Hierarchical conflict detection and resolution method and device for cross-document knowledge synthesis, computer device and storage medium

Abstract

The application discloses a hierarchical conflict detection and adjudication method and device for cross-document knowledge synthesis, electronic equipment and a storage medium. The method comprises the following steps: obtaining structured knowledge objects from different document sources, and extracting information masking features; analyzing the correlation possibility between the knowledge objects based on the information masking features, and generating suspected correlation information; determining relationship confirmation features according to the suspected correlation information, and generating a verification strategy to obtain a verification result; determining an evolution mapping relationship according to the verification result; performing synthesis adjudication processing on attribute descriptions, correlation relationships and state information based on the evolution mapping relationship, determining information retention relationships and generating a target knowledge structure. The application can identify the reason for the difference between cross-document knowledge, confirm the evolution correlation relationship between knowledge objects, and perform information retention and structure reorganization according to the knowledge subject change process, thereby improving the accuracy and integrity of cross-document knowledge synthesis.

Description

Technical Field

[0001] This application relates to the field of knowledge document processing technology, and in particular to a hierarchical conflict detection and adjudication method, apparatus, computer equipment, and storage medium for cross-document knowledge synthesis. Background Technology

[0002] As digital document resources accumulate, information about the same object or business entity may be scattered across documents created from different sources and at different stages. Due to differences in the background, recording methods, and content of different documents, the same object may correspond to different names, descriptions, attribute information, status information, and related information in different documents. Therefore, in the process of multi-source document knowledge processing, it is necessary to perform correlation analysis and organization of knowledge content scattered across different documents to achieve effective synthesis of knowledge from different sources.

[0003] Existing cross-document knowledge processing methods typically rely on document content matching, object feature comparison, or association rule analysis to associate knowledge content across different documents. However, in practical applications, due to the different stages of document creation, the knowledge descriptions corresponding to the same object may differ, leading to inconsistencies or difficulty in mapping some knowledge content. For example, the same object may have different state descriptions, different associated objects, or different attribute behaviors at different stages, making it difficult to directly correlate knowledge content across different documents. Furthermore, relying solely on a single descriptive feature can easily result in inaccurate knowledge association results, affecting the effectiveness of subsequent knowledge synthesis.

[0004] Furthermore, in the process of cross-document knowledge synthesis, for knowledge content that differs across documents, it is necessary to further determine the correspondence between different knowledge descriptions and how to retain them in the synthesis result. Existing methods typically focus on filtering, covering, or selecting the differing content, making it difficult to analyze the reasons for the differences in conjunction with the formation process of knowledge objects at different stages. This results in the synthesized knowledge structure failing to accurately express the relationships and changes between different knowledge contents. Therefore, a method is needed that can analyze the differences in the cross-document knowledge synthesis process and determine the knowledge content processing method based on the analysis results. Summary of the Invention

[0005] This application aims to address the problems in existing cross-document knowledge synthesis processes, where differences in knowledge descriptions formed by different document sources for the same knowledge subject make it difficult to accurately determine the correspondence between different knowledge objects, and where it is difficult to determine a reasonable processing method for different knowledge content during the synthesis process.

[0006] Specifically, in documents from different sources, the same knowledge subject may form knowledge objects with different manifestations due to differences in the stage of formation, descriptive perspective, or recording status. When there are differences in object characteristics, state descriptions, and related expressions between different knowledge objects, some knowledge objects may not be able to establish a correspondence through direct matching, leading to problems such as uncertain associations or inconsistent content during the knowledge synthesis process.

[0007] Furthermore, when synthesizing different knowledge objects, the differences between them do not necessarily indicate an error in the description of a particular knowledge object. Rather, they may be due to differences arising from the knowledge subject being at different stages or under different conditions. Therefore, it is necessary to analyze the relationships between different knowledge objects, determine the basis for these differences, and, in conjunction with the changes in the knowledge subject, judge how different knowledge content is processed in the synthesis result to improve the accuracy and completeness of cross-document knowledge synthesis.

[0008] To achieve the above objectives, this application provides a hierarchical conflict detection and adjudication method for cross-document knowledge synthesis, comprising: Obtain structured knowledge objects from different document sources, and extract corresponding information masking features based on the differences in object features, state descriptions, and related expressions formed when describing the same knowledge subject in different document sources; Based on the information masking features, the possibility of association between different structured knowledge objects is analyzed to generate suspected association information. The suspected association information is used to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis. Based on the suspected association information, relationship confirmation features are determined to distinguish different association relationships, and corresponding verification strategies are generated based on the relationship confirmation features to obtain verification results for the association relationships between the structured knowledge objects. Based on the verification results, the evolutionary mapping relationship between different structured knowledge objects is determined. The evolutionary mapping relationship is used to characterize the knowledge object change relationship formed by the same knowledge subject in different document stages or under different state conditions. Based on the evolutionary mapping relationship, the attribute descriptions, associations and state information between different structured knowledge objects are synthesized and adjudicated to determine the information retention relationship between knowledge objects and generate the target knowledge structure.

[0009] Optionally, the step of obtaining structured knowledge objects from different document sources includes: Perform subject association analysis on the knowledge description content in different document sources to identify the corresponding knowledge subject scope between different document sources; Based on the scope of the knowledge subject, analyze the correspondence between attribute descriptions, status descriptions, and association descriptions formed for the same knowledge subject in different document sources, and form a knowledge object description structure that includes subject correspondence and stage description relationships.

[0010] Optionally, the step of extracting corresponding information masking features based on differences in object features, state descriptions, and association expressions formed when describing the same knowledge subject in different document sources includes: Based on the differences in object characteristics, state descriptions, and related expressions, the scope of the related influence of different difference factors is analyzed to form information describing the impact of differences. Based on the difference in the description information, the occlusion effect of knowledge description at different stages on the judgment of the relationship between knowledge subjects is identified, and the information occlusion feature is formed according to the occlusion effect.

[0011] Optionally, the step of extracting the corresponding information masking features includes: Based on the description of the differences in the impact of the differences, we analyze the occlusion patterns corresponding to the differences in the characteristics, state descriptions, and related expressions of different objects. Based on the synergistic relationship between different forms of occlusion, an occlusion feature description is formed to characterize the reasons for the concealment of the relationship between knowledge subjects, and the information occlusion feature is improved based on the occlusion feature description.

[0012] Optionally, the step of analyzing the likelihood of association between different structured knowledge objects based on the information masking features and generating suspected association information includes: Based on the stage judgment impact description corresponding to the information masking features, identify the subject retention features that remain across state changes between different structured knowledge objects; The formation stages, manifestations, and associated effects of the characteristics of multiple subjects are analyzed to form cross-stage retention description information. Based on the cross-stage preservation description information, suspected association information is generated to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis.

[0013] Optionally, the step of generating associated suspected information includes: Based on the source stage and associated role of the cross-stage retention description information, multiple candidate association explanations are constructed. The support status of multiple candidate related interpretations in knowledge description at different stages is analyzed to form the trajectory of related interpretation changes; Based on the change trajectory of the association interpretation, the association basis corresponding to different candidate association interpretations is organized to form the suspected association information.

[0014] Optionally, the step of determining the relationship confirmation features for distinguishing different relationships based on the suspected association information includes: Based on the direction of the relationship to be confirmed corresponding to the suspected related information, multiple candidate confirmation features are obtained; The retention, distinguishability, and correlation support of different candidate confirmation features before and after changes in the state of the knowledge subject are analyzed to form feature contribution description information. The combined contribution relationship between different candidate confirmation features is analyzed based on the feature contribution description information to form a confirmation basis combination, and the relationship confirmation feature is determined based on the confirmation basis combination.

[0015] Optionally, the step of generating a corresponding verification strategy based on the relationship confirmation features to obtain verification results for the relationships between the structured knowledge objects includes: Based on the degree of state preservation and association differentiation function corresponding to the relationship confirmation features, candidate verification objects reflecting the differences of different association relationships are determined; The response changes of different candidate verification objects under the corresponding knowledge subject state conditions are analyzed to form verification differentiation description information; The verification path is selected based on the verification distinction description information, and the verification response result is obtained according to the verification path to form a verification result for the relationship between the structured knowledge objects.

[0016] Optionally, the step of obtaining the verification results for the relationships between the structured knowledge objects includes: Based on the verification response results and the direction of the unconfirmed relationship corresponding to the associated suspected information, a matching analysis is performed to form verification deviation description information; Based on the verification deviation description information, analyze whether the verification response corresponding to the current verification path meets the associated confirmation requirements, and adjust the corresponding verification object and verification path when the verification response does not meet the associated confirmation requirements. The verification response result is re-acquired based on the adjusted verification path to update the verification result.

[0017] Optionally, the step of determining the evolutionary mapping relationship between different structured knowledge objects based on the verification result includes: Based on the relationship confirmation status corresponding to the verification results, analyze the persistent subject characteristics between different structured knowledge objects to form subject continuity description information; Based on the subject continuity description information, analyze the changing state characteristics, change stages, and change sequence between different structured knowledge objects to form stage change description information; Based on the correspondence between the subject continuity description information and the stage change description information, an evolutionary mapping relationship is established to characterize the state change process of the knowledge subject.

[0018] Optionally, the step of establishing an evolutionary mapping relationship to characterize the state change process of a knowledge subject includes: Based on the described stage change information, analyze the state change trajectory corresponding to attribute changes, state changes, and association changes among knowledge objects at different stages; Based on the state change trajectory analysis, the reasons for the changes between knowledge objects at different stages are formed, the stage change explanation relationship is formed, and the change process description in the evolutionary mapping relationship is supplemented.

[0019] Optionally, the step of dynamically reconstructing the attribute descriptions, relationships, and state information between different structured knowledge objects based on the evolutionary mapping relationship to generate the target knowledge structure includes: Based on the evolutionary mapping relationship, identify the subject continuity relationship that needs to be maintained between knowledge objects at different stages, as well as the stage change relationship formed with state changes; Based on the subject continuity relationship and the stage change relationship, the information retention relationship between knowledge objects is determined, and a stage knowledge retention structure is established based on the information retention relationship. The attribute descriptions, associations and state information between different structured knowledge objects are reorganized to form a target knowledge structure that includes the evolution process of the knowledge subject.

