Cross-document same procurement object identification method and system for power plant bidding procurement

Abstract

The application discloses a cross-document same procurement object recognition method for power plant bidding procurement and belongs to the technical field of data processing. The method comprises the following steps: obtaining procurement object information, extracting procurement object records from the procurement object information; constructing an object attribute closed hypergraph according to the procurement object records to obtain a same procurement object relationship state; performing counterfactual branch contribution calculation on the same procurement object relationship state, mapping the contribution value into support cost and exclusion cost according to the contribution value, and generating a same procurement object relationship cost; performing multi-cut division according to the same procurement object relationship cost to form procurement object grouping; performing attribute binding according to the procurement object grouping to form procurement object summary records and attribute conflict items, and outputting a recognition result. The method realizes cross-document procurement object relationship structured expression through the object attribute closed hypergraph, can uniformly judge multiple states to reduce the risk of misjudgment and misdivision, and provides stable intermediate representation for subsequent global grouping.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, specifically to a method, system, device, and medium for identifying the same procurement object across documents in power plant bidding and procurement. Background Technology

[0002] The bidding and procurement process in power plants spans multiple stages, with procurement object information scattered across documents such as technical specifications, drawings, purchase requisitions, and quotation lists, circulating among various parties. Different documents have different focuses; for example, technical specifications emphasize functional requirements, drawings emphasize tag numbers and drawing numbers, purchase requisitions emphasize procurement names, and quotation lists emphasize the quoting objects. Power plant process systems are complex, and procurement objects are diverse, potentially appearing repeatedly in different professional volumes and business forms with inconsistent granularity of name recording. Some documents also suffer from missing items, misalignments, and inconsistent terminology. The bidding and procurement review process requires identifying the same procurement object across multiple documents, summarizing relevant information, and verifying technical requirements to support subsequent work. Current methods revolve around document parsing, field extraction, and object comparison. These methods include manual review, where reviewers compare and merge documents manually; rule-based processing, which involves segmenting and locating various documents and matching records according to multiple conditions; converting procurement records into structured entries; calculating record correspondences using multiple methods; and some engineering management systems introducing tags, codes, and numbers for database management. However, existing methods, which mainly rely on similar names, overlapping fields, or pairwise record matching, have many shortcomings. They are difficult to represent the closed relationship of multiple procurement records forming a single procurement object, cannot uniformly determine complementary records between different documents within the same relational object, and can easily confuse missing document items with missing actual information. It is also difficult to link and distinguish tag numbers, drawing numbers, host connection relationships, and entry / exit relationships. Furthermore, the unified scoring method compresses information that supports merging and promoting splitting, leading to the mis-merging of similar-named but different items and the mis-segmentation of complementary records across documents, which affects the consistency review of procurement object summaries and technical requirements. Summary of the Invention

[0003] To address the problems in existing technologies where procurement records primarily rely on pairwise matching based on similar names, making it difficult to represent closed relationships, prone to confusion and missing items, and difficult to perform linkage judgments, resulting in errors in record merging and classification during scoring, this invention provides a method for identifying the same procurement object across documents in power plant bidding and procurement. This method achieves a structured representation of cross-document procurement object relationships through a closed hypergraph of object attributes, enabling unified judgment of multiple states to reduce the risk of mismerging and misclassification, and providing a stable intermediate representation for subsequent global grouping.

[0004] To achieve the above objectives, the present invention provides the following technical solution.

[0005] In a first aspect, the present invention provides a method for identifying the same procurement object across documents in power plant bidding and procurement, comprising: Obtain procurement object information and extract procurement object records from the procurement object information; Construct a closed hypergraph of object attributes based on the procurement object records, forming hyperedges of relationships within the same procurement object; calculate the relationship status of the same procurement object based on the hyperedges of relationships within the same procurement object. The counterfactual branch contribution calculation is performed on the relationship status of the same procurement object to obtain the contribution value of the same procurement object and the segmentation into different procurement objects. The contribution value of the same procurement object and the segmentation into different procurement objects is mapped to the support cost and the rejection cost. The relationship cost of the same procurement object is generated based on the support cost and the rejection cost. Based on the relationship cost of the same procurement object, the closed hypergraph of object attributes is divided into high-order relationship multi-cut partitions to form procurement object groups; Based on the procurement objects, attribute binding is performed to form a summary record of procurement objects and attribute conflict items; Based on the summary record of procurement objects and attribute conflict items, determine the consistency results of technical requirements, and output the procurement object grouping, the summary record of procurement objects, and the consistency results of technical requirements.

[0006] As a further improvement of the present invention, the step of obtaining procurement object information and extracting procurement object records from the procurement object information includes: Obtain information on the procurement targets from the technical specifications, drawings, purchase requisitions, and quotation lists in power plant bidding processes; Extract the object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit, and document source from the procurement object information to form a procurement object record.

[0007] As a further improvement of the present invention, the step of constructing a closed hypergraph of object attributes based on the procurement object record to form a hyperedge of the same procurement object relationship; and calculating the relationship state of the same procurement object based on the hyperedge of the same procurement object relationship, includes: Construct a closed hypergraph of object attributes based on the procurement object record. Use the procurement object record as the record node, and use system ownership, host connection relationship, entry and exit relationship, tag number and map number, parameter unit and document source as attribute nodes. Combine the record nodes according to the complementary conditions of object category correspondence, tag number and map number correspondence, host connection relationship connection, entry and exit relationship connection and document source to form the same procurement object relationship hyperedge. The relationship status of the same procurement object is calculated based on the hyperedge of the relationship of the same procurement object.

[0008] As a further improvement of the present invention, the counterfactual branch contribution calculation for the relationship state of the same procurement object yields contribution values ​​for the same procurement object and for segments into different procurement objects. These contribution values ​​are mapped to support costs and rejection costs. The relationship cost for the same procurement object is generated based on the support costs and rejection costs, including: Counterfactual branch contribution calculation is performed on the relationship status of the same procurement object to obtain the combined evaluation value of the same procurement object and the evaluation value of the different procurement objects. The combined evaluation value of the same procurement object and the evaluation value of the different procurement objects are then used to form the benchmark evaluation value. The contribution value of the corresponding branch pair when it is merged into the same procurement object and the contribution value of the corresponding branch pair when it is divided into different procurement objects are calculated based on the difference between the benchmark evaluation value and the replacement evaluation value. Based on the attribute coverage status, tag number and map number correspondence status, and connection connectivity status of the same procurement object relationship status, determine the contribution values ​​corresponding to attribute coverage, tag number and map number correspondence, and connection connectivity, and map them as support costs. The system's ownership branch, entry / exit relationship branch, and parameter unit branch are divided into branches corresponding to different procurement objects and mapped to exclusion costs. Generate the relationship cost for the same procurement object based on the support cost and the exclusion cost.

[0009] As a further improvement of the present invention, the step of performing high-order relation multi-cut partitioning on the closed hypergraph of object attributes based on the relation cost of the same procurement object to form procurement object grouping includes: Based on the closed hypergraph of object attributes, read the procurement object record, the hyperedge of the same procurement object relationship, and the cost of the same procurement object relationship, and summarize the cost of the same procurement object relationship based on the hyperedge of the same procurement object relationship to form a list of costs of the same procurement object relationship; Based on the cost list of relationships for the same procurement object, a segmentation decision variable is assigned to the super-edge of each relationship for the same procurement object to form a segmentation solution structure; Based on the segmentation solution structure, a segmentation consistency constraint set is constructed. Based on the segmentation consistency constraint set, a higher-order relation multi-cut solution is performed on the segmentation decision variables to obtain the segmentation decision result. Based on the segmentation determination results, the connectivity of the closed hypergraph of object attributes is reconstructed to form group labels, and the procurement objects are grouped according to the group labels.

[0010] As a further improvement of the present invention, the step of binding attributes according to the grouping of procurement objects to form a summary record of procurement objects and attribute conflict items includes: Extract procurement object records within the same group based on the procurement object grouping, and form a grouping field alignment result; Attribute binding is applied to the alignment results of grouped fields to create attribute conflict items; Based on the field binding values, write the object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit, and document source into the procurement object summary record; Output a summary record of procurement objects, and output any attribute conflicts corresponding to the summary record of procurement objects.

[0011] As a further improvement of the present invention, the step of determining the consistency result of technical requirements based on the summary record of procurement objects and attribute conflict items, and outputting the consistency result of procurement object grouping, summary record of procurement objects and technical requirements, includes: Consistency determination is performed on the summary records of procurement objects and the conflict items of attributes respectively. Conflicts with different values ​​are determined as inconsistency determination conditions, and conflicts without different values ​​are determined as consistency determination conditions. Based on the inconsistency and consistency criteria, generate technical requirement consistency results and associate the technical requirement consistency results with the procurement object summary record. Output the grouping of procurement objects, the summary records of procurement objects, and the consistency results of technical requirements.

[0012] Secondly, the present invention provides a cross-document same procurement object identification system for power plant bidding and procurement, comprising: The procurement object record module is used to obtain procurement object information and extract procurement object records from the procurement object information. The Procurement Object Status Module is used to construct a closed hypergraph of object attributes based on procurement object records, forming hyperedges representing relationships within the same procurement object; and to calculate the relationship status of the same procurement object based on these hyperedges. Procurement Object Cost Module: Used to perform counterfactual branch contribution calculation on the relationship status of the same procurement object, to obtain the contribution value of the same procurement object and the segmentation into different procurement objects, and to map the contribution value of the same procurement object and the segmentation into different procurement objects into support cost and rejection cost, and to generate the relationship cost of the same procurement object based on the support cost and rejection cost. The procurement object grouping module is used to perform high-order relation multi-cut partitioning on the closed hypergraph of object attributes based on the relation cost of the same procurement object, thereby forming procurement object groups. Object summary record module: used to bind attributes according to the grouping of procurement objects, and form a summary record of procurement objects and attribute conflict items; Procurement Object Identification Module: This module is used to determine the consistency of technical requirements based on the procurement object summary record and attribute conflict items, and outputs the procurement object grouping, procurement object summary record, and technical requirement consistency results.

[0013] Thirdly, the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for identifying the same procurement object across documents for power plant bidding and procurement.

[0014] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the method for identifying the same procurement object across documents in power plant bidding and procurement.

[0015] Fifthly, the present invention provides a computer program product, including computer instructions, which, when executed by a processor, implement the aforementioned method for identifying the same procurement object across documents in power plant bidding and procurement.

[0016] Compared with the prior art, the present invention has the following beneficial effects: This invention constructs a closed hypergraph of object attributes, forming record nodes from procurement object records and organizing hyperedges representing relationships within the same procurement object. This achieves a structured representation of cross-document procurement object relationships, effectively overcoming the shortcomings of traditional methods and clearly presenting complex procurement object relationships. Secondly, by calculating various relationship states of the same procurement object, such as attribute coverage and missing attribute states, this invention ensures that issues like complementary cross-document records, differences in record scope leading to missing items, tag number location correspondence, and connection chain consistency can all be uniformly identified within the same relationship object. This significantly reduces misjudgments caused by relying solely on name similarity or field overlap, providing an accurate and reliable foundation for subsequent operations and guaranteeing the quality of procurement data. Subsequently, counterfactual branch contribution calculations are performed on the relationship states of the same procurement object to obtain contribution values ​​for the same procurement object and for segments into different procurement objects. These values ​​are then mapped to support and rejection costs, generating a relationship cost for the same procurement object. Based on this cost, higher-order relationship multi-slicing is performed to form procurement object groups. This refined calculation and segmentation method provides a scientific basis for procurement decisions and helps in the rational planning of procurement processes and resource allocation. Attached Figure Description

[0017] The accompanying drawings described herein are for illustrative purposes only and are not intended to limit the scope of the invention in any way. In the drawings: Figure 1 This is a flowchart illustrating a method for identifying the same procurement object across documents in power plant bidding and procurement, according to the present invention. Figure 2 This is a schematic diagram illustrating the specific process of a method for identifying the same procurement object across documents in power plant bidding and procurement according to the present invention. Figure 3This is a schematic diagram of the specific process of step S1 in the method for identifying the same procurement object across documents for power plant bidding and procurement of the present invention; Figure 4 This is a schematic diagram of the specific process of step S2 in the method for identifying the same procurement object across documents for power plant bidding and procurement of the present invention; Figure 5 This is a schematic diagram of the specific process of step S3 in the cross-document same procurement object identification method for power plant bidding and procurement of the present invention; Figure 6 This is a schematic diagram of the specific process of step S4 in the cross-document same procurement object identification method for power plant bidding and procurement of the present invention; Figure 7 This is a schematic diagram of the specific process of step S5 in the cross-document same procurement object identification method for power plant bidding and procurement of the present invention; Figure 8 This is a schematic diagram of the specific process of step S6 in the cross-document same procurement object identification method for power plant bidding and procurement of the present invention. Figure 9 This is a schematic diagram of the algorithm processing flow and electronic device deployment structure in an embodiment of the present invention; Figure 10 This is a schematic diagram of the hypergraph composed of record nodes and attribute nodes in an embodiment of the present invention; Figure 11 These are the pairwise similarity heatmaps recorded in the embodiments of the present invention; Figure 12 This is a schematic diagram of the relationship lines between records in an embodiment of the present invention; Figure 13 This is a schematic diagram of the current grouping results obtained by multiple cuts in an embodiment of the present invention; Figure 14 This is a colored contour plot of the cost function in the parameter space in an embodiment of the present invention; Figure 15 This is a schematic diagram of the cross-document unified procurement object identification interface in an embodiment of the present invention; Figure 16 This is a schematic diagram comparing the grouping scheme, the counterfactual branching scheme, and the cost of default in an embodiment of the present invention; Figure 17 This is a cost surface diagram showing the positional relationship between the current solution and the counterfactual solution in an embodiment of the present invention; Figure 18 This is a map showing the uncertainty of grouping results in an embodiment of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are merely some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this invention.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] Definitions: Procurement object information: This refers to the field-based information used to characterize the procurement object, obtained from technical specifications, drawings, purchase requisitions, and quotation lists.

