A data governance system based on semantic analysis

Abstract

The application discloses a kind of data governance systems based on semantic analysis, it is related to data governance technical field, including: data governance foundation construction module, for obtaining data governance associated data and standardization processing, generates data governance basic dataset;Semantic anchor point generation module, for generating data object semantic anchor point, standard semantic anchor point and rule semantic anchor point;Anchor point connection construction module, for generating standard matching connection and rule adaptation connection;Semantic anchor point graph construction module, for constructing data governance semantic anchor point graph;Standard rule matching module, for generating data standard matching result and governance rule adaptation result;Governance problem identification module, for generating data governance problem identification result;Feedback update module, for updating data governance semantic anchor point graph.The application has the advantages of high standard matching accuracy, high rule adaptation automation degree and sustainable optimization of governance result.

Description

Technical Field

[0001] This invention relates to the field of data governance technology, and in particular to a data governance system based on semantic analysis. Background Technology

[0002] In recent years, with the development of enterprise digital systems, data platforms, metadata management, data standard management, and artificial intelligence semantic analysis technologies, data governance surrounding multi-source business data has gradually become an important direction for enterprise data asset management. Existing technologies typically involve collecting database table structures, field names, field comments, data standards, governance rules, and historical governance records to clean, classify, and standardize data objects. Then, methods such as field name matching, manual tagging, rule template configuration, or keyword retrieval are used to map data objects to data standards. Furthermore, governance rules such as integrity checks, format checks, consistency checks, and compliance checks are configured for different types of data objects to identify issues such as non-standard field naming, missing standards, missing rules, and data quality anomalies.

[0003] However, existing technologies still face challenges in data governance, including inconsistent semantic representation of data objects, low accuracy of standard matching, and difficulty in automatically adapting governance rules. On one hand, field names, field comments, data table descriptions, and business terminology often contain abbreviations, aliases, synonyms, and contextual differences across different business systems. Simply relying on field names or keyword matching is insufficient to accurately determine the corresponding data standard for a data object, easily leading to missed or incorrect matches. On the other hand, there is a lack of semantically based connections between data standards and governance rules. Standard constraints, rule conditions, and applicable scope are difficult to establish stable relationships, causing governance rule configuration to rely on manual experience. Furthermore, historical governance results are difficult to promptly reflect in subsequent standard matching and rule adaptation processes, thus affecting the consistency and sustainable update capabilities of data governance results. Summary of the Invention

[0004] One objective of this invention is to propose a data governance system based on semantic analysis. This invention fully utilizes semantic analysis and data governance semantic anchor graph construction technology, and describes in detail the semantic matching, problem identification and feedback update process between data objects, data standards and governance rules. It has the advantages of high accuracy in standard matching, high degree of automation in rule adaptation and sustainable optimization of governance results.

[0005] A data governance system based on semantic analysis according to an embodiment of the present invention includes: The data governance infrastructure module is used to acquire data governance-related data and perform standardized processing to generate a basic data governance dataset. The semantic anchor generation module is used to perform semantic parsing on the basic dataset for data governance, and generate semantic anchors for data objects, standard semantic anchors, and rule-based semantic anchors respectively. The anchor connection building module is used to determine the standard matching connection between the semantic anchor of the data object and the standard semantic anchor, as well as the rule adaptation connection between the standard semantic anchor and the rule semantic anchor. The semantic anchor graph construction module is used to construct a data governance semantic anchor graph based on data object semantic anchors, standard semantic anchors, rule semantic anchors, standard matching connections, and rule adaptation connections. The standard rule matching module is used to determine the standard semantic anchor and rule semantic anchor corresponding to the semantic anchor of the data object based on the data governance semantic anchor graph, and generate data standard matching results and governance rule adaptation results. The governance issue identification module is used to identify semantic conflicts, missing standards, and rule mismatches based on the data standard matching results and governance rule adaptation results, and generate data governance issue identification results. The feedback update module is used to generate data governance tasks based on the data governance problem identification results, update the data governance semantic anchor graph based on the processing results of the data governance tasks, and generate updated data standard matching results and updated governance rule adaptation results.

[0006] Optionally, the data governance associated data includes data object information, data standard information, governance rule information, and historical governance information. The standardization process includes unified object identification, organized field hierarchy, unified standard name, organized standard definition, rule type marking, scope of application marking, processing result marking, hit status marking, and associated storage processing.

[0007] Optionally, the semantic anchor generation module includes: Separate data object information, data standard information, and governance rule information from the basic dataset for data governance; The data object information is segmented by field name, parsed by field annotation, and extracted by contextual semantics. The parsing results are semantically encoded to generate semantic anchors for the data object. The data standard information is parsed for standard name, standard definition and standard caliber, and the parsing results are semantically encoded to generate standard semantic anchors; The governance rule information is parsed by rule conditions, rule objects, and rule results. The parsing results are semantically encoded to generate rule semantic anchors. Write data object semantic anchors, standard semantic anchors, and rule semantic anchors into the data governance basic dataset according to the object identifier.

[0008] Optionally, the anchor point connection building module includes: Extract the semantic vector and semantic keywords of the data object from the semantic anchor points of the data object, and extract the standard semantic vector and standard keywords from the standard semantic anchor points; The similarity between the semantic vector of the data object and the standard semantic vector is calculated, and a standard matching score is generated by combining the overlap between semantic keywords and standard keywords. Connect the semantic anchors of data objects whose standard matching scores reach the standard connection threshold and whose business domains and constraint contents are consistent with the standard semantic anchors to generate standard matching connections. Extract the standard constraint content from the standard semantic anchor point, extract the rule condition content from the rule semantic anchor point, perform consistency calculation on the standard constraint content and the rule condition content, and generate a rule adaptation score. Connect the standard semantic anchor points that have reached the rule connection threshold and have the same applicable scope to the rule semantic anchor points to generate rule adaptation connections.

[0009] Optionally, the semantic anchor graph construction module includes: Write the semantic anchors of data objects, standard semantic anchors, and rule semantic anchors into the graph node set, and record the anchor type, anchor number, and semantic vector for each graph node; Write the standard matching connections and rule-adapted connections into the graph edge set, and record the connection direction, connection type and connection weight for each graph edge; Organize the graph edges according to the order of data object semantic anchor points pointing to standard semantic anchor points, and standard semantic anchor points pointing to rule semantic anchor points, and generate anchor point connection paths; Repeated paths are merged and broken paths are marked for anchor connection paths, and anchor connection paths whose connection weights reach the corresponding connection thresholds are retained. The graph node set, graph edge set, and anchor connection path are associated and stored to construct a data governance semantic anchor graph.

