Content form perception-based rule document parsing and extraction method and device, computer device and storage medium

Abstract

The application discloses a content form perception rule document analysis and extraction method. The method acquires multiple document content units in a rule document, extracts structural features, text expression features and context association features, determines corresponding form attribution strength and a target form mode based on multiple candidate form modes, and obtains form semantic relationships according to a differential processing strategy corresponding to the target form mode. Further, the semantic association features between different content forms are fused to generate an associated knowledge structure, and a retrieval adaptation unit is generated based on the associated knowledge structure, and a retrieval result is determined according to query semantic features and query context features corresponding to a retrieval request. The application can perform differential processing on different content forms such as rules, definitions, processes, cases and aids, establish knowledge association relationships between contents, and improve the accuracy and applicability of rule document processing and retrieval.

Description

Technical Field

[0001] This application relates to the field of information retrieval technology, and in particular to a method, apparatus, computer device, and storage medium for content-form-aware rule-based document parsing and extraction. Background Technology

[0002] With the widespread application of business rules, management standards, operational procedures, and policy documents in various business scenarios, rule documents have gradually become an important information carrier for business constraints, execution requirements, and processing basis. These documents typically have strong structured characteristics, including not only rule clauses defining behavioral requirements, but also definitions explaining professional concepts, process descriptions of execution procedures, and case studies illustrating specific applications. Although different types of content coexist within the same rule document, their information expression functions differ.

[0003] Current methods for processing rule documents typically focus on parsing the text content itself or matching information based on keywords and semantic similarity. However, the different contents in a rule document do not function solely through their literal meaning. For example, the scope of application of a rule clause may need to be understood in conjunction with its corresponding definition, and the applicable conditions required by a rule may need to be determined in conjunction with the specific circumstances in a case. Furthermore, some auxiliary descriptive content, while not directly constituting rule requirements, may affect the overall understanding of the rule content. Therefore, processing based solely on textual representation can easily lead to insufficient reflection of the relationships between different contents, resulting in inaccurate matching scope and insufficient evidence of association during rule understanding and retrieval.

[0004] Therefore, a method is needed that can identify the functions of different content in a rule document and differentiate them according to their characteristics, so that different content in the rule document can form effective relationships and support a more accurate retrieval process based on these relationships. Summary of the Invention

[0005] This application aims to address the problem that, in the existing rule document processing, different content has different information expression functions, making it difficult to accurately identify the relationship between content functions when using a unified text processing method.

[0006] Specifically, existing technologies typically focus on the matching relationships between textual content in rule documents, making it difficult to further identify the knowledge roles of rule content, definition content, process content, case content, and auxiliary content. This results in the inability to effectively establish relationships between different types of content. For example, when searching solely based on the text of rule clauses, it may be impossible to accurately determine the scope of the object by combining the definition content, or to understand the applicable conditions of the rule by combining the case content, thus affecting the consistency between search results and actual application needs.

[0007] To this end, this application provides a content-form-aware rule document parsing and extraction method. By identifying the form patterns corresponding to different content in the rule document, adopting corresponding processing strategies according to different form patterns, forming knowledge association relationships between different content, and generating retrieval adaptation units based on the associated knowledge structure, the retrieval process can combine query semantic features and query context features to determine the corresponding results.

[0008] To achieve the above objectives, this application provides a content-format-aware rule-based document parsing and extraction method, including: Obtain multiple document content units from the rule document, and extract the structural features, textual expression features, and contextual association features corresponding to each document content unit; Based on the structural features, text expression features, and context association features, the semantic attribution degree of each document content unit in multiple candidate morphological modes is analyzed to determine the morphological attribution strength of each document content unit, and the target morphological mode corresponding to the document content unit is determined according to the morphological attribution strength. Based on the differentiated processing strategy corresponding to the target morphological pattern, the document content unit is subjected to content morphological semantic representation to obtain the corresponding morphological semantic relationship; Based on the morphological semantic relationships corresponding to different document content units, the semantic association features between different content forms are fused and analyzed to determine the knowledge association relationships corresponding to different content forms and generate the associated knowledge structure corresponding to the rule document. Based on the aforementioned knowledge structure, a retrieval adaptation unit is generated. In response to a retrieval request, the retrieval adaptation unit is matched with the query semantic features and query context features corresponding to the retrieval request, and the corresponding retrieval results are extracted.

[0009] Optionally, the acquisition of multiple document content units in the rule document includes: Based on the chapter structure, clause relationships, and text context relationships in the rule document, the content of the rule document is analyzed to identify multiple candidate semantic fragments; For each candidate semantic fragment, based on its scope of application in the rule document and its degree of association with other semantic fragments, a document content unit is determined to carry the same content form expression.

[0010] Optionally, the extraction of structural features, textual expression features, and contextual association features corresponding to each document content unit includes: Based on the organizational position, clause level, and referencing relationship of the document content unit in the rule document, determine the structural features used to reflect the role of the content; Based on the restrictive expressions, definitional expressions, procedural expressions, and contextual expressions in the document content units, determine the textual expression features that reflect the semantic function of the content; Based on the interpretive, constraint, and applicability relationships between the document content units and other content units, contextual association features that reflect the content association effect are determined.

[0011] Optionally, determining the morphological attribution strength corresponding to each of the document content units includes: Based on the content expression rules corresponding to each candidate morphological pattern, the structural features, text expression features and contextual association features corresponding to the document content unit are matched and analyzed to determine the morphological matching result between the document content unit and each candidate morphological pattern. Based on the morphological matching results, further analyze the semantic role, association scope, and knowledge expression purpose of the document content unit in the rule document, and determine the morphological attribution strength used to characterize the degree to which the document content unit undertakes the corresponding content morphological function.

[0012] Optionally, determining the target morphological pattern corresponding to the document content unit based on the morphological attribution strength includes: When the document content unit corresponds to multiple candidate morphological patterns, multiple candidate semantic interpretation paths are determined according to the morphological attribution strength corresponding to each candidate morphological pattern. For each of the candidate semantic interpretation paths, analyze the corresponding knowledge association results and their impact on the subsequent retrieval process; Based on the association effects corresponding to different candidate semantic interpretation paths, the target morphological pattern corresponding to the document content unit is determined, and subsequent processing is performed according to the processing path corresponding to the target morphological pattern.

[0013] Optionally, processing the document content unit according to the differentiated processing path corresponding to the target form pattern includes: When the target form mode is a rule form mode, the restriction objects, restriction behaviors, triggering conditions and exception conditions in the document content unit are analyzed for correlation to determine the constraint range corresponding to the rule. The applicable boundaries of the rule are determined based on the reference and applicability relationships between the constraint scope and other related content in the rule document; Based on the applicable boundaries, rule retrieval association conditions are established so that the corresponding rule content can be matched according to the target object and applicable conditions during the subsequent retrieval process.

[0014] Optionally, processing the document content unit according to the differentiated processing path corresponding to the target form pattern includes: When the target morphological pattern is a defined morphological pattern, the concept name, explanation, limiting conditions and associated objects in the document content unit are parsed to determine the semantic limitation scope of the corresponding concept; Based on the semantic definition and the reference relationship between the concept and other content in the rule document that involves the concept, determine the scope of applicability of the concept in different rule scenarios; Based on the aforementioned scope of application, the object matching conditions in the retrieval process are modified to avoid erroneous retrieval results caused by the expansion of the concept scope.

[0015] Optionally, processing the document content unit according to the differentiated processing path corresponding to the target form pattern includes: When the target form mode is a process form mode, the association between process nodes is determined based on the execution subject, execution action, action sequence and condition transformation relationship in the document content unit; Based on the relationships between the process nodes, process semantic relationships are formed to characterize the changing paths of the rule execution process; When responding to a query request that includes process status or processing stage, the corresponding process node and associated rule content are determined based on the process semantic relationship.

[0016] Optionally, processing the document content unit according to the differentiated processing path corresponding to the target form pattern includes: When the target form mode is a case form mode, the event subject, behavior purpose, occurrence conditions, environmental factors and processing results in the document content unit are analyzed to determine the application context of the corresponding event. Based on the correlation between the application scenarios and the rule application scope corresponding to the rule form patterns, identify the scenario change factors that affect the rule application results; Contextual association information is established based on the aforementioned contextual change factors to describe the reasons why the same rule produces different processing results under different application conditions.

[0017] Optionally, after establishing the context association information, the method further includes: Based on the context change factors in the context association information, the triggering conditions and exception conditions in the scope of application of the rule are adjusted to determine the context adaptation relationship of the rule. The rule retrieval conditions are updated based on the rule context adaptation relationship, so that the corresponding rule content can be filtered according to the actual context of the query object during the query process. The final retrieval result is determined based on the matching result between the rule context adaptation relationship and the query context.

[0018] Optionally, processing the document content unit according to the differentiated processing path corresponding to the target form pattern includes: When the target form mode is an auxiliary form mode, for document content units that do not directly form rule constraint relationships, concept limitation relationships, or process execution relationships, the reference relationships, positional association relationships, and semantic inheritance relationships between the document content unit and other document content units are obtained. Based on the aforementioned citation relationships, positional association relationships, and semantic continuity relationships, analyze whether the document content unit participates in the formation of rule application conditions, context judgment criteria, or knowledge association paths; When it is determined that the document content unit affects the rule understanding process, the corresponding implicit correlation factors are identified.