[0021] Object Category: The value of the field in the procurement object record used to characterize the type of procurement object.

[0022] System Affiliation: This refers to the value of the field in the procurement object record that represents the system to which the procurement object belongs.

[0023] Host connection relationship: This refers to the value of the field in the procurement object record that represents the connection position relationship between the procurement object and the host.

[0024] Entry / Exit Relationship: This refers to the value of the field in the procurement object record that represents the interface direction relationship of the procurement object.

[0025] Location Number / Drawing Number: This is the value of the field in the procurement object record used to identify the location of the procurement object in the drawing or list.

[0026] Parameter Unit: The value of the field used to characterize the unit of measurement of the parameter in the procurement object record.

[0027] Document Source: The field values ​​in the procurement object record that represent the source of procurement object information from technical specifications, drawings, purchase requisitions, or quotation lists.

[0028] Procurement object record: a structured record formed according to object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit, and document source.

[0029] Object attribute closed hypergraph: A hypergraph structure based on record nodes consisting of procurement object records and attribute nodes consisting of system affiliation, host connection relationship, entry and exit relationship, tag number, parameter unit and document source, and organized through hyperedges of the same procurement object relationship.

[0030] Record node: A node in the closed hypergraph of object attributes that consists of procurement object records and carries the object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit and document source.

[0031] Attribute node: A node in the closed hypergraph of object attributes, consisting of field values ​​for system affiliation, host connection relationship, entry and exit relationship, tag number, parameter unit, and document source, and connected to the record node.

[0032] The conditions for matching object category, tag number, host connection, entry and exit relationship, and document source complementarity are a set of continuous filtering conditions used to combine record nodes and form a superedge of the same procurement object relationship.

[0033] Hyperedge of the same procurement object relationship: It is a hyperedge formed by the combination of record nodes corresponding to multiple procurement object records in the closed hypergraph of object attributes, which meet the conditions of object category correspondence, tag number correspondence, host connection relationship connection, entry and exit relationship connection and document source complementarity.

[0034] Attribute coverage status: This refers to the status of whether object categories, system affiliations, host connection relationships, entry and exit relationships, tag numbers, drawing numbers, and parameter units are covered in the same procurement object relationship superedge; Missing Attribute Status: This status refers to the state formed by combining the missing attributes in the super-edge of the same procurement object relationship with the corresponding record range of the document source.

[0035] The status of the tag number and drawing number correspondence is the status formed for the correspondence between tag numbers and drawing numbers in the super-edge of the same procurement object.

[0036] Connectivity status: This refers to the consistency of the connection chain formed by the host connection relationship and the entry / exit relationship in the hyperedge of the same procurement object relationship.

[0037] Conflict state: This refers to a situation where there are different values ​​for system affiliation, entry and exit relationships, and parameter units in the superedge of the same procurement object relationship, and the document sources are not complementary.

[0038] The relationship status of the same procurement object is a set of states that combines attribute coverage status, attribute missing status, tag number / map number correspondence status, connection status, and conflict status to represent the relationship status of the super-edge relationship of the same procurement object.

[0039] Counterfactual branch contribution calculation: This involves setting up system affiliation branches, host connection relationship branches, entry and exit relationship branches, tag number and map number branches, parameter unit branches, object category branches, and document source branches based on the relationship status of the same procurement object. The centralized evaluation unit then replaces the output of each branch to calculate the contribution value.

[0040] System Affiliation Branch, Host Connection Relationship Branch, Ingress / Exit Relationship Branch, Tag Number / Map Number Branch, Parameter Unit Branch, Object Category Branch, and Document Source Branch: These are the branch calculation paths for receiving corresponding field information from the relationship status of the same procurement object and outputting the branch output.

[0041] The centralized evaluation unit replaces the output of a single branch and combines it with the output of an unreplaced branch: This process replaces the selected single branch output with a preset replacement value and inputs it into the centralized evaluation unit for evaluation calculation while keeping the output of the unreplaced branch unchanged.

[0042] The contribution value of each branch to merging into the same procurement object and splitting into different procurement objects: This is a value obtained based on the difference between the benchmark evaluation value and the replacement evaluation value, used to characterize the marginal effect of a single branch on the evaluation value of merging into the same procurement object and the evaluation value of splitting into different procurement objects.

[0043] Mapping the contribution values ​​corresponding to attribute coverage, tag number correspondence, and connection connectivity to support costs: This involves converting the contribution values ​​corresponding to attribute coverage status, tag number correspondence status, and connection connectivity status into the cost value used to support merging them into the same procurement object.

[0044] Mapping the contribution values ​​corresponding to system conflicts, entry / exit conflicts, and parameter conflicts to exclusion costs: This involves converting the contribution values ​​corresponding to system affiliation, entry / exit relationships, and parameter unit conflicts in a conflict state into cost values ​​used to support the segmentation into different procurement objects.

[0045] Cost of the same procurement object relationship: It is the cost formed by the combination of support cost and exclusion cost and used for the division of higher-order relationships.

[0046] High-order relation multi-cut partitioning: Under the constraints of single ownership of procurement object records and the same procurement object relation splitting constraint, the partitioning process is performed on the closed hypergraph of object attributes based on the relation cost of the same procurement object to form procurement object grouping.

[0047] Single attribution constraint for procurement object records: a constraint that ensures each procurement object record corresponds to only one procurement object group.

[0048] Constraints on the segmentation of relationships within the same procurement object: These constraints ensure that the super-edges of relationships within the same procurement object satisfy the constraint of segmentation consistency when sharing procurement object records.

[0049] Procurement Object Grouping: This is the grouping result output by the multi-slice division of higher-order relations, which is formed by aggregating procurement object records according to grouping labels.

[0050] Attribute binding: This is a process that aligns and determines the values ​​of objects, systems, hosts, entry and exit points, tag numbers, drawing numbers, parameter units, and document sources within a group of procurement objects to form a summary record of procurement objects.

[0051] Summary record of procurement objects: A summary record formed after binding attributes to procurement object records within the same procurement object group.

[0052] Attribute conflict items: These are conflict information that is generated and recorded when there are inconsistencies in system affiliation, entry / exit relationship, tag number, and parameter unit values.

[0053] Technical Requirements Consistency Result: This is the result of a consistency determination based on the procurement object summary records and attribute conflict items, which is performed on the system affiliation, entry and exit relationships, tag numbers, drawing numbers, and parameter units.

[0054] To address the problems in existing technologies where procurement records primarily rely on pairwise matching based on similar names, making it difficult to represent closed relationships, easily leading to confusion and missing items, and posing challenges to linkage identification and scoring, this invention provides a method for identifying the same procurement object across documents in power plant bidding and procurement. Figure 1 As shown, it includes: S100: Obtain procurement object information and extract procurement object records from the procurement object information; S200: Construct a closed hypergraph of object attributes based on the procurement object records to form hyperedges of relationships within the same procurement object; calculate the relationship status of the same procurement object based on the hyperedges of relationships within the same procurement object. S300: Perform counterfactual branch contribution calculation on the relationship status of the same procurement object to obtain the contribution value of the same procurement object and the segmentation into different procurement objects. Map the contribution value of the same procurement object and the segmentation into different procurement objects to support cost and rejection cost. Generate the relationship cost of the same procurement object based on the support cost and rejection cost. S400: Based on the relationship cost of the same procurement object, perform high-order relation multi-cut partitioning on the closed hypergraph of object attributes to form procurement object groups; S500: Bind attributes to groups of procurement objects to generate a summary record of procurement objects and attribute conflict items; S600: Determine the consistency result of technical requirements based on the summary record of procurement objects and attribute conflict items, and output the consistency result of procurement object grouping, summary record of procurement objects and technical requirements.

[0055] This method achieves a structured representation of cross-document procurement object relationships through a closed hypergraph of object attributes. It can uniformly identify multiple states to reduce the risk of mismerging and misclassification, and provide a stable intermediate representation for subsequent global grouping.

[0056] The present invention will be further explained and described below with reference to the accompanying drawings.

[0057] A method for identifying the same procurement object across documents in power plant bidding and procurement, such as Figure 2 As shown, it includes the following steps: S1: Obtain procurement object information from technical specifications, drawings, purchase requisitions and quotation lists, extract object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit and document source, and form procurement object record; like Figure 3 As shown, specifically: Obtain procurement target information from technical specifications, drawings, purchase requisitions and quotation lists, and segment continuous text into sentence groups, tables into rows and columns, and drawing annotations into annotation boxes according to document source to obtain segmented procurement target information; The segmented procurement object information is positioned and aligned by field location, determining the object name field, system field, connection field, tag number field, and parameter field, while maintaining the corresponding position of the connection field and the tag number field, to obtain the procurement object information after field alignment; Extract the object category, system affiliation, host connection relationship, inlet and outlet relationship, tag number and parameter unit from the field-aligned procurement object information, and keep the parameter unit corresponding to the parameter field to obtain the procurement object information after extracting attributes; The extracted procurement object information and document source are written in the same order according to the unified fields of object category, system affiliation, host connection relationship, entry and exit relationship, tag number and drawing number, parameter unit and document source. Procurement object information with the same tag number and drawing number and the same host connection relationship are merged to form a procurement object record.

[0058] The procurement object records generated by S1 have uniformly extracted object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit and document source, but the procurement object records still maintain the form of being recorded in a scattered manner according to documents.

[0059] In the bidding and procurement scenarios of thermal power plants, the same procurement object often appears separately in technical specifications, drawings, purchase requisitions, and quotation lists. Some procurement object records mainly provide system affiliation and host connection relationships, while others mainly provide tag numbers, drawing numbers, and parameter units. Document sources can also lead to missing attributes. If procurement object records are directly partitioned into higher-order relationships, records can usually only establish ordinary relationships based on similar names, overlapping fields, or partial similarities. Connection conditions corresponding to host connection relationships and entry / exit relationships, location conditions corresponding to tag numbers and drawing numbers, and missing conditions corresponding to document sources are difficult to include in the same relational object. When merging them into the same procurement object, it is easy to separate complementary records or link procurement object records with similar names but different systems or connection locations together.

[0060] S2: Construct a closed hypergraph of object attributes based on the procurement object records. Use the procurement object records as record nodes and system affiliation, host connection relationship, entry and exit relationship, tag number / map number, parameter unit, and document source as attribute nodes. Combine record nodes according to the conditions of object category correspondence, tag number / map number correspondence, host connection relationship connection, entry and exit relationship connection, and document source complementarity to form a hyperedge of the same procurement object relationship. Calculate the attribute coverage status, attribute missing status, tag number / map number correspondence status, connection connectivity status, and conflict status of the hyperedge of the same procurement object relationship to obtain the relationship status of the same procurement object. like Figure 4 As shown, specifically: Based on the object category, system affiliation, host connection relationship, entry and exit relationship, tag number and drawing number, parameter unit and document source in the procurement object record, a record node is constructed, and the system affiliation, host connection relationship, entry and exit relationship, tag number and drawing number, parameter unit and document source are written into the corresponding attribute node respectively, so that each field in each procurement object record corresponds to the corresponding attribute node, forming a record node and attribute node; The system affiliation, host connection relationship, entry and exit relationship, tag number, parameter unit, and document source in the record node are matched field by field with the attribute node. A connection is established when the field values ​​are consistent. If the field is missing but the record range corresponding to the document source does not include the missing field, a complementary document source connection is established to form a connection between the record node and the attribute node.

[0061] Based on the connection between record nodes and attribute nodes, the corresponding object category, host connection relationship, and entry and exit relationship are used as continuous filtering conditions. Multiple procurement object records are combined to form the same procurement object relationship hyperedge when the tag number and map number correspondence is true or the tag number and map number are missing but the document source is complementary. For the same procurement object relationship hyperedge, the object category, system affiliation, host connection relationship, entry and exit relationship, tag number, map number, and parameter unit are covered and statistically analyzed. The number of times each attribute is covered, the source of coverage, and the situation where multiple document sources jointly supplement it are calculated to form the attribute coverage status. Based on the attribute coverage status, identify the uncovered attributes in the hyperedge of the same procurement object relationship, and compare the uncovered attributes with the recording scope corresponding to the document source. If the uncovered attribute does not belong to the recording scope of the document source, it is determined to be an attribute missing due to document source complementarity. If the uncovered attribute belongs to the recording scope of the document source, it is determined to be an attribute missing, thus obtaining the attribute missing status. The tag number and map number in the hyperedge of the same procurement object relationship are compared accordingly, and the connection direction of the host connection relationship and the interface direction of the inlet and outlet relationship are compared for consistency. When the tag number and map number correspond and the connection direction and the interface direction are consistent, a tag number and map number correspondence state and a connection connectivity state are formed. Conflict comparisons are performed on the system affiliation, entry / exit relationship, and parameter units in the hyperedge of the same procurement object relationship. A conflict state is formed when there are different values ​​and the document sources are not complementary. The attribute overlay state, attribute missing state, tag number / map number correspondence state, connection connectivity state, and conflict state are combined in a fixed order of system affiliation, host connection relationship, entry / exit relationship, tag number / map number, parameter unit, object category, and document source to obtain the relationship state of the same procurement object.

[0062] When forming a hyperedge for the same procurement object relationship, the document sources corresponding to the combined procurement object records are limited to at least two different document sources. The number of procurement object records included in the hyperedge for the same procurement object relationship is compared with a threshold parameter. If the number does not exceed the threshold parameter, a hyperedge for the same procurement object relationship is formed. If there are missing tag numbers or drawing numbers in the combined procurement object records, the missing tag numbers or drawing numbers are determined to be outside the scope of the document source based on the complementary connection of document sources. A hyperedge for the same procurement object relationship is formed if there are procurement object records with tag numbers or drawing numbers in the same combination.