[0010] Optionally, the standard rule matching module includes: Select semantic anchors of data objects to be matched from the semantic anchor graph of data governance, find standard semantic anchors that are connected to the semantic anchors of data objects to be matched along the standard matching connection, determine the standard semantic anchors whose connection weight reaches the standard connection threshold as the target standard semantic anchor set, and generate data standard matching results. Search for rule semantic anchors along the rule adaptation connection corresponding to the target standard semantic anchor. Determine the rule semantic anchors whose connection weight reaches the rule connection threshold and whose applicable scope is consistent as the target rule semantic anchor set, and generate governance rule adaptation results. The semantic anchors of the data object, the set of target standard semantic anchors, the set of target rule semantic anchors, the connection weights, and the matching status are stored together.

[0011] Optionally, the governance problem identification module includes: Obtain data standard matching results and governance rule adaptation results, and perform matching status checks on the target standard semantic anchors in the data standard matching results; When the semantic anchors of the data object do not form the target standard semantic anchor set, or when the target standard semantic anchor set is empty, a standard missing result is generated; When the semantic anchor point of the same data object corresponds to multiple target standard semantic anchor points, and the standard constraint content between the multiple target standard semantic anchor points is inconsistent, a semantic conflict result is generated. Perform an adaptation status check on the set of semantic anchor points of the target rule in the governance rule adaptation results; When the target standard semantic anchor points do not form a set of target rule semantic anchor points, a rule missing result is generated; When the rule condition content in the target rule semantic anchor set is inconsistent with the constraint description corresponding to the data object semantic anchor, a rule mismatch result is generated. The results of missing standards, semantic conflicts, missing rules, and rule mismatches are associated with object identifiers to generate data governance problem identification results.

[0012] Optionally, the feedback update module includes: Obtain the data governance problem identification results, and determine the corresponding data object semantic anchor, target standard semantic anchor, and target rule semantic anchor according to the object identifier; Data governance tasks are generated based on standard missing results, semantic conflict results, rule missing results, and rule mismatch results. The problem type, pending anchor point, processing status, and processing requirements are written for the data governance tasks. After the data governance task is completed, obtain the processing result. When the processing result is a confirmed match, increase the connection weight of the corresponding standard matching connection or rule adaptation connection. When the processing result is a negative match, reduce the connection weight of the corresponding standard matching connection or rule adaptation connection. When the processing result contains new standards or new rules, new standard semantic anchors or rule semantic anchors are generated and written into the data governance semantic anchor graph. Based on the updated data governance semantic anchor graph, the updated data standard matching results and the updated governance rule adaptation results are regenerated.

[0013] The beneficial effects of this invention are: This invention standardizes data governance-related data through a data governance infrastructure module, generating a basic data governance dataset containing data object information, data standard information, governance rule information, and historical governance information. This enables data from different business systems, naming conventions, and governance sources to participate in subsequent processing with a unified structure. A semantic anchor generation module generates semantic anchors for data objects, standards, and rules, respectively. This allows field names, field annotations, contextual semantics, standard definitions, standard definitions, and rule conditions to be converted into computable and connectable semantic representations, thereby reducing mismatches and missed matches that occur when relying solely on field names or manual labels for data standard matching.

[0014] This invention uses an anchor point connection construction module to determine standard matching connections and rule adaptation connections. A semantic anchor point graph construction module then organizes the three types of semantic anchor points and two types of connections into a data governance semantic anchor point graph, creating continuous semantic connection paths between data objects, data standards, and governance rules. The standard rule matching module generates data standard matching results and governance rule adaptation results based on the data governance semantic anchor point graph. This allows for the determination of the corresponding data standard for a data object, while simultaneously identifying the appropriate governance rules. This reduces the workload of manually searching for standards, configuring rules, and verifying the scope of application during data governance, improving the consistency between data standard matching and governance rule adaptation.

[0015] This invention utilizes a governance problem identification module to identify semantic conflicts, missing standards, and rule mismatches based on data standard matching results and governance rule adaptation results. This transforms issues such as inconsistent semantic expressions of data objects, inconsistent constraints across multiple data standards, and inconsistencies between governance rules and data object constraint descriptions into clear data governance problem identification results. A feedback update module generates data governance tasks based on the identified data governance problem results and updates the data governance semantic anchor graph based on the processing results. This allows standard matching connections and rule adaptation connections to be continuously corrected with historical governance information, further generating updated data standard matching results and updated governance rule adaptation results. This enhances the automation, continuous adaptability, and stability of the data governance process. Attached Figure Description

[0016] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of a data governance system based on semantic analysis proposed in this invention; Figure 2 This is a schematic diagram illustrating the construction of a data governance semantic anchor graph for a data governance system based on semantic analysis proposed in this invention. Figure 3 This diagram illustrates the generation of data governance problem identification results in a data governance system based on semantic analysis proposed in this invention. Detailed Implementation

[0017] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0018] refer to Figures 1-3 A data governance system based on semantic analysis includes: The data governance infrastructure module is used to acquire data governance-related data and perform standardized processing to generate a basic data governance dataset. The semantic anchor generation module is used to perform semantic parsing on the basic dataset for data governance, and generate semantic anchors for data objects, standard semantic anchors, and rule-based semantic anchors respectively. The anchor connection building module is used to determine the standard matching connection between the semantic anchor of the data object and the standard semantic anchor, as well as the rule adaptation connection between the standard semantic anchor and the rule semantic anchor. The semantic anchor graph construction module is used to construct a data governance semantic anchor graph based on data object semantic anchors, standard semantic anchors, rule semantic anchors, standard matching connections, and rule adaptation connections. The standard rule matching module is used to determine the standard semantic anchor and rule semantic anchor corresponding to the semantic anchor of the data object based on the data governance semantic anchor graph, and generate data standard matching results and governance rule adaptation results. The governance issue identification module is used to identify semantic conflicts, missing standards, and rule mismatches based on the data standard matching results and governance rule adaptation results, and generate data governance issue identification results. The feedback update module is used to generate data governance tasks based on the data governance problem identification results, update the data governance semantic anchor graph based on the processing results of the data governance tasks, and generate updated data standard matching results and updated governance rule adaptation results.

[0019] In this embodiment, the data governance associated data includes data object information, data standard information, governance rule information, and historical governance information. The standardization process includes object identification unification, field hierarchy organization, standard name unification, standard definition organization, rule type marking, scope of application marking, processing result marking, hit status marking, and associated storage processing.

[0020] The data governance-related data comes from data object information in three business systems, comprising 120 data tables, 3680 fields, 86 data standards, 240 governance rules, and historical governance records for the past 180 days. Data object information includes database name, table name, field name, field comment, field type, field length, business domain, and sample field value.