[0020] Optionally, the step of generating the target knowledge structure includes: Evolutionary integrity description information is formed based on the subject continuity relationship and stage change relationship in the target knowledge structure; Based on the evolutionary integrity description information, analyze whether the target knowledge structure simultaneously retains the continuous and changing relationships of the knowledge subject at different stages; When there is a lack of continuous subject relationships or abnormal stage change relationships, the evolutionary mapping relationship is adjusted according to the corresponding abnormal reason, and the target knowledge structure is updated.

[0021] To achieve the above objectives, this application also provides a hierarchical conflict detection and adjudication device for cross-document knowledge synthesis, comprising: The knowledge object acquisition module is used to acquire structured knowledge objects from different document sources. It extracts the corresponding information masking features based on the differences in object features, state descriptions, and related expressions when describing the same knowledge subject in different document sources. The association analysis module is used to analyze the association probability between different structured knowledge objects based on the information masking features, and generate suspected association information. The suspected association information is used to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis. The relationship verification module is used to determine relationship confirmation features for distinguishing different relationships based on the suspected association information, and to generate corresponding verification strategies based on the relationship confirmation features in order to obtain verification results for the relationships between the structured knowledge objects. An evolutionary analysis module is used to determine the evolutionary mapping relationship between different structured knowledge objects based on the verification results. The evolutionary mapping relationship is used to characterize the knowledge object change relationship formed by the same knowledge subject in different document stages or under different state conditions. The knowledge reconstruction module is used to perform synthetic adjudication processing on the attribute descriptions, associations and state information between different structured knowledge objects based on the evolutionary mapping relationship, determine the information retention relationship between knowledge objects, and generate the target knowledge structure.

[0022] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, it implements the steps of the above method.

[0023] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method.

[0024] Compared with the prior art, this application has at least the following beneficial effects: This application acquires structured knowledge objects from different document sources and analyzes the differences in object features, state descriptions, and related expressions formed when describing the same knowledge subject in different document sources. It extracts the corresponding information masking features, enabling the cross-document knowledge synthesis process to identify the influencing factors that cause differences between different knowledge objects and avoid judging the relationship between knowledge objects based solely on surface description differences.

[0025] Furthermore, this application analyzes the possibility of association between different structured knowledge objects based on the information masking features, forms suspected association information, and generates a verification strategy in combination with relationship confirmation features to verify the association relationship between different knowledge objects. This enables inconsistent relationships generated during cross-document knowledge synthesis to be analyzed based on corresponding association criteria, thereby improving the reliability of knowledge object association judgment.

[0026] Furthermore, this application analyzes the retention, distinguishability, and correlation support of different candidate confirmation features before and after changes in the state of the knowledge subject, forms a combination of confirmation criteria, and determines the relationship confirmation features based on the combination of confirmation criteria, so that the difference relationships between different knowledge objects can be identified according to different levels of correlation criteria, and realizes hierarchical analysis of difference relationships in the process of cross-document knowledge synthesis.

[0027] Furthermore, this application determines the evolutionary mapping relationship between different structured knowledge objects based on the verification results, so that knowledge objects formed in different document stages or under different state conditions can establish corresponding change relationships, thereby distinguishing the difference between description differences caused by changes in knowledge subjects and actual association errors, and providing a basis for subsequent knowledge synthesis adjudication.

[0028] Furthermore, this application performs synthetic adjudication processing on the attribute descriptions, associations, and state information between different structured knowledge objects based on the aforementioned evolutionary mapping relationship, determines the information retention relationship between knowledge objects, and generates a target knowledge structure. This enables the knowledge synthesis process to no longer rely solely on a single knowledge description for coverage or selection, but rather to combine the knowledge subject's change process to determine the organization of different knowledge content, thereby improving the target knowledge structure's ability to express the continuous relationship and stage change relationship of the knowledge subject. Attached Figure Description

[0029] Figure 1 Flowchart of a hierarchical conflict detection and adjudication method for cross-document knowledge synthesis provided in this application embodiment; Figure 2 A schematic diagram of the structure of a hierarchical conflict detection and adjudication device for cross-document knowledge synthesis provided in an embodiment of this application; Figure 3 A schematic diagram of an electronic device structure provided in an embodiment of this application; Figure 4 This is a schematic diagram of a computer-readable storage medium structure provided in an embodiment of this application. Detailed Implementation

[0030] The technical solution of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the following embodiments are only used to illustrate this application and are not intended to limit the scope of protection of this application.

[0031] In this application, a hierarchical conflict detection and adjudication method for cross-document knowledge synthesis is used to process document content from different sources, at different stages of formation, or with different description methods. By constructing structured knowledge objects corresponding to different document sources, the method analyzes the differences between different knowledge objects, determines the basis for the association between different knowledge objects, and, based on the changing relationships formed by knowledge subjects at different stages, synthesizes and adjudicates the knowledge content with differences, thereby generating a target knowledge structure that can express the continuous relationship and stage change relationship of knowledge subjects.

[0032] It should be noted that in the process of cross-document knowledge synthesis, the knowledge descriptions formed by different document sources for the same knowledge subject are not necessarily consistent. Due to differences in document creation time, recording purpose, application scenario, and descriptive perspective, the same knowledge subject may be represented as a knowledge object in different documents. For example, for a certain equipment object, in the technical document formed during the R&D phase, the equipment can be described through design parameters, composition structure, and R&D status; in the test document formed during the testing phase, the equipment can be described through test indicators, test results, and abnormal situations; in the operation and maintenance document formed during the operation and maintenance phase, the equipment can be described through operating status, maintenance records, and fault association information. Although the knowledge objects formed in the above different stages have different descriptive content, they may all correspond to the same knowledge subject. Therefore, this application does not simply merge the knowledge content in different documents, but first analyzes the reasons for the differences between different knowledge objects, identifies whether the differences originate from the change process of the knowledge subject, and determines the association relationship between different knowledge objects and the synthesis processing method based on the analysis results.

[0033] like Figure 1 As shown, this embodiment provides a hierarchical conflict detection and adjudication method for cross-document knowledge synthesis, including the following steps: S1. Obtain structured knowledge objects from different document sources. Extract corresponding information masking features based on the differences in object features, state descriptions, and related expressions when describing the same knowledge subject in different document sources.

[0034] In this embodiment, document data from different document sources is first acquired, and the knowledge descriptions from these sources are then structured to generate corresponding structured knowledge objects. The document source represents the origin or formation environment of the knowledge content and can be determined based on the specific application scenario. For example, in an enterprise technical data management scenario, the document source may include product design documents, experimental testing documents, production record documents, and equipment maintenance documents; in a legal knowledge management scenario, the document source may include legal provisions, case analysis documents, and court judgments.

[0035] For knowledge descriptions from different document sources, this application first identifies the scope of the knowledge subjects contained therein, and then structures and organizes the different descriptions surrounding the same knowledge subject to form corresponding structured knowledge objects. The structured knowledge object represents the organized knowledge description unit, which may include subject description information, attribute description information, state description information, and association description information. Specifically, subject description information characterizes the scope of the object corresponding to the knowledge object; attribute description information characterizes the features of the knowledge subject; state description information characterizes the performance state of the knowledge subject at a specific stage or under a specific environment; and association description information characterizes the relationship between the knowledge subject and other knowledge objects.

[0036] For example, for a specific model of equipment, different document sources may generate multiple structured knowledge objects. In a design document, this knowledge object may include the equipment model, structural components, and design parameters; in a test document, it may include the test environment, test results, and performance changes; and in a maintenance document, it may include runtime, fault types, and maintenance measures. Although these knowledge objects contain different types of information, they may all correspond to the same equipment entity.

[0037] Preferably, after obtaining the structured knowledge objects corresponding to different document sources, this application further analyzes the descriptive differences between different structured knowledge objects. Specifically, it analyzes the differences in object features, state descriptions, and relational expressions formed when describing the same knowledge subject in different document sources. The differences in object features are used to characterize the differences in describing the basic characteristics of the knowledge subject in different document sources. For example, the same device may be described using different numbers, abbreviations, or classification methods in different documents. This type of difference does not necessarily indicate that the corresponding knowledge objects belong to different subjects, but may be due to different document formation stages or application purposes. The differences in state descriptions are used to characterize the state changes of the same knowledge subject at different stages or under different operating conditions. For example, the same device corresponds to the design state in the design stage, the test state in the testing stage, and the operating state in the operation stage. There are differences between different state descriptions, but these differences may reflect the development process of the knowledge subject itself. The differences in relational expressions are used to characterize the differences in how different document sources describe the relationship between the knowledge subject and other objects. For example, in a research and development document, a device may be described as a "component of the control module"; in a maintenance document, the device may be described as a "fault detection object." Although the two expressions are different, they may reflect different relationships formed by the same knowledge subject at different application stages.

[0038] Based on the above difference analysis results, this application further determines the scope of influence of different difference factors on the judgment of the association of knowledge subjects, and forms corresponding information masking features. The information masking features are used to characterize the factors that prevent knowledge objects that originally had the possibility of association from directly showing a corresponding relationship due to the description differences between different document sources.

[0039] For example, a piece of equipment might be described using an internal R&D number during the R&D phase and an asset management number during the operational phase. If matching is done solely based on name or number, a connection between the two knowledge objects might not be established. However, by analyzing the changes in document phase, attribute description, and associated objects corresponding to the number changes, a potential relationship can be discovered. In this case, the number changes, phase changes, and description changes that prevent the connection from being directly apparent constitute the corresponding information masking feature. It should be noted that the information masking feature in this application does not indicate a lack of knowledge information, but rather that existing knowledge information, due to different formation conditions, different expression methods, or different state changes, makes the potential relationship between knowledge subjects difficult to identify through direct matching.

[0040] Through S1 processing, this application can further analyze knowledge objects that exhibit surface differences during cross-document knowledge synthesis, distinguish between inconsistent relationships caused by different knowledge subjects and different relationships formed by the same knowledge subject at different stages, and provide a basis for subsequent generation of related suspected information and hierarchical conflict detection.

[0041] S2. Based on the information masking features, analyze the possibility of association between different structured knowledge objects and generate suspected association information. The suspected association information is used to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis.