[0063] S2 first restructures the procurement object records to construct a closed hypergraph of object attributes. In this hypergraph, procurement object records are set as record nodes, and system affiliation, host connection relationships, entry / exit relationships, tag numbers / map numbers, parameter units, and document sources are set as attribute nodes. Record nodes are then combined according to the conditions of object category correspondence, tag number / map number correspondence, host connection relationship connection, entry / exit relationship connection, and document source complementarity to form a hyperedge representing the same procurement object relationship. This hyperedge no longer expresses whether two procurement object records are similar, but rather whether multiple procurement object records can together form a complete procurement object. The attribute coverage state, attribute missing state, tag number / map number correspondence state, connection connectivity state, and conflict state are further calculated around the hyperedge to obtain the relationship state of the same procurement object. Attribute coverage status indicates the degree to which multiple procurement object records supplement object category, system affiliation, host connection relationship, entry / exit relationship, tag number / map number, and parameter unit. Attribute missing status distinguishes whether missing items originate from document boundaries or from insufficient object information, based on document source. Tag number / map number correspondence status fixes the location correspondence between procurement object records. Connection connectivity status indicates whether the connection chains formed by system affiliation, host connection relationship, and entry / exit relationship are consistent. Conflict status retains system conflicts, entry / exit conflicts, and parameter conflicts. The relationship status of the same procurement object becomes the direct input object of S3. Only based on the relationship status of the same procurement object can the centralized evaluation unit calculate the contribution value of each branch to merging into the same procurement object and splitting into different procurement objects under unified constraints.

[0064] S3: Perform counterfactual branch contribution calculation on the relationship status of the same procurement object, and set up system affiliation branch, host connection relationship branch, entry and exit relationship branch, tag number and map number branch, parameter unit branch, object category branch and document source branch. The centralized evaluation unit replaces the output of a single branch and combines it with the output of the unreplaced branch to calculate the contribution value of each branch to merging into the same procurement object and splitting into different procurement objects. The contribution values ​​of attribute coverage, tag number and map number correspondence and connection connectivity are mapped to support costs. The contribution values ​​of system conflict, entry and exit conflict and parameter conflict are mapped to exclusion costs. The relationship cost of the same procurement object is generated based on the support cost and exclusion cost. like Figure 5 As shown, specifically: Extract system affiliation information, host connection relationship information, entry and exit relationship information, tag number and drawing number information, parameter unit information, object category information, and document source information based on the relationship status of the same procurement object, and form system affiliation branch input, host connection relationship branch input, entry and exit relationship branch input, tag number and drawing number branch input, parameter unit branch input, object category branch input, and document source branch input respectively; Branch calculations are performed on the system affiliation branch input, host connection relationship branch input, entry / exit relationship branch input, tag number / map number branch input, parameter unit branch input, object category branch input, and document source branch input respectively to obtain the system affiliation branch output, host connection relationship branch output, entry / exit relationship branch output, tag number / map number branch output, parameter unit branch output, object category branch output, and document source branch output; The system affiliation branch output, host connection relationship branch output, entry and exit relationship branch output, tag number and map number branch output, parameter unit branch output, object category branch output, and document source branch output are concatenated in a fixed order to form a joint branch output; Input the relationship status of the joint branch output with the same procurement object into the centralized evaluation unit, calculate the evaluation value of merging into the same procurement object and the evaluation value of dividing into different procurement objects, and form the benchmark evaluation value by combining the evaluation value of merging into the same procurement object and the evaluation value of dividing into different procurement objects. For the joint branch output, select individual branch outputs in sequence and replace them with preset replacement values ​​to form replacement joint branch outputs. Input the relationship status of the replacement joint branch outputs with the same procurement object into the centralized evaluation unit to calculate the replacement evaluation value. Based on the difference between the benchmark evaluation value and the replacement evaluation value, calculate the contribution value of the corresponding branch pair when merged into the same procurement object and the contribution value of the corresponding branch pair when divided into different procurement objects. Based on the attribute coverage status, tag number / map number correspondence status, and connection connectivity status in the same procurement object relationship status, determine the contribution values ​​corresponding to attribute coverage, tag number / map number correspondence, and connection connectivity and map them as support costs. Based on the conflict status in the same procurement object relationship status, determine the contribution values ​​corresponding to system conflicts, entry / exit conflicts, and parameter conflicts and map them as exclusion costs. Then, generate the same procurement object relationship cost based on the support costs and exclusion costs.

[0065] The default replacement value is an all-zero replacement value of the same dimension as the output of a single branch; the contribution values ​​of the corresponding branches of the system affiliation branch, entry and exit relationship branch and parameter unit branch that are divided into different procurement objects are mapped as exclusion costs; the contribution values ​​of the corresponding branches of the tag number and map number branch, host connection relationship branch and object category branch that are merged into the same procurement object are mapped as support costs; and the contribution values ​​of the corresponding branches of the document source branch that are merged into the same procurement object are included in the support cost mapping.

[0066] The relationship status of the same procurement object obtained by S2 has organized multiple procurement object records into a unified relationship object around the relationship hyperedge of the same procurement object. However, the relationship status of the same procurement object cannot be directly entered into the high-order relationship multi-cut partitioning.

[0067] The input required for high-order relation multi-slice partitioning is the relation cost of the same procurement object. However, in the bidding and procurement scenario of thermal power plants, the relation cost of the same procurement object cannot be directly given by a single total score. The reason is that attribute coverage status, tag number and map number correspondence status, and connection connectivity status reflect complementary filling and connection consistency, while system conflict, entry and exit conflict, and parameter conflict reflect object boundary exclusion. Document source also affects the interpretation direction of attribute missing status. Existing processing usually compresses all factors into a uniform weight or uniform score. As a result, even when tag numbers and map numbers are already corresponding and host connection relationships are connected, procurement object records with different system affiliations may still be brought closer together; when system affiliations are consistent but parameter units have strong conflicts, local proximity relationships may continue to be retained. Therefore, the relation cost cannot accurately serve the single affiliation constraint of procurement object records and the relation segmentation constraint of the same procurement object.

[0068] S3 incorporates the counterfactual branch contribution calculation into the cost formation process for the same procurement object relationship, and directly maps branch settings to the constraint sources in the same procurement object relationship state. System affiliation branches, host connection relationship branches, entry / exit relationship branches, tag number / map number branches, parameter unit branches, object category branches, and document source branches each read their respective branch association information from the same procurement object relationship state. The centralized evaluation unit replaces the output of a single branch and combines it with the output of unreplaced branches to calculate the contribution value of a single branch change to merging into the same procurement object and splitting into different procurement objects. The contribution value is not a general branch weight, but rather a marginal action obtained under the condition of maintaining consistency in the same procurement object relationship state. Therefore, it can distinguish which type of information supports the same procurement object relationship and which type of information drives the splitting. Subsequently, the contribution values ​​corresponding to attribute coverage, tag number / map number correspondence, and connection connectivity are mapped to support costs, and the contribution values ​​corresponding to system conflicts, entry / exit conflicts, and parameter conflicts are mapped to repulsion costs. Based on the support and repulsion costs, the cost of the same procurement object relationship is generated. Document source branches and attribute missing status jointly participate in contribution value calculation. Missing items formed by record boundaries between technical specifications, drawings, purchase requisitions, and quotation lists will not be directly converted into exclusion costs. The relationship cost of the same procurement object is used as the input object of S4 to enter the higher-order relation multi-cut partitioning. The single ownership constraint of procurement object record and the relationship splitting constraint of the same procurement object can thus be executed around the relationship cost decomposed by contribution value.

[0069] S4: Perform high-order relation multi-cut partitioning on the closed hypergraph of object attributes based on the relation cost of the same procurement object, and form procurement object groups under the single ownership constraint of procurement object records and the relation splitting constraint of the same procurement object; like Figure 6 As shown, specifically: Based on the closed hypergraph of object attributes, read the procurement object record, the hyperedge of the same procurement object relationship, and the cost of the same procurement object relationship, and summarize the cost of the same procurement object relationship according to the hyperedge of the same procurement object relationship to form a list of costs of the same procurement object relationship; Based on the cost list of the same procurement object relationship, assign a segmentation decision variable to each superedge of the same procurement object relationship, and establish a correspondence between the segmentation decision variable and the cost of the same procurement object relationship to form a segmentation solution structure; Based on the single attribution constraint of the procurement object record, a set of record attribution constraints is constructed for the partitioning solution structure, so that each procurement object record corresponds to only one grouping label, forming a partitioning solution structure with single attribution constraint. Based on the segmentation constraints of the same procurement object relationship, a segmentation consistency constraint set is constructed for the segmentation solution structure with single affiliation constraint, so that the segmentation decision variables corresponding to the hyperedges of the same procurement object relationship satisfy the segmentation consistency under the condition of sharing procurement object records, thus forming a consistent constraint segmentation solution structure; Under the constraints of the record attribution constraint set and the partition consistency constraint set, the partitioning decision variable is solved by high-order relation multi-cutting, and the partitioning decision result is obtained by taking the total cost corresponding to the cost list of the same procurement object relationship as the preset minimum criterion to meet the solution objective. Based on the segmentation determination results, the connectivity relationship of the closed hypergraph of object attributes is reconstructed. The procurement object records corresponding to the hyperedges of the same procurement object relationship that are not segmented are aggregated to form group labels, and procurement object groups are formed based on the group labels.

[0070] Before performing high-order relation multi-cutting on the segmentation decision variables, the difference result is calculated based on the difference between the support cost and the rejection cost. The difference result is compared with the threshold parameter. If the rejection cost is greater than the support cost and the difference result meets the threshold parameter, the corresponding segmentation decision variable is determined to be segmented. If the support cost is greater than the rejection cost and the difference result meets the threshold parameter, the corresponding segmentation decision variable is determined not to be segmented. High-order relation multi-cutting is then performed on the segmentation decision variables that are not determined to be segmented or not segmented to form procurement object groups.

[0071] S5: Bind attributes according to procurement objects to form a summary record of procurement objects, and generate attribute conflict items when there are inconsistencies in system affiliation, entry and exit relationship, tag number, drawing number, and parameter unit values. like Figure 7 As shown, specifically: Extract procurement object records within the same group based on the procurement object group, and align the object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit, and document source by field to form a grouped field alignment result; For each field in the grouping field alignment result, filter non-empty values. If all non-empty values ​​in the same field are consistent, determine the non-empty value as the field binding value. If there are multiple different non-empty values ​​in the same field, form an attribute conflict item and write the different non-empty values ​​into the attribute conflict item. Based on the field binding values, write the object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit, and document source into the procurement object summary record; Output a summary record of procurement objects, and output any attribute conflicts corresponding to the summary record of procurement objects.

[0072] S6: Determine the consistency result of technical requirements based on the summary record of procurement objects and attribute conflict items, and output the consistency result of procurement object grouping, summary record of procurement objects and technical requirements.

[0073] like Figure 8 As shown, specifically: Based on the procurement object summary record, read the system affiliation, entry / exit relationship, tag number / map number, and parameter unit, and determine the system affiliation conflict, entry / exit relationship conflict, tag number / map number conflict, and parameter unit conflict based on the attribute conflict items; Consistency determination is performed on system ownership conflicts, entry and exit relationship conflicts, tag number and map number conflicts, and parameter unit conflicts respectively. Conflicts with different values ​​are determined as inconsistency determination conditions, and conflicts without different values ​​are determined as consistency determination conditions. Based on the inconsistency and consistency criteria, generate technical requirement consistency results and associate the technical requirement consistency results with the procurement object summary record. Output the grouping of procurement objects, the summary records of procurement objects, and the consistency results of technical requirements.

[0074] In summary, this invention addresses the problem of difficulty in identifying the same procurement object due to scattered records, complementary attributes, missing fields, and conflicts in technical specifications, drawings, purchase requisitions, and quotation lists. This invention extracts procurement object records, constructs a closed hypergraph of object attributes containing record nodes, attribute nodes, and hyperedges representing relationships with the same procurement object, calculates attribute coverage, missing attributes, tag number / map number correspondence, connectivity, and conflict states, generates the relationship cost of the same procurement object based on counterfactual branch contributions, and groups procurement objects through high-order relation multi-cut partitioning, completing attribute binding, attribute conflict identification, and technical requirement consistency determination. This invention can be used for procurement object aggregation, technical requirement verification, and consistency review in power plant bidding and procurement scenarios.

[0075] In this method, the object attribute closure hypergraph transforms procurement object records into hyperedges of relationships within the same procurement object and states of relationships within the same procurement object. These states of relationships serve as continuous intermediate objects between the procurement object record and the cost of the relationship within the same procurement object. The counterfactual branch contribution calculation directly references the state of the relationship within the same procurement object, performing branched contribution calculations on supporting and excluding information, and writing the results into the cost of the relationship within the same procurement object. This cost of the relationship within the same procurement object is further referenced by higher-order relations to form procurement object groups. The object attribute closure hypergraph solves the problem that procurement object records cannot be organized around relationships within the same procurement object, while the counterfactual branch contribution calculation solves the problem that the cost of the relationship within the same procurement object cannot distinguish between supporting and excluding factors. The continuous reference chain thus extends to the procurement object summary record, attribute conflict items, and the results of technical requirement consistency.

[0076] The present invention will be further explained and illustrated below with reference to specific embodiments.