[0021] Object identifiers uniformly adopt a three-level encoding method of "business system code - data table code - field code" to generate a 32-bit object identifier; when the same field appears in different sources, the object identifier is used as a unique index.

[0022] When organizing the field hierarchy, a field hierarchy structure is established according to three levels: business domain, data table, and field. The system writes each field into the corresponding business domain and data table, and records the field type, field length, field comments, and sample values, so that the field can correspond to its business domain and data table.

[0023] When standardizing data information by unifying standard names and standard definitions, a mapping table between standard names and aliases is established. Standard names with the same meaning from different sources are unified into the same standard name. For each data standard, the standard number, standard name, data type, field length, value format, allowed null state, and applicable business domain are organized.

[0024] When marking the rule type and scope of application for governance rule information, the governance rules are marked according to integrity verification, format verification, consistency verification, duplication verification and compliance verification, and the applicable business domain, data table scope and field semantic type of each governance rule are recorded.

[0025] When marking the processing results and hit status of historical governance information, the data governance task records within the set time range are read, and the processing results are marked as confirmed match, negative match, manual correction and rule addition, and the hit status is marked as hit, miss and conflict hit.

[0026] After processing, the data is associated and stored using the object identifier as the primary key. Data object information, data standard information, governance rule information, and historical governance information are written into the same data governance base dataset. Each record includes at least the object identifier, field level, standard number, rule number, processing result flag, and hit status flag, which serve as the basis for subsequent semantic anchor generation.

[0027] In this embodiment, the semantic anchor generation module includes: Separate data object information, data standard information, and governance rule information from the basic dataset for data governance; The data object information is segmented by field name, parsed by field annotation, and extracted by contextual semantics. The parsing results are semantically encoded to generate semantic anchors for the data object. The generation of semantic anchors for data objects involves: obtaining object identifiers, field names, field comments, table names, field types, field lengths, business domains, and sample values ​​from the data governance foundation dataset. Field names are limited to 1-128 characters, field comments to 1-256 characters, and a maximum of 20 sample values ​​are retained. Field names are segmented using underscores, hyphens, case boundaries, numeric boundaries, and Chinese / English boundaries. When parsing field comments, character uniformity, whitespace merging, and invalid symbol cleanup are performed. Comment keywords, business entities, attribute content, and constraint descriptions are identified based on the business dictionary and data standard dictionary. If a field comment is missing, the segmented field name, table name, and adjacent field names are combined to form a substitute comment text, and the comment source marker is recorded. When extracting contextual semantics, the system extracts the field names and field comments of the five fields before and after the current field in the same table, and also extracts the table name, table comments, business domain, and primary key field. The length of the context text is controlled within 512 characters. If the length is exceeded, the content is retained in the order of current field, adjacent fields, table-level description, and business domain description. The system encodes the field name segmentation result, field annotation parsing result, and context semantic extraction result into name semantic vector, annotation semantic vector, and context semantic vector, respectively. The name semantic vector, annotation semantic vector, and context semantic vector are merged with weights of 0.35, 0.40, and 0.25 to generate a data object semantic vector. The object identifier, name keywords, annotation keywords, constraint description, context window number, and data object semantic vector are combined to generate a data object semantic anchor. Each data object semantic anchor corresponds to a field-level data object, and the anchor type is marked as a data object anchor. The data standard information is parsed for standard name, standard definition and standard caliber, and the parsing results are semantically encoded to generate standard semantic anchors; The generation of standard semantic anchors involves obtaining the standard number, standard name, standard alias, standard definition, applicable business domain, and standard version from the data governance foundation dataset. The standard name length is controlled between 1 and 128 characters, the standard definition length is controlled between 1 and 512 characters, and the maximum number of standard aliases is 20. When parsing the standard name, character unification, word segmentation, and synonym merging are performed to generate standard name keywords, and the standard name keywords are matched with the standard number. When parsing the standard definition, invalid symbols are cleaned up, stop words are removed, and business terms are identified to extract the business subject, attribute meaning, and scope of use in the standard definition. When the standard definition is missing, the standard name, standard alias, and applicable business domain are combined into supplementary definition text, and the definition source mark is recorded. When parsing the standard scope, the data type, field length, value format, allowed null state, enumeration range, and sensitivity level are obtained from the data governance basic dataset. The data type is unified to character, numeric, date, enumeration, and boolean types. The field length is converted to a uniform length representation. The allowed null state is converted to a required or nullable mark. The value format, enumeration range, and sensitivity level are combined into a standard constraint description. The system encodes the standard name resolution result, standard definition resolution result, and standard caliber resolution result into standard name semantic vectors, standard definition semantic vectors, and standard caliber semantic vectors, respectively. These three semantic vectors are then merged with weights of 0.30, 0.40, and 0.30 to generate a standard semantic vector. The standard number, standard name keywords, definition keywords, standard constraint description, applicable business domain, standard version, and standard semantic vector are combined to generate standard semantic anchors. Each standard semantic anchor corresponds to a data standard, and the anchor type is marked as a standard anchor. The governance rule information is parsed by rule conditions, rule objects, and rule results. The parsing results are semantically encoded to generate rule semantic anchors. The generation of rule semantic anchors specifically involves: obtaining the rule number, rule name, rule type, rule conditions, rule objects, rule results, applicable scope, and rule version from the data governance foundation dataset. The rule name length is controlled between 1 and 128 characters, the rule condition length between 1 and 512 characters, the maximum number of rule objects is 50, and the rule result length is controlled between 1 and 256 characters. When parsing rule conditions, character unification, logical symbol organization, and condition fragment segmentation are performed. Judgment fields, comparison symbols, threshold ranges, format expressions, and triggering conditions in the rule conditions are identified, generating rule condition keywords and condition constraint descriptions. When parsing rule objects, the applicable business domain, data table range, field semantic type, and standard number are extracted. The rule object is bound to the corresponding applicable scope, generating the rule object parsing result. When parsing rule results, the post-execution processing action, problem type, risk level, and output status are identified. Processing actions are unified as prompt, interception, correction, desensitization, and reassignment, and output status is unified as passed, abnormal, and pending confirmation. The system encodes the rule condition parsing results, rule object parsing results, and rule result parsing results into rule condition semantic vectors, rule object semantic vectors, and rule result semantic vectors, respectively. These three semantic vectors are then fused with weights of 0.40, 0.35, and 0.25 to generate a rule semantic vector. The system then combines the rule number, rule type, rule condition keywords, condition constraint description, scope of application, problem type, processing action, rule version, and rule semantic vector to generate rule semantic anchors. Each rule semantic anchor corresponds to a governance rule, and the anchor type is marked as a rule anchor. Write data object semantic anchors, standard semantic anchors, and rule semantic anchors into the data governance basic dataset according to the object identifier.

