the two adjacent features, and regenerate a plurality of feature sets; Step S205: randomly select an abnormal feature, determine the key set where the selected abnormal feature is located, and divide all abnormal features into a plurality of abnormal feature sets of the corresponding document of the selected audit log according to the distribution of each abnormal feature in each feature set. 4.The artificial intelligence-based document compliance auditing data management method of claim 3, wherein: The step S500 includes the following steps: Step S501: real-time monitor the document audit process, perform entity recognition and feature extraction on the document data in the real-time document, compare the audit policy database and the feature attribute library to generate the audit information set of each feature in the real-time document; obtain the abnormal mark of the audit information set of each feature to obtain a plurality of abnormal features of the real-time document; Step S502: randomly select two abnormal features from the plurality of abnormal features, extract the audit time point in the audit information set of the selected two abnormal features, generate an expected causal feature group according to the chronological order of the audit time point, and obtain a plurality of expected causal feature groups of the real-time document Step S503: retrieve the causal feature traceability network and compare it with the plurality of expected causal feature groups; if the plurality of expected causal feature groups are continuously connected in the causal feature traceability network, generate an expected traceability chain, obtain the traceability features of each abnormal feature, and perform abnormal repair reminders on each traceability feature; and perform abnormal feature identification on the repaired real-time document again until the real-time document is completed.

5. A document compliance audit data management system for performing the artificial intelligence based document compliance audit data management method of any one of claims 1-4, characterized by: The management system comprises a document data acquisition module, a document feature identification module, a causal correlation simulation module, a causal correlation generation module, and an audit anomaly capture module. The document data acquisition module is configured to record the compliance audit process of any document, generate a corresponding audit log, analyze the document data in any audit log, generate an audit result of the corresponding document, and provide feedback. The document feature identification module is configured to extract the corresponding document of any audit log, extract abnormal features of the corresponding document according to a preset feature type, and divide the corresponding document to obtain an abnormal feature set of the corresponding document. The causal correlation simulation module is configured to analyze the correlation between different features based on the audit result of any audit log, and simulate the causal relationship between the correlated features based on the correlation between the features in different audit logs. The causal correlation generation module is configured to analyze the causal simulation of any correlated feature in any audit log to obtain an effective causal feature pair, and generate an actual causal relationship between the features based on the audit result of any effective causal feature pair in different audit logs. The audit anomaly capture module is configured to identify abnormal features existing in the actual document audit process, and perform abnormal traceability of the abnormal features based on the actual causal relationship between the abnormal features.

6. The document compliance audit data management system of claim 5, wherein: The document data acquisition module comprises a data acquisition processing unit and an audit result feedback unit. The data acquisition processing unit is configured to record the compliance audit process of any document, generate a corresponding audit log, and the audit result feedback unit is configured to analyze the document data in any audit log, generate an audit result of the corresponding document, and provide feedback.

7. The document compliance audit data management system of claim 5, wherein: The causal correlation simulation module comprises a feature correlation analysis unit and a causal relationship simulation unit. The feature correlation analysis unit is configured to analyze the correlation between different features based on the audit result of any audit log, and the causal relationship simulation unit is configured to simulate the causal relationship between the correlated features based on the correlation between the features in different audit logs.

8. The document compliance audit data management system of claim 5, wherein: The causal correlation generation module comprises a relationship simulation verification unit and a feature causal generation unit. The relationship simulation verification unit is configured to analyze the causal simulation of any correlated feature in any audit log to obtain an effective causal feature pair, and the feature causal generation unit is configured to generate an actual causal relationship between the features based on the audit result of any effective causal feature pair in different audit logs.

Images