[0077] Example This embodiment receives procurement object records obtained from the parsing of technical specifications, drawings, purchase requisitions, and quotation lists. Each procurement object record is encoded as a 192-dimensional input vector, which consists of 16 dimensions for object category, 32 dimensions for system affiliation, 24 dimensions for host connection relationship, 16 dimensions for entry / exit relationship, 48 dimensions for tag number / map number, 40 dimensions for parameter unit, and 16 dimensions for document source. The attribute nodes in the object attribute closed hypergraph consist of a 64-dimensional input vector composed of the corresponding attribute value and document source, which is then mapped to 128 dimensions through an input projection layer. The procurement object records and attribute nodes together serve as the input objects for the object attribute closed hypergraph, unifying the scattered connection conditions, location conditions, and document source conditions recorded in the documents into a unified data format.

[0078] The main body of the high-order relation multi-cut partitioning is set on the object attribute closed hypergraph. The input projection layer performs fully connected mapping on the 192-dimensional procurement object records and 64-dimensional attribute nodes respectively, resulting in 128-dimensional record node features and 128-dimensional attribute node features. Each fully connected neuron contains a weight corresponding to the input dimension and one bias, and the nonlinear unit adopts ReLU. Three consecutive hypergraph convolutional layers are set on the object attribute closed hypergraph, each containing 128 adjacency aggregation neurons. The adjacency aggregation neurons aggregate the features of connected nodes according to the connection relationship of record nodes, attribute nodes, and hyperedges of the same procurement object relationship, and output the 128-dimensional update result. This allows the correspondence of object category, the correspondence of tag number and map number, the connection relationship of host, the connection relationship of entry and exit, and the complementarity of document source to jointly enter the relation construction stage of high-order relation multi-cut partitioning.

[0079] The scenario-based transformation of high-order relationships occurs during the relationship construction and relationship scoring input formation stages. The object attribute closure hypergraph does not directly create ordinary edges between two procurement object records. Instead, it combines record nodes that satisfy the conditions of object category correspondence, tag number / graph number correspondence, host connection relationship connection, entry / exit relationship connection, and document source complementarity into a hyperedge for the same procurement object relationship. Then, it aggregates the record node features and attribute node features of each hyperedge for the same procurement object relationship, forming a 256-dimensional aggregation vector. This vector is then used to generate the relationship state for the same procurement object through 160 fully connected neurons. The 160-dimensional relationship state for the same procurement object is expanded into 32 dimensions of attribute coverage, 16 dimensions of attribute missing, 32 dimensions of tag number / graph number correspondence, 32 dimensions of connection connectivity, and 48 dimensions of conflict. Internally, the 160-dimensional relationship state is further segmented according to 24 dimensions of system affiliation, 24 dimensions of host connection, 16 dimensions of entry / exit relationship, 40 dimensions of tag number / graph number, 24 dimensions of parameter units, 16 dimensions of object category, and 16 dimensions of document source. This modification distinguishes procurement object records with similar names but different system affiliations or connection locations from procurement object records that are complementary across documents and places them within the same procurement object relationship state.

[0080] The counterfactual branch contribution calculation is embedded within the cost calculation stage of the same procurement object relationship. The system affiliation branch receives a 24-dimensional system affiliation segment, the host connection relationship branch receives a 24-dimensional host connection relationship segment, the entry / exit relationship branch receives a 16-dimensional entry / exit relationship segment, the tag number / map number branch receives a 40-dimensional tag number / map number segment, the parameter unit branch receives a 24-dimensional parameter unit segment, the object category branch receives a 16-dimensional object category segment, and the document source branch receives a 16-dimensional document source segment. Each branch consists of two fully connected layers: the first layer has 32 neurons, and the second layer has 16 neurons. Each neuron in each layer uses fully connected weights, biases, and ReLU. The branch outputs an 8-dimensional branch vector. The centralized evaluation unit concatenates the seven 8-dimensional branch vectors with the 160-dimensional relationship state of the same procurement object to form a 216-dimensional input. This input then passes through a fully connected layer with 128 neurons and a fully connected layer with 32 neurons. Finally, two output neurons provide the combined evaluation value for the same procurement object and the evaluation values ​​for different procurement objects. The centralized evaluation unit performs a same-dimensional all-zero replacement on the output of a single branch, and calculates the contribution value by combining the output of the unreplaced branch. Then, the contribution values ​​corresponding to attribute coverage, tag number and map number correspondence, and connection connectivity are compressed into a 1-dimensional support cost. The contribution values ​​corresponding to system conflict, entry and exit conflict, and parameter conflict are compressed into a 1-dimensional exclusion cost. The support cost and exclusion cost are combined to form a 1-dimensional relationship cost for the same procurement object.

[0081] The higher-order relation multi-slice partitioning layer receives the cost of the same procurement object relationship and performs global partitioning. In a fixed embodiment, each time 128 procurement object records, 320 attribute nodes, and 256 hyperedges of the same procurement object relationship are processed; the higher-order relation multi-slice partitioning layer solves the 256 one-dimensional costs of the same procurement object relationship, and outputs 128 one-dimensional grouping labels under the single ownership constraint of the procurement object record and the partitioning constraint of the same procurement object relationship. The 128 one-dimensional grouping labels directly correspond to the procurement object grouping. The attribute binding unit aggregates the procurement object records in the same group according to the grouping labels, forming no more than 128 procurement object summary records. Each procurement object summary record maintains a 192-dimensional attribute structure; the conflict determination unit performs item-by-item comparison of system ownership, entry and exit relationships, tag number, map number, and parameter unit, and outputs four-dimensional attribute conflict items and one-dimensional technical requirement consistency results. The procurement object grouping, procurement object summary records, and technical requirement consistency results are directly obtained from the same model link.

[0082] In this embodiment, step S1 specifically includes: Record the technical specifications, drawings, purchase requisitions, and quotation lists in the following order: ,in Indicates the number of documents. This indicates a document index.

[0083] For each document Add document source tags , ,in Indicates technical specifications, Representing drawings, This indicates a purchase order. This indicates a price list. The document source is... At that time, continuous text is segmented into sentence groups consisting of periods, semicolons, colons, and line breaks; the document source is... or When splitting the table by row and column units, the column order within the same row should be maintained; document source: When doing so, divide the drawing annotations according to the annotation box, and ensure that the annotation characters correspond to the leader lines. The segmented units obtained from splitting a document are denoted as follows: ,in Indicates the first The number of segmentation units in a document. This indicates the segment cell index.

[0084] For each segment unit Perform field positioning and field alignment to form .in, The field represents the object name. Represents a system field. Indicates the connection field. This indicates the tag number / drawing number field. This refers to parameter fields. In technical specifications, field locations are defined by keywords within the same sentence group; in purchase requisitions and quotation lists, field locations are defined by the same table row and corresponding column header; in drawings, field locations are defined by the same annotation box and the area pointed to by the same leader line. When multiple sets of connection fields and tag number / drawing number fields exist within a segmented unit, they are first sorted from left to right and top to bottom according to the reading order, and then multiple fields are formed according to the same position. , so that each With the same order Maintain corresponding positions; when multiple connection fields and multiple tag number fields appear in the drawing annotation, sort them from near to far according to the distance from the center of the annotation character box to the leader endpoint, and then pair them to ensure that the correspondence between the connection fields and the tag number fields is fixed during the field alignment stage.

[0085] From each Extract the attributes of the procurement object and record them sequentially as follows: ,in This indicates the number of procurement object information items after attribute extraction. This represents an index of information about the procurement targets.

[0086] Each Written as .in, Indicates the object category, Indicates system affiliation. Indicates the host connection relationship. Indicates the relationship between entry and exit. Indicates the tag number or diagram number. Indicates the unit of the parameter. Indicates the document's source.

[0087] object category Depend on After standardization, the standardization rules are as follows: remove serial numbers, quantities, material codes, specifications, parentheses, and extra spaces, retaining only the name phrase representing the type of procurement object; system affiliation. Depend on After removing spaces and unifying the capitalization of English letters, write directly; host connection relationship Depend on The connection is formed by concatenating character fragments representing the hostname, connection location, and connection direction; the ingress / exgress relationship. Depend on The directional words in the text are determined; when they include "entrance," "import," or "inhalation," they are written as... When it contains "export", "discharge", or "send out", it is written as When both entry and exit terms are included, it is written as If no direction term is present, write an empty value; position number / map number Depend on Letters, numbers, hyphens, forward slashes, and periods are retained in their original order in the parameter field. When multiple parameters are expressed, they are written in the order of their appearance as follows: ,in Indicates the first The parameters in the procurement object information express the quantity. This indicates that the parameter expresses the index, and then from each Extracting unit strings And connected by semicolons in the same index order to form This ensures that the parameter units correspond to the parameter fields; (Document source) Direct write .

[0088] Will Write the data in the following order: object category, system affiliation, host connection relationship, entry / exit relationship, tag number / map number, parameter unit, and document source. Merge the data within the same document source. First, write the data in the following order: Grouping, and then within each document source group, grouping the tag number and figure number. Identical and host connection relationship Identical procurement object information is grouped into the same consolidation group. Within the consolidation group, object category, system affiliation, entry / exit relationship, and parameter unit are merged at the field level only if "field values ​​are the same" or "one field value is empty and the other field value is not empty." If the same field has two different non-empty values, merging is not performed, and each value is retained separately. This results in the procurement object record. ,in Indicates the quantity of the procurement object record. This represents the index of the procurement object records. Each procurement object record is written as... Field order and Maintain consistency.

[0089] Furthermore, to enable procurement object records to be directly used as inputs to record nodes in the closed hypergraph of object attributes, each procurement object record... Encoded as a 192-dimensional input vector .in, Segment the object category into 16 dimensions. The system is assigned to a 32-dimensional segment. The host connection relationship is segmented into 24 dimensions. The ingress-exgress relationship is segmented into 16 dimensions. The position number is divided into 48 dimensions. The parameter unit is divided into 40-dimensional segments. The document source is segmented into 16 dimensions. Each segment uses fixed-length character position encoding: first, full-width to half-width characters, English to uppercase characters, consecutive space compression, and irrelevant punctuation are performed on the corresponding field; then, a fixed length is extracted according to character order, and zero characters are added to the end if the length is insufficient.

[0090] The object category segment retains the first 8 characters, with each character written into a 2D code point to form a 16-dimensional segment; the system affiliation segment retains the first 16 characters to form a 32-dimensional segment; the host connection relationship segment retains the first 12 characters to form a 24-dimensional segment; the entry / exit relationship segment retains the first 8 characters to form a 16-dimensional segment; the tag number / map number segment retains the first 24 characters to form a 48-dimensional segment; the parameter unit segment retains the first 20 characters to form a 40-dimensional segment; and the document source segment retains the first 8 characters to form a 16-dimensional segment. The 2D code point for each character consists of the high 8 bits and low 8 bits of the code point corresponding to the standardized character. Null value fields are filled with zeros. Index The procurement object records correspond to the 192-dimensional input vector. The 192-dimensional input vector is arranged in the order of record index and used to construct the record nodes in the closed hypergraph of object attributes.

[0091] Figure 9 The upper part shows the functional modules such as data acquisition, preprocessing, object attribute extraction, hypergraph modeling, multi-cut, counterfactual evaluation, and result output in chronological order, as well as the data flow between them; the lower part illustrates the deployment of these modules in electronic devices, including processors, memory, communication interfaces, etc., used to execute the above programs.

[0092] In this embodiment, step S2 specifically includes: Read procurement object records , .variable Represents object category, variable Indicates system affiliation, variable Indicates host connection relationships, variables Indicates the relationship between inlet and outlet, variables Indicates tag number, variable Indicates the unit of the parameter, variable Indicates the document's source.

[0093] Record the corresponding 192-dimensional input vector for the procurement object. Write it directly as the record node input. The document source record range table is denoted as... .when hour, Includes object category, system affiliation, host connection relationship, entry / exit relationship, and parameter unit; when hour, Includes object category, host connection relationship, ingress / exgress relationship, and tag number / map number; when hour, Includes object category, system affiliation, tag number, and parameter unit; when hour, Includes object category, tag number, and parameter unit.

[0094] Each 192-dimensional input vector Write as the initial features of the record node .from , , , , and Extract non-empty field instances and assign a field type code to each field instance. Field type code Indicates system affiliation, field type code Indicates host connection relationship, field type code Indicates the ingress / exgress relationship, field type code Indicates the tag number and drawing number; field type code. Indicates the parameter unit, field type code Indicates the document source. Standardization is applied to each field value, including full-width to half-width conversion, capitalization of English letters, compression of consecutive spaces, and removal of irrelevant punctuation.

[0095] Field instances with identical "field type code + standardized field value + document source" are merged into a single attribute node. Each attribute node is encoded as a 64-dimensional input vector. , .

[0096] variable This represents a 48-dimensional attribute value segment. Each segment is a string concatenated from the field type code and the standardized field value, truncated into 24 characters and written with each character having 2 dimensions. (Variable) This represents a 16-dimensional document source segment, which is formed by truncating the document source string into 8-character segments and writing each character with 2-dimensional code points. Each non-empty field instance corresponds to only one attribute node, and each non-empty field in the record node maintains a field-level correspondence with its corresponding attribute node.

[0097] The system affiliation, host connection relationship, entry / exit relationship, tag number / map number, parameter unit, and document source in the record nodes are matched field by field with the attribute nodes. If the field types, standardized field values, and document sources are the same, a connection with consistent field values ​​is established, and a connection index is assigned to each connection. The field value is empty and the field type does not belong to... In cases where a field is missing, it is determined that the document source's corresponding record range does not include the missing field; in object categories Within the same record node range, retrieve attribute nodes with the same field type, different document sources, and non-empty field values, and establish complementary document source links. Each link must include the link field type, link mode, and the document sources involved. The link mode can only be either "join with consistent field values" or "join with complementary document sources."