[0028] In this embodiment, the anchor point connection construction module includes: Extract the semantic vector and semantic keywords of the data object from the semantic anchor points of the data object, and extract the standard semantic vector and standard keywords from the standard semantic anchor points; The similarity between the semantic vector of the data object and the standard semantic vector is calculated, and a standard matching score is generated by combining the overlap between semantic keywords and standard keywords. The standard matching score is generated as follows: Semantic vectors of the data object are retrieved from the semantic anchor points of the data object, and standard semantic vectors are retrieved from the standard semantic anchor points. Both vectors are normalized to convert the values ​​of each dimension to a uniform scale, and the directional similarity between the two vectors in corresponding dimensions is calculated to generate a vector matching value, represented by a value between 0 and 1. Semantic keyword sets are retrieved from the semantic anchor points of the data object, and standard keyword sets are retrieved from the standard semantic anchor points. Each set retains a maximum of 20 keywords. Synonym merging and stop word removal are performed on the two keyword sets respectively. The number of identical keywords in the two sets and the number of identical keywords after synonym merging are counted to obtain the keyword overlap. The keyword overlap value is calculated based on the keyword overlap value and the total number of keywords after merging the two sets, represented by a value between 0 and 1. The vector matching value and the keyword overlap value are weighted and fused, with the weight of the vector matching value set to 0.75 and the weight of the keyword overlap value set to 0.25, to generate the standard matching score. Connect the semantic anchors of data objects whose standard matching scores reach the standard connection threshold and whose business domains and constraint contents are consistent with the standard semantic anchors to generate standard matching connections. The standard connection threshold is set as follows: The system extracts standard matching records marked with completed processing results from historical governance information. Records marked as confirmed matches are used as positive samples, and records marked as negative matches are used as negative samples. Standard matching scores are calculated for positive and negative samples respectively, forming positive and negative sample score distributions. The interval from 0 to 1 is segmented and statistically analyzed at a score interval of 0.01, and the retention of positive samples and the exclusion of negative samples are calculated segment by segment at different score positions. The score position that allows both the retention of positive samples and the exclusion of negative samples to simultaneously meet the connection requirements is selected as the standard connection threshold. In this embodiment, the historical governance information uses standard matching records from the most recent 180 days, with a sample size of no less than 1000 records, a score statistical interval of 0.01, and a standard connection threshold of 0.80. When the standard matching score reaches 0.80 and the business domain and constraint content are consistent, the system generates a standard matching connection. When the standard matching score does not reach 0.80, the system does not generate a standard matching connection and writes the corresponding matching record into the pending confirmation record. Extract the standard constraint content from the standard semantic anchor point, extract the rule condition content from the rule semantic anchor point, perform consistency calculation on the standard constraint content and the rule condition content, and generate a rule adaptation score. The generation of the rule adaptation score is as follows: Standard constraint content is retrieved from the standard semantic anchor, and rule condition content is retrieved from the rule semantic anchor; the standard constraint content and rule condition content are converted into unified condition items, with the number of condition items controlled between 5 and 20. Each condition item includes a condition name, condition value, and condition source; consistency comparison is performed on condition values ​​under the same condition name. Complete consistency is recorded as a consistent condition, a partially consistent condition if a compatible range exists, and an inconsistent condition if no match is found; the system calculates the condition consistency value based on the number of consistent conditions, partially consistent conditions, and inconsistent conditions, and generates a range consistency value by considering whether the target object of the rule condition content covers the applicable range of the standard constraint content; the condition consistency value and the range consistency value are weighted and merged, with the condition consistency value weight set to 0.70 and the range consistency value weight set to 0.30, to generate the rule adaptation score. Connect the standard semantic anchor points that have reached the rule connection threshold and have the same applicable scope to the rule semantic anchor points to generate rule adaptation connections.

[0029] The rule connection threshold is set as follows: The system extracts rule adaptation records that have been processed and marked from historical governance information. Records marked as confirmed adaptation are used as positive samples, and records marked as negative adaptation are used as negative samples. Rule adaptation scores are calculated for positive and negative samples respectively, forming positive and negative sample score distributions. The interval from 0 to 1 is segmented and statistically analyzed according to a score interval of 0.01, and the retention of positive samples and the exclusion of negative samples are calculated for each segment at different score positions. A selection is made that allows both positive sample retention and negative sample exclusion to occur simultaneously. The score position that meets the connection requirements is used as the rule connection threshold. In this implementation, historical governance information uses rule adaptation records from the most recent 180 days, with a sample size of no less than 1000 records, a score statistical interval of 0.01, and a rule connection threshold of 0.75. When the rule adaptation score reaches 0.75, and the applicable scope of the rule semantic anchor point covers the applicable business domain and standard constraint content corresponding to the standard semantic anchor point, the system generates a rule adaptation connection. When the rule adaptation score does not reach 0.75, the system does not generate a rule adaptation connection and writes the corresponding adaptation record into the pending confirmation record. In this embodiment, the semantic anchor graph construction module includes: Write the semantic anchors of data objects, standard semantic anchors, and rule semantic anchors into the graph node set, and record the anchor type, anchor number, and semantic vector for each graph node; Write the standard matching connections and rule-adapted connections into the graph edge set, and record the connection direction, connection type and connection weight for each graph edge; The connection direction and connection weight are determined as follows: The connection direction is determined according to the governance processing order. The starting anchor point of the standard matching connection is the data object semantic anchor point, and the ending anchor point is the standard semantic anchor point. The connection direction is marked as the data object pointing to the data standard. The starting anchor point of the rule adaptation connection is the standard semantic anchor point, and the ending anchor point is the rule semantic anchor point. The connection direction is marked as the data standard pointing to the governance rule. The connection weight is written according to the score when the connection is generated. The connection weight of the standard matching connection is taken from the standard matching score, and the connection weight of the rule adaptation connection is taken from the rule adaptation score. The connection weight is represented by a value between 0 and 1 and is retained to four decimal places. When there are multiple graph edges of the same connection type between the same set of anchor points, the system retains the graph edge with the highest connection weight and writes the remaining graph edges into the duplicate connection record. When the standard matching score reaches 0.80 and the business domain and constraint content are consistent, the corresponding standard matching connection is written into the graph edge set. When the rule adaptation score reaches 0.75 and the applicable scope is consistent, the corresponding rule adaptation connection is written into the graph edge set. Organize the graph edges in the order of data object semantic anchor points pointing to standard semantic anchor points, and standard semantic anchor points pointing to rule semantic anchor points, and generate anchor point connection paths; Anchor connection paths are merged for duplicate paths and marked for broken paths, retaining anchor connection paths whose connection weights reach the corresponding connection thresholds. Specifically, when multiple anchor connection paths have the same data object semantic anchor, standard semantic anchor, and rule semantic anchor, they are determined to be duplicate paths. The connection weights of standard matching connections and rule-adapted connections in the duplicate paths are compared, and the anchor connection path with the highest weighted result of the two types of connection weights is retained, while the remaining anchor connection paths are added to the duplicate path record. When an anchor connection path lacks a standard matching connection from a data object semantic anchor to a standard semantic anchor, it is marked as a duplicate path. This is recorded as a standard broken path. When an anchor connection path lacks a rule-adapted connection from a standard semantic anchor to a rule semantic anchor, it is marked as a rule broken path. The connection weights of the retained anchor connection paths are verified. When the connection weight of the standard matching connection reaches the standard connection threshold and the connection weight of the rule-adapted connection reaches the rule connection threshold, the anchor connection path is written into the valid path set. Anchor connection paths that do not reach the corresponding connection threshold are written into the path set to be confirmed. After processing, the valid path set, duplicate path records, broken path markers, and path set to be confirmed are written into the data governance semantic anchor graph. The graph node set, graph edge set, and anchor connection path are associated and stored to construct a data governance semantic anchor graph.