[0019] Optionally, after determining the implicit correlation factor, the method further includes: Based on the relationship between the implicit related factors and the rule application boundaries, concept limitation scope, and context-related information, determine the knowledge supplementation path corresponding to the implicit related factors; Adjust the association method between the corresponding document content units according to the knowledge supplementation path, and associate the implicit association factors with the corresponding retrieval knowledge units; When responding to a query request, the search results are further filtered based on the degree of matching between the implicit related factors and the query context.

[0020] Optionally, determining the knowledge association relationship based on the morphological semantic relationship corresponding to different morphological patterns includes: Obtain the relationships between rule semantic relationships, concept semantic relationships, process semantic relationships, contextual association information, and implicit association factors; Based on the scope of application of different semantic relationships and the correspondence between related objects, a comprehensive knowledge association relationship is determined to describe the association path between rule content, application conditions and knowledge basis. Based on the comprehensive knowledge relationships, a related knowledge structure is constructed to support rule-based retrieval.

[0021] Optionally, after constructing the associated knowledge structure based on the comprehensive knowledge relationship, the method further includes: When content change information in the rule document is detected, the set of affected morphological semantic relationships is determined based on the document location, morphological pattern, and associated path corresponding to the content change information. Based on the set of affected morphological semantic relationships, the scope of influence of the relationships among the rule semantic relationships, concept semantic relationships, process semantic relationships, contextual association information and implicit association factors is analyzed to determine the association knowledge paths that need to be adjusted. Based on the dependencies corresponding to the associated knowledge paths, the associated objects, relationships, and retrieval adaptation relationships in the associated knowledge structure are updated synchronously.

[0022] Optionally, the step of generating a retrieval adaptation unit based on the associated knowledge structure includes: Based on the association paths between rule semantic relationships, concept semantic relationships, process semantic relationships, and contextual association information in the comprehensive knowledge association relationship, complete semantic units that maintain the complete knowledge expression scope are determined. Based on the local knowledge associations in the complete semantic unit corresponding to different query requirements, a local semantic unit for performing retrieval matching is generated; Establish a mapping relationship between the local semantic units and the complete semantic units so that the corresponding complete knowledge context can be obtained based on the local matching results during the retrieval process.

[0023] Optionally, determining the search results based on the search request includes: The retrieval request is semantically parsed to determine the target object, query purpose, and application context corresponding to the retrieval request. Based on the target object, the query purpose, and the application context, the candidate search results are determined by matching the rule semantic relationship, contextual association information, and implicit association factors corresponding to the local semantic unit. Based on the mapping relationship between the local semantic units and the complete semantic units, the complete knowledge context corresponding to the candidate search results is associated and obtained to generate the final search results.

[0024] Optionally, after generating the final search results, the method further includes: Based on the target rule knowledge corresponding to the final retrieval result, the rule semantic relationship, concept semantic relationship, process semantic relationship, contextual association information and implicit association factors associated with the target rule knowledge are traced back along the associated knowledge structure. Based on the association paths between the rule semantic relationships, the concept semantic relationships, the process semantic relationships, the contextual association information, and the implicit association factors, the knowledge basis chain supporting the final retrieval results is determined; Based on the knowledge chain, the corresponding document content unit is obtained, and associated traceability information is generated to characterize the basis for the formation of the final search result.

[0025] To achieve the above objectives, this application also provides a content-format-aware rule-based document parsing and extraction apparatus, comprising: The document feature acquisition module is used to acquire multiple document content units in a rule document and extract the structural features, text expression features, and contextual features corresponding to each document content unit. The morphological pattern determination module is used to analyze the semantic belonging degree of each document content unit in multiple candidate morphological patterns based on the structural features, the text expression features and the context association features, determine the morphological belonging strength of each document content unit, and determine the target morphological pattern of the document content unit according to the morphological belonging strength. The differentiation processing module is used to perform content morphological semantic representation on the document content unit according to the differentiation processing strategy corresponding to the target morphological pattern, and obtain the corresponding morphological semantic relationship. The knowledge association construction module is used to perform fusion analysis on the semantic association features between different content forms based on the morphological semantic relationships corresponding to different document content units, determine the knowledge association relationships corresponding to different content forms, and generate the associated knowledge structure corresponding to the rule document. The retrieval processing module is used to generate a retrieval adaptation unit based on the associated knowledge structure, and in response to the retrieval request, match the retrieval adaptation unit with the query semantic features and query context features corresponding to the retrieval request, and extract the corresponding retrieval results.

[0026] 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.

[0027] 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.

[0028] Compared with the prior art, this application has at least the following beneficial effects: This application performs multi-dimensional feature analysis on document content units in rule documents, which can identify the knowledge roles undertaken by different content and adopt differentiated processing methods according to different form patterns. This avoids processing all content in the rule document in a uniform text format and improves the accuracy of understanding the rule document content.

[0029] Furthermore, this application constructs knowledge associations between different forms of content, enabling rule content, definition content, process content, case content, and auxiliary content to jointly participate in the rule understanding process. Specifically, the specific conditions in the case content can be used to analyze the scope of rule application, and the related information in the auxiliary content can be used to refine the knowledge associations, thereby avoiding the information loss caused by relying solely on explicit rule text in traditional retrieval processes.

[0030] Furthermore, this application generates retrieval adaptation units based on associated knowledge structures and performs matching by combining query semantic features and query context features. This enables the retrieval process to not only focus on the query text itself, but also to determine rule content that better meets the needs based on actual application conditions, thereby improving the accuracy and applicability of rule document retrieval results. Attached Figure Description

[0031] Figure 1 Flowchart of the rule document parsing and extraction method provided in the embodiments of this application; Figure 2 A schematic diagram of the rule document parsing and extraction device provided in the embodiments 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

[0032] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] like Figure 1 As shown, this application provides a content-form-aware method for parsing and extracting rule documents. This method can be applied to scenarios involving content analysis, knowledge organization, and information retrieval of rule documents. The rule documents may include, but are not limited to, business management specifications, operating procedures, institutional documents, legal rules, technical specifications, and other document types containing rule constraints, execution requirements, or explanatory content.

[0034] In practical applications, rule documents are typically not composed of a single type of text content, but rather a combination of various elements with different expressive purposes. For example, some content clarifies the behavioral requirements that objects must adhere to, some explains related concepts or defines the scope of the object, some describes the business execution process, some illustrates the applicability of the rule through specific cases, and some, while not directly specifying behavioral requirements, provides assistance in rule understanding, content association, and retrieval judgment. All of these different contents exist within the same rule document, but because they serve different informational functions, they cannot be processed in a completely uniform manner.

[0035] Preferably, the content form awareness in this application refers to identifying the form patterns corresponding to different content based on the knowledge representation function of the rule document content, and applying corresponding processing methods to different form patterns. The form patterns characterize the type of role of the rule document content in the knowledge organization process, including rule form, definition form, process form, case form, and auxiliary form. By differentiating the processing of content corresponding to different form patterns, different content in the rule document can form semantic relationships matching their functions, further supporting subsequent construction of related knowledge structures and retrieval processing.

[0036] Specifically, such as Figure 1 As shown, the content-form-aware rule document parsing and extraction method provided in this application includes the following steps: S1, obtain multiple document content units in the rule document, and extract the structural features, text expression features and context association features corresponding to each document content unit.

[0037] In this embodiment, the rule document to be processed is first obtained, and the rule document is parsed to determine multiple document content units. The document content unit is used to represent a content object in the rule document that has a relatively complete knowledge expression function. It can correspond to a complete clause, a group of related paragraphs, a case description fragment, or other text range that can independently express a certain knowledge function.

[0038] Preferably, the document content units in this application are not determined by fixed character length, fixed number of sentences, or simple paragraph division, but are determined after correlation analysis based on the organizational characteristics of the rule document itself. Since the knowledge expression in rule documents is usually continuous, the true meaning of a certain content may need to be understood in conjunction with adjacent content. For example, a rule requirement may include the applicable object, execution conditions, and exceptions, where the applicable object may be located at the beginning of the clause, and the exception conditions may be located in the subsequent explanation section. If the document is divided only according to natural paragraphs or sentences, it is easy to break up information within the scope of the same rule, thereby affecting subsequent morphological pattern judgment.

[0039] Therefore, in one embodiment, this application can perform correlation analysis on continuous content in a rule document based on the chapter structure, clause hierarchy, content reference relationships, and contextual relationships in the rule document, thereby determining document content units with complete semantic scope. For example, when it is detected that a certain piece of content is used to explain the object scope in a previous rule, that piece of content can be associated with the aforementioned rule content and processed as the same knowledge expression scope.

[0040] After identifying multiple document content units, this application further extracts the structural features, textual expression features, and contextual features corresponding to each document content unit. The structural features describe the organizational relationship of the document content unit within the entire rule document, such as its chapter, clause level, sequential position, and citation relationships. The textual expression features describe the linguistic characteristics of the document content unit itself, such as binding, explanatory, procedural, and descriptive expressions. The contextual features describe the semantic relationships between the document content unit and other content units, such as the explanatory relationship between a rule and its corresponding definition, the applicability relationship between a case study and its corresponding rule, and the association between auxiliary content and other content.

[0041] Through the above methods, this application can obtain comprehensive features of document content units from three dimensions: document structure, text expression, and content association, providing a data foundation for subsequent determination of the morphological patterns corresponding to different document content units.