[0098] Record nodes sharing at least one attribute node connection or sharing at least one complementary document source connection are grouped into the same candidate group. Each candidate group is filtered in the following order: object category correspondence, host connection relationship, entry / exit relationship, and tag number / map number. Object category correspondence is determined using... Complete consistency of standardization results is used as a criterion for judgment. When connecting host relationships, start from... Remove "inlet," "outlet," "import," "inhale," "exhale," and "export" along with their corresponding English direction terms, while retaining the host name, interface name, and connection order to obtain the host interface identifier. If all non-empty host interface identifiers within the candidate combination are identical, the host connection is considered established. The connection direction code is denoted as... The interface direction code is denoted as .

[0099] when When it contains directional words related to entry, ; when When it contains export-related directional terms, ; when When it contains both entry-type directional terms and exit-type directional terms ; when When there is no directional word, .

[0100] variable According to the same rules Extract from [the list]. All non-empty candidate combinations. The values ​​are the same, or all are not empty. Only by and Composed of, or all of which are non-empty Only by and The composition establishes the connection between the entrance and exit points.

[0101] The tag number and map number conditions use two determination paths: all not empty. When the standardization results are completely consistent, it is determined that the tag number and map number correspondence is valid; if null values ​​exist... At that time, only when the null value field type does not belong to Candidate combinations are retained if there are non-empty position numbers or figure numbers in the candidate combination, or if there are complementary links in the document source where position number or figure number fields exist in the link.

[0102] The number of document sources for the candidate combination must be at least 2, and the number of procurement object records included in the candidate combination is denoted as [missing information]. , satisfy When the same procurement object relationship is formed, the boundary is super-border. .variable This represents the threshold for the number of records exceeding the boundary of the same procurement object relationship, which is set to 6 in a fixed embodiment.

[0103] Record the initial characteristics of the node Attribute node input vector The connection relationships and hyperedges related to the same procurement object are fed into the object attribute closed hypergraph. The input projection layer of the object attribute closed hypergraph performs a fully connected mapping on the 192-dimensional record node input and outputs 128-dimensional record node features, and performs a fully connected mapping on the 64-dimensional attribute node input and outputs 128-dimensional attribute node features. The nonlinear unit of the input projection layer is ReLU. Three hypergraph convolutional layers are sequentially set on the object attribute closed hypergraph, each containing 128 adjacency aggregation neurons. Each hypergraph convolutional layer first aggregates the connected record node features and attribute node features for each hyperedge related to the same procurement object, and generates 128-dimensional hyperedge aggregation features by averaging dimensionally; then it aggregates the connected attribute node features and connected hyperedge aggregation features for each record node to generate new 128-dimensional record node features; then it aggregates the connected record node features and connected hyperedge aggregation features for each attribute node to generate new 128-dimensional attribute node features. After the three hypergraph convolutional layers are completed, the updated 128-dimensional record node features and updated 128-dimensional attribute node features are obtained. For each hyperedge of the same procurement object relationship, the updated record node features of the associated nodes are averaged one dimension at a time to form a 128-dimensional record aggregation component, and the updated attribute node features of the associated nodes are averaged one dimension at a time to form a 128-dimensional attribute aggregation component. Then, they are concatenated in a fixed order to form a 256-dimensional aggregation vector.

[0104] For each relationship exceeding the boundary of the same procurement object, coverage statistics, missing item identification, correspondence comparison, consistency comparison, and conflict comparison are performed. The attribute coverage status adopts a 32-dimensional writing method. The first 24 dimensions are arranged in the order of object category, system affiliation, host connection relationship, entry and exit relationship, tag number / map number, and parameter unit. Each type of attribute occupies 4 dimensions. The content of the 4 dimensions is, in order, the number of non-empty coverage times, the number of coverage sources, the multi-document joint supplementation mark, and the covered mark within the recorded scope. The last 8 dimensions are written in the following order: the number of covered attribute types, the number of uncovered attribute types, the number of multi-document joint supplementation attribute types, the number of covered attribute types within the recorded scope, the object category coverage mark, the system affiliation coverage mark, the host connection relationship coverage mark, and the tag number / map number coverage mark. The attribute missing status is written using a 16-dimensional method. The first 12 dimensions are arranged in the order of object category, system affiliation, host connection relationship, entry and exit relationship, tag number / map number, and parameter unit. Each attribute type occupies 2 dimensions, and the content of the 2 dimensions is, in order, the attribute missing marker formed by document source complementarity and the attribute missing marker. The last 4 dimensions are written in the order of the number of attribute missing types formed by document source complementarity, the number of attribute missing types, the number of records involved, and the number of document sources involved. The tag number / map number corresponding status is written using a 32-dimensional method. The first 8 dimensions are written in the order of the number of non-empty tag number / map number records, the number of non-empty tag number / map number document sources, the number of unique tag number / map number, the tag number / map number completely consistent marker, the tag number / map number missing marker, the tag number / map number missing and document source complementarity established marker, the number of records participating in tag number / map number complementarity, and the number of document sources participating in tag number / map number complementarity. The last 24 dimensions are arranged in the order of... , , and The document sources are arranged in order, with each type of document source occupying 6 dimensions. The 6 dimensions are, in order, whether the document source appears, whether the document source provides a tag number or map number, whether the document source is missing a tag number or map number, whether the document source participates in complementarity, whether the tag number or map number of the document source is consistent with the one in the combination, and whether the tag number or map number of the document source conflicts. The connection connectivity state is written using a 32-dimensional writing method, with the first 8 dimensions written sequentially. middle , , and The number of unique host interface identifiers, the host interface identifier complete consistency flag, the direction compatibility flag, and the direction consistency flag are written sequentially in the middle 8 dimensions. middle , , and The number of documents, the number of document sources, the unified marker for non-empty interface directions, the complementary supplement marker, and the direction conflict marker; the last 16 dimensions are... , , and The documents are arranged in order, with each document source occupying 4 dimensions. The 4 dimensions are, in order, whether the document source appears, whether the direction is readable, whether the connection direction with the host is consistent, and whether the connection direction with the interface is consistent.

[0105] Conflict status is written using a 48-dimensional method. System attribution conflicts, entry / exit relationship conflicts, and parameter unit conflicts each occupy 16 dimensions. Each 16-dimensional segment sequentially writes the number of different values, the number of conflict records, the number of conflict document sources, the document sources not constituting complementary markers, the maximum number of repetitions, the minimum number of repetitions, the number of primary value records, and the number of secondary value records. Whether to participate in the conflict Whether to participate in the conflict Whether to participate in the conflict Whether to participate in a conflict, the number of cross-document conflicts, the number of intra-document conflicts, whether the conflict can be resolved, and whether the conflict retains its marker.

[0106] The attribute overriding status, attribute missing status, tag number / map number correspondence status, connection connectivity status, and conflict status are rearranged in a fixed order according to system affiliation, host connection relationship, entry / exit relationship, tag number / map number, parameter unit, object category, and document source, forming a 160-dimensional relationship status of the same procurement object. .

[0107] variable This represents a 24-dimensional system attribution segment, which consists of coverage information, missing information, and conflicting system attribution information corresponding to system attribution; variables This represents a 24-dimensional host connectivity segment, which consists of coverage information, missing information, and connectivity information corresponding to the host connectivity relationships; variables This represents a 16-dimensional entry-exit relationship segment, which consists of coverage information, missing information, connectivity information, and entry-exit relationship conflict information corresponding to the entry-exit relationship; variables This represents a 40-dimensional position and map number segment, which consists of the coverage information, missing information, and position and map number correspondence information corresponding to the position and map number; variables This represents a 24-dimensional parametric unit segment, which consists of coverage information, missing information, and parametric unit conflict information corresponding to the parametric units; variables This represents a 16-dimensional object category segment, which consists of coverage and missing information corresponding to the object category; variables This represents a 16-dimensional document source segment, which consists of overlay source information and complementary document source information. Each hyperedge representing a relationship with the same procurement object corresponds to a 160-dimensional state vector representing the same procurement object relationship.

[0108] In this embodiment, step S3 specifically includes: For the Read the status of the same procurement object relationship in 160 dimensions. , ,variable This indicates the number of edges in the same procurement object relationship.

[0109] variable This represents a 24-dimensional system attribution segment, which consists of 4-dimensional system attribution coverage information, 2-dimensional system attribution missing information, and 18-dimensional system attribution conflict information. The 4-dimensional system attribution coverage information includes, in order, the number of system attribution coverage sources, multiple document-shared supplementary markers, and covered markers within the recorded scope. The 2-dimensional system attribution missing information includes, in order, attribute missing markers formed by complementary document sources and attribute missing markers. The 18-dimensional system attribution conflict information includes, in order, the number of different values, the number of conflicting records, the number of conflicting document sources, markers indicating that document sources do not constitute complementarity, the maximum number of repetitions, the minimum number of repetitions, the number of primary value records, and the number of secondary value records. Participate in tagging, Participate in tagging, Participate in tagging, Participation flags, number of cross-document conflicts, number of intra-document conflicts, conflict resolvable flags, conflict retention flags, valid flags for system attribution coverage, and valid flags for missing system attribution.

[0110] variable This represents a 24-dimensional host connectivity segment, which consists of 4-dimensional host connectivity coverage information, 2-dimensional missing host connectivity information, and 18-dimensional connectivity information. The 18-dimensional connectivity information comprises, in order, the number of unique host interface identifiers, a marker indicating completely identical host interface identifiers, and... middle quantity, middle quantity, middle quantity, middle quantity, middle quantity, middle quantity, middle quantity, middle Quantity, direction compatibility marker, direction consistency marker, Direction readable marker Direction readable marker Direction readable marker Directional readability markers, multi-document complementary supplementation markers, and valid linking chain markers.

[0111] variable This represents a 16-dimensional entry-exit relationship segment, which consists of 4-dimensional entry-exit relationship coverage information, 2-dimensional entry-exit relationship missing information, 6-dimensional interface direction information, and 4-dimensional entry-exit relationship conflict information. The 6-dimensional interface direction information is as follows: quantity, quantity, quantity, The number of non-empty interface directions are uniformly marked, and the interface direction is complementary and supplementary. The 4-dimensional entry and exit relationship conflict information is in the following order: the number of values ​​in different directions, the number of conflict records, the number of conflict document sources, and the conflict retention mark.

[0112] variable This represents a 40-dimensional position and map number segmentation. Each segment consists of 4-dimensional position and map number coverage information, 2-dimensional position and map number missing information, and 34-dimensional position and map number correspondence information. The 34-dimensional position and map number correspondence information includes the number of non-empty position and map number records, the number of non-empty position and map number document sources, the number of unique position and map numbers, markers indicating complete position and map number consistency, markers indicating missing position and map numbers, markers indicating missing position and map numbers with complementary document sources, the number of records participating in position and map number complementarity, the number of document sources participating in position and map number complementarity, and information categorized by position and map number. , , , The 24-dimensional document source tag number comparison information is unfolded sequentially. In the 24-dimensional document source tag number comparison information, each type of document source occupies 6 dimensions. The 6 dimensions are, in order, document source occurrence marker, tag number and map number providing marker, missing tag number and map number missing marker, participation complementary marker, tag number and map number consistency marker, and tag number and map number conflict marker.

[0113] variable This represents a 24-dimensional parameter unit segment, which consists of 4-dimensional parameter unit coverage information, 2-dimensional parameter unit missing information, and 18-dimensional parameter unit conflict information. The order of the 18-dimensional parameter unit conflict information is consistent with the system's attribution conflict information.

[0114] variable This represents a 16-dimensional object category segment, which consists of 12-dimensional object category coverage information and 4-dimensional object category missing information. The 12-dimensional object category coverage information includes, in order: object category coverage count, number of object category coverage sources, multi-document joint supplementation markers, covered markers within the recorded scope, number of covered attribute types, number of uncovered attribute types, number of multi-document joint supplementation attribute types, and number of covered attribute types within the recorded scope. Object category appears marked, The object category is marked. Object category appears marked and The object category is marked; the missing information of the 4-dimensional object category is as follows: attribute missing mark formed by document source complementarity, attribute missing mark, number of attribute missing types formed by document source complementarity, and number of attribute missing types.

[0115] variable This represents a 16-dimensional document source segment, which consists of 8-dimensional overlay source information and 8-dimensional complementary document source information; the 8-dimensional overlay source information is as follows: Marking appears Marking appears Marking appears The document source complementarity information includes: occurrence markers, number of covered sources, number of cross-source sources, number of same-source sources, and valid markers for covered sources; the 8-dimensional document source complementarity information is as follows: Complementary participation markers, Complementary participation markers, Complementary participation markers, Complementary participation markers, number of document source complementarities, valid markers for document source complementarities, number of missing items outside the scope of the record, and number of missing items within the scope of the record.

[0116] Will , , , , , and Send the data to the system affiliation branch, host connection relationship branch, entry / exit relationship branch, tag number / map number branch, parameter unit branch, object category branch, and document source branch, respectively.

[0117] The seven branches employ a consistent branch computation structure. Each branch first performs a fully connected computation of 32 neurons on the input segment, followed by a fully connected computation of 16 neurons. Both hidden layers use ReLU, and the 16-dimensional hidden result is then processed by 8 output neurons to form an 8-dimensional branch output.

[0118] The system's branch output is denoted as The host connection relationship branch output is denoted as The inlet and outlet relationship branch output is denoted as The output of the position number diagram branch is denoted as The parameter unit branch output is denoted as The output of the object category branch is denoted as The output of the document source branch is denoted as Each 8-dimensional branch output is generated dimension-wise by 8 output neurons, and each dimension is a scalar that can be directly used in subsequent evaluation calculations.

[0119] Will , , , , , and The output is concatenated into a 56-dimensional joint branch output in a fixed order based on system affiliation, host connection relationship, entry / exit relationship, tag number / map number, parameter unit, object category, and document source. .