[0030] In this embodiment, the standard rule matching module includes: Select semantic anchors of data objects to be matched from the semantic anchor graph of data governance, find standard semantic anchors that are connected to the semantic anchors of data objects to be matched along the standard matching connection, determine the standard semantic anchors whose connection weight reaches the standard connection threshold as the target standard semantic anchor set, and generate data standard matching results. Search for rule semantic anchors along the rule adaptation connection corresponding to the target standard semantic anchor. Determine the rule semantic anchors whose connection weight reaches the rule connection threshold and whose applicable scope is consistent as the target rule semantic anchor set, and generate governance rule adaptation results. The generation of governance rule adaptation results is as follows: Starting from each target standard semantic anchor in the target standard semantic anchor set, read the graph edges whose starting anchor number matches the target standard semantic anchor number and whose connection type is rule adaptation connection. The system then searches for the corresponding rule semantic anchor based on the ending anchor number in the graph edge, and retrieves the rule number, rule type, rule condition content, applicable scope, processing action, and rule version from the rule semantic anchor. The found rule semantic anchor is then subjected to connection weight verification and applicable scope verification. The connection weight uses the rule adaptation score recorded in the graph edge, and the rule connection threshold is set to 0.7 in this embodiment. 5. When the connection weight reaches 0.75 and the applicable scope of the rule semantic anchor covers the applicable business domain, field semantic type, and standard constraint content corresponding to the target standard semantic anchor, the rule semantic anchor is written into the target rule semantic anchor set; when the connection weight does not reach 0.75 or the applicable scope is inconsistent, the rule semantic anchor is written into the set of rules to be confirmed; after the target rule semantic anchor set is formed, the system organizes the governance rule adaptation results according to the rule type and processing action. The governance rule adaptation results include the target standard semantic anchor number, the target rule semantic anchor set, the connection weight, the rule type, the processing action, and the adaptation status. The semantic anchors of the data object, the set of target standard semantic anchors, the set of target rule semantic anchors, the connection weights, and the matching status are stored together.

[0031] In this embodiment, the governance problem identification module includes: Obtain data standard matching results and governance rule adaptation results, and perform matching status checks on the target standard semantic anchors in the data standard matching results; When the semantic anchors of the data object do not form the target standard semantic anchor set, or when the target standard semantic anchor set is empty, a standard missing result is generated; When the semantic anchor point of the same data object corresponds to multiple target standard semantic anchor points, and the standard constraint content between the multiple target standard semantic anchor points is inconsistent, a semantic conflict result is generated. The generation of semantic conflict results is as follows: When the same data object semantic anchor corresponds to multiple target standard semantic anchors, the system retrieves the object identifier corresponding to the data object semantic anchor and searches for all target standard semantic anchors associated with that object identifier in the data standard matching results; when the number of target standard semantic anchors is greater than 1, the system reads the standard number, standard name, data type, field length, value format, allowed null state, sensitivity level, and applicable business domain from each target standard semantic anchor, and organizes the above content into standard constraint content; the standard constraint content between multiple target standard semantic anchors is compared item by item. If the data type is inconsistent, the field length difference exceeds the set length difference range, the value format is inconsistent, or the allowed null state is not specified, the system will resolve the conflict. If the empty states are opposite, the sensitivity levels are inconsistent, or the applicable business domains are inconsistent, the corresponding target standard semantic anchors are marked as conflicting standard semantic anchors. In this embodiment, the field length difference range is set to 10%. When the field length difference between two target standard semantic anchors exceeds 10%, it is determined to be a field length conflict. The system writes the standard number, conflict item, conflict content, standard matching score, and object identifier of the conflicting standard semantic anchor into the semantic conflict record and marks the semantic conflict record as a semantic conflict result. When the standard constraint content between multiple target standard semantic anchors is consistent, the system retains the target standard semantic anchor with the highest standard matching score as a valid matching result and writes the remaining target standard semantic anchors into the candidate standard record. Perform an adaptation status check on the set of semantic anchor points of the target rule in the governance rule adaptation results; When the target standard semantic anchor points do not form a set of target rule semantic anchor points, a rule missing result is generated; When the rule condition content in the target rule semantic anchor set is inconsistent with the constraint description corresponding to the data object semantic anchor, a rule mismatch result is generated. The generation of rule mismatch results specifically involves: retrieving the object identifier, constraint description, field type, field length, and business domain corresponding to the data object semantic anchor; and retrieving the rule number, rule type, rule condition content, applicable scope, and processing action corresponding to each rule semantic anchor in the target rule semantic anchor set; converting the constraint description and rule condition content into unified condition items, each including condition name, condition value, and condition source; and comparing the condition values ​​under the same condition name item by item. If the condition values ​​in the data object semantic anchor correspond to the condition values ​​in the rule semantic anchor, then the rule conditions are considered consistent. If the condition value in the semantic anchor of a rule does not cover the condition value in the semantic anchor of a data object, or if the scope of application in the semantic anchor of a rule does not cover the business domain to which the semantic anchor of a data object belongs, then the rule condition is determined to be inconsistent. When any target rule semantic anchor is determined to be inconsistent, the system generates a rule mismatch record, which includes the object identifier, rule number, inconsistent condition item, inconsistent content, rule type, and processing action. The system writes the rule mismatch record into the data governance problem identification result and transfers the corresponding target rule semantic anchor from the governance rule adaptation result to the set of rules to be confirmed. The results of missing standards, semantic conflicts, missing rules, and rule mismatches are associated with object identifiers to generate data governance problem identification results.