[0042] After analyzing the differences between different structured knowledge objects in S1, this application further analyzes the uncertainty of association caused by differences in object features, state descriptions, and related expressions. It should be noted that in the process of cross-document knowledge synthesis, differences between different knowledge objects do not necessarily indicate the absence of associations between them; some differences may stem from different manifestations of the knowledge subject at different stages. Therefore, this application does not directly determine the existence of conflicts between knowledge objects based on the difference results, but rather further analyzes whether there are associations hidden by the difference descriptions between different knowledge objects based on the aforementioned information masking characteristics.

[0043] Specifically, this application identifies subject-preserving features in different structured knowledge objects that can continuously exhibit characteristics across state changes, based on the stage-based impact description corresponding to the information masking features. These subject-preserving features represent the continuous characteristics of a knowledge subject across different document stages or state conditions. They can originate from attribute descriptions, compositional relationships, functional relationships, business affiliation relationships, or other descriptive content that reflects the stability of the subject within the knowledge object.

[0044] For example, in equipment lifecycle management scenarios, the same equipment may form different structured knowledge objects in design documents, test documents, and maintenance documents. The design document records the equipment's design model, components, and intended use; the test document records the equipment's test number, test items, and performance; and the maintenance document records the equipment's asset number, maintenance status, and fault handling records. Because different documents focus on different content, the numbers, statuses, and some attribute descriptions of these knowledge objects may differ significantly. Matching solely by name or number may fail to determine whether knowledge objects in different documents correspond to the same equipment. However, by analyzing the relationships between components, functions, and associated equipment information in different knowledge objects, it can be found that certain features consistently revolve around the same equipment entity. These cross-stage maintained descriptions can be considered as entity-retained features.

[0045] Preferably, after identifying multiple subject retention features, this application further analyzes the formation process, manifestation, and correlation of different subject retention features in different document stages to form cross-stage retention description information. Specifically, for multiple structured knowledge objects formed by the same knowledge subject in different document stages, this application first analyzes the document stage in which each subject retention feature first appears, the scope of stages in which it persists, and the changes in different stages to determine whether the subject retention feature runs through the development process of the knowledge subject. For example, for a certain equipment object, it can analyze whether the relationship between its core components exists simultaneously in the design document, test document, and operation document, and determine whether the component relationship remains unchanged in different stages or is adjusted accordingly with changes in the equipment state; furthermore, this application analyzes the specific expression of the same subject retention feature in different documents to determine whether there is a semantic correspondence between different descriptive forms. For example, the same device might be described as a "control module connection object" in a design document and as a "fault detection associated module" in a maintenance document. Although they use different expressions, they both point to the same control function association. Furthermore, this application analyzes the supporting role of different subject-retention features in determining the association of knowledge objects. For instance, a certain attribute information can only indicate that two knowledge objects have similar descriptions, while a certain compositional relationship, functional relationship, or business affiliation relationship can further support that two knowledge objects belong to the same knowledge subject. By comprehensively analyzing the formation stage, expression changes, and association support degree of subject-retention features, it is possible to determine whether there are cross-stage continuity relationships between different structured knowledge objects and to form cross-stage retention description information used to describe candidate association directions and association basis.

[0046] Preferably, based on the cross-stage retention description information, this application generates suspected association information. This suspected association information does not directly determine that different structured knowledge objects belong to the same knowledge subject, but rather serves to carry candidate relationships with a high probability of association at the current stage. In specific implementation, this application organizes the possible association directions between different structured knowledge objects according to the stage distribution, change trajectory, and association support degree corresponding to the subject retention characteristics contained in the cross-stage retention description information, to form multiple descriptions of relationships to be confirmed. For example, when two structured knowledge objects originate from the product design stage and the operation and maintenance stage respectively, this application can generate candidate association relationships indicating that the two knowledge objects may correspond to different stages of the same product subject based on the continuous compositional relationship, functional relationship, and business affiliation relationship between them; when there is an attribute change relationship between two structured knowledge objects, a candidate association relationship indicating that one knowledge object may have been formed by the other knowledge object through state changes can be generated; when there is a continuous change in the associated object between two knowledge objects, a candidate association relationship indicating that there may be a continuation of the relationship between them can be generated. The relationship direction represents the possible association types and evolution paths between different knowledge objects. For example, different knowledge objects may correspond to the same knowledge subject's manifestation objects at different stages, or there may be state changes, attribute inheritance, or continuation of association relationships between different knowledge objects. The association basis records the specific evidence sources supporting the formation of the above-mentioned candidate association relationships, including subject-maintained characteristics, attribute description changes, state change trajectories, and association description changes that persist in different stages. By organizing the relationship direction and association basis accordingly, this application forms suspected association information that reflects "why two knowledge objects are considered to be related," enabling the subsequent relationship confirmation process to verify specific candidate relationships rather than judging solely based on the similarity between knowledge objects. By generating suspected association information, this application transforms the differences in knowledge objects that appear "inconsistent" during cross-document knowledge synthesis into candidate association relationships that can be further analyzed, enabling the subsequent conflict detection process to judge the source of the relationship between different knowledge objects, rather than simply determining the conflict result based on whether the descriptions are consistent.

[0047] Furthermore, in practical applications, the degree of suspected association between different knowledge objects may vary. For example, some knowledge objects share multiple cross-stage preservation features, indicating a high probability of association; while others share only a few similar descriptions, suggesting a low probability of association. Therefore, this application can hierarchically organize suspected association information based on the number, stability, and association effects of preservation features for different subjects, enabling candidate associations at different levels to enter the subsequent verification process, thereby achieving hierarchical conflict detection in the cross-document knowledge synthesis process. Specifically, the hierarchical organization is not simply a ranking of candidate relationships based on the numerical value of association probability, but rather a distinction based on the information foundation upon which the association judgment between different knowledge objects depends. For example, when two structured knowledge objects simultaneously possess cross-stage maintained subject features, continuously changing state descriptions, and stable association support, it indicates that they not only have surface-level descriptive similarities but can also establish a correspondence through the knowledge subject evolution process. Such candidate associations can be prioritized for subsequent verification as high-level associations to be confirmed. When two structured knowledge objects only have similar attribute descriptions or consistent local association expressions, although there is a certain possibility of association, there is a lack of complete stage-maintaining evidence. Such candidate associations can be further supplemented as intermediate-level associations. When two structured knowledge objects only have similar names, tags, or single descriptive content but cannot form cross-stage maintained descriptions, they are considered low-level candidate associations for restrictive verification. Through the above hierarchical approach, this application enables candidate associations of different degrees to adopt verification methods that match their association evidence, avoiding the use of the same judgment standard to handle all knowledge object differences. At the same time, it allows subsequent relationship confirmation processes to conduct targeted analysis around different levels of association evidence, achieving hierarchical conflict detection in the cross-document knowledge synthesis process. Through the above processing, this application completes the transformation from information masking features to related suspected information, enabling the hidden relationships between knowledge objects formed at different document stages to be explicitly expressed, providing a foundation for subsequent conflict confirmation based on relationship confirmation features and synthetic adjudication based on evolutionary mapping relationships.

[0048] S3. Determine relationship confirmation features for distinguishing different relationships based on the suspected association information, and generate corresponding verification strategies based on the relationship confirmation features to obtain verification results for the relationships between the structured knowledge objects.

[0049] After generating the suspected association information in S2, this application further confirms the candidate association relationships between different structured knowledge objects. It should be noted that since the suspected association information is formed based on the analysis of information masking features and subject-preserving features, it can characterize the possibility of association between different knowledge objects and the corresponding basis for association, but it cannot be directly used as the final association result. This is because in the process of cross-document knowledge synthesis, different knowledge subjects may also have some similar features. For example, different models of equipment may have the same functional modules, and different business objects may have the same attribute descriptions. If association judgment is based solely on partial preservation features, it is easy to incorrectly associate different knowledge subjects. Therefore, this application further determines relationship confirmation features based on the suspected association information and distinguishes the authenticity of different association relationships through a verification process.

[0050] Specifically, this application obtains multiple candidate confirmation features based on the direction of the relationship to be confirmed corresponding to the suspected association information. These candidate confirmation features are used to describe whether a stable association exists between knowledge objects from different dimensions. They can originate from state changes, attribute changes, association changes, and cross-stage retention between different knowledge objects. For example, in a device knowledge synthesis scenario, two structured knowledge objects both contain the association description "control module," but this association description alone cannot determine whether the two knowledge objects correspond to the same device entity. In this case, candidate confirmation features such as device number change relationships, composition structure change relationships, functional purpose change relationships, and operational stage change relationships can be further obtained. By analyzing the performance of different candidate confirmation features at different stages, it can be determined whether they can effectively distinguish the association relationship between the two knowledge objects.

[0051] Preferably, after obtaining multiple candidate confirmation features, this application further analyzes the retention, distinguishability, and association support of different candidate confirmation features before and after changes in the knowledge subject's state, forming feature contribution description information. Here, retention describes whether a candidate confirmation feature can persist as the knowledge subject's stage changes; distinguishability describes whether a candidate confirmation feature can distinguish between knowledge objects formed by the same knowledge subject at different stages and knowledge objects formed by different knowledge subjects; and association support describes the strength of support a candidate confirmation feature provides for the current suspected association relationship.

[0052] For example, a piece of equipment may have different serial numbers during the design and operation phases, but its core components and functional uses remain consistent. In this case, while the serial number change can reflect the phase change, it alone cannot confirm the relationship between two knowledge objects. However, the structural and functional relationships remain consistent across different phases and can exclude other equipment objects, thus providing a higher degree of correlation support. Through the above analysis, feature contribution description information corresponding to different candidate confirmation features can be generated.

[0053] Preferably, based on the feature contribution description information, this application further analyzes the combined contribution relationship between different candidate confirmation features to form a confirmation criterion combination, and determines the relationship confirmation feature based on the confirmation criterion combination. Unlike the single feature judgment method, this application combines multiple candidate confirmation features to enable the relationship confirmation process to comprehensively consider multiple dimensions of information during the knowledge subject's change process. For example, for the confirmation of a knowledge object's association relationship, the subject structure maintenance relationship, attribute change trajectory, and changes in associated objects can be considered simultaneously. When multiple candidate confirmation features form a consistent support relationship across different stages, it can be considered that this set of confirmation criteria can effectively support the existence of an association between two knowledge objects; when there are obvious contradictions between different candidate confirmation features, further verification is required to distinguish the association relationship.