[0120] Output the 56-dimensional joint branch Relationship status with the same procurement object in dimension 160 The input is assembled in a fixed order to form a 216-dimensional evaluation input. ,variable This represents the input vector of the centralized evaluation unit. The centralized evaluation unit employs a fully connected structure consisting of two hidden layers and one output layer: the first hidden layer has 128 neurons, the second hidden layer has 32 neurons, both hidden layers use ReLU, and the output layer has two output neurons. The first output neuron outputs the combined evaluation value for the same procurement object. The second output neuron provides evaluation values ​​for different procurement objects. . and All are 1-dimensional scalars. Constituting the first The benchmark evaluation value for the relationship between the same procurement object and the boundary.

[0121] The preset replacement value is denoted as , , This represents an 8-dimensional all-zero replacement value with the same dimension as the output of a single branch.

[0122] For the For each item in the same procurement object relationship, a single-branch replacement evaluation is performed seven times sequentially. When the system's assigned branch is selected, it is used... replace Host connection relationship branch output Inlet / outlet relationship, branch output , tag number, diagram number, branch output Parameter unit branch output , Object category branch output and document source branch output Keep the baseline value unchanged; when the host connection relationship branch, entry and exit relationship branch, tag number and map number branch, parameter unit branch, object category branch and document source branch are selected, the same processing rules shall be applied.

[0123] After each replacement, the output is reassembled to obtain a 56-dimensional replacement joint branch, which is then compared with the same 160-dimensional relationship status of the same procurement object. The input is concatenated into a 216-dimensional replacement evaluation input. The centralized evaluation unit performs a forward calculation on the 216-dimensional replacement evaluation input to obtain the replaced and merged evaluation values ​​for the same procurement object. The replaced segmentation is divided into evaluation values ​​for different procurement objects. superscript This indicates the branch that was replaced.

[0124] variable This is recorded as the contribution value of the corresponding branch to the merger into the same procurement object, specifically taken from the benchmark evaluation value. Replacement evaluation value The difference; variable This is denoted as the contribution value of the corresponding branch to the segmentation into different procurement objects, and the specific value is taken from the benchmark evaluation value. Replacement evaluation value The difference.

[0125] This forms , , , , , , and , , , , , , .

[0126] All contribution values ​​originate from the scalar output difference of the same centralized evaluation unit, and all contribution values ​​maintain the same dimension, being dimensionless evaluation differences.

[0127] The underlying numerical markers used for cost mapping are read from fixed positions within the 160-dimensional relationship state of the same procurement object. The object category overlay marker in the object category segment is denoted as... The effective marker for joint supplementation by multiple documents in the document source segment is denoted as: The tag number and map number in the tag number and map number segment are completely consistent and marked as ; The missing reference numbers and map numbers in the reference number / map number segment, and the fact that the document sources are complementary, are marked as follows: The direction consistency marker in the host connection relationship segment is denoted as... The direction compatibility flag in the host connection relationship segment is denoted as... The system affiliation conflict retention flag in the system affiliation segment is denoted as... The entry / exit relationship conflict retention flag in the entry / exit relationship segment is denoted as... The parameter unit conflict retention flag in the parameter unit segment is denoted as... All nine markers are 0 or 1 values ​​that can be directly used for calculations.

[0128] The contribution values ​​of the corresponding branches of the system affiliation branch, entry / exit relationship branch, and parameter unit branch, which are divided into different procurement objects, are mapped to exclusion costs. The contribution values ​​of the corresponding branches of the tag number / map number branch, host connection relationship branch, and object category branch, which are merged into the same procurement object, are mapped to support costs. The contribution values ​​of the corresponding branches of the document source branch, which are merged into the same procurement object, are also included in the support cost mapping. Exclusion costs are expressed as fractions on the left side of the formula, support costs are expressed as fractions on the right side of the formula, and the cost of the same procurement object relationship is calculated by subtracting the support cost from the exclusion cost.

[0129] In the formula, Indicates the first The cost of the same procurement object relationship corresponding to the superedge of the same procurement object relationship; Indicates a super-edge index representing a relationship within the same procurement object; Indicates a system ownership conflict retention flag; A marker is reserved to indicate a conflict between the entry and exit points; Indicates that a conflict in the units of the parameters is preserved by a flag; This indicates the contribution value of the system's affiliated branch to the segmentation into different procurement objects; This indicates the contribution value of the entry and exit relationship branches to the segmentation into different procurement objects; This indicates the contribution value of the parameter unit branch to the segmentation into different procurement objects; Indicates the object category overriding tag; This indicates that multiple documents can jointly supplement valid tags; This indicates that the tag number and drawing number are completely identical. This indicates that the tag number or figure number is missing and the document source is complementary. Indicates a consistent direction; Indicates orientation compatibility; This indicates the contribution value of the object category branch to the merging into the same procurement object; This indicates the contribution value of the document source branch to the merged procurement object; This indicates the contribution value of the tag number / drawing number branch to the merger into the same procurement object; This indicates the contribution value of the host connection relationship branch to the merged purchase object; This represents a constant used to prevent the denominator from being zero. ; Subscript Indicates the contribution value merged into the same procurement object; subscript This indicates that the contribution value is divided into different procurement objects; superscript Indicates the branch to which the system belongs; superscript Indicates a branch in the host connection relationship; superscript Indicates the branching relationship between entry and exit; superscript Indicates the tag number, diagram number, and branch; superscript Indicates a branch with a parameter unit; superscript Indicates a branch of an object category; superscript Indicates the branch from which the document originates.

[0130] In this embodiment, step S4 specifically includes: Closing hypergraph from object properties Read the procurement object record Excessive boundaries in the relationship between the same procurement object Support Costs Exclusion cost Cost of relationships with the same procurement target .variable Indicates the index of the procurement object record, variable Indicates a super-edge index representing a relationship within the same procurement object. , ,variable Indicates the quantity of records for the procurement object, variable This indicates the number of overflow edges in the same procurement object relationship. The [number]th [item / section]... The relationship between the same procurement object and the boundary is written as dimensional hyperedge belonging vector , No. dimension Directly corresponds to the first One procurement target record, Indicates the first The procurement object record belongs to the first The relationship between the same procurement object exceeds the boundary. Indicates the first The procurement object record does not belong to the first The relationship between the same procurement object exceeds the boundary. According to... Extracting the value of the first The record indexes of procurement objects contained in the superedge of the same procurement object relationship form a record index set. Sort all by superedge index. At that time, the first The position directly corresponds to the first... Each relationship with the same procurement object extends beyond the boundary, thus forming a cost list for relationships with the same procurement object; all are written synchronously in the same order. , , and This forms the input for solving the high-order relation multi-cut partitioning layer. Specifically, each time the high-order relation multi-cut partitioning layer receives 256 one-dimensional same-purchase-object relation costs, 256 one-dimensional support costs, 256 one-dimensional exclusion costs, and 256 128-dimensional hyperedge attribution vectors, and provides 128 one-dimensional grouping labels.

[0131] Assign a 1-dimensional partitioning decision variable to each superedge of the same procurement object relationship. , Indicates segmentation, This indicates that it is not divided.

[0132] Will support the cost With the cost of rejection Perform the absolute difference calculation to obtain the difference result. .

[0133] The difference results With threshold parameter Compare and generate fixed judgment values ​​simultaneously. and solution indicator value .

[0134] when and Not less than At that time, the first The relationship between the same procurement object is directly fixed as a segment and written into the code. , and ; when and Not less than At that time, the first The relationship between the same procurement object is directly fixed as not to be split, and written into the code. , and ; when At that time, the first The relationship between the same procurement object and the boundary is kept in a state to be solved, and written into... , and retain As a candidate splitting decision variable for high-order relation multi-cut partitioning layer.

[0135] After the threshold prediction is completed, all the segmentation decision variables are kept in correspondence with the cost list of the same procurement object according to the hyperedge index order, forming a segmentation solution structure.

[0136] The single ownership constraint of the procurement object record is achieved through 1-dimensional grouping labels. Build. Variables Indicates the first Grouping tags for each procurement object record. During initialization, a unique tag value is written to each procurement object record. Read the first... The set of record indexes corresponding to the superedges of the same procurement object relationship At that time, if or current candidate Then Minimum tag value overwrite All procurement object records ;like or current candidate Then keep The existing label values ​​of all procurement object records remain unchanged. When both fixed judgment values ​​and candidate segmentation judgment variables exist, the fixed judgment value is written first. If a procurement object record has two different label values ​​in the same round of candidate assignment, it is directly determined to violate the single ownership constraint of the procurement object record, and the current candidate assignment is revoked. The constraint check results obtained according to this writing rule constitute the record ownership constraint set.

[0137] The constraint for splitting relationships within the same procurement object is constructed through consistency checks on shared procurement object records. Variables and This represents two distinct hyperedge indices representing the same procurement object relationship. For any two hyperedges representing the same procurement object relationship... and Read record index set and .

[0138] when and There is a common record index, and and All are 0, or the current candidate and When both are 0, and All procurement object records covered are written to the same group label; when , , or When a value of 1 exists, it is prohibited to use the common record index to rewrite the super-edge of the same procurement object relationship that has been split into a single group label.

[0139] After the candidate assignment is completed, if the record index set corresponding to the super edge of a certain segmented procurement object relationship is... If all procurement object records are assigned the same group label, it is directly determined to violate the same procurement object relationship segmentation constraint, and the current candidate assignment is revoked. The constraint check results obtained according to this check rule constitute the segmentation consistency constraint set.

[0140] Higher-order relations are divided into layers by multiple cuts. , , , , , , , and As input, use the stabilized segmentation to determine the variable. and group tags This is the output.

[0141] The higher-order relation multi-cut partitioning layer does not contain trainable parameters. It operates in the following order: threshold prediction, candidate assignment, constraint checking, total cost comparison, and label reconstruction. The higher-order relation multi-cut partitioning layer first considers the difference results... Arrange the hyperedges of the same procurement object relationship in descending order of the unsolved state. Then, for each hyperedge of the same procurement object relationship in the unsolved state, construct a splitting candidate value of 1 and a non-splitting candidate value of 0 in sequence. After verifying that the set of attribution constraints and the set of splitting consistency constraints pass the check, calculate the total cost according to the following formula. and retain the total cost Smaller candidate values:

[0142] In the formula, This represents the total cost corresponding to a multi-cut partition of a higher-order relation; Indicates a super-edge index representing a relationship within the same procurement object; Indicates the number of overflow edges in the same procurement object relationship; Indicates the first The solution indicator value for the superedge of the same procurement object relationship; Indicates the first Candidate splitting decision variables for the super-edge of the same procurement object relationship; Indicates the first A fixed judgment value for exceeding the boundary of a relationship with the same procurement object; Indicates the first The cost of cross-border exclusion in the relationship of the same procurement object; Indicates the first The support cost of a cross-border relationship within the same procurement object. Variable , and All are 0 or 1 values, variables and All are one-dimensional values, therefore the products, sums, and comparisons in the formula maintain the same dimensions. At that time, the total cost Read the candidate splitting decision variables of the state to be solved at the corresponding position. ;when At that time, the total cost The threshold prediction value is already fixed at the corresponding position. When the candidate value is a split, the total cost is... Include support costs in the corresponding locations. When the candidate value is not split, the total cost is... The exclusion cost is calculated at the corresponding position. The higher-order relationship is divided into multiple layers, and the splitting selection of each hyperedge of the same procurement object relationship is uniformly written into the same total cost target.

[0143] According to the total cost After completing one round of candidate assignment, a consistency check is performed again on the cross-edge relationships of the same procurement object for all shared procurement object records. When a label conflict occurs, the conflicting splitting decision variable is reread, and splitting candidate value 1 and non-splitting candidate value 0 are reconstructed according to the current label state. Constraint checks and comparisons with the total cost are then performed again. The current round of label writing results are stable and the total cost is [not specified]. When the value stops decreasing, write the current value back. The connectivity reconstruction phase reads all stable data. and all ; At that time, All procurement object records are aggregated into the same group label; At that time, keep The existing grouping labels for all procurement object records remain unchanged. After traversing all hyperedges related to the same procurement object, the entire... Perform compressed renumbering to map identical tag values ​​to the same consecutive number. (The result is the compressed renumbered version.) Directly corresponding to procurement object grouping. In a fixed embodiment, the high-order relation multi-segmentation layer outputs 128 one-dimensional grouping labels. The 128 one-dimensional grouping labels correspond one-to-one with the 128 procurement object records. Grouping labels with the same value correspond to the same procurement object group, and grouping labels with different values ​​correspond to different procurement object groups.

[0144] like Figure 10 As shown, each original record is modeled as a record node, and common information such as system affiliation, host connection, tag number, parameter unit, and document source are modeled as attribute nodes. Multiple records are connected through the same attribute, forming a hyperedge of the same procurement object relationship. The dashed-line enclosed area in the diagram represents the relationship triggered by shared attributes across documents.

[0145] Figure 11 As shown, this is a pairwise similarity matrix recording R1 to R8. Rows and columns are arranged in the order of grouping: Group A (R1, R3, R5), Group B (R2, R6), and Group C (R4, R7, R8). Dark squares indicate high similarity, and light squares indicate low similarity. The block-like structures within the thick boxes correspond to the three natural groups.

[0146] In this embodiment, step S5 specifically includes: Read 1D group labels and procurement object records ,in , This indicates the quantity of the purchased item records. Different values ​​can be used to... After removing duplicates and sorting by label value in ascending order, we get Grouping of procurement objects, variables Indicates the number of groups of procurement objects, variable This represents the grouping index for procurement objects. For those that meet the following conditions... The record indexes are sorted in ascending order to form an ordered record index sequence. ,variable Indicates the first The number of purchase object records contained in each purchase object group, variable Indicates the first Index of record location within each procurement object group. Read the object category, system affiliation, host connection relationship, entry / exit relationship, tag number / map number, parameter unit, and document source in sequence, and write them into rows with fields aligned. .variable Indicates the first Group of procurement objects The object category field value of each record, variable Indicates the value of the system-owned field, variable Represents the value of the host connection relationship field, variable Indicates the value of the field representing the inlet / outlet relationship, variable Indicates the value of the tag number / drawing number field, variable Indicates the unit field value of the parameter, variable This indicates the value of the document source field. (Click) The sequentially arranged rows of aligned fields constitute the grouped field alignment result, which maintains the original correspondence and field order of the seven fields within the same procurement object record.