[0032] In this embodiment, the feedback update module includes: Obtain the data governance problem identification results, and determine the corresponding data object semantic anchor, target standard semantic anchor, and target rule semantic anchor according to the object identifier; Data governance tasks are generated based on results of missing standards, semantic conflicts, missing rules, and rule mismatches. Each data governance task is then assigned a problem type, pending anchor point, processing status, and processing requirements. Specifically, the object identifier, problem type, problem content, and associated anchor point number are retrieved from the data governance problem identification results, and a task number is generated for each problem record. The task number is generated by combining the business domain code, date, and sequence number, with the sequence number using a 6-digit incrementing sequence. The system determines the anchor points to be processed and the processing requirements based on the problem type: When the problem type is "standard missing," the semantic anchor point of the data object is used as the anchor point to be processed, and the processing requirement is to supplement the target standard semantic anchor point or confirm that no matching standard is needed; when the problem type is "semantic conflict," the semantic anchor point of the conflicting standard is used as the anchor point to be processed, and the processing requirement is to confirm the valid standard semantic anchor point and mark invalid connections; when the problem type is "rule missing," the semantic anchor point of the target standard is used as the anchor point to be processed, and the processing requirement is to supplement the target rule semantic anchor point set or confirm that no rule adaptation is needed; when the problem type is "rule mismatch," the semantic anchor point of the mismatched rule is used as the anchor point to be processed, and the processing requirement is to confirm the rule condition content or adjust the scope of application. Each data governance task is written with a task number, object identifier, problem type, anchor point to be processed, processing requirements, creation time, and processing status. The initial processing status is marked as "pending"; the creation time is accurate to the second.

[0033] After a data governance task is generated, the system writes the task to the data governance task queue and pushes it to the data governance workbench. The data governance workbench pulls 100 tasks from the task queue every 5 minutes. The system determines the task processing flow based on the problem type: standard missing results and semantic conflict results enter the standard processing flow, while rule missing results and rule mismatch results enter the rule processing flow. Each data governance task displays its task number, object identifier, field name, field comment, problem type, pending anchor, candidate standard semantic anchor, candidate rule semantic anchor, and processing requirements in the data governance workbench. The task status is initially marked as pending. When a data governance task enters the standard processing flow, for results with missing standards, the system selects the standard semantic anchor with the highest connection weight and consistent applicable business domain from the candidate standard semantic anchors, or generates a new standard request if no candidate standard semantic anchor exists. For results with semantic conflicts, the system retains the standard semantic anchor with the highest connection weight and the latest standard version from the conflicting standard semantic anchors, and marks the remaining conflicting connections as invalid. When a data governance task enters the rule processing flow, for results with missing rules, the system selects the rule semantic anchor with a connection weight reaching the rule connection threshold and consistent applicable scope from the candidate rule semantic anchors, or generates a new rule request if no candidate rule semantic anchor exists. For results with rule mismatches, the system performs consistency checks on the rule condition content and the constraint description corresponding to the data object semantic anchor, and generates a confirmation match or negative match processing result based on the check result. After generating the processing result, the system updates the task status to completed or returned. After the processing results are generated, they are written into the historical governance information and stored in association with the task number, object identifier, processing flow, processing time, problem type, and processing description. After the data governance task is completed, the processing result is obtained. When the processing result is a confirmed match, the connection weight of the corresponding standard matching connection or rule adaptation connection is increased. In this embodiment, the connection weight is increased by 0.03, and the updated connection weight is limited to between 0 and 1. When the processing result is a negative match, the connection weight of the corresponding standard matching connection or rule adaptation connection is reduced; in this embodiment, the connection weight is reduced by 0.05, and the updated connection weight is limited to between 0 and 1. When the processing result contains new standards or new rules, new standard semantic anchors or rule semantic anchors are generated and written into the data governance semantic anchor graph. When the processing result contains a new standard, the system extracts the object identifier, field name, field comment, field type, field length, business domain, constraint description, and standard processing result from the processing result to generate a new standard record. The system assigns a standard number to the new standard record. The standard number is generated by combining the business domain code, standard category code, and 6-digit sequence number. The field name, field comment, and constraint description are organized into a standard name, standard definition, and standard scope. The standard name length is controlled between 1 and 128 characters, the standard definition length is controlled between 1 and 512 characters, and the standard scope includes at least the data type, field length, value format, allowed null states, and applicable business domain. According to the standard semantic anchor generation method, the new standard record is parsed for standard name, standard definition, and standard scope, and semantically encoded to generate a new standard semantic anchor. After the new standard semantic anchor is generated, it is written into the graph node set of the data governance semantic anchor graph, and the standard matching score is recalculated with the corresponding data object semantic anchor as the starting point. When the standard matching score reaches the standard connection threshold and the business domain and constraint content are consistent, a new standard matching connection is generated and written into the graph edge set. When the processing result contains a new rule, the system extracts the standard number, rule type, rule condition content, scope of application, processing action, and rule processing result from the processing result to generate a new rule record. The system assigns a rule number to the new rule record, which is generated by combining the rule type code, business domain code, and 6-digit sequence number. The length of the rule condition content is controlled between 1 and 512 characters, and the scope of application must include at least the business domain, field semantic type, and standard number. The processing action is determined from prompting, interception, correction, desensitization, and dispatch. The system performs rule condition parsing, rule object parsing, and rule result parsing on the new rule record according to the rule semantic anchor point generation method, and performs semantic encoding to generate a new rule semantic anchor point. After the new rule semantic anchor point is generated, the system writes it into the graph node set of the data governance semantic anchor point graph, and recalculates the rule adaptation score with the corresponding standard semantic anchor point as the starting point. When the rule adaptation score reaches the rule connection threshold and the scope of application is consistent, a new rule adaptation connection is generated and written into the graph edge set. Based on the updated data governance semantic anchor graph, the updated data standard matching results and the updated governance rule adaptation results are regenerated.

[0034] Example 1: To verify the feasibility of this invention in practice, it was applied to a data governance scenario of a retail enterprise. This enterprise simultaneously runs a membership management system, an order transaction system, an after-sales service system, a financial settlement system, and a marketing campaign system. These systems have different development cycles, and their field naming rules and data standard maintenance methods are inconsistent. The membership management system contains fields such as customer phone number, contact number, and mobile phone number; the order transaction system contains fields such as delivery phone number and recipient contact information; and the after-sales service system contains fields such as service phone number and follow-up number. While these fields overlap in their business meanings, their names, annotations, and the business domains they belong to are not entirely consistent. Traditional governance methods mainly rely on field name retrieval and manual rule configuration, which can easily match fields with different business meanings to the same data standard and may overlook format validation, integrity validation, and sensitive information processing rules.