[0054] Preferably, after determining the relationship confirmation features, this application generates a corresponding verification strategy based on the relationship confirmation features. The verification strategy is used to select appropriate verification objects and verification paths for different relationships to obtain verification response results that reflect the relationship status of knowledge objects. Specifically, this application determines candidate verification objects that can reflect the differences between different relationships based on the degree of state retention and the distinguishing effect of the relationship confirmation features. The candidate verification objects can be attribute information used to verify the continuity of the subject, state information used to verify stage changes, or association information used to verify the continuity of the relationship.

[0055] For example, when determining whether two equipment knowledge objects belong to the same equipment entity, the equipment composition structure, functional relationships, changes in operating status, and changes in related components can be selected as candidate verification objects for comprehensive analysis. If two knowledge objects, despite having different corresponding numbers, names, or management identifiers, can form a continuous correspondence in their composition, function, core component relationships, and associated equipment relationships, and if the state changes between the two knowledge objects conform to the evolutionary pattern of the equipment lifecycle, then it can be considered that there is a high probability of a subject association between them. Such differences are more likely to originate from description changes formed in different management stages rather than corresponding to different equipment subjects. If two knowledge objects have similar names and some attribute descriptions, but their composition, function, and associated objects are completely different, it indicates that there is only surface-level description consistency between them, lacking continuous evidence to support the relationship of the same subject, and their association probability needs to be reduced. Furthermore, in some cases, two knowledge objects may only be consistent in some dimensions, such as having the same equipment model but with adjustments to the composition, or having the same equipment function but with changes in associated components. In this case, it cannot be directly determined that they belong to the same knowledge subject. Instead, it is necessary to combine the attribute change trajectory, state change process, and changes in the associated relationship to further determine whether they belong to the same equipment formed in different upgrade stages, or belong to a similar description relationship formed between different equipment. By analyzing the response changes of different candidate verification objects under different association hypotheses, this application can distinguish the differential relationships formed by the same knowledge subject at different stages, the similar relationships between different knowledge subjects, and the object relationships with evolutionary inheritance relationships, thereby providing a basis for subsequent verification path selection and association relationship confirmation.

[0056] Preferably, after determining the candidate verification objects, this application further analyzes the response changes of different candidate verification objects under the corresponding knowledge subject state conditions to form verification differentiation description information. The verification differentiation description information is used to describe the differentiated performance of different verification objects under different association hypotheses, thereby providing a basis for verification path selection. For example, for two candidate associated knowledge objects, analysis can be performed according to two scenarios: "performance of the same knowledge subject at different stages" and "similar descriptions of different knowledge subjects." If, under the same subject hypothesis, the state changes, attribute changes, and association changes between the two knowledge objects can form a continuous change relationship, but cannot form a reasonable explanation under different subject hypotheses, then the verification differentiation description information can support the confirmation of the association relationship.

[0057] Preferably, based on the verification differentiation description information, this application selects a corresponding verification path and obtains the verification response result according to the verification path to form a verification result for the relationship between the structured knowledge objects. In this embodiment, the verification path is not limited to a single form and can be adjusted according to the difference type corresponding to different suspected association information. For example, for uncertain relationships caused by differences in object features, the change pattern of the subject identifier can be verified first; for uncertain relationships caused by differences in state description, the state change process can be verified first; for uncertain relationships caused by differences in association expression, the change relationship between associated objects can be verified first. Through the above processing, this application completes the transformation from suspected association information to verification result, enabling inconsistencies in cross-document knowledge synthesis to be verified through multiple dimensions, rather than simply judging based on similarity or a single rule.

[0058] Furthermore, from the perspective of hierarchical conflict detection, stage S1 primarily identifies the influencing factors leading to knowledge discrepancies, stage S2 further forms candidate objects with potential relationships, and stage S3 further distinguishes these candidate relationships through relationship confirmation features and verification strategies. Therefore, the conflict detection in this application does not merely judge the final result, but analyzes the process of knowledge object discrepancies step-by-step, enabling different levels of knowledge discrepancies to obtain corresponding judgment criteria and improving the accuracy of associations in the cross-document knowledge synthesis process. Through the S3 processing, this application can obtain verification results for the relationships between different structured knowledge objects, providing a reliable foundation for subsequently establishing knowledge subject evolution mapping relationships and performing synthesis adjudication processing.

[0059] S4. Determine the evolutionary mapping relationship between different structured knowledge objects based on the verification results. The evolutionary mapping relationship is used to characterize the knowledge object change relationship formed by the same knowledge subject in different document stages or under different state conditions.

[0060] After completing the association verification in S3, this application further analyzes the changing process of the relationship between different structured knowledge objects based on the verification results, so as to determine the changing relationship of knowledge objects corresponding to the same knowledge subject in different document stages or under different state conditions.

[0061] It should be noted that in the process of cross-document knowledge synthesis, even if two structured knowledge objects have been verified to have a relationship, all information in the two knowledge objects cannot be directly merged. This is because the same knowledge subject may exhibit different attributes, different state information, and different relationships at different stages. Without analysis of the knowledge object's evolution process, direct merging may result in historical stage information overlapping with current stage information, or an inability to accurately represent the knowledge subject's development process. Therefore, this application further determines the evolutionary mapping relationship between different structured knowledge objects based on the verification results. This evolutionary mapping relationship is not used to simply describe whether two knowledge objects are related, but rather to describe the corresponding changes formed between different knowledge objects as the knowledge subject changes from one stage state to another.

[0062] Specifically, this application first confirms the state based on the relationship corresponding to the verification results, analyzes the persistent subject characteristics between different structured knowledge objects, and forms subject continuity description information. This subject continuity description information is used to represent the characteristic relationships that can continuously characterize the same knowledge subject at different document stages or under different state conditions. These characteristics are not required to remain completely consistent across different stages, but rather can reflect the continuous existence of the knowledge subject through changing relationships.

[0063] For example, in an equipment lifecycle management scenario, the knowledge objects corresponding to a piece of equipment during the design phase include its structural parameters, design number, and relationships between its components. During the operation phase, the knowledge objects include its asset number, operating status, and maintenance relationships. Although the design number and asset number differ, and the operating status changes, the core structural components, functional uses, and associated control modules of the equipment remain continuous across different stages. By analyzing these persistent key characteristics, a continuous descriptive information about the equipment's core components can be generated.

[0064] Furthermore, this application analyzes the changing state characteristics, change stages, and change sequence between different structured knowledge objects based on the aforementioned subject continuity description information to form stage change description information. State characteristics represent the state changes of a knowledge subject at different stages; for example, equipment changes from a design state to a testing state, and then to an operating state; attribute descriptions change from design parameters to operating parameters; and relationships change from design composition relationships to operation and maintenance relationships. Change stages represent the document formation stage or knowledge state stage corresponding to different knowledge objects, used to determine the sequential relationship between different knowledge objects. Change sequence represents the evolution direction formed when a knowledge subject changes from one state to another, so that knowledge objects at different stages no longer appear as independent data sets, but rather form a change process with temporal or state logic.

[0065] For example, for a certain technical product, the knowledge objects in the early design documents may record the initial structural scheme, design parameters, and relationships between components; the knowledge objects in the mid-term testing documents may record structural adjustment information, performance change information, and test verification results based on test feedback; and the knowledge objects in the later operation documents may record the actual application status, changes in operating parameters, and maintenance-related information. Although the knowledge objects in the above different stages differ significantly in terms of attribute description, state information, and relationships—for example, the design stage focuses on the theoretical structure of the product, the testing stage focuses on the performance after structural adjustments, and the operation stage focuses on state changes during actual application—there are still persistent core characteristics among the knowledge objects in different stages, such as the core component relationships, functional and usage relationships, and key component relationships. Based on the above continuous core relationships, this application further analyzes the formation process corresponding to attribute changes, state changes, and relationship changes among knowledge objects in different stages. For example, it determines whether a change in a structural parameter originates from test optimization, whether a change in an operating state originates from design adjustments, and whether a change in a related component belongs to a replacement relationship in the product upgrade process. By mapping the persistent main characteristics with the information on stage changes, we can determine the continuous change process of the product knowledge subject from "design formation - testing and verification - structural optimization - operation and application", and establish the evolutionary mapping relationship between knowledge objects at different stages. This allows the knowledge descriptions in different documents to no longer be independent information units, but to form a related knowledge structure that reflects the product life cycle change process.

[0066] Preferably, after forming the subject continuity description information and the stage change description information, this application further establishes an evolutionary mapping relationship based on the correspondence between the two. Specifically, the subject continuity description information is used to determine the continuity relationship between different knowledge objects "whether they belong to the same knowledge subject," while the stage change description information is used to determine the evolutionary relationship "how the same knowledge subject changes." By associating the two, an evolutionary mapping relationship describing the change process of a knowledge subject can be formed. For example, a certain equipment knowledge subject forms a first structured knowledge object in the design phase, a second structured knowledge object in the testing phase, and a third structured knowledge object in the operation phase. Through verification results, it can be determined that the three knowledge objects all correspond to the same equipment subject; further analysis of the state changes, attribute changes, and association changes among the three knowledge objects can establish an evolutionary mapping relationship between the first structured knowledge object, the second structured knowledge object, and the third structured knowledge object.

[0067] Preferably, the evolutionary mapping relationship can include stage correspondence, state change relationship, attribute change relationship, and association change relationship between different knowledge objects. For example, it can represent that a certain attribute description changes from a theoretical parameter in the design stage to a measurement parameter in the testing stage, and then to an actual parameter in the operation stage; it can also represent that a certain association relationship changes from a composition relationship in the design stage to a maintenance relationship in the operation stage. By establishing the evolutionary mapping relationship, this application further clarifies the reasons for the formation of differences between different knowledge objects in the cross-document knowledge synthesis process. That is, this application does not simply determine the difference between two knowledge objects as a conflict, but further determines whether the difference belongs to the stage difference generated in the evolution of the knowledge subject. If the difference can be explained by the evolutionary mapping relationship, it means that the difference belongs to the synthesizable difference; if the difference cannot form a reasonable evolutionary relationship, it can be treated as an abnormal relationship in the subsequent adjudication process.