[0147] Perform non-empty value filtering on each of the seven fields in the grouping field alignment result. Read directly during comparison. , , , , , and Existing field values ​​will not be used to introduce new derived fields. The object category field, system affiliation field, host connection relationship field, entry / exit relationship field, tag number / map number field, and parameter unit field use the same binding rules: along The system sequentially scans the corresponding field positions. If a field value is not empty, it is recorded as a non-empty value; if a field value is empty, the current field position is skipped. If only one different non-empty value appears in the same field, that different non-empty value is determined as the field-bound value. If all values ​​in the same field are empty, the corresponding field-bound value is written as empty. If two or more different non-empty values ​​appear in the same field, the corresponding field-bound value is written as empty. The document source field is used to record the source coverage of the same procurement object group and does not use single-value consistency rules; along... After reading all non-empty document sources in sequence, first remove duplicates, then... , , and The data is arranged in a fixed order and connected by semicolons to form the document source field, which is then bound to retrieve values. Indicates technical specifications, Representing drawings, This indicates a purchase order. This indicates a price list.

[0148] The following fields are bound to values ​​in a fixed order: object category, system affiliation, host connection relationship, entry / exit relationship, tag number / map number, parameter unit, and document source. These are then written as a summary record for the procurement object. .variable Represents a summary field for object categories, variables Indicates the system's aggregate field, variable This field represents a summary of host connection relationships; variables. This field represents the summary field of the inbound and outbound relationship, and the variable is... This represents a summary field for tag numbers and drawing numbers; variables. This indicates a summary field for parameter units, and the variable is... This indicates the document source summary field. When the object category field, system affiliation field, host connection relationship field, entry / exit relationship field, tag number / map number field, and parameter unit field have a unique non-empty value, the corresponding field's bound value is directly written to the procurement object summary record; when the corresponding field does not have a unique non-empty value, the corresponding field remains empty in the procurement object summary record; the document source field always writes the deduplicated and concatenated document source field bound value. The procurement object summary record maintains a 7-field structure, with the field order completely consistent with the procurement object record.

[0149] Attribute conflict entries are constructed only for the system affiliation field, entry / exit relationship field, tag number / map number field, and parameter unit field. The number of distinct non-empty values ​​is counted for each of these fields; when two or more distinct non-empty values ​​are counted, an attribute conflict entry is formed for that field. Each attribute conflict entry is constructed along... The system sequentially scans for the first occurrence of different non-null values ​​and writes them in the format "field value@document source"; when the same field value comes from multiple document sources, the multiple document sources are sorted according to... , , and The fixed order of values ​​is written after the same field value, with multiple document sources connected by forward slashes and different field values ​​connected by semicolons; when the number of different non-empty values ​​is less than 2, the corresponding position for the attribute conflict item is written as an empty value. This forms a structure similar to the first... Attribute conflict items corresponding to each procurement object group .variable Indicates system affiliation conflict items, variables Indicates the conflict term in the entry-exit relationship; variable Indicates a conflict between the tag number and the map number; variable This indicates any conflicting unit parameters. Each procurement object group corresponds to one procurement object summary record. and a conflicting attribute The summary records of procurement objects and items with conflicting attributes are grouped and indexed according to the same procurement object. Each item corresponds to the previous one.

[0150] Figure 12 Using R1 to R8 as record nodes, directed or undirected lines represent constraints such as belonging to the same procurement object or not belonging to the same procurement object. The diagram presents a chain-like and hierarchical structure, reflecting the implicit sequential and pairing relationships in the document, which will be used as constraint edge inputs in subsequent multi-cutting.

[0151] like Figure 13This section demonstrates the grouping results output by the multi-cut algorithm after running it on the aforementioned relational graph / hypergraph. R1 to R8 are divided into three procurement object groups: A, B, and C, with accompanying textual examples. It emphasizes that each group corresponds to a single procurement object, and records within each group are treated as the same item.

[0152] Figure 14 In the diagram, the horizontal axis represents the number of groups k, the vertical axis represents the constraint weight λ, and the background grayscale represents the total default cost under the corresponding parameter combination: light gray indicates low cost, and dark gray indicates high cost. The elliptical lines represent the contour lines of the cost, the dashed rectangles represent the recommended parameter regions, and the black dots represent the optimal parameters used by the current algorithm. The diagram illustrates that parameter selection is not arbitrary but is obtained through cost function optimization.

[0153] In this embodiment, step S6 specifically includes: Grouping indexes for each procurement object Summary records from procurement targets Only read the system-attributed summary field Summary fields of inlet and outlet relationships , Tag Number / Drawing Number Summary Field Summary field of parameter units and from attribute conflict items Only read system ownership conflict items Conflicts in the relationship between entry and exit , Item number conflict Parameter unit conflict terms .

[0154] variable This indicates the number of procurement object groups. The system affiliation summary field, entry / exit relationship summary field, tag number / drawing number summary field, and parameter unit summary field are used to fix the field positions for determining technical requirement consistency. The system affiliation conflict item, entry / exit relationship conflict item, tag number / drawing number conflict item, and parameter unit conflict item provide direct evidence of whether there are conflicting values. Attribute conflict items are only generated during the attribute binding phase when two or more different non-empty values ​​appear in the same field. Therefore, without rescanning procurement object records within the same procurement object group, only checking whether the corresponding attribute conflict item is empty is sufficient to determine whether there are conflicting values ​​for the corresponding field.

[0155] To convert attribute conflict items into directly determinate discrete tags, conflict existence tags are generated for each of the four attribute conflict items. , , and .

[0156] variable Indicates a system ownership conflict exists; variable Indicates a conflict in the entry / exit relationship; the variable is marked. Indicates a flag indicating a conflict between the tag number and the map number; variable This indicates that there is a conflict in the units of the parameters.

[0157] System ownership conflict items When it is an empty value, Write it as 0; System ownership conflict item When it is a non-null value, Write it as 1.

[0158] Conflicts in the relationship between entry and exit , Item number conflict Parameter unit conflict terms Using the same write rules, we obtained respectively , and .

[0159] The four conflict markers directly correspond to the system attribution inconsistency judgment criteria, entry / exit relationship inconsistency judgment criteria, tag number / map number inconsistency judgment criteria, and parameter unit inconsistency judgment criteria. System attribution summary field. Summary fields of inlet and outlet relationships , Tag Number / Drawing Number Summary Field Or parameter unit summary field When the value is empty and the corresponding conflicting attribute is also empty, it is still treated as if there is no conflicting value, because the consistency determination object is whether there is a conflicting value, not whether the field is empty.

[0160] Based on the conflict presence marker, consistency determination markers are generated for each of the four fields. , , and .

[0161] variable Indicates a system ownership consistency determination flag, variable Indicates the consistency of the inlet and outlet relationship as a flag, variable Indicates the tag number / map number consistency determination flag, variable This indicates a flag indicating whether the units of the parameters are consistent.

[0162] At that time, Write it as 1; At that time, Write it as 0.

[0163] , and Using the same mapping rules, we obtained respectively , and Therefore, the existence of conflicts with different values ​​is determined as the inconsistency judgment condition, and the absence of conflicts with different values ​​is determined as the consistency judgment condition. The field-by-field consistency judgment of system affiliation, entry and exit relationship, tag number, and parameter unit is all written as directly usable 0 or 1 flags.

[0164] Consistency results are generated based on four consistency criteria for the marker generation technique. .variable Indicates the first Consistency results of 1D technical requirements for each procurement object group This indicates that the technical requirements are consistent. This indicates a discrepancy in technical requirements.

[0165] Only when , , and When all values ​​are 1, Write it as 1; if only , , and If there is a value of 0, then... Write it as 0.

[0166] Technical Requirements Consistency Results Index grouped by purchase object Summary record of procurement targets Maintaining the correlation between each item, the grouping of procurement objects retains its original grouping results. This results in a field-level consistency determination flag arranged according to the procurement object grouping index. , , , Consistency sequence of technical requirements .

[0167] like Figure 15 As shown, records related to the same equipment extracted from various engineering documents such as technical specifications, drawings, purchase requisitions, and quotation lists are automatically identified and merged into a unified procurement object, circulating water pump P-101A, in the system interface. The upper part shows the original scattered records, and the lower part shows the unified summary records, along with the attribute consistency check results.

[0168] like Figure 16As shown in the diagram, the upper half displays the groups A, B, and C obtained from the current multi-slice; the lower half presents several counterfactual branches: for example, moving some records from group A to group B, or merging group B with group C. Below each branch are the number of violated business constraints and the total cost of violation, used to compare the merits of different group modification schemes and select the scheme with the lowest total cost.

[0169] like Figure 17 The diagram illustrates the three-dimensional surface of the cost function C in the parameter space (k, λ). The gray slope represents a valley with a concave center, signifying a low-cost region. Black dots represent the optimal solution selected by the current algorithm, hollow dots represent the cost position corresponding to a counterfactual solution, and dashed boxes enclose the recommended low-cost region. This diagram visually shows whether the counterfactual solution is moving uphill or downhill relative to the current solution.

[0170] like Figure 18 As shown, the matrix on the left is a mean plot, with rows representing records R1 to R8 and columns representing groups A / B / C. Black cells indicate a high weight for a record in a particular group, while light gray indicates a low weight. The bars on the right are a variance plot, with the shade of gray representing the degree of uncertainty a record has in a group. Records with high means and low variance are the most reliable groups, while records with high variance require more intensive manual review.

[0171] In summary, power plant procurement processes span multiple stages, including preliminary design, construction drawing design, equipment requisition, tender clarification, supplier quotation, and technical clarification. Information about the procurement targets is typically scattered across technical specifications, drawings, purchase requisitions, and quotation lists, continuously circulating among design units, procurement departments, suppliers, and reviewers. Technical specifications emphasize functional requirements, system configuration, interface boundaries, and parameter requirements; drawings focus on tag numbers, layout locations, host connection relationships, and inlet / outlet directions; purchase requisitions emphasize the procurement name, system affiliation, tag number, and parameter summary; and quotation lists emphasize the quotation targets, specifications, and quotation range. Power plant process systems are multi-layered, with long equipment connection chains and complex interdisciplinary relationships. Procurement targets include not only main equipment but also supporting valves, filters, heat exchangers, interface components, and auxiliary parts. Procurement targets may appear repeatedly in different professional documents and business forms, often under various names such as system name plus component name, tag number name, interface name, drawing number name, specification name, or supply scope name, indicating inconsistent granularity. Some documents only cover the system and connection location, while others only cover tag numbers and parameter units. Furthermore, some documents suffer from missing items, misalignments, and inconsistent terminology due to differences in table structure, annotation methods, field order, and directional terminology. The bidding and procurement review process requires identifying the same procurement object across existing documents, then summarizing system affiliation, connection relationships, tag numbers, drawing numbers, and parameter units, and further verifying the consistency of technical requirements. This is essential to support the confirmation of procurement boundaries, verification of technical clauses, review of supply interfaces, and subsequent contract execution.

[0172] Existing technical solutions typically revolve around document parsing, field extraction, and object comparison. Some solutions employ manual review, where reviewers compare and merge documents one by one based on object name, tag number, drawing number, system name, interface direction, and parameter table. Other solutions use rule-based processing, segmenting technical specification texts, table lists, and drawing annotations, locating keywords, extracting fields, and aligning headers, then matching records based on criteria such as name consistency, tag number consistency, system consistency, interface consistency, or parameter similarity. Still other solutions convert procurement records into structured entries, combining natural language processing, string similarity, field weighted scoring, general drawing relationship matching, or knowledge base alignment methods to calculate the correspondence between records and output candidate merging results, object directories, or consistency verification reports. Some engineering management systems also introduce document source tags, professional codes, and ledger numbers to manage the parsed procurement object records in separate databases, perform list-based verification, and conduct manual review.

[0173] Existing technical solutions often rely on similar names, overlapping fields, or pairwise record matching, making it difficult to represent the closed-loop relationship where multiple procurement records together form a single procurement object. Complementary records between technical specifications, drawings, purchase requisitions, and quotation lists often cannot be uniformly judged within the same relational object. Missing items due to different document scopes are easily confused with missing actual information, and tag numbers, host connection relationships, and entry / exit relationships are also difficult to identify in conjunction. A unified scoring method can also easily compress information that supports merging and conflicting information that drives splitting into the same score, leading to the mismerging of similar-named but different items and the misclassification of complementary records across documents, ultimately affecting the summary of procurement objects and the review of consistency with technical requirements.