[0042] S2, based on the structural features, the text expression features, and the context association features, analyze the semantic belonging degree of each document content unit in multiple candidate morphological modes, determine the morphological belonging strength corresponding to each document content unit, and determine the target morphological mode corresponding to the document content unit according to the morphological belonging strength.

[0043] In this embodiment, after obtaining the structural features, textual expression features, and contextual association features corresponding to the document content unit, this application further analyzes the semantic attribution degree of each document content unit under multiple candidate morphological patterns to determine which morphological pattern is more suitable for processing the document content unit. It should be noted that the morphological patterns in this application are not used to represent simple text classification results, but rather to represent the expressive role played by the document content unit in the rule-based knowledge organization process. Since the same document content unit may simultaneously contain multiple expressive roles, this application does not directly limit it to a single category, but rather analyzes the degree of association between it and multiple candidate morphological patterns.

[0044] Preferably, the candidate format patterns may include rule formats, definition formats, process formats, case formats, and auxiliary formats. Different format patterns correspond to different content expression characteristics and processing requirements. Among them, the rule format is mainly used to represent content with a binding effect, such as content used to limit behavioral requirements, execution conditions, or prohibited items; the definition format is mainly used to represent conceptual explanations or object scope limitations; the process format is mainly used to represent content with execution order, processing steps, or state change relationships; the case format is mainly used to represent the specific event occurrence process and corresponding application conditions; and the auxiliary format is mainly used to represent content that, although it does not directly form rule constraints, conceptual limitations, process steps, or case descriptions, can assist in rule understanding, relationship establishment, and retrieval judgment.

[0045] In one embodiment, this application can establish morphological recognition rules based on the expressive features corresponding to different morphological patterns, and use the morphological recognition rules to perform matching analysis on document content units. For example, for rule-based morphological patterns, recognition can be performed on expressions containing constraint requirements; for definition-based morphological patterns, recognition can be performed on expressions used to explain the meaning or scope of an object; for process-based morphological patterns, recognition can be performed on expressions reflecting the sequence of actions or the execution process; for case-based morphological patterns, recognition can be performed on expressions describing specific events, processing procedures, and result states; and for auxiliary morphological patterns, analysis can be performed by combining the relationships between content and the contextual relationships.

[0046] Preferably, the morphological recognition rules in this application can be implemented using regular expressions. Unlike scenarios requiring complex semantic reasoning on open text, rule documents typically possess strong expressive norms and format stability, with different knowledge functions often corresponding to relatively fixed language expression patterns. For example, rule content often contains expressions with binding meanings such as "should," "must not," "prohibited," and "must"; definition content often contains expressions used for conceptual explanation such as "refers to," "includes," and "hereinafter referred to as"; process content often contains expressions reflecting process relationships such as "first," "then," "execute," "submit," and "review"; and case content often contains expressions used to describe specific application scenarios such as "when," "under certain circumstances," and "for example." Therefore, by establishing regular expressions for the above stable expression patterns, rapid recognition of the morphological content of rule documents can be achieved.

[0047] Compared to processing all text content using general semantic understanding methods, morphological recognition using regular expressions makes the recognition process more interpretable and traceable. On the one hand, rule-based documents usually have clear expression specifications, and recognition methods based on expression patterns can more accurately capture their linguistic characteristics. On the other hand, when a document content unit is identified as a certain morphological pattern, its recognition basis can be directly determined according to the corresponding matching rules, facilitating the subsequent tracing of the formation process of morphological semantic relationships.

[0048] Furthermore, to avoid misjudgments based solely on a single expressive feature, this application combines structural features and contextual association features corresponding to document content units when determining the degree of morphological attribution. For example, although a piece of content may contain case description keywords, if this content is located within a rule clause and is directly related to the preceding and following rule constraints, it may also serve as an explanation of the rule's application. Similarly, although a piece of explanatory text may not directly contain rule qualifiers, if it follows a rule clause and explains the scope of application of that rule, it may possess auxiliary morphological features. Therefore, this application comprehensively judges the degree of semantic attribution between document content units and each candidate morphological pattern by integrating multi-dimensional features.

[0049] Based on the above analysis, this application determines the morphological attribution strength corresponding to each document content unit. The morphological attribution strength characterizes the degree to which a document content unit assumes the knowledge function corresponding to a certain morphological pattern, and can be represented by numerical values, levels, intervals, or other forms that can indicate the degree of association. For example, when a document content unit simultaneously possesses rule-expressing features and case-expressing features, the attribution strength of its corresponding rule-form and case-form can be determined separately, and the corresponding target morphological pattern can be selected based on the subsequent knowledge association construction requirements.

[0050] Furthermore, this application determines the target morphological pattern based on the morphological attribution strength. When a document content unit corresponds to a single morphological pattern with a high attribution strength, that morphological pattern can be directly determined as the target morphological pattern; when a document content unit corresponds to multiple candidate morphological patterns, the target morphological pattern suitable for subsequent processing can be determined by further combining the semantic interpretation results corresponding to different candidate morphological patterns and the subsequent knowledge association construction requirements.

[0051] By employing the above method, this application avoids the problem of simply dividing document content into a single category in traditional text classification methods, enabling the content in the rule document to select the corresponding processing path according to its actual knowledge role, thus providing a foundation for subsequent differentiated processing and the construction of related knowledge structures.

[0052] S3. Based on the differentiated processing strategy corresponding to the target morphological pattern, perform content morphological semantic representation on the document content unit to obtain the corresponding morphological semantic relationship.

[0053] In this embodiment, after determining the target form pattern corresponding to each document content unit, this application further performs content form semantic representation on the document content units according to the differentiated processing strategies corresponding to different target form patterns, so as to obtain the form semantic relationship corresponding to their knowledge function. It should be noted that this application does not merely label the document content units, but extracts and organizes the key semantic elements contained within them according to the knowledge function undertaken by different content, so that different forms of content can form a structured expression that can participate in the construction of knowledge association.

[0054] Preferably, since different forms and patterns correspond to different knowledge expression purposes, their corresponding differentiated processing strategies also differ. For example, the rule form focuses on rule constraint relationships, the definition form focuses on concept limitation relationships, the process form focuses on process change relationships, the case form focuses on specific events and applicable conditions, and the auxiliary form focuses on auxiliary understanding and related support relationships. By adopting different processing methods for different forms, the problem of weakening the role of some key content when using a uniform semantic processing method can be avoided.

[0055] Specifically, when the target form pattern is a rule form, this application can analyze the constraint objects, constraint conditions, execution requirements, and exception conditions in the document content unit to form rule semantic relationships used to characterize the scope of the rule. For example, a rule may contain the expression "a certain subject should perform a certain operation when a certain condition is met." This application can determine the relationship between the rule's applicable object, triggering condition, and constraint behavior, so that the rule content no longer exists only in text form, but can participate in subsequent processing as a knowledge relationship with a clear boundary of action.

[0056] When the target form pattern is a defined form, this application can analyze the conceptual objects, explanatory content, scope of limitation, and associated objects in the document content unit to form a defined semantic relationship used to characterize the semantic boundary of the concept. In actual rule documents, the same concept may have different interpretations in different application environments. For example, a certain object name may correspond to different business entities or different management scopes. If the search is based solely on the text name, inaccurate matching is likely to occur. Therefore, this application determines the association between the concept and its explanatory scope through a defined form processing method, enabling subsequent retrieval processes to combine the semantic boundary of the concept for matching.

[0057] When the target form pattern is a process form, this application can analyze the executing subject, operation actions, execution sequence, state changes, and condition transition relationships in the document content units to form process semantic relationships used to characterize process evolution relationships. The process content in rule documents is usually not a simple set of actions, but rather consists of multiple interdependent processing steps. For example, a business processing procedure may include multiple stages such as application, review, confirmation, and execution. If only one action keyword is extracted, the complete process cannot be accurately understood. Therefore, this application uses process form processing to preserve the sequential relationships and state change relationships between different actions.

[0058] When the target form is a case form, this application focuses on analyzing the specific events and corresponding contextual factors contained in the case content to form a case semantic relationship used to characterize the applicability of the case. Preferably, the case form is not merely used to assist in explaining the rule content, but rather participates in the rule application analysis as an important component of the rule knowledge system. In practical applications, rule texts are usually used to describe general requirements, while specific cases can demonstrate the actual role of the rule under specific conditions. For example, the same behavior may correspond to different processing results under different subjects, different environments, or different purposes. Therefore, this application can extract contextual factors such as the case subject, behavioral process, occurrence conditions, environmental factors, and result status from the case content, and establish a correlation between the case content and the corresponding rule content, so that the specific context in the case can participate in subsequent rule understanding and retrieval matching.

[0059] Furthermore, in the processing of case formats, this application does not simply use case content as search keywords, but rather analyzes the factors in the case that affect the application of rules. For example, in a case description, whether a certain behavior violates a rule may depend not only on the behavior itself, but also on the subject's state, the environment, or special conditions at the time the behavior occurred. By identifying the above factors, the contextual information corresponding to the case content can be determined, and this information can be included as part of the case's semantic relationship, enabling the rule retrieval process to make judgments in conjunction with specific application conditions.