[0068] From the perspective of graded conflict detection and adjudication, S4 is responsible for the stage of conflict relationship interpretation. In S1, this application identifies the influencing factors that cause differences between different knowledge objects; in S2, this application forms suspected related information based on the influencing factors of differences; in S3, this application confirms whether there is a relationship between different knowledge objects through relationship confirmation features and verification strategies; and in S4, this application further answers the question of "why are different" between different knowledge objects, that is, determines whether the differences originate from the change process of the knowledge subject itself.

[0069] For example, two knowledge objects record the "design capacity" and "actual operating capacity" of a device. If only the numerical values ​​are compared, there might be a significant inconsistency, easily leading to the conclusion of a conflict. However, through an evolutionary mapping relationship, it can be discovered that the design capacity corresponds to the device's design stage state, while the actual operating capacity corresponds to the device's operational stage state. Both belong to the same knowledge subject, representing different state descriptions at different stages. Therefore, this difference is not erroneous information that needs to be eliminated, but rather should be preserved based on its stage attribute during subsequent synthetic adjudication.

[0070] Through S4 processing, this application establishes an intermediate technical object that connects conflict detection and synthetic adjudication, enabling the difference relationship between different knowledge objects to be further transformed from "whether there is a conflict" to "how it changes", providing a basis for subsequent determination of information retention relationships and generation of target knowledge structures.

[0071] S5. Based on the evolutionary mapping relationship, perform synthetic adjudication processing on the attribute descriptions, associations and state information between different structured knowledge objects to determine the information retention relationship between knowledge objects and generate the target knowledge structure.

[0072] After establishing the evolutionary mapping relationship in S4, this application further performs synthetic adjudication processing on the attribute descriptions, associations, and state information between different structured knowledge objects based on the evolutionary mapping relationship, so as to determine the organization method of different knowledge content in the final target knowledge structure.

[0073] It should be noted that in the cross-document knowledge synthesis process, the multiple structured knowledge objects whose relationships have been confirmed through the aforementioned steps do not mean that the information they contain can be directly superimposed or overwritten. Since different structured knowledge objects may correspond to the same knowledge subject's manifestations at different document stages or under different state conditions, some information within different knowledge objects may have different functions. For example, one attribute description might be used to express the state of the knowledge subject in a historical stage, while another attribute description might be used to express the state in the current stage; one relationship might be used to represent the compositional relationships in the formation process of the knowledge subject, while another relationship might be used to represent the business relationships in the operation process of the knowledge subject. If a simple fusion method is directly adopted, it is easy for information from different stages to overlap, making it impossible for the target knowledge structure to accurately express the changes in the knowledge subject.

[0074] Based on this, this application analyzes the attribute descriptions, associations, and state information of different structured knowledge objects according to the evolutionary mapping relationship, determines the stage correspondence and change relationship between different knowledge contents, and further performs synthetic adjudication processing.

[0075] In this embodiment, the synthetic adjudication process is not a simple deletion or selection of information that differs between different knowledge objects, but rather a determination of the role of different information in the target knowledge structure based on the evolutionary process of the knowledge subject. Specifically, this application analyzes the continuous relationship and stage change relationship of the subject recorded in the evolutionary mapping relationship to determine whether the information in different knowledge objects belongs to information that continuously exists in the knowledge subject, information that is generated in stages, or information formed due to changes in the description method, and determines the corresponding information processing method based on the judgment result. For information reflecting the continuous existence of knowledge subjects, such as object identification relationships, core component relationships, functional association relationships, and business affiliation relationships, this application identifies such information as continuous subject information because it maintains subject consistency across different stages of knowledge objects, and establishes cross-stage maintenance relationships during the synthesis process. For information generated as the state of knowledge subjects changes, such as parameter descriptions formed during the design phase, performance indicators formed during the testing phase, and status records formed during the operation phase, this application does not replace such information with each other because it reflects the specific manifestation of knowledge subjects at different stages. Instead, it determines the corresponding stage based on the evolutionary mapping relationship and organizes and retains it according to the stage change relationship. For information differences caused by different document sources using different expression methods, such as the same relationship being expressed with different names, different descriptive angles, or different business terms in different documents, this application analyzes whether they belong to different expressions of the same knowledge content based on the corresponding relationship confirmation results and evolutionary mapping relationship, and performs association and merging processing on them. Through the above methods, this application avoids simply covering information formed at different stages and avoids misidentifying description differences as knowledge conflicts. Instead, it determines the retention, association, and organization of different information based on the evolution process of the knowledge subject, forming information retention relationships to support the construction of the target knowledge structure.

[0076] For example, in a device knowledge synthesis scenario, the structured knowledge object formed during the design phase includes the design model, design parameters, and initial composition relationships; the structured knowledge object formed during the testing phase includes test parameters, adjustment results, and test status; and the structured knowledge object formed during the operation phase includes asset number, operation parameters, and maintenance relationships. These pieces of information do not have simple substitution relationships; design parameters reflect the initial design state of the device, test parameters reflect the verification stage state, and operation parameters reflect the actual application state. Retaining only information from one stage would result in a loss of lifecycle information for the knowledge subject. Therefore, this application determines the stage role of different information based on evolutionary mapping relationships and determines the retention method for various types of information in the target knowledge structure.

[0077] Specifically, this application identifies the subject continuity relationship that needs to be maintained between knowledge objects at different stages and the stage change relationship formed with state changes based on the evolutionary mapping relationship, and determines the information retention relationship between knowledge objects based on the subject continuity relationship and the stage change relationship.

[0078] The subject continuity relationship is used to represent the common associations that need to be maintained between knowledge objects at different stages. It reflects the core relationships that remain consistent across different document stages or state conditions. Specifically, the subject continuity relationship does not require all descriptive content in knowledge objects at different stages to remain unchanged, but rather it is used to identify stability characteristics that can prove that different knowledge objects belong to the same knowledge subject. For example, in the equipment lifecycle management scenario, the same equipment may have different numbers, different operating parameters, and different state descriptions in the design, testing, and operation stages. However, the core structural components, functional uses, production objects, and key related component relationships of the equipment can usually persist across multiple stages. The aforementioned persistent information constitutes the subject continuity relationship. By determining the subject continuity relationship, this application can identify the basic associations that need to be retained between different knowledge objects, avoiding the splitting of the same knowledge subject into multiple independent knowledge objects due to stage differences or description differences during the synthesis process. The stage change relationship is used to represent the information changes that occur as the knowledge subject changes with stage changes. It reflects the state transitions, attribute adjustments, and changes in associations that occur during the lifecycle evolution of the same knowledge subject. Specifically, the stage change relationship is used not only to record the result of a certain attribute value changing from one stage to another, but also to describe the formation stage, direction of change, and cause of the change. For example, for a certain technical product, the theoretical parameters formed in the design stage may be adjusted according to experimental results in the testing stage, and form practical application parameters in the operation stage; the equipment composition relationship may involve component replacement with functional upgrades, and the operating state may change from the initial operating state to the maintenance state with changes in the usage environment. The above change information does not indicate that the knowledge subject has changed, but rather reflects the different manifestations of the same knowledge subject under different stage conditions. By combining the subject continuity relationship with the stage change relationship, this application can distinguish between stable knowledge content that needs to be maintained and changed knowledge content that needs to be organized according to stages, and further determine the information retention method between different structured knowledge objects, so that the target knowledge structure can maintain the consistency of the knowledge subject and fully express the evolution process of the knowledge subject with changes in time or state.

[0079] By combining the continuity relationship of the subject and the relationship of stage change, this application can determine the information retention strategy in different knowledge objects. For example, attribute information that can reflect the continuity of the knowledge subject's identity can be retained as stable information in the target knowledge structure; information that reflects the stage change process of the knowledge subject can be retained as stage information; and information that is formed due to changes in document description methods but whose actual correspondence remains the same can be uniformly organized according to evolutionary mapping relationships.

[0080] Preferably, based on the information retention relationship, this application further establishes a staged knowledge retention structure. This staged knowledge retention structure represents the organizational relationship between different knowledge contents after synthesis and adjudication. It does not simply store information from different document sources, but rather reorganizes the effective information in knowledge objects at different stages according to the evolution process of the knowledge subject. Specifically, when generating the staged knowledge retention structure, this application first determines the retention method of information in different structured knowledge objects within the target knowledge structure based on the information retention relationship, and then organizes different types of information hierarchically according to the continuity relationship and stage change relationship of the knowledge subject. Specifically, for information that can characterize the continuous existence of the same knowledge subject, such as subject identification relationships, core component relationships, functional association relationships, and stable business attributes, the stage knowledge retention structure associates and saves them as cross-stage shared information, enabling knowledge objects formed at different stages to establish connections based on a unified subject. For information generated as the state of the knowledge subject changes, such as design parameters, test results, operating status, and maintenance records, the stage knowledge retention structure records its formation stage, direction of change, and preceding and following relationships according to the corresponding stage change relationship, enabling information at different stages to be organized according to the evolution process of the knowledge subject. For differences in expression due to different document sources, such as the same attribute being described with different names or the same association being expressed with different business terms, the stage knowledge retention structure determines its corresponding relationship based on the evolutionary mapping relationship and the association confirmation result, and associates and merges information with the same knowledge meaning but different forms of expression. Therefore, the stage-based knowledge retention structure does not horizontally splice knowledge objects from multiple documents. Instead, while maintaining the continuity of the knowledge subject, it reorganizes the knowledge content formed at different stages according to its formation reasons, change process, and interrelationships. This allows the target knowledge structure to simultaneously express both the "unchanging parts" and the "parts that change with each stage" of the knowledge subject. For example, for the same equipment object, the structural parameters formed during the design stage, the performance adjustment information formed during the testing stage, and the status information formed during the operation stage are not independent fields that overlap in the stage-based knowledge retention structure. Instead, they are stored as related information from different stages in the equipment evolution process, thus forming a knowledge organization relationship that reflects the entire process of equipment from design formation, testing optimization to actual operation. By establishing a stage-based knowledge retention structure, this application realizes the transformation from simple knowledge content merging to structured synthesis based on knowledge evolution relationships, providing a foundation for the subsequent generation of a target knowledge structure that can fully express the continuous and changing relationships of the knowledge subject.

[0081] For example, for a specific product, the phase knowledge retention structure can simultaneously store the structural parameters formed during the product design phase, the performance change information formed during the testing phase, and the state information formed during the operation phase. It can also describe the change process between the above information through the phase relationships, enabling users to understand the development process of the product from design to actual operation based on the same knowledge structure.