[0174] Therefore, the core technical problem this application needs to solve is, in the power plant bidding and procurement scenario, how to establish a cross-document relationship mechanism for the same procurement object that can uniformly represent attribute complementarity, missing item sources, connection correspondence, and conflict exclusion, relying only on the field information and document scope constraints provided by technical specifications, drawings, purchase requisitions, and quotation lists, so that the relationship cost of the same procurement object can be reliably determined, thereby completing the closed-loop processing of procurement object grouping, attribute aggregation, and technical requirement consistency review. This invention achieves a structured representation of cross-document procurement object relationships through a closed hypergraph of object attributes. Under the condition of only having access to technical specifications, drawings, purchase requisitions, and quotation lists, this invention constructs procurement object records as record nodes and attribute nodes for system affiliation, host connection relationships, entry / exit relationships, tag numbers / drawing numbers, parameter units, and document sources. It then organizes the hyperedges of the same procurement object relationship based on object category correspondence, tag number / drawing number correspondence, host connection relationship connection, entry / exit relationship connection, and document source complementarity. This invention further calculates attribute coverage status, attribute missing status, tag number / drawing number correspondence status, connection connectivity status, and conflict status to form the relationship status of the same procurement object. This allows for unified judgment of missing sources due to cross-document complementary records, differences in record scope, tag number location correspondence, and connection chain consistency within the same relationship object, thereby reducing the risk of mismerging and misclassification caused by relying solely on name similarity or field overlap, and providing a stable intermediate representation for subsequent global grouping. Secondly, this invention uses counterfactual branch contribution calculation to separate and support merging information and drive segmentation information. This invention sets up system affiliation branches, host connection relationship branches, entry / exit relationship branches, tag number / map number branches, parameter unit branches, object category branches, and document source branches based on the relationship status of the same procurement object, forming a joint branch output. The output of the centralized evaluation unit is then merged into an evaluation value for the same procurement object and segmented into evaluation values ​​for different procurement objects. This invention performs a preset replacement value replacement on the output of a single branch while keeping the output of unreplaced branches unchanged. It calculates the branch contribution value based on the difference between the benchmark evaluation value and the replacement evaluation value, and maps the contribution values ​​corresponding to attribute coverage, tag number / map number correspondence, and connection connectivity to support costs. It maps the contribution values ​​corresponding to system conflicts, entry / exit conflicts, and parameter conflicts to exclusion costs, generating the relationship cost for the same procurement object. Compared to unified weighting or unified total score methods, this invention can exclude system affiliation conflicts in scenarios where tag numbers / map numbers are already corresponding and connection relationships are already connected, and it can also segment parameter unit conflicts in scenarios where system affiliation is consistent. Simultaneously, it allows document source and recording scope to constrain the direction of missing interpretation, making the relationship cost more consistent with the business mechanism of complementary recording and conflict coexistence in power plant bidding and procurement documents. Furthermore, this invention outputs results that can be directly used for review through high-order relation multi-cut partitioning and attribute binding. Based on the relation cost of the same procurement object, this invention performs high-order relation multi-cut partitioning on the closed hypergraph of object attributes. Under the constraints of single ownership of procurement object records and consistency of relation segmentation within the same procurement object, procurement object groups are formed. Furthermore, by predicting the threshold difference between support and rejection costs, some relations are fixed, reducing uncertain branches in the global solution. Based on the procurement object grouping, this invention performs attribute binding to generate a summary record of procurement objects. Under conditions of inconsistencies in system ownership, entry / exit relationships, tag numbers, and parameter unit values, attribute conflict items are generated, thereby forming a consistent technical requirement result.This invention enables the identification, attribute summarization, conflict localization, and consistency conclusion output of the same procurement object across documents to be connected by the same processing link, which meets the actual needs of power plant bidding and procurement scenarios to realize the verification of procurement objects and the review of technical requirements based on existing documents. The second objective of this invention is to propose A cross-document same procurement object identification system for power plant bidding and procurement includes: The procurement object record module is used to obtain procurement object information and extract procurement object records from the procurement object information. The Procurement Object Status Module is used to construct a closed hypergraph of object attributes based on procurement object records, forming hyperedges representing relationships within the same procurement object; and to calculate the relationship status of the same procurement object based on these hyperedges. Procurement Object Cost Module: Used to perform counterfactual branch contribution calculation on the relationship status of the same procurement object, to obtain the contribution value of the same procurement object and the segmentation into different procurement objects, and to map the contribution value of the same procurement object and the segmentation into different procurement objects into support cost and rejection cost, and to generate the relationship cost of the same procurement object based on the support cost and rejection cost. The procurement object grouping module is used to perform high-order relation multi-cut partitioning on the closed hypergraph of object attributes based on the relation cost of the same procurement object, thereby forming procurement object groups. Object summary record module: used to bind attributes according to the grouping of procurement objects, and form a summary record of procurement objects and attribute conflict items; Procurement Object Identification Module: This module is used to determine the consistency of technical requirements based on the procurement object summary record and attribute conflict items, and outputs the procurement object grouping, procurement object summary record, and technical requirement consistency results.

[0175] A third objective of this invention is to provide an electronic device comprising a processor, a memory, and a display screen. The memory and display screen are both connected to the processor, such as via a bus. Optionally, the electronic device may further include a transceiver. It should be noted that in practical applications, the transceiver is not limited to a single unit, and the structure of this electronic device does not constitute a limitation on the embodiments of this application.

[0176] The processor can be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor can also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0177] A bus can include a pathway for transmitting information between the aforementioned components. The bus can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc.

[0178] The memory may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited to these.

[0179] The memory stores the application code that executes the solution of this application, and its execution is controlled by the processor. The processor executes the application code stored in the memory to implement the content shown in the foregoing method embodiments.

[0180] A fourth objective of this invention is to provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, performs the aforementioned functions. Figure 1 The illustrated method embodiments include various processes. For example, a memory may include instructions that can be executed by a processor of an electronic device to perform the described method.

[0181] A computer-readable storage medium can be a tangible device that holds and stores instructions used by an instruction execution device. A computer-readable storage medium can be, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any combination thereof. Specifically, a computer-readable storage medium can be a portable computer disk, a hard disk, a USB flash drive, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), staging random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory stick, floppy disk, optical disk, magnetic disk, mechanical encoding device, or any combination thereof.

[0182] A fifth objective of this invention is to provide a computer program product comprising computer instructions that, when executed by a processor, implement the above-described... Figure 1 The various processes of the method embodiments shown can achieve the same technical effect, and will not be described again here to avoid repetition.

[0183] Many embodiments and applications beyond the examples provided will be apparent to those skilled in the art upon reading the foregoing description. Therefore, the scope of this teaching should not be determined by reference to the foregoing description, but rather by reference to the foregoing claims and the full scope of their equivalents. For purposes of completeness, all articles and references, including patent applications and publications, are incorporated herein by reference. The omission of any aspect of the subject matter disclosed herein in the foregoing claims is not intended as a waiver of that subject matter, nor should it be construed as an indication that the applicant has not considered that subject matter as part of the disclosed inventive subject matter.

[0184] The above content provides a further detailed description of the present invention. It should not be construed that the specific embodiments of the present invention are limited to this. For those skilled in the art, several simple deductions or substitutions can be made without departing from the concept of the present invention, and all such deductions or substitutions should be considered as falling within the scope of protection of the present invention as defined by the submitted claims.

Claims

1. A method for identifying the same procurement object across documents in power plant bidding and procurement, characterized in that, include: Obtain procurement object information and extract procurement object records from the procurement object information; Construct a closed hypergraph of object attributes based on the procurement object records, forming hyperedges representing relationships within the same procurement object; The relationship status of the same procurement object is calculated based on the hyperedge of the relationship of the same procurement object. The counterfactual branch contribution calculation is performed on the relationship status of the same procurement object to obtain the contribution value of the same procurement object and the segmentation into different procurement objects. The contribution value of the same procurement object and the segmentation into different procurement objects is mapped to the support cost and the rejection cost. The relationship cost of the same procurement object is generated based on the support cost and the rejection cost. Based on the relationship cost of the same procurement object, the closed hypergraph of object attributes is divided into high-order relationship multi-cut partitions to form procurement object groups; Based on the procurement objects, attribute binding is performed to form a summary record of procurement objects and attribute conflict items; Based on the summary record of procurement objects and attribute conflict items, determine the consistency results of technical requirements, and output the procurement object grouping, the summary record of procurement objects, and the consistency results of technical requirements.

2. The method for identifying the same procurement object across documents in power plant bidding and procurement as described in claim 1, characterized in that, The process of obtaining procurement object information and extracting procurement object records from the procurement object information includes: Obtain information on the procurement targets from the technical specifications, drawings, purchase requisitions, and quotation lists in power plant bidding processes; Extract the object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit, and document source from the procurement object information to form a procurement object record.

3. The method for identifying the same procurement object across documents in power plant bidding and procurement according to claim 2, characterized in that, The process involves constructing a closed hypergraph of object attributes based on the procurement object records, forming hyperedges representing relationships within the same procurement object. The relationship status of the same procurement object is calculated based on the hyperedge of the same procurement object relationship, including: Construct a closed hypergraph of object attributes based on the procurement object record. Use the procurement object record as the record node, and use system ownership, host connection relationship, entry and exit relationship, tag number and map number, parameter unit and document source as attribute nodes. Combine the record nodes according to the complementary conditions of object category correspondence, tag number and map number correspondence, host connection relationship connection, entry and exit relationship connection and document source to form the same procurement object relationship hyperedge. The relationship status of the same procurement object is calculated based on the hyperedge of the relationship of the same procurement object.

4. The method for identifying the same procurement object across documents in power plant bidding and procurement according to claim 1, characterized in that, The counterfactual branch contribution calculation for the relationship status of the same procurement object yields contribution values ​​for the same procurement object and for segments into different procurement objects. These contribution values ​​are mapped to support costs and repulsion costs. The relationship cost for the same procurement object is generated based on the support costs and repulsion costs, including: Counterfactual branch contribution calculation is performed on the relationship status of the same procurement object to obtain the combined evaluation value of the same procurement object and the evaluation value of the different procurement objects. The combined evaluation value of the same procurement object and the evaluation value of the different procurement objects are then used to form the benchmark evaluation value. The contribution value of the corresponding branch pair when it is merged into the same procurement object and the contribution value of the corresponding branch pair when it is divided into different procurement objects are calculated based on the difference between the benchmark evaluation value and the replacement evaluation value. Based on the attribute coverage status, tag number and map number correspondence status, and connection connectivity status of the same procurement object relationship status, determine the contribution values ​​corresponding to attribute coverage, tag number and map number correspondence, and connection connectivity, and map them as support costs. The system's ownership branch, entry / exit relationship branch, and parameter unit branch are divided into branches corresponding to different procurement objects and mapped to exclusion costs. Generate the relationship cost for the same procurement object based on the support cost and the exclusion cost.

5. The method for identifying the same procurement object across documents in power plant bidding and procurement according to claim 1, characterized in that, The step of performing high-order relation multi-cut partitioning on the closed hypergraph of object attributes based on the relation cost of the same procurement object to form procurement object grouping includes: Based on the closed hypergraph of object attributes, read the procurement object record, the hyperedge of the same procurement object relationship, and the cost of the same procurement object relationship, and summarize the cost of the same procurement object relationship based on the hyperedge of the same procurement object relationship to form a list of costs of the same procurement object relationship; Based on the cost list of relationships for the same procurement object, a segmentation decision variable is assigned to the super-edge of each relationship for the same procurement object to form a segmentation solution structure; Based on the segmentation solution structure, a segmentation consistency constraint set is constructed. Based on the segmentation consistency constraint set, a higher-order relation multi-cut solution is performed on the segmentation decision variables to obtain the segmentation decision result. Based on the segmentation determination results, the connectivity of the closed hypergraph of object attributes is reconstructed to form group labels, and the procurement objects are grouped according to the group labels.

6. The method for identifying the same procurement object across documents in power plant bidding and procurement according to claim 1, characterized in that, The step of binding attributes based on the grouping of procurement objects to form a summary record of procurement objects and attribute conflict items includes: Extract procurement object records within the same group based on the procurement object grouping, and form a grouping field alignment result; Attribute binding is applied to the alignment results of grouped fields to create attribute conflict items; Based on the field binding values, write the object category, system affiliation, host connection relationship, entry and exit relationship, tag number, drawing number, parameter unit, and document source into the procurement object summary record; Output a summary record of procurement objects, and output any attribute conflicts corresponding to the summary record of procurement objects.

7. The method for identifying the same procurement object across documents in power plant bidding and procurement according to claim 1, characterized in that, The process of determining the consistency of technical requirements based on the summary records of procurement objects and attribute conflict items, and outputting the procurement object grouping, the summary records of procurement objects, and the consistency results of technical requirements, includes: Consistency determination is performed on the summary records of procurement objects and the conflict items of attributes respectively. Conflicts with different values ​​are determined as inconsistency determination conditions, and conflicts without different values ​​are determined as consistency determination conditions. Based on the inconsistency and consistency criteria, generate technical requirement consistency results and associate the technical requirement consistency results with the procurement object summary record. Output the grouping of procurement objects, the summary records of procurement objects, and the consistency results of technical requirements.

8. A cross-document same procurement object identification system for power plant bidding and procurement, based on the cross-document same procurement object identification method for power plant bidding and procurement as described in any one of claims 1-7, characterized in that, include: The procurement object record module is used to obtain procurement object information and extract procurement object records from the procurement object information. The procurement object status module is used to construct a closed hypergraph of object attributes based on procurement object records, forming hyperedges representing relationships within the same procurement object. The relationship status of the same procurement object is calculated based on the hyperedge of the relationship of the same procurement object. Procurement Object Cost Module: Used to calculate the counterfactual branch contribution of the relationship state of the same procurement object, obtain the contribution value of the same procurement object and the segmentation into different procurement objects, map the contribution value of the same procurement object and the segmentation into different procurement objects to support cost and rejection cost, and generate the relationship cost of the same procurement object based on the support cost and rejection cost. The procurement object grouping module is used to perform high-order relation multi-cut partitioning on the closed hypergraph of object attributes based on the relation cost of the same procurement object, thereby forming procurement object groups. Object summary record module: used to bind attributes according to the grouping of procurement objects, and form a summary record of procurement objects and attribute conflict items; Procurement Object Identification Module: This module is used to determine the consistency of technical requirements based on the procurement object summary record and attribute conflict items, and outputs the procurement object grouping, procurement object summary record, and technical requirement consistency results.

9. An electronic device, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the method for identifying the same procurement object across documents for power plant bidding and procurement as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method for identifying the same procurement object across documents for power plant bidding and procurement, as described in any one of claims 1-7.

Images