[0060] When the target form is an auxiliary form, this application analyzes the content in the rule document that does not directly constitute rule constraints, concept definitions, process steps, or case descriptions, but can assist in rule understanding and knowledge association establishment, forming auxiliary semantic relationships. In actual rule documents, some content may appear as background information, supplementary explanations, related hints, or explanatory descriptions. While these may not superficially belong to the core rule content, they may affect the understanding of the rule's scope of application, its formation reasons, or related relationships. For example, a certain explanatory text may explain the background of a rule's formulation, and a related description may indicate the connection between two rules. Ignoring such content may lead to an incomplete knowledge structure.

[0061] Preferably, the auxiliary format processing in this application does not simply retain all non-main content, but rather analyzes the relationship between the auxiliary content and other content units to determine whether it can affect rule understanding, knowledge association, or retrieval judgment. When the auxiliary content has a strong correlation with a certain rule content, definition content, process content, or case content, the correlation factors can be extracted to form corresponding auxiliary semantic relationships.

[0062] Through the above-mentioned differentiated processing methods, this application can establish corresponding morphological semantic relationships for different knowledge functions in the rule document, so that the rule form, definition form, process form, case form and auxiliary form content can all be converted into knowledge expressions that can be associated and processed, providing a foundation for the semantic association and fusion between different content forms in the future.

[0063] S4. Based on the morphological semantic relationships corresponding to different document content units, perform fusion analysis on the semantic association features between different content forms, determine the knowledge association relationships corresponding to different content forms, and generate the associated knowledge structure corresponding to the rule document.

[0064] In this embodiment, after obtaining the morphological semantic relationships corresponding to different document content units, this application further performs a fusion analysis on the semantic association features between different content forms to establish knowledge association relationships between different content forms and generate the associated knowledge structure corresponding to the rule document. It should be noted that this application does not store rule forms, definition forms, process forms, case forms, and auxiliary forms as independent information objects, but rather further analyzes the interaction relationships between different content forms so that different content can jointly support rule understanding and retrieval judgment.

[0065] In actual rule documents, different content formats often have interdependencies. For example, a rule format defines whether a certain behavior is allowed, but the scope of the object in the rule may need to be determined in conjunction with the definition format; a process format describes the specific execution process, but the processing conditions in the process may need to be constrained in conjunction with the rule format; a case format describes a specific event and its occurrence conditions, and its value lies not only in recording the event itself, but also in reflecting how a rule applies in a specific context; while auxiliary formats do not directly form rule requirements, the explanatory content, related hints, or background information they contain may affect the understanding of the relationship between rules, cases, or processes.

[0066] Therefore, in one embodiment, this application analyzes the semantic association features between different content forms based on the morphological semantic relationships corresponding to different document content units. These semantic association features can be used to represent explanatory, constraining, applicable, referential, and auxiliary support relationships between different content forms. For example, when a definitional content is used to explain the meaning of an object in a rule form, it can be determined that there is a conceptual limitation relationship between the definitional content and the corresponding rule form content; when a case content describes the execution result of a rule under specific conditions, it can be determined that there is a case application relationship between the case content and the corresponding rule form content; when an auxiliary content is used to explain the background or basis for the formation of a rule, it can be determined that there is an auxiliary association relationship between the auxiliary content and the corresponding rule form content.

[0067] Preferably, the knowledge association relationships in this application are not limited to direct reference relationships between content, but may also include indirect association relationships obtained based on content function analysis. In some rule documents, different content may not have obvious textual references, but because their knowledge functions are related, it is still necessary to establish an association. For example, a case description may not directly reference a rule clause, but the behavioral conditions, subject status, and processing results in the case are highly relevant to the scope of application of a certain rule. In this case, an applicable association can be determined based on the matching relationship between the semantic relationship of the case and the semantic relationship of the rule.

[0068] Furthermore, this application generates a related knowledge structure for the rule document based on the determined knowledge relationships corresponding to different content forms. The related knowledge structure represents the organizational relationships between different content units in the rule document, and may include content unit objects, morphological semantic relationships, and relationships between different content forms. By constructing a related knowledge structure, the rule document can be transformed from a traditional linear text organization into a knowledge expression form containing multi-dimensional relationships.

[0069] For example, the corresponding knowledge structure for a business processing rule may include: rule-based content representing the restrictions on business processing; definition-based content defining the scope of the business; process-based content representing the steps involved in business processing; case-based content representing the applicable results of the rule in different practical situations; and auxiliary content providing relevant explanations or supporting evidence. These contents, connected through different knowledge relationships, can collectively form a complete knowledge representation of the rule.

[0070] Preferably, when constructing the associated knowledge structure, this application can also determine the strength of the association relationship based on the degree of association between different content forms to represent the closeness of the association between different knowledge objects. For example, definition content directly related to the rule content can have a high degree of association, while auxiliary content indirectly related to the rule content but capable of influencing the judgment result can have a corresponding degree of association. By recording the degree of association, a more suitable knowledge path can be selected according to the query requirements during subsequent retrieval processes.

[0071] By employing the methods described above, this application integrates and connects different content formats within the rule document, ensuring that rules, definitions, processes, cases, and supporting content no longer exist as isolated texts but rather collectively form a connected knowledge structure capable of supporting rule understanding and retrieval judgment. Compared to methods that rely solely on text similarity matching, this application leverages the knowledge relationships formed between different content formats to improve information organization and retrieval accuracy in complex rule scenarios.

[0072] S5. Based on the associated knowledge structure, a retrieval adaptation unit is generated, and in response to the retrieval request, the retrieval adaptation unit is matched with the query semantic features and query context features corresponding to the retrieval request to extract the corresponding retrieval results.

[0073] In this embodiment, after generating the associated knowledge structure corresponding to the rule document, this application further generates a retrieval adaptation unit based on the associated knowledge structure to support content matching in subsequent retrieval processes. It should be noted that the retrieval adaptation unit does not simply correspond to the original text fragments in the rule document, but rather is a retrieval object formed by adapting and organizing the knowledge content in the rule document based on the knowledge relationships formed between different content forms. By generating the retrieval adaptation unit, the subsequent retrieval process can simultaneously utilize the relationships between rule content, definition content, process content, case content, and auxiliary content, rather than matching based solely on a single text content.

[0074] Preferably, when generating retrieval adaptation units, this application can combine content with common knowledge functions or related influences based on the knowledge association relationships between different content units in the associated knowledge structure. For example, for a certain rule, its corresponding definition, process, case, and auxiliary content can be organized according to their association relationships to form a retrieval adaptation unit that fully reflects the rule's scope of application and conditions. In this way, when a user enters a retrieval request, not only can the rule text itself be matched, but other knowledge content related to the rule can also be found through association relationships.

[0075] In one embodiment, the retrieval adaptation unit may include rule-related information, definition-related information, process-related information, case-related information, and auxiliary-related information. Rule-related information represents rule requirements and constraints; definition-related information represents the scope of related concepts; process-related information represents the rule execution process; case-related information represents the applicability of the rule in a specific event; and auxiliary-related information represents supplementary evidence that can influence the understanding of the rule. Through the combination of the above information, the retrieval adaptation unit can describe rule knowledge from multiple dimensions.

[0076] Furthermore, upon receiving a retrieval request, this application first analyzes the retrieval request to determine the corresponding query semantic features and query context features. The query semantic features are used to represent the object, behavior, rule requirements, or processing matters that the user wishes to query; the query context features are used to represent the specific application conditions corresponding to the user's query, such as the state of the query object, the environment in which the event occurs, the stage of business processing, or special applicable conditions.

[0077] In traditional rule-based retrieval methods, search results are typically determined based on keyword matching between the query text and the rule text. However, in practical applications, user questions often correspond to specific application scenarios rather than just a single rule text. For example, for the same behavior, different subject states, different occurrence conditions, or different processing purposes may lead to different rule application results. If matching is based solely on keywords in the query text, it may return rule content that is literally relevant but actually inapplicable.

[0078] Therefore, when performing retrieval matching, this application not only analyzes the semantic content of the query request, but also combines query context features with case association information, rule association information, and auxiliary association information in the retrieval adaptation unit for comprehensive matching. For example, when a query request involves a business processing problem under a specific situation, this application can match the case association information in the retrieval adaptation unit according to the query context features to determine the case context corresponding to that specific situation, and further associate the corresponding rule content, thereby improving the consistency between the retrieval results and actual application needs.

[0079] Preferably, the query context features in this application can be determined based on object information, event information, condition information, and result requirement information in the query request. For example, when a user enters a query request such as "Can a certain entity perform a certain action under special circumstances?", this application not only focuses on the textual information of "perform a certain action", but also further analyzes contextual factors such as "a certain entity" and "special circumstances", and matches them with case-type content in the associated knowledge structure to determine search results that better match the current application conditions.

[0080] Furthermore, after determining the matching relationship between the retrieval adaptation unit and the query request, this application can extract the corresponding retrieval results based on the degree of matching. The retrieval results may include rule content, related definition content, process content, case content, and auxiliary related content corresponding to the query request, enabling users to understand the current query results from multiple knowledge dimensions.

[0081] For example, in a business rule query scenario, a user queries whether a certain business operation meets the rule requirements. This application can first determine the corresponding rule range based on the query semantic features, then combine the query context features to match the specific application conditions in the case form, and further call the related information in the auxiliary form for judgment, finally returning the rule result that matches the current business context. At this time, the search result does not come from a single rule text, but is formed by the joint support of rule form, definition form, case form, and auxiliary form.