[0082] Furthermore, based on the aforementioned stage knowledge retention structure, this application reorganizes the attribute descriptions, relationships, and state information between different structured knowledge objects to generate a target knowledge structure. This target knowledge structure represents the final knowledge organization result after cross-document knowledge synthesis. It not only includes knowledge content from different document sources but also retains the basis for the formation of relationships between different knowledge contents and the process of knowledge subject change. For example, in traditional knowledge synthesis methods, multiple descriptions of the same device in different documents may only form a single knowledge record containing multiple attribute fields, failing to distinguish the formation stages corresponding to different attributes. The target knowledge structure generated by this application can organize the attribute descriptions, state information, and relationships formed at different stages according to the evolutionary mapping relationship, enabling the complete expression of information changes formed by the same knowledge subject at different stages.

[0083] Preferably, based on the information retention relationship, this application further establishes a staged knowledge retention structure. This staged knowledge retention structure represents the organizational relationship between different knowledge contents after synthesis and adjudication. It does not simply store information from different document sources, but rather reorganizes the effective information in knowledge objects at different stages according to the knowledge subject's evolution process. Specifically, when establishing the staged knowledge retention structure, this application first determines the retention methods corresponding to attribute descriptions, relationships, and state information in different structured knowledge objects based on the information retention relationship, and then organizes different knowledge contents according to their roles in the knowledge subject's evolution process. For information that exists across multiple stages and reflects the continuity of the knowledge subject, this application treats it as a unified, retained part of the subject, enabling knowledge objects formed at different stages to establish stable connections around the same knowledge subject. For information that appears only at a specific stage and reflects changes in the state of the knowledge subject, this application determines its corresponding stage position based on evolutionary mapping relationships and retains it according to the order of stage changes, ensuring that information at different stages reflects the development process of the knowledge subject. For information whose expression differs due to different document sources, descriptive perspectives, or business scenarios, this application determines its correspondence based on the association confirmation results and information retention relationships, and organizes the correspondence between different expressions through association mapping. Thus, the stage-based knowledge retention structure does not simply arrange knowledge content from different document sources according to the order of acquisition, nor does it simply select information from a particular document as the final result. Instead, it reorganizes information formed at different stages while maintaining the continuity of the knowledge subject, ensuring that each type of knowledge content retains its corresponding role in the target knowledge structure. For example, for the same device, structural parameters in the design document describe the design state during the device formation stage, adjustment parameters in the test document describe the optimization process during the device verification stage, and operating parameters in the operation document describe the device's performance during actual application. Although these pieces of information correspond to different stages, they are all valid information in the evolution of the device. Through a stage knowledge retention structure, the information from these different stages can be associated according to the changes in the device, enabling the target knowledge structure to express both the current state of the device and trace its change path from design and testing to operation. Furthermore, the stage knowledge retention structure can also serve as a transitional structure in the synthesis and adjudication process, transforming the original document description relationships between different knowledge objects into organizational relationships based on the evolution of knowledge subjects. This avoids problems such as loss of historical stage information, inability to express state changes, or incorrect overwriting of association relationships caused by direct merging.

[0084] For example, if the target knowledge structure only retains the current running status information and lacks the information corresponding to the design and testing phases, it can be determined that the target knowledge structure does not fully express the change process of the knowledge subject. In this case, the corresponding evolution mapping relationship can be re-examined to determine whether there are any omissions in the phase association, and the information retention relationship can be further adjusted to supplement the missing phase information.

[0085] Through S5 processing, this application completes the transformation from knowledge object association confirmation to knowledge structure generation, enabling the difference information in the cross-document knowledge synthesis process to be rationally organized based on the knowledge subject evolution process. Compared with the approach of simply treating differences as conflicts and deleting or overwriting them, this application determines the information retention relationship through evolutionary mapping, so that the knowledge content formed at different stages can maintain a reasonable expression in the target knowledge structure, thereby realizing hierarchical conflict detection and adjudication for cross-document knowledge synthesis.

[0086] The following section uses a product lifecycle knowledge management scenario as an example to further illustrate the hierarchical conflict detection and adjudication method for cross-document knowledge synthesis proposed in this application.

[0087] In product lifecycle management, the same product typically goes through multiple stages, including design, testing, production, operation, and maintenance. Different business departments usually create corresponding documents for each stage. For example, the R&D department creates product design documents, the testing department creates experimental test documents, the production department creates production record documents, and the maintenance department creates operation and maintenance documents. Because different departments focus on different technical aspects, the documents at each stage typically use different descriptive methods to record the product object, resulting in the same product appearing as multiple distinct knowledge objects in different documents.

[0088] For example, for a certain model of industrial equipment, the design document records the first knowledge object, which includes the equipment design number, structural composition information, design parameters, and design purpose; the test document records the second knowledge object, which includes the test number, test environment, test indicators, and performance change information; and the maintenance document records the third knowledge object, which includes the asset number, operating status, maintenance records, and fault association information. If we directly perform text matching on these three knowledge objects, since the design number, test number, and asset number belong to different management stages and there is no direct relationship between the three numbers, it is easy to mistakenly identify the three knowledge objects as different objects. At the same time, the design parameters, test indicators, and operating parameters may also have numerical differences. If we judge solely based on whether the attribute values ​​are consistent, we may further misjudge information from different stages of the same equipment as conflicting information.

[0089] To address the above issues, this application first structures the knowledge descriptions from different document sources, forming corresponding structured knowledge objects. It then analyzes the subject description, attribute description, state description, and association description information within these three knowledge objects to identify differences in descriptions of the same knowledge subject across different document sources. Specifically, the subject description and attribute description information primarily reflect changes in the object characteristics of the knowledge subject; the state description information reflects changes in the state of the knowledge subject at different stages or under different conditions; and the association description information reflects changes in the way relationships are expressed between the knowledge subject and other objects. Based on the above analysis, this application further identifies the differences in object characteristics, state descriptions, and association expressions among different structured knowledge objects, and analyzes the causes of these differences and their impact on the judgment of associations between knowledge subjects.

[0090] During the analysis of object characteristic differences, this application found that although the design number in the first knowledge object, the test number in the second knowledge object, and the asset number in the third knowledge object have different names, they are all used to identify the management identifiers of the same equipment object in different business stages. Therefore, the change in number does not directly indicate a change in the object subject, but rather a difference in object characteristics due to the use of different identification methods in different business stages. Furthermore, this application analyzed the document stage, equipment composition relationship, and associated attributes corresponding to the different numbers to determine that the above-mentioned number changes are descriptive changes formed by the knowledge subject at different stages of its life cycle, rather than object conflicts arising between different knowledge subjects.

[0091] During the state description difference analysis, this application found that the "design state" in the first knowledge object, the "test state" in the second knowledge object, and the "operational state" in the third knowledge object correspond to different stages in the device lifecycle. While there are significant differences between these state descriptions, these differences reflect the state changes of the device from design to actual application.

[0092] During the analysis of the differences in related expressions, this application found that in the first knowledge object, a design composition relationship is formed between the equipment and its components; in the second knowledge object, a testing relationship is formed between the equipment and the test items; and in the third knowledge object, a maintenance relationship is formed between the equipment and the maintenance tasks. Although the related objects in the three stages are different, they all revolve around the same equipment entity.

[0093] Based on the above analysis, this application identifies the impact of different discrepancy factors on the determination of knowledge object associations and forms corresponding information masking features. For example, object identifier masking features corresponding to changes in management number, state masking features corresponding to changes in lifecycle stage, and association expression masking features corresponding to changes in business focus.

[0094] Furthermore, this application analyzes whether there is a potential relationship between the three knowledge objects based on the aforementioned information masking features. In this process, this application does not directly determine that the three knowledge objects belong to the same equipment entity, but rather extracts entity-retaining features that can persist across different stages. For example, the core component structure, functional uses, control relationships, and production line information of the equipment remain consistent across different document stages. Although the specific expression of the above information may differ in different documents, its descriptive objects and associated effects remain continuous. By analyzing the formation stages, manifestations, and associated effects corresponding to multiple entity-retaining features, this application forms cross-stage retention descriptive information and generates suspected associated information based on this descriptive information.

[0095] The suspected association information can indicate that: there may be a relationship between the first knowledge object and the second knowledge object that is the same device entity; there may be a relationship between the second knowledge object and the third knowledge object that is a state change relationship from the testing phase to the operation phase; and the first knowledge object, the second knowledge object, and the third knowledge object may form a stage association relationship within the same product lifecycle. However, the above suspected association information is still a relationship to be confirmed. This is because in practical applications, different devices may have the same components or the same functional purpose. For example, multiple devices in the same production line may all use the same control module. If the judgment is based solely on the relationship between control modules, it may lead to incorrect associations. Therefore, this application further determines relationship confirmation features based on the suspected association information.

[0096] Preferably, during the relationship confirmation process, this application can acquire the equipment composition structure retention relationship, functional purpose retention relationship, state change trajectory, and related object change as candidate confirmation features, and analyze the contribution of different candidate confirmation features to distinguishing different relationships. Specifically, this application can determine corresponding feature contribution description information for each candidate confirmation feature, wherein the feature contribution description information can be expressed in the form of contribution value, contribution level, or contribution range, etc., to characterize the supporting role of different candidate confirmation features in the current relationship confirmation process. For example, this application can determine the corresponding stable contribution value based on the retention frequency, continuous existence range, and consistency of change of the candidate confirmation feature in knowledge objects at different stages; determine the corresponding relationship differentiation contribution value based on the differentiation effect of the candidate confirmation feature in the relationship between the same knowledge subject and the similarity relationship between different knowledge subjects; and determine the corresponding relationship support contribution value based on the matching degree between the candidate confirmation feature and the relationship basis in the suspected relationship information. Further, this application can combine and analyze the above stable contribution value, relationship differentiation contribution value, and relationship support contribution value to form feature contribution description information to represent the comprehensive contribution level of the candidate confirmation feature. For example, regarding the structural relationship of equipment components, if this structural relationship exists simultaneously in the design, testing, and operation phases, and no other different equipment objects have the same structural combination, then this candidate confirmation feature can obtain a high stability contribution value and a relationship differentiation contribution value. For name-description relationships that appear only in a single document, due to the lack of cross-phase retention evidence and the inability to effectively distinguish similar objects, the corresponding contribution value is relatively low. By quantifying the contribution degree of different candidate confirmation features, this application can determine the combination of confirmation evidence that has a stronger supporting effect on the current relationship from multiple candidate confirmation features, and further generate relationship confirmation features used to distinguish different relationships.