[0035] In this scenario, the system accesses data governance data from five business systems, including 120 data tables, 3680 fields, 86 data standards, 240 governance rules, and historical governance information from the past 180 days. Data object information includes database name, table name, field name, field comment, field type, field length, business domain, and sample values. The system standardizes this data to generate a basic data governance dataset. For each field, the system retains a maximum of 20 sample values. Field names are limited to 1-128 characters in length, and field comments are limited to 1-256 characters in length. An object identifier is generated using a combination of business system code, data table code, and field code, allowing different representations of the same field in the metadata directory, business descriptions, and historical governance information to be grouped under the same object.

[0036] The system generates data object semantic anchors, standard semantic anchors, and rule semantic anchors from the data governance foundation dataset. For fields with complete annotations, the system generates data object semantic anchors by combining the field name, field annotation, data table name, business domain, and adjacent field content. For fields with missing annotations, the system uses the field name segmentation result, data table name, and the field names of the five fields before and after the field to form alternative annotation text. Standard semantic anchors are generated from standard names, standard definitions, and standard definitions, while rule semantic anchors are generated from rule conditions, rule objects, and rule results. Subsequently, the system calculates the standard matching score between data object semantic anchors and standard semantic anchors, establishing standard matching connections for anchors with a score of 0.80 and consistent business domain and constraint content; it also calculates the rule adaptation score between standard semantic anchors and rule semantic anchors, establishing rule adaptation connections for anchors with a score of 0.75 and consistent applicable scope, and constructs a data governance semantic anchor graph accordingly.

[0037] In actual governance, 38 contact information fields in the member domain were organized into the same semantic processing chain. A significant proportion of these fields had abbreviations, incomplete comments, or ambiguous contexts. Instead of matching solely by field name, the system combined field comments, adjacent fields, the business domain, and standard definitions to generate data standard matching results. This ensured that customer contact numbers, delivery contact numbers, and after-sales follow-up numbers were linked to their respective data standards. For the delivery contact number field in the order transaction system, the system linked it to the order delivery contact information standard based on contextual fields such as delivery address, tracking number, and order number in its data table. It also adapted contact information format validation, null value validation, and sensitive information processing rules to prevent misclassification into the basic member contact information standard.

[0038] In the after-sales service system, it was discovered that some status fields were simultaneously connected to multiple standard semantic anchors. For example, after-sales work order status and customer account status had similar field names, but their value formats and applicable business domains in the standard constraints were inconsistent. The system identified this as a semantic conflict and generated a data governance task. The task included the object identifier, the conflicting standard semantic anchor, the conflicting item, and the processing requirements. After the task was completed, the system increased the connection weight of valid standard matching connections and decreased the connection weight of invalid connections based on the processing results, enabling subsequent fields of the same type to complete matching along the updated data governance semantic anchor graph.

[0039] In the financial settlement system, the system identifies that monetary fields have formed a set of target standard semantic anchors, but some fields have not formed a set of target rule semantic anchors, thus generating a rule missing result. Based on the rule conditions and applicable scope of the candidate rule semantic anchors, the system supplements rules for monetary format verification, precision consistency verification, and null value verification, and generates new rule semantic anchors after adding the new rules, writing them into the data governance semantic anchor graph. After continuous governance cycles, the number of times the enterprise records repeated confirmations of field standard matching decreased, the number of times rules were missing and supplemented decreased, and the number of records awaiting confirmation due to missing field annotations decreased. This demonstrates that the present invention can improve the accuracy of data standard matching, enhance the automatic adaptation capability of governance rules, and continuously optimize governance results through feedback updates in data governance scenarios with multiple systems, multiple standards, and multiple naming methods.

[0040] Table 1 Comparison of the effects of data governance methods based on semantic analysis

[0041] As shown in Table 1, the data standard matching accuracy of the method of this invention increased from 84.6% to 91.8%, the false match rate decreased from 8.2% to 4.6%, and the missed match rate decreased from 7.2% to 3.6%. The main reason is that this invention does not rely solely on field names, but instead converts field names, field comments, contextual semantics, and standard definitions into semantic anchors for matching. In cases where field abbreviations, field comments are incomplete, or naming conventions differ across business systems, semantic anchors provide more complete matching information than keywords, thus reducing some false and missed matches.

[0042] Regarding governance rule adaptation, the accuracy of governance rule adaptation using the method of this invention increased from 81.5% to 89.4%, and the omission rate of governance rules decreased from 11.6% to 6.8%. This result demonstrates that by establishing rule adaptation connections between standard semantic anchors and rule semantic anchors, the system can continue to match applicable governance rules after determining data standards, reducing omissions that occur when manually configuring rules item by item. Especially for rules such as format validation, integrity validation, and compliance validation, which are strongly related to standard constraints, this invention can establish rule adaptation connections based on the consistency between standard constraint content and rule condition content, thereby improving the completeness of rule configuration.

[0043] In terms of problem identification, the accuracy of semantic conflict identification using the method of this invention has increased from 74.8% to 84.9%, and the accuracy of rule mismatch identification has increased from 76.3% to 86.1%. This improvement mainly comes from the connection path check and constraint content comparison in the data governance semantic anchor graph. For cases where the same data object is connected to multiple standard semantic anchors, the system can further compare the data types, field lengths, value formats, and applicable business domains among the multiple standard semantic anchors to identify semantic conflicts caused by inconsistencies in standard constraint content. For cases where governance rules and data object constraint descriptions are inconsistent, the system can generate rule mismatch results through consistency checks between rule condition content and constraint descriptions.

[0044] Overall, the method of this invention has more stable data standard matching capabilities and governance rule adaptation capabilities compared to traditional methods. The performance improvement is due to the fact that this invention transforms data objects, data standards, and governance rules into a computable and connectable structure through semantic anchors, and constructs a data governance semantic anchor graph through standard matching connections and rule adaptation connections. This allows the data governance process to move beyond simply matching field names and configuring fixed rules, and instead make comprehensive judgments based on contextual semantics, standard definitions, and rule conditions. Therefore, it achieves better results in terms of accuracy, mismatch control, missed match control, and problem identification.