[0082] Through the above methods, this application realizes a complete processing flow from rule document content form recognition, differentiation processing, knowledge association construction to retrieval matching, enabling different types of content in rule documents to participate in retrieval judgment and improving the accuracy and applicability of retrieval results in complex rule scenarios.

[0083] In this embodiment, for document content units whose target form pattern is determined to be a rule form, this application further analyzes the rule constraint relationships contained therein to form corresponding rule semantic relationships.

[0084] In actual rule documents, rule content is typically used to define the actions that specific objects should, are allowed, or are prohibited from performing under specific conditions. While such content usually has a relatively clear constraint expression, its complete meaning is not determined by a single keyword, but rather by multiple factors such as the constrained object, triggering conditions, execution requirements, and exceptions. For example, a rule might describe "a certain type of subject should complete a certain operation when certain conditions are met," where subject information identifies the object of the rule, condition information defines the scope of application, operation information defines the rule requirements, and exception information limits the boundaries of the rule's application.

[0085] Preferably, when processing the rule content, this application can first identify the constraining subject, constraining behavior, constraining condition, and limitation scope in the rule content, and construct the rule semantic relationship based on the correlation between the above information. The constraining subject represents the object to which the rule applies, and may include personnel, organizations, equipment, business objects, or other entities that need to be constrained by the rule; the constraining behavior represents the action that the rule requires or prohibits; the constraining condition represents the environmental conditions or preconditions that the rule must meet to take effect; and the limitation scope represents the boundary range to which the rule applies.

[0086] For example, a business management rule may contain the following: "When the application materials are incomplete, the reviewer shall notify the applicant to supplement the materials within the specified period." This application does not store the above-mentioned rules directly as ordinary text. Instead, it identifies the reviewers as the binding subject, notifying the applicant to supplement materials as the binding action, incomplete materials as the triggering condition, and the specified time limit as the binding scope, thereby forming a rule semantic relationship to describe "what subject performs what action under what conditions".

[0087] Furthermore, this application can further determine the boundaries of rule application by combining the document structure and contextual features of the rule content. In some rule documents, the rule clauses themselves may only describe general requirements, while the specific scope of application may be determined by the preceding definitions, subsequent exceptions, or related clauses. Therefore, this application can improve the semantic relationship of the rules based on the relationship between the rule's formal content and other formal content.

[0088] For example, a rule might stipulate that "relevant personnel shall not enter restricted areas," but the specific scope of "relevant personnel" may be limited by the preceding definition, and the scope of "restricted areas" may be explained by supplementary explanations. In this case, if only the rule text itself is processed, the scope of the rule's constraints cannot be accurately determined. This application supplements the semantic relationships of the rules by associating the semantic relationships of definitions and supplementary semantic relationships in the knowledge structure, enabling the rule content to form a more complete applicable boundary.

[0089] Preferably, the rule format processing results in this application can be used for rule matching in subsequent retrieval processes. For example, when a user submits a query request, the object information, behavior information, and condition information in the query request can be matched with the constraint subject, constraint behavior, and constraint condition in the rule semantic relationship to determine whether the rule is applicable to the current query scenario.

[0090] Through the above methods, this application differentiates the content of the rule form, transforming the rule content from the original text form into a semantic expression with a clear scope and constraint relationship, thus providing a foundation for subsequent knowledge association construction and retrieval result determination.

[0091] In this embodiment, for document content units whose target form pattern is determined to be a defined form, this application further analyzes the conceptual interpretation relationships contained therein to form corresponding definitional semantic relationships.

[0092] In rule documents, definitions are used to explain the specific meaning of an object, term, or business concept. Their purpose is not simply to interpret the text, but to define the scope of application of the object in subsequent rule interpretation. Since different rule documents may be formulated for different business environments, even objects with the same name may have different actual meanings. Therefore, in the rule retrieval process, matching solely based on object names can easily lead to inaccurate definition scope.

[0093] Preferably, when processing definitional content, this application can identify the relationships between the definition object, definition description, limiting conditions, and associated objects, and form definition semantic relationships based on the above relationships. Specifically, the definition object represents the conceptual entity being explained; the definition description represents the specific meaning corresponding to the concept; the limiting conditions represent the scope requirements that the concept must satisfy to be valid; and the associated objects represent other knowledge objects that are related to the concept.

[0094] For example, in a rule document, "staff" might be defined as "relevant personnel who participate in the business processing procedure and have operational authority." In this case, "staff" is not a simple reference to all personnel, but is limited by conditions such as "participating in the business processing procedure" and "having operational authority." This application, through morphological definition processing, can determine the semantic boundaries corresponding to this concept and associate it with relevant rule content.

[0095] Furthermore, this application can associate semantic relationships between definitions and semantic relationships between rules, enabling objects within the rule content to have a more accurate scope of interpretation based on the definition content. For example, when a rule restricts "staff members shall not perform a certain operation," this application can determine the actual scope of "staff members" based on the corresponding semantic relationship of the definition, thereby avoiding the incorrect inclusion of objects that do not belong to that scope in the search results.

[0096] Furthermore, in complex rule documents, definitions may be related to case studies, process flows, and supporting documentation. For example, the definition of a business object may affect the scope of the subject in case analysis, and the operational object in a process node may require the definition to determine its specific meaning. Therefore, this application enables definitions to participate in the overall rule understanding process by forming semantic relationships between their morphological content and incorporating them into a related knowledge structure.

[0097] By employing the above methods, this application avoids the problem of blurred concept boundaries caused by focusing solely on keyword matching in traditional rule retrieval, enabling the definition content in the rule document to play its actual role in knowledge organization and retrieval judgment.

[0098] In this embodiment, for document content units whose target form pattern is determined to be a process form, this application further analyzes the process change relationships contained therein to form corresponding process semantic relationships.

[0099] In rule documents, process content typically describes the multiple processing stages a task undergoes from triggering to completion. The focus is not on individual actions themselves, but rather on the sequential, execution, and state change relationships between different actions. If only keyword matching is used to process process content, individual operations within the process are easily isolated, leading to an inaccurate understanding of the complete business process.

[0100] For example, a business processing rule might involve a series of steps such as "submitting application materials, reviewing application information, confirming review results, and performing subsequent processing." The terms "submit," "review," "confirm," and "execute" represent different actions, but only by associating these actions in the order of execution can the complete business processing flow be reflected. Therefore, this application, focusing on the content of the process structure, not only identifies the action information within the process but also further analyzes the sequential relationships, conditional relationships, and state transition relationships between actions.

[0101] Preferably, when processing process content, this application can extract the executing subject, executing action, executing sequence, and state change conditions from the process content unit, and construct process semantic relationships based on the above information. Specifically, the executing subject represents the object performing a certain process action; the executing action represents the specific operation in the process; the executing sequence represents the sequential relationship between different actions; and the state change conditions represent the conditions that must be met for the process to move from one stage to the next.

[0102] For example, a device maintenance process may include steps such as "detecting device operating status, determining fault type, developing maintenance plan, and performing maintenance operations." This application can determine the process relationships between detection, judgment, development, and execution based on the above content, and form process semantic relationships to describe changes in the device maintenance process.

[0103] Furthermore, this application can also constrain the semantic relationships of processes by combining the content of rule forms. In actual rule documents, processes usually do not exist independently, but are restricted by rule conditions. For example, a certain process step may only be executed after certain conditions are met, or a certain operation must be performed by a specific subject. Therefore, this application can associate process semantic relationships with rule semantic relationships, so that the process content not only represents "how to execute", but also "under what rule conditions to execute".

[0104] Furthermore, the content of the process may also be related to the content of the case. For the same process, different case scenarios may correspond to different execution paths. For example, under normal circumstances, it may be processed according to the standard process, while under special circumstances, an exception process may be triggered. By associating the semantic relationship between the process and the semantic relationship between the case, this application enables the selection of more suitable process content based on specific case conditions during subsequent retrieval.

[0105] Through the above methods, this application differentiates the content of process forms, transforming the process description in the rule document from a simple set of actions into a process knowledge expression that includes subjects, actions, sequences, and conditional relationships, thus providing a foundation for the construction of related knowledge structures.

[0106] In this embodiment, for document content units whose target form pattern is determined to be a case form, this application further analyzes the specific events and corresponding contextual factors contained in the case content to form corresponding case semantic relationships.

[0107] In rule documents, case studies typically illustrate the execution of a rule in a specific application scenario. Traditional rule processing methods often treat case studies as supplementary reading material outside of the rule clauses, solely to help users understand the rule's meaning, without incorporating the specific conditions, subject states, and changes in outcomes contained within the case studies into the rule knowledge organization process. However, in practical applications, case studies often contain applicable conditions not directly expressed in the rule text, playing a crucial role in determining whether a rule applies to the current situation.

[0108] Preferably, the case format in this application is not equivalent to simple example text, but rather treats cases as knowledge objects that can reflect the actual process of rule operation. Specifically, this application can analyze the case subject, behavioral process, occurrence conditions, environmental factors, and outcome state in the case content, and form case semantic relationships based on the above factors.

[0109] For example, a management rule might state that "entry into restricted areas is prohibited without authorization." From a rule-form perspective, this clearly establishes the constraint of prohibition. However, in practice, whether an individual's entry into a restricted area violates the rule can be influenced by specific circumstances. For instance, factors such as whether the individual has an emergency rescue purpose, is in a special work situation, or has received temporary authorization can all affect the final judgment.