[0097] For example, two equipment knowledge objects both contain "control module A," but one equipment corresponds to production line 1, and the other to production line 2. Judging solely based on the control module relationship would not effectively distinguish the two equipment objects. However, by further considering production line affiliation, changes in equipment structural parameters, and continuity of maintenance records, it can be determined that one of the knowledge objects has a higher probability of being associated with the other.

[0098] Preferably, after forming relationship confirmation features based on the combination of candidate confirmation features, this application generates a corresponding verification strategy and obtains the verification response results. Unlike judgment based on single feature similarity, the verification strategy in this application is used to select verification objects with discriminative capabilities for different candidate associations, and to determine whether the current candidate association can be established by analyzing the response differences generated by different verification objects under different association assumptions. Specifically, this application first determines candidate verification objects that can reflect the source of differences between knowledge objects based on the different candidate confirmation criteria contained in the relationship confirmation features, and establishes corresponding verification paths for different candidate verification objects. For example, when the equipment composition structure relationship is used as the main confirmation criterion, this application can select equipment components, component hierarchical relationships, and functional connection relationships as verification objects, and analyze whether the components in different documents can form a continuous change relationship according to the equipment evolution process; if the control module recorded in the design phase, the adjustment module recorded in the testing phase, and the control unit recorded in the operation phase can establish a corresponding relationship, it indicates that the structural changes between the three knowledge objects conform to the equipment evolution law, rather than corresponding to different equipment bodies. Furthermore, when the state change trajectory is used as the primary confirmation basis, this application can select equipment lifecycle states, performance parameter changes, and operational status records as verification objects to analyze whether there is a change process that conforms to the actual evolution law between the design state, test state, and operational state. For example, whether the theoretical parameters formed in the design stage can be mapped to the actual parameters in the operational stage through the adjustment process in the test stage. When the changes in related objects are used as the primary confirmation basis, this application can select the changes in the relationship between the knowledge subject and external related objects as verification objects to determine whether the changes in related objects in different stages conform to the development process of the knowledge subject. For example, whether the change in the relationship between the equipment and components in the design stage to maintenance objects in the operational stage can be reasonably explained through the stage change relationship. Furthermore, this application can obtain corresponding verification response results for multiple candidate verification objects and analyze the degree of matching between the verification response results under different verification paths and the direction of the relationship to be confirmed in the suspected related information to form verification differentiation description information. For example, when a verification path enables two knowledge objects to form continuous state changes, stable subject-maintained relationships, and reasonable association migration relationships, the verification response result corresponding to that path can effectively support the association relationship of the same knowledge subject. Conversely, when a verification path can only prove that two knowledge objects have partial descriptive similarity but cannot form a stage-continuous relationship, it indicates that the verification path is insufficient to support that the two belong to the same knowledge subject. Through the above verification strategy, this application does not directly determine the association result based on existing features, but actively constructs a verification process for candidate relationships, enabling different association hypotheses to produce distinguishable response results under the corresponding verification path, thereby improving the accuracy of cross-document knowledge object association confirmation.When the verification results show that multiple knowledge objects can form a continuous changing relationship, it indicates that the differences between different knowledge objects are more likely to come from the performance differences formed by the same knowledge subject at different stages, rather than the erroneous association between different subjects.

[0099] Building upon this foundation, this application further establishes an evolutionary mapping relationship. For example, the following transformation relationship can be established: Design Knowledge Object → Testing Knowledge Object → Operation and Maintenance Knowledge Object. Here, the design parameters in the design knowledge object correspond to information from the product formation stage, the testing indicators in the testing knowledge object correspond to information from the product verification stage, and the operational status in the operation and maintenance knowledge object corresponds to information from the product application stage. Through this evolutionary mapping relationship, this application can explain the reasons for differences between different knowledge objects. For example, differences between design parameters and operational parameters do not necessarily indicate that one parameter is incorrect, but rather that they correspond to the product design stage and the actual operation stage, respectively; different equipment numbers do not necessarily indicate different equipment, but rather represent changes in identification due to different management stages. Finally, this application performs a synthetic adjudication process based on the evolutionary mapping relationship.

[0100] In this process, this application determines the information that needs to be retained based on the continuity of the main body, such as the equipment main body identifier, core component relationships, and functional usage information; it determines the information that needs to be retained according to the stage change relationship, such as design parameters, test results, and operating status information; for information that is formed only due to changes in the description method, it is uniformly organized according to the corresponding relationship, thereby forming the target knowledge structure. This target knowledge structure not only includes knowledge content from different document sources, but also expresses the formation stages, change relationships, and correlation basis between different knowledge contents. For example, when querying information about a certain equipment, not only can the current operating status be obtained, but the complete change process of the equipment from design formation, testing and verification to operation and maintenance can also be traced.

[0101] As can be seen from the above embodiments, the hierarchical conflict detection in this application does not simply judge the inconsistency on the surface of knowledge content, but analyzes the source of difference, the basis for association, and the process of change layer by layer; at the same time, the adjudication process in this application does not simply select a certain knowledge content as the correct result, but determines the retention method of different knowledge content in the target knowledge structure according to the evolution relationship of knowledge subjects, thereby achieving accurate organization for cross-document knowledge synthesis.

[0102] This embodiment further illustrates the hierarchical conflict detection process in this application. It should be noted that in the cross-document knowledge synthesis process, the inconsistencies between different document sources have different causes. If all inconsistencies are treated as conflicts, the underlying formation mechanisms of these differences may be overlooked, making it difficult to accurately determine the processing method in subsequent knowledge synthesis processes. Therefore, this application performs hierarchical conflict detection based on the hierarchical relationships arising from differences between different knowledge objects. Specifically, the hierarchical conflict detection in this application does not simply divide conflicts according to the number, degree, or level of conflict, but rather analyzes potential conflict relationships layer by layer based on different levels of association in the formation process of knowledge object differences.

[0103] In the first-level conflict detection process, this application analyzes the differences in object characteristics between different structured knowledge objects to identify surface conflicts arising from changes in object description methods. In practical applications, different document sources may use different identification methods to describe the same knowledge subject. For example, the same equipment may be described using a research and development number during the R&D phase, a product number during the production phase, and an asset number during the operation and maintenance phase. Because different numbering systems are used at different stages, knowledge objects in different documents differ at the object identification level. If matching is done directly based on object identification, different numbers may be considered to correspond to different knowledge subjects, leading to incorrect conflict judgments. Therefore, this application first analyzes the formation environment of the differences in object characteristics to determine whether the differences conform to the identification change pattern of the knowledge subject in different document stages. When the differences in object characteristics can be explained through stage change relationships, the differences are treated as associative differences rather than directly judged as knowledge conflicts.

[0104] For example, a piece of equipment might be recorded as "Equipment R&D Number X01" in the R&D document, "Product Number P01" in the production document, and "Asset Number A01" in the maintenance document. Although the three numbers are completely different, by analyzing the corresponding document stages, associated attributes, and equipment composition relationships, it can be determined that all three numbers describe the same equipment entity. In this case, the numbering variation is due to differences in object characteristics, rather than a conflict of entities.

[0105] In the second-layer conflict detection process, this application further analyzes the differences in state descriptions between different structured knowledge objects to identify stage conflicts arising from changes in the state of the knowledge subject. In practical applications, the same knowledge subject may be in different states throughout its lifecycle. For example, equipment may go through design, manufacturing, testing, operation, and maintenance states. The state descriptions corresponding to different stages may be completely different, or even contradictory. For example, a design document records that a certain piece of equipment is "not yet in operation," while a maintenance document records that the equipment is "operating." If analyzed without considering the time stage, the two descriptions are clearly inconsistent. However, if analyzed in conjunction with the document formation stage, it can be found that they correspond to different stages in the equipment's lifecycle and there is no factual conflict. Therefore, this application analyzes the impact of stage changes corresponding to differences in state descriptions based on information masking features, and combines subject retention features to determine whether different state descriptions belong to continuous changes of the same knowledge subject. When different state descriptions can form a reasonable change process, this application identifies them as stage change relationships and retains their change information; when different state descriptions cannot form a continuous change relationship, the application proceeds to the subsequent relationship confirmation and verification process.

[0106] In the third-layer conflict detection process, this application further analyzes the differences in the expression of associations between different structured knowledge objects to identify relationship conflicts arising from changes in descriptive perspective. Because different document sources have different recording purposes, the same knowledge subject may form different types of associations with different related objects. For example, in a design document, a device may form a compositional relationship with a "control module"; in a test document, the device may form a verification relationship with a "test project"; and in a maintenance document, the device may form a management relationship with a "maintenance task." If only the names of the related objects are compared, it might be assumed that there are relationship conflicts between knowledge objects in different documents. However, in reality, the above-mentioned associations reflect the different roles of the knowledge subject at different stages. Therefore, this application further analyzes the relationship change process corresponding to the differences in association expression to determine whether different associations can be explained through the evolution process of the knowledge subject.

[0107] Through the above three levels of analysis, this application achieves a step-by-step detection process from surface differences to deep relationships. The first level is used to determine whether differences in object descriptions originate from different identification methods; the second level is used to determine whether differences in state descriptions originate from changes in the knowledge subject stage; and the third level is used to determine whether differences in association expression originate from changes in relationships under different application stages.

[0108] Based on the aforementioned hierarchical detection process, this application avoids simply classifying all knowledge differences as conflicts. For example, for the same equipment, "the R&D number differs from the asset number" is an object-level difference; "the design state differs from the operating state" is a state-level difference; and "the composition relationship differs from the maintenance relationship" is an association-level difference. All of these differences can be explained through the knowledge subject evolution process and further proceed to the synthesis and adjudication stage.