[0045] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A data governance system based on semantic analysis, characterized in that, include: The data governance infrastructure module is used to acquire data governance-related data and perform standardized processing to generate a basic data governance dataset. The semantic anchor generation module is used to perform semantic parsing on the basic dataset for data governance, and generate semantic anchors for data objects, standard semantic anchors, and rule-based semantic anchors respectively. The anchor connection building module is used to determine the standard matching connection between the semantic anchor of the data object and the standard semantic anchor, as well as the rule adaptation connection between the standard semantic anchor and the rule semantic anchor. The semantic anchor graph construction module is used to construct a data governance semantic anchor graph based on data object semantic anchors, standard semantic anchors, rule semantic anchors, standard matching connections, and rule adaptation connections. The standard rule matching module is used to determine the standard semantic anchor and rule semantic anchor corresponding to the semantic anchor of the data object based on the data governance semantic anchor graph, and generate data standard matching results and governance rule adaptation results. The governance issue identification module is used to identify semantic conflicts, missing standards, and rule mismatches based on the data standard matching results and governance rule adaptation results, and generate data governance issue identification results. The feedback update module is used to generate data governance tasks based on the data governance problem identification results, update the data governance semantic anchor graph based on the processing results of the data governance tasks, and generate updated data standard matching results and updated governance rule adaptation results.

2. The data governance system based on semantic analysis according to claim 1, characterized in that, The data governance associated data includes data object information, data standard information, governance rule information, and historical governance information. The standardization process includes unified object identification, organized field hierarchy, unified standard name, organized standard definition, rule type marking, scope of application marking, processing result marking, hit status marking, and associated storage processing.

3. A data governance system based on semantic analysis according to claim 1, characterized in that, The semantic anchor point generation module includes: Separate data object information, data standard information, and governance rule information from the basic dataset for data governance; The data object information is segmented by field name, parsed by field annotation, and extracted by contextual semantics. The parsing results are semantically encoded to generate semantic anchors for the data object. The data standard information is parsed for standard name, standard definition and standard caliber, and the parsing results are semantically encoded to generate standard semantic anchors; The governance rule information is parsed by rule conditions, rule objects, and rule results. The parsing results are semantically encoded to generate rule semantic anchors. Write data object semantic anchors, standard semantic anchors, and rule semantic anchors into the data governance basic dataset according to the object identifier.

4. A data governance system based on semantic analysis according to claim 1, characterized in that, The anchor point connection construction module includes: Extract the semantic vector and semantic keywords of the data object from the semantic anchor points of the data object, and extract the standard semantic vector and standard keywords from the standard semantic anchor points; The similarity between the semantic vector of the data object and the standard semantic vector is calculated, and a standard matching score is generated by combining the overlap between semantic keywords and standard keywords. Connect the semantic anchors of data objects whose standard matching scores reach the standard connection threshold and whose business domains and constraint contents are consistent with the standard semantic anchors to generate standard matching connections. Extract the standard constraint content from the standard semantic anchor point, extract the rule condition content from the rule semantic anchor point, perform consistency calculation on the standard constraint content and the rule condition content, and generate a rule adaptation score. Connect the standard semantic anchor points that have reached the rule connection threshold and have the same applicable scope to the rule semantic anchor points to generate rule adaptation connections.

5. A data governance system based on semantic analysis according to claim 1, characterized in that, The semantic anchor graph construction module includes: Write the semantic anchors of data objects, standard semantic anchors, and rule semantic anchors into the graph node set, and record the anchor type, anchor number, and semantic vector for each graph node; Write the standard matching connections and rule-adapted connections into the graph edge set, and record the connection direction, connection type and connection weight for each graph edge; Organize the graph edges in the order of data object semantic anchor points pointing to standard semantic anchor points, and standard semantic anchor points pointing to rule semantic anchor points, and generate anchor point connection paths; Repeated paths are merged and broken paths are marked for anchor connection paths, and anchor connection paths whose connection weights reach the corresponding connection thresholds are retained. The graph node set, graph edge set, and anchor connection path are associated and stored to construct a data governance semantic anchor graph.

6. A data governance system based on semantic analysis according to claim 1, characterized in that, The standard rule matching module includes: Select semantic anchors of data objects to be matched from the semantic anchor graph of data governance, find standard semantic anchors that are connected to the semantic anchors of data objects to be matched along the standard matching connection, determine the standard semantic anchors whose connection weight reaches the standard connection threshold as the target standard semantic anchor set, and generate data standard matching results. Search for rule semantic anchors along the rule adaptation connection corresponding to the target standard semantic anchor. Determine the rule semantic anchors whose connection weight reaches the rule connection threshold and whose applicable scope is consistent as the target rule semantic anchor set, and generate governance rule adaptation results. The semantic anchors of the data object, the set of target standard semantic anchors, the set of target rule semantic anchors, the connection weights, and the matching status are stored together.

7. A data governance system based on semantic analysis according to claim 1, characterized in that, The governance problem identification module includes: Obtain data standard matching results and governance rule adaptation results, and perform matching status checks on the target standard semantic anchors in the data standard matching results; When the semantic anchors of the data object do not form the target standard semantic anchor set, or when the target standard semantic anchor set is empty, a standard missing result is generated; When the semantic anchor point of the same data object corresponds to multiple target standard semantic anchor points, and the standard constraint content between the multiple target standard semantic anchor points is inconsistent, a semantic conflict result is generated. Perform an adaptation status check on the set of semantic anchor points of the target rule in the governance rule adaptation results; When the target standard semantic anchor points do not form a set of target rule semantic anchor points, a rule missing result is generated; When the rule condition content in the target rule semantic anchor set is inconsistent with the constraint description corresponding to the data object semantic anchor, a rule mismatch result is generated. The results of missing standards, semantic conflicts, missing rules, and rule mismatches are associated with object identifiers to generate data governance problem identification results.

8. A data governance system based on semantic analysis according to claim 1, characterized in that, The feedback update module includes: Obtain the data governance problem identification results, and determine the corresponding data object semantic anchor, target standard semantic anchor, and target rule semantic anchor according to the object identifier; Data governance tasks are generated based on standard missing results, semantic conflict results, rule missing results, and rule mismatch results. The problem type, pending anchor point, processing status, and processing requirements are written for the data governance tasks. After the data governance task is completed, obtain the processing result. When the processing result is a confirmed match, increase the connection weight of the corresponding standard matching connection or rule adaptation connection. When the processing result is a negative match, reduce the connection weight of the corresponding standard matching connection or rule adaptation connection. When the processing result contains new standards or new rules, new standard semantic anchors or rule semantic anchors are generated and written into the data governance semantic anchor graph. Based on the updated data governance semantic anchor graph, the updated data standard matching results and the updated governance rule adaptation results are regenerated.

Images