[0110] In the aforementioned situations, relying solely on the rule text itself for retrieval may only yield general rules such as "entry into restricted areas is prohibited," without being able to determine whether there are applicable differences in special circumstances. This application, by processing the content of case examples, can identify contextual factors such as the subject's state, behavioral purpose, conditions of occurrence, and changes in outcome from the cases, and establish an applicable relationship between the case content and the rule content.

[0111] Furthermore, the semantic relationships of cases in this application can be used to describe how specific situational factors in a case affect the application of rules. For example, if a case records an entity performing a certain action in a specific environment and ultimately obtaining a certain result, this application can associate the conditions of occurrence, the process of action, and the state of the result in that case, making it an important basis for judging the application of rules.

[0112] Preferably, the focus in the case processing is not on all the descriptive content in the case text, but rather on selecting key factors that can influence the understanding and application of the rules. For example, some background descriptions may only be used to enrich the case narrative, while certain subject states, time conditions, environmental factors, or behavioral purposes may directly affect whether the rules apply. Therefore, this application can determine the situational factors with judgment value in the case based on the degree of correlation between the case content and the rule content.

[0113] Through the above methods, this application transforms case content from supplementary explanatory material in traditional rule documents into important objects participating in the construction of knowledge associations. Cases are no longer merely supplementary materials for explaining rules, but can supplement the application boundaries of rules through the specific contextual factors they contain, enabling the subsequent retrieval process to determine more accurate retrieval results by combining them with actual application scenarios.

[0114] In this embodiment, for document content units whose target form pattern is determined to be auxiliary form, this application further analyzes the auxiliary information contained therein and its relationship with other content to form corresponding auxiliary semantic relationships.

[0115] In rule documents, not all knowledge-valuable content is directly expressed as rule requirements, concept definitions, execution procedures, or specific cases. Some content may appear as background information, explanations, related hints, the basis for the rule's formulation, or other supplementary information. Although this content usually does not directly stipulate whether a certain action should be performed, nor does it directly limit the scope of a certain object, it may affect the understanding of the rule's meaning, scope of application, and knowledge relationships. Therefore, if only content with obvious rule characteristics is retained while the aforementioned supplementary content is ignored, it may lead to gaps in the rule's knowledge structure.

[0116] Preferably, the auxiliary form in this application is used to represent content in the rule document that can support the understanding, association establishment, or retrieval judgment of other content. The auxiliary form does not refer to all non-rule subject content, but rather to content that, after association analysis, is determined to have knowledge-supporting value. For example, a descriptive text, although not directly forming a rule constraint, explains the background and reasons for the creation of a rule; a relational description, although not directly defining the scope of the object, explains the connection between a concept and other business objects; supplementary content, although not constituting a process step, reveals the special conditions applicable to a certain processing procedure. All of the above content may affect the subsequent rule understanding process.

[0117] Specifically, when processing auxiliary content, this application can analyze the relationship between auxiliary content units and rule-based, definition-based, process-based, and case-based content, and determine the auxiliary semantic relationship corresponding to the auxiliary content based on the relationship results. For example, if an auxiliary content is located after a rule clause and further explains the special conditions in that rule clause, then it can be determined that there is an explanatory relationship between the auxiliary content and the corresponding rule content; if an auxiliary content appears together with multiple case contents and is used to explain the reasons for the case, then it can be determined that there is a supporting relationship between the auxiliary content and the case contents.

[0118] In one implementation, a rule document contains the following content: "Due to special circumstances in the business process, in order to ensure consistent results, it is necessary to make a comprehensive judgment based on the actual situation." The aforementioned content does not directly specify what actions the business entity must perform, nor does it clearly define a particular process step; therefore, it does not appear to be a rule or process in itself. However, this content reveals the need to consider special circumstances when applying rules. If this content is ignored in subsequent searches, only general rule requirements may be returned, failing to reflect the requirement to analyze specific situations in the actual judgment process.

[0119] Based on this, this application analyzes the relationship between this type of content and other knowledge objects through auxiliary morphological processing, and incorporates factors that can affect rule understanding into the associated knowledge structure. For example, the auxiliary content can be associated with the corresponding rule content, so that the retrieval process can not only obtain the rule itself, but also the auxiliary basis that affects the judgment of rule applicability.

[0120] Furthermore, the auxiliary morphological processing in this application can also be analyzed in conjunction with contextual association features. Since auxiliary content typically does not have explicit constraints like rule content, its knowledge function is more reflected in its relationships with other content. Therefore, this application can determine whether auxiliary content needs to be included in the knowledge structure based on its location, associated objects, semantic continuity, and connections with other morphological content.

[0121] For example, in a rule document, a passage might be located between a rule clause and a case description. Although this passage doesn't directly specify the behavioral requirements, it explains why the case applies to the rule. If judged solely based on the textual expression, this passage might be considered ordinary explanatory text. However, through the auxiliary morphological analysis of this application, it can be found that this passage connects the rule form and the case form, playing a crucial role in understanding the applicability of the case. Therefore, it can form a corresponding auxiliary semantic relationship and be incorporated into the related knowledge structure.

[0122] Preferably, auxiliary content can also be used to improve the connection paths between different content formats. In some complex rule documents, there may not be a direct correspondence between rule content and case content, but rather a connection is established through certain explanatory content, background information, or related descriptions. By identifying the above auxiliary relationships, originally scattered knowledge objects can be formed into a complete chain of connections, improving the integrity of the rule knowledge structure.

[0123] By employing the above methods, this application avoids the problem of traditional rule document processing focusing only on explicit rule content, enabling auxiliary content in rule documents to participate in the knowledge organization process according to its actual function. The auxiliary form does not simply retain non-subject text in the document, but rather identifies information with supporting value by analyzing its relationship with other knowledge objects, and transforms it into auxiliary semantic relationships that can participate in retrieval judgment.

[0124] In this embodiment, after completing the differentiation processing of different content forms, this application further performs a fusion analysis on the semantic association features between different content forms based on the morphological semantic relationship corresponding to each document content unit, so as to determine the knowledge association relationship corresponding to different content forms and generate the associated knowledge structure corresponding to the rule document.

[0125] In practical rule document processing, rule content, definition content, process content, case content, and auxiliary content do not exist independently but play a collective role in the rule understanding process. For example, the rule form describes whether a certain behavior meets the constraint requirements; the definition form determines the specific meaning of the relevant objects in the rule; the process form explains the operational relationships in the rule execution process; the case form demonstrates the actual performance of the rule under specific application conditions; and the auxiliary form provides relevant evidence that affects rule understanding. Only by establishing interrelationships among these different types of content can a complete expression of rule knowledge be formed.

[0126] Preferably, the associated knowledge structure in this application is used to represent the organizational relationships between different document content units in a rule document. It may include the morphological semantic relationships corresponding to the document content units, the association relationships between different morphological semantic relationships, and the function information corresponding to the association relationships. By constructing an associated knowledge structure, a linear rule document originally organized according to chapters and clauses can be transformed into an associated structure organized based on knowledge function relationships, enabling the subsequent retrieval process to obtain information related to the query request from multiple dimensions.

[0127] Specifically, this application can determine the semantic association features between content forms based on the morphological semantic relationships corresponding to different document content units. For example, when a rule form contains a specific business object, and the definition form in another document content unit is used to explain the meaning of that business object, this application can determine that there is a conceptual association between the rule form and the definition form based on the correspondence between the rule object and the definition object.

[0128] For example, when a rule form specifies that a business process needs to meet certain conditions, and a process form describes the execution steps of that business process, this application can determine the execution relationship between the rule form and the process form based on the correspondence between the rule constraints and the process execution, so that the corresponding process content can be further obtained according to the rule requirements during the subsequent retrieval process.

[0129] Furthermore, regarding the content of case forms, this application can establish a case application relationship between case forms and rule forms based on the correlation between situational factors in the case and the conditions for rule application. Specifically, the subject's state, behavioral process, occurrence conditions, and outcome state in the case content may reflect the actual application of the rule in specific circumstances, and therefore can serve as important connecting factors between case content and rule content.

[0130] For example, a rule might stipulate that a certain type of entity is prohibited from performing a certain operation, while a case describes an entity performing that operation under specific conditions and obtaining different results. In this case, this application does not simply determine whether the case content and the rule content are textually similar, but rather analyzes whether the specific situational factors in the case change the applicable conditions of the rule. If the specific conditions in the case affect the rule judgment, a corresponding case application relationship can be established, allowing the case content to participate in subsequent search judgments.

[0131] For supplementary content, this application can establish supplementary associations based on its supporting relationship with other content forms. For example, if a supplementary piece of content is used to explain the reasons for the formulation of a rule or to illustrate the background of a case, then the supplementary content can be associated with the corresponding rule content or case content, making it a component of the complete knowledge chain.

[0132] Preferably, when determining the knowledge association relationships between different content formats, this application can further analyze the strength of the association relationships. The association strength is used to represent the degree of closeness of the association between different knowledge objects. For example, directly explaining the definition of a rule object can have a high association strength; auxiliary content that indirectly affects the understanding of a rule can have a corresponding association strength. By recording the association strength, a more suitable knowledge association path can be selected according to the query requirements during subsequent retrieval processes.

[0133] Furthermore, this application generates a related knowledge structure based on the aforementioned determined knowledge relationships. The related knowledge structure can be represented using a graph structure, a relational structure, or other data structures capable of representing knowledge objects and their relationships. In one embodiment, different document content units can be used as knowledge nodes, and the semantic relationships between different content forms can be used as connection relationships between nodes, thereby forming a related knowledge structure used to describe the overall knowledge relationships of the rule document.

[0134] For example, the associated knowledge structure of a business management rules document may include: rule nodes, representing specific management requirements; definition nodes, representing explanations of related business objects; process nodes, representing execution steps; case nodes, representing actual application scenarios; and auxiliary nodes, representing related explanatory content. These nodes, connected through rule associations, definition associations, process associations, case associations, and auxiliary associations, can form a knowledge network that reflects the complete meaning of the rules.

[0135] By employing the above method, this application breaks through the traditional approach of organizing content in rule documents solely according to textual order or keyword relationships. It enables different content formats within the rule document to establish connections based on their actual knowledge applications. Consequently, rule content no longer exists in isolation, case studies are no longer merely reading examples, and auxiliary content is no longer simply filtered. Instead, they all participate in the construction of the rule knowledge structure, providing a foundation for subsequent retrieval adaptation unit generation and query matching.

[0136] In this embodiment, after generating the associated knowledge structure corresponding to the rule document, this application further generates a retrieval adaptation unit based on the associated knowledge structure and extracts the corresponding retrieval results according to the retrieval request input by the user.

[0137] In traditional rule document retrieval, the text content of the rule document is typically used directly as the search object, and the returned results are determined based on the degree of keyword matching or semantic similarity between the query text and the document text. However, due to the diverse nature of knowledge representation in rule documents, users' actual query needs usually do not correspond to a single isolated text fragment, but rather to a comprehensive judgment process comprised of rule requirements, scope, execution process, specific cases, and supporting explanations. For example, when a user queries whether a certain business activity can be performed, they may not only need to know whether the rule allows the behavior, but also need to further understand the scope of subjects involved, applicable conditions, execution process, and handling methods in special circumstances. Therefore, matching based solely on text content can easily lead to search results lacking necessary relevant information.

[0138] Preferably, this application generates retrieval adaptation units based on associated knowledge structures. The retrieval adaptation unit represents the retrieval object after content-form association processing. It is not a simple copy of the original text in the rule document, but rather an adaptation structure obtained by organizing relevant knowledge content based on the knowledge association relationships formed between different content forms.

[0139] Specifically, when generating retrieval adaptation units, this application can combine document content units that share common knowledge functions or have judgmental associations based on the association paths in the associated knowledge structure. For example, for a certain rule-type content, it can be associated with its corresponding definition-type content to clarify the specific meaning of the object in the rule; it can be associated with its corresponding process-type content to clarify the rule execution process; it can be associated with its corresponding case-type content to reflect the application of the rule in specific situations; and it can also be associated with its corresponding auxiliary-type content to provide supplementary evidence affecting the understanding of the rule. Through the above organization method, the resulting retrieval adaptation units can more completely express rule knowledge.

[0140] For example, in a business approval rule scenario, the rule form can indicate "approval can be processed after the conditions are met," the definition form can further clarify the scope of the "applicant entity," the process form can describe the submission, review, and confirmation steps in the approval process, the case form can illustrate the approval processing method in special circumstances, and the auxiliary form can provide relevant explanations. By organizing the above content based on a related knowledge structure, a retrieval adaptation unit capable of adapting to complex query needs can be formed.

[0141] Furthermore, upon receiving a retrieval request, this application analyzes the retrieval request to determine the corresponding query semantic features and query context features. The query semantic features represent the target object, target behavior, target rule, or target processing matter of the user's query; the query context features represent the specific application conditions corresponding to the user's query, such as the status of the involved entity, the business stage, the environment in which the query occurs, special conditions, or result requirements.

[0142] Preferably, the query context features in this application are specifically designed to address the discrepancy between "general rules" and "specific applicable situations" during rule retrieval. In practical applications, users often do not simply inquire about the content of a rule, but rather determine whether a certain behavior meets the rule requirements in a specific scenario. For example, the same behavior may be restricted by a rule under normal circumstances, but different handling methods may exist in emergency situations, special authorization situations, or specific business stages. If the retrieval process only focuses on behavioral keywords, it cannot accurately determine whether the current query falls within the scope of rule application.

[0143] Based on this, this application associates and matches query semantic features and query context features with search adaptation units during retrieval matching. Specifically, it first determines the range of possible association rules based on query semantic features, and then further matches case association information, auxiliary association information, and other relevant knowledge relationships in the search adaptation units based on query context features, thereby determining search results that better meet the current query requirements.

[0144] For example, when a user enters a query request such as "Can a certain staff member enter a restricted area under special circumstances?", this application will not only return relevant rules prohibiting entry based on the keyword "entering a restricted area", but will further analyze contextual factors such as "staff member" and "special circumstances" in the query, and determine the scope of staff members by combining the definition content in the related knowledge structure, judge whether there are applicable differences under special circumstances by combining the case content, and finally determine the final applicable requirements by combining the rule content.

[0145] Furthermore, after determining the search results, this application can also trace the process of forming the search results based on the knowledge associations corresponding to the search results. In one embodiment, rule semantic relationships, definition semantic relationships, case semantic relationships, and auxiliary semantic relationships related to the search results can be obtained along the association paths in the associated knowledge structure, thereby determining the association basis supporting the current search results.

[0146] For example, when a search result originates from the content of a certain rule, this application can further obtain the object definition information, applicable case information, and auxiliary explanation information corresponding to the rule, so that users can not only obtain the rule text itself, but also understand the knowledge relationships involved in the formation of the search result.

[0147] Through the above methods, this application generates retrieval adaptation units based on associated knowledge structures and performs matching by combining query semantic features and query context features. This transforms the rule document retrieval process from traditional text matching to a knowledge association matching method based on content form relationships. Therefore, rule forms, definition forms, process forms, case forms, and auxiliary forms can all participate in retrieval judgment, improving the accuracy and applicability of rule document retrieval in complex application scenarios.

[0148] In this embodiment, the specific implementation process of morphological pattern recognition in step S2 is further described. In this application, the purpose of morphological pattern recognition is not to perform general text classification of rule documents, but rather to identify the knowledge representation role of different content within the rule documents. Since rule documents typically have strong normative characteristics, their different types of content have formed relatively stable expression patterns over long-term use. Therefore, recognition rules can be established based on the linguistic expression characteristics corresponding to different content forms, and regular expressions can be used to match and analyze document content units.

[0149] It should be noted that this application uses regular expressions for morphological pattern recognition, rather than relying on simple keyword searching. Instead, it identifies representative expression patterns based on the expression structure, semantic position, and contextual features in the rule document. As a pattern-matching processing method, regular expressions can quickly identify deterministic textual expression rules and can clearly record the specific textual basis that triggers the recognition results, thus possessing strong interpretability and traceability.

[0150] Preferably, when establishing morphological recognition rules, this application can establish corresponding expression pattern sets for different morphological patterns. Specifically, the expression pattern set corresponding to the rule morphology can include expressions for representing constraints, prohibitions, and execution requirements; the expression pattern set corresponding to the definition morphology can include expressions for representing concept explanations, scope limitations, and object descriptions; the expression pattern set corresponding to the process morphology can include expressions for representing execution order, operation procedures, and state changes; the expression pattern set corresponding to the case morphology can include expressions for representing specific events, occurrence processes, and application conditions; and auxiliary morphologies can be identified by combining content association and contextual relationships.

[0151] For example, rule-based content typically contains expressions with clear binding force, such as "should," "must," "must not," "prohibited," and "permitted." This application can construct regular expressions corresponding to rule-based content based on these expressions and match them against document content units. When the above expression pattern appears in a content unit, it can be preliminarily determined that the content unit has rule-based characteristics.

[0152] For defining morphological content, rule documents typically use phrases such as "refers to," "includes," "hereinafter referred to as," or "defined as" to explain concepts. This application can identify the defined object and its corresponding scope of interpretation based on the above expression patterns. For example, when the expression structure "an object refers to..." is detected, it can be determined that the content unit has definitional morphological characteristics, and the relationship between the defined object and the definition description can be further extracted.

[0153] For process-oriented content, since it typically reflects the execution relationship between multiple actions, this application can identify expression patterns indicating process changes such as "first," "then," "further," "after completion," "submit," "review," and "confirm." Furthermore, it can combine the arrangement relationship between multiple actions and the conditional transformation relationship to determine whether the content unit possesses process-oriented characteristics.

[0154] Regarding case format content, this application can identify expression patterns such as "for example," "under certain circumstances," "when," and "a certain subject performs..." which describe specific events and conditions of occurrence. Furthermore, this application does not directly use the matching results of the above expression patterns as the case format judgment result, but rather conducts a comprehensive analysis combining factors such as the subject, behavior, conditions, and results in the case content. For example, although a piece of content contains the word "for example," if it is only used to explain a concept and does not contain a specific event process, its case format classification can be reduced by combining contextual association features.

[0155] For auxiliary content, since it typically lacks a fixed constrained expression, this application cannot rely solely on a single regular expression for judgment. Instead, it requires comprehensive analysis based on content relationships. For example, if a piece of content does not contain rule keywords but is located between rule clauses and case descriptions, and explains the reasons for the case's applicability, then it can be determined to have auxiliary morphological characteristics based on its position, contextual coherence, and relevance to other content.

[0156] Furthermore, after obtaining preliminary matching results based on regular expressions, this application can also refine the matching results by combining structural features and contextual features. For example, if a content unit simultaneously satisfies both rule-based and case-based expression patterns, its semantic attribution degree under different morphological patterns can be calculated separately, and the target morphological pattern can be determined based on the morphological attribution strength.

[0157] Preferably, the morphological attribution strength can be determined based on a combination of multiple factors, including but not limited to: the degree of matching between textual expression features and the expression pattern corresponding to the target morphological pattern; the degree of association between document structural features and the organizational position corresponding to the target morphological pattern; and the degree of conformity between contextual features and the knowledge function corresponding to the target morphological pattern.

[0158] For example, if a text contains rule expressions such as "shall" or "shall not," and is located in a rule clause of a chapter, and is related to the preceding and following constraints, then its rule-form attribution strength is relatively high; if a text contains case expressions such as "for example," but its main function is to explain the conditions for the application of the rule, then its case-form attribution strength can be further increased.

[0159] By employing the methods described above, this application leverages the inherent expressive patterns of rule documents to achieve stable and interpretable morphological pattern recognition through regular expressions. It also combines structural features and contextual association features to avoid misjudgments caused by simple matching. Compared to methods that directly rely on open semantic reasoning, this application is more suitable for rule document processing scenarios with clear expressive norms and stable language patterns, enabling the morphological pattern recognition results to further support subsequent differential processing, knowledge association construction, and retrieval processes.

[0160] This embodiment further illustrates the role of case formats and auxiliary formats in the rule document knowledge organization process. Compared to rule formats, definition formats, and process formats, case formats and auxiliary formats typically do not directly represent rule constraints or execution requirements, but the information they contain may affect the understanding of the rule's scope of application, judgment criteria, and relationships. Therefore, this application identifies and processes case formats and auxiliary formats to enable the aforementioned content to participate in the rule knowledge construction and retrieval judgment process.

[0161] In actual rule documents, the rules are usually described in an abstract way as general requirements, while actual business processes are often complex. The same rule requirement may produce different applicable results in different subjects, environments, and under different conditions. If we rely solely on the rule text itself for understanding, we may easily overlook the changing factors of the rule in specific application scenarios. Therefore, this application treats case studies as an independent content form, and establishes the applicable relationship between cases and rules by analyzing the specific events and corresponding contextual factors contained in the case content.

[0162] Preferably, when processing the content of a case, this application first determines the case object corresponding to the case content and further extracts the key contextual factors in the case. These contextual factors may include information such as the case subject, subject state, behavioral process, occurrence conditions, environmental factors, and outcome state. Specifically, the case subject represents the object participating in the event; the subject state represents the state of the object when the event occurs; the behavioral process represents the specific actions performed by the object; the occurrence conditions represent the triggering conditions of the event; environmental factors represent the external conditions affecting the handling of the event; and the outcome state represents the final processing result of the event.

[0163] For example, a rule document may contain rules restricting a certain type of behavior. Based on the rule's format, it can be determined that the rule stipulates the constraints on a certain behavior under normal circumstances. However, in practical application, if a case describes an entity performing this behavior under special circumstances and ultimately obtaining a different outcome than in the general case, then the special conditions in that case may constitute a significant factor affecting the scope of the rule's applicability.

[0164] In the above situations, this application does not simply treat the case content as an adjunct to the rule content, but further analyzes the relationship between the situational factors in the case and the rule constraints. For example, it can determine whether the state of the subject in the case falls within the scope of the rule, whether the occurrence conditions in the case trigger exceptions in the rule, and whether the outcome state in the case reflects the boundaries of rule application. Through the above analysis, semantic relationships between cases can be formed, enabling the case content to participate in the rule understanding process.

[0165] Furthermore, this application can determine the influence of case content on rule application judgment based on the degree of correlation between the semantic relationship of the case and the semantic relationship of the rule. For example, when the behavior object, occurrence conditions, and processing result in a case all correspond to the content of a rule, a strong case application correlation can be established; when the case content only provides general background information and does not affect the rule judgment, its correlation can be reduced.

[0166] Through the above methods, this application enables the case study format to not only serve an explanatory function, but also to become an important knowledge object connecting abstract rules with specific applications. The rule content provides general constraints, while the case study content provides specific applicable conditions; the combination of the two can form a more complete expression of rule application.

[0167] Furthermore, this application also addresses auxiliary content. Within rule documents, there exists a category of content that, while not explicitly serving the functions of rule formulation, concept explanation, process description, or case illustration, may still influence the rule understanding process. For example, some content explains the background of rule formulation, some explains the reasons for the connections between different clauses, and some indicates special handling methods. If judged solely by its textual form, this content might easily be considered ordinary explanatory text and overlooked, but it may actually contain important factors influencing rule interpretation.

[0168] Therefore, this application provides an auxiliary format for identifying and processing the aforementioned supporting content. Preferably, in the processing of the auxiliary format, this application does not rely on whether the content contains a specific keyword for judgment, but rather conducts a comprehensive analysis based on the location of the auxiliary content, its contextual relationships, and its association with other content formats.

[0169] For example, a rule clause might be followed by explanatory text that doesn't directly specify the behavioral requirements but explains the reason for the rule and further clarifies the need for different handling methods in certain special cases. If only the rule clause itself is analyzed, this special handling basis might not be discovered; however, through auxiliary morphological analysis, a supplementary relationship between the explanatory text and the rule content can be identified and incorporated into the related knowledge structure.

[0170] For example, a background description precedes a case description, introducing the business environment in which the case occurred and the relevant constraints. Although this part does not directly describe the case behavior, it may affect the understanding of the case context factors. Therefore, this application can identify it as auxiliary content based on its degree of relevance to the case content, and form a corresponding auxiliary semantic relationship.

[0171] Preferably, the auxiliary form and the case form can also form a synergistic relationship. The case form is mainly used to describe specific events and the situational factors in the events, while the auxiliary form can be used to explain the background of the case, supplement the basis for judgment, or connect the case with the rules. For example, a case may describe a subject performing a certain behavior under special circumstances, while the auxiliary content further explains why this special circumstance affects the application of the rules. In this case, the case form provides "what happened," and the auxiliary form provides "why this understanding is needed," and the two together support the rule judgment.

[0172] By employing the above methods, this application avoids the problem of traditional rule document processing focusing only on explicit rule content, enabling both case studies and auxiliary content to participate in the construction of a related knowledge structure based on their actual knowledge applications. Specifically, case studies connect rule requirements with specific application scenarios, while auxiliary content refines the connection paths between different knowledge objects, thereby enhancing the overall completeness of the knowledge expression in the rule document.

[0173] In this embodiment, the specific implementation process of maintaining knowledge relationships and tracing retrieval results based on the associated knowledge structure is further explained.

[0174] In practical applications, rule documents are not static but may be updated as business adjustments, management requirements change, processes are optimized, or standards are revised. When the content of a rule document changes, the changes may affect not only the text itself but also the definitions, processes, cases, and supporting relationships associated with that content. If only the changed text is replaced without adjusting the corresponding knowledge relationships, the associated knowledge structure may become inconsistent with the actual rule content, thus affecting subsequent search results.

[0175] Therefore, after generating the associated knowledge structure, this application can further update the associated knowledge structure based on changes in the content of the rule document. Preferably, when a change is detected in a document content unit in the rule document, this application can first determine the morphological pattern and corresponding morphological semantic relationship corresponding to the changed content, and then determine the scope of association that needs to be adjusted synchronously based on the knowledge association relationship between the morphological semantic relationship and other content forms.

[0176] For example, when the constraints in a rule change, the impact may not be limited to the rule itself, but may also affect case studies that rely on the rule for judgment, processes executed according to the rule, and definitions that interpret the scope of the rule's objects. In this case, this application can determine the affected knowledge objects based on the rule relationships in the related knowledge structure and update the corresponding relationships.

[0177] Furthermore, when the form and content of a case change, this application can reassess its applicability to the rule based on changes in the contextual factors within the case content. For example, a case originally used to illustrate the application of a rule under specific conditions may have its supporting role in the rule's application boundaries affected if the conditions or outcome of that case change. Therefore, this application can re-analyze the semantic relationships of cases and update the corresponding case relationships.

[0178] When the content of auxiliary forms changes, this application can determine whether the associated knowledge structure needs to be adjusted based on the degree of correlation between the auxiliary content and other content forms. For example, if a piece of auxiliary explanatory content was originally used to explain the reason for the application of a rule, and the modification of the explanatory content may affect the rule understanding process, then the corresponding auxiliary association path can be updated based on its auxiliary semantic relationship.

[0179] In this way, this application does not simply update the text content in the rule document, but adjusts the knowledge relationships in the related knowledge structure in sync with the content changes, so that the content of the rule document is consistent with the knowledge structure.

[0180] Furthermore, this application can also achieve retrieval of search results based on the associated knowledge structure. In traditional retrieval methods, users typically only obtain text results that match the query request, without understanding the judgment criteria involved in the formation of the result. When the rule document content is