[0109] Furthermore, after completing the hierarchical conflict detection, this application forms corresponding adjudication criteria based on the detection results. Specifically, if a certain difference relationship can be explained through an evolutionary mapping relationship, it indicates that the difference belongs to the information difference formed during the change of the knowledge subject, and it needs to be retained according to its stage role in the subsequent adjudication process; if a certain difference relationship cannot form a reasonable evolutionary relationship, it can be treated as an abnormal relationship for further processing. Therefore, the hierarchical conflict detection in this application is not intended to eliminate all differences, but to identify the causes of different differences, so that the subsequent adjudication process can make judgments based on the laws of change of the knowledge subject.

[0110] In summary, this application addresses the challenges of difficult association confirmation, conflict differentiation, and accurate representation of the evolutionary process in knowledge synthesis results caused by differences in description stages, expression methods, and state conditions among knowledge objects formed from different document sources for the same knowledge subject. This is achieved by constructing a hierarchical conflict detection and adjudication mechanism for cross-document knowledge synthesis. First, this application analyzes the differences in object features, state descriptions, and association expressions formed when describing the same knowledge subject in different document sources. Information masking features are extracted to characterize the causes of these differences, enabling the cross-document knowledge processing to distinguish between superficial differences arising from variations in description methods and genuine conflicts arising from different knowledge subjects. This avoids erroneous associations based solely on text consistency or single feature matching. Furthermore, based on information masking features, this application further identifies subject features maintained across different knowledge objects at different stages, forming suspected association information. Candidate association relationships are actively verified through relationship confirmation features and verification strategies. This transforms the association judgment between different knowledge objects from simple matching into a relationship confirmation process based on multi-dimensional verification criteria, improving the reliability of cross-document knowledge object association analysis.

[0111] Furthermore, based on the verification results, this application establishes an evolutionary mapping relationship between different structured knowledge objects. This transforms knowledge objects formed at different document stages from independent information units into an associative structure that reflects the changing process of the knowledge subject. This allows for the reasonable explanation of attribute changes, state changes, and association changes, avoiding the misidentification of normal changes in the knowledge subject's lifecycle as knowledge conflicts. Simultaneously, this application determines the subject's continuity relationship and stage change relationship based on the evolutionary mapping relationship, further forming information retention relationships and stage knowledge retention structures. This enables the knowledge content formed at different stages to be organized and retained according to its role in the knowledge subject's evolution, rather than using simple overwriting, deletion, or merging methods to handle differing information. Therefore, the target knowledge structure generated by this application not only retains the effective information formed by the knowledge subject at different stages but also fully expresses the continuity relationship and change process of the knowledge subject, improving the completeness, accuracy, and interpretability of the cross-document knowledge synthesis results.

[0112] Furthermore, this application transforms the handling of differences between knowledge objects from "judging conflicts after discovering differences" to "analyzing the source of differences, confirming the relationship, explaining the change process, and determining the information retention method" through a hierarchical conflict detection and adjudication process. This approach can adapt to the knowledge fusion needs of different sources, different stages, and different description systems, reduce knowledge loss, erroneous fusion, and misjudgment of relationships caused by document heterogeneity, and provide a more stable and reliable technical implementation path for cross-document knowledge synthesis in complex scenarios.

[0113] Please see Figure 2 This application also provides a hierarchical conflict detection and adjudication device for cross-document knowledge synthesis, comprising: The knowledge object acquisition module is used to acquire structured knowledge objects from different document sources. It extracts the corresponding information masking features based on the differences in object features, state descriptions, and related expressions when describing the same knowledge subject in different document sources. The association analysis module is used to analyze the association probability between different structured knowledge objects based on the information masking features, and generate suspected association information. The suspected association information is used to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis. The relationship verification module is used to determine relationship confirmation features for distinguishing different relationships based on the suspected association information, and to generate corresponding verification strategies based on the relationship confirmation features in order to obtain verification results for the relationships between the structured knowledge objects. An evolutionary analysis module is used to determine the evolutionary mapping relationship between different structured knowledge objects based on the verification results. The evolutionary mapping relationship is used to characterize the knowledge object change relationship formed by the same knowledge subject in different document stages or under different state conditions. The knowledge reconstruction module is used to perform synthetic adjudication processing on the attribute descriptions, associations and state information between different structured knowledge objects based on the evolutionary mapping relationship, determine the information retention relationship between knowledge objects, and generate the target knowledge structure.

[0114] like Figure 3 As shown, an electronic device according to an embodiment of this application includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the hierarchical conflict detection and adjudication method for cross-document knowledge synthesis as described above in this application.

[0115] like Figure 4 As shown, this is a computer-readable storage medium according to an embodiment of the present application. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the hierarchical conflict detection and adjudication method for cross-document knowledge synthesis as described above in this application.

[0116] The above Figures 2-4 For an exemplary description, please refer to the above. Figure 1 This will not be elaborated upon here.

Claims

1. A hierarchical conflict detection and adjudication method for cross-document knowledge synthesis, characterized in that, include: Obtain structured knowledge objects from different document sources, and extract corresponding information masking features based on the differences in object features, state descriptions, and related expressions formed when describing the same knowledge subject in different document sources; Based on the information masking features, the possibility of association between different structured knowledge objects is analyzed to generate suspected association information. The suspected association information is used to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis. Based on the suspected association information, relationship confirmation features are determined to distinguish different association relationships, and corresponding verification strategies are generated based on the relationship confirmation features to obtain verification results for the association relationships between the structured knowledge objects. Based on the verification results, the evolutionary mapping relationship between different structured knowledge objects is determined. The evolutionary mapping relationship is used to characterize the knowledge object change relationship formed by the same knowledge subject in different document stages or under different state conditions. Based on the evolutionary mapping relationship, the attribute descriptions, associations and state information between different structured knowledge objects are synthesized and adjudicated to determine the information retention relationship between knowledge objects and generate the target knowledge structure.

2. The hierarchical conflict detection and adjudication method for cross-document knowledge synthesis according to claim 1, characterized in that, The steps for obtaining structured knowledge objects from different document sources include: Perform subject association analysis on the knowledge description content in different document sources to identify the corresponding knowledge subject scope between different document sources; Based on the scope of the knowledge subject, analyze the correspondence between attribute descriptions, status descriptions, and association descriptions formed for the same knowledge subject in different document sources, and form a knowledge object description structure that includes subject correspondence and stage description relationships.

3. The hierarchical conflict detection and adjudication method for cross-document knowledge synthesis according to claim 1, characterized in that, The step of extracting corresponding information masking features based on differences in object features, state descriptions, and association expressions formed when describing the same knowledge subject in different document sources includes: Based on the differences in object characteristics, state descriptions, and related expressions, the scope of the related influence of different difference factors is analyzed to form information describing the impact of differences. Based on the difference in the description information, the occlusion effect of knowledge description at different stages on the judgment of the relationship between knowledge subjects is identified, and the information occlusion feature is formed according to the occlusion effect.

4. The hierarchical conflict detection and adjudication method for cross-document knowledge synthesis according to claim 3, characterized in that, The step of extracting the corresponding information masking features further includes: Based on the description of the differences in the impact of the differences, we analyze the occlusion patterns corresponding to the differences in the characteristics, state descriptions, and related expressions of different objects. Based on the synergistic relationship between different forms of occlusion, an occlusion feature description is formed to characterize the reasons for the concealment of the relationship between knowledge subjects, and the information occlusion feature is improved based on the occlusion feature description.

5. A hierarchical conflict detection and adjudication method for cross-document knowledge synthesis according to claim 1, characterized in that, The step of analyzing the likelihood of association between different structured knowledge objects based on the information masking features and generating suspected association information includes: Based on the stage judgment impact description corresponding to the information masking features, identify the subject retention features that remain across state changes between different structured knowledge objects; The formation stages, manifestations, and associated effects of the characteristics of multiple subjects are analyzed to form cross-stage retention description information. Based on the cross-stage preservation description information, suspected association information is generated to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis.

6. A hierarchical conflict detection and adjudication method for cross-document knowledge synthesis according to claim 5, characterized in that, The step of generating the suspected association information based on the cross-stage preserved description information to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis includes: Based on the source stage and associated role of the cross-stage retention description information, multiple candidate association explanations are constructed. The support status of multiple candidate related interpretations in knowledge description at different stages is analyzed to form the trajectory of related interpretation changes; Based on the change trajectory of the association interpretation, the association basis corresponding to different candidate association interpretations is organized to form the suspected association information.

7. A hierarchical conflict detection and adjudication method for cross-document knowledge synthesis according to claim 1, characterized in that, The step of determining the relationship confirmation features used to distinguish different relationships based on the suspected association information includes: Based on the direction of the relationship to be confirmed corresponding to the suspected related information, multiple candidate confirmation features are obtained; The retention, distinguishability, and correlation support of different candidate confirmation features before and after changes in the state of the knowledge subject are analyzed to form feature contribution description information. The combined contribution relationship between different candidate confirmation features is analyzed based on the feature contribution description information to form a confirmation basis combination, and the relationship confirmation feature is determined based on the confirmation basis combination.

8. A hierarchical conflict detection and adjudication device for cross-document knowledge synthesis, characterized in that, include: The knowledge object acquisition module is used to acquire structured knowledge objects from different document sources. It extracts the corresponding information masking features based on the differences in object features, state descriptions, and related expressions when describing the same knowledge subject in different document sources. The association analysis module is used to analyze the association probability between different structured knowledge objects based on the information masking features, and generate suspected association information. The suspected association information is used to characterize the direction of the relationship to be confirmed between different structured knowledge objects and the corresponding association basis. The relationship verification module is used to determine relationship confirmation features for distinguishing different relationships based on the suspected association information, and to generate corresponding verification strategies based on the relationship confirmation features in order to obtain verification results for the relationships between the structured knowledge objects. An evolutionary analysis module is used to determine the evolutionary mapping relationship between different structured knowledge objects based on the verification results. The evolutionary mapping relationship is used to characterize the knowledge object change relationship formed by the same knowledge subject in different document stages or under different state conditions. The knowledge reconstruction module is used to perform synthetic adjudication processing on the attribute descriptions, associations and state information between different structured knowledge objects based on the evolutionary mapping relationship, determine the information retention relationship between knowledge objects, and generate the target knowledge structure.

9. An electronic device, characterized in that, The electronic device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the hierarchical conflict detection and adjudication method for cross-document knowledge synthesis as claimed in any one of claims 1 to 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 steps of the hierarchical conflict detection and adjudication method for cross-document knowledge synthesis as described in any one of claims 1 to 7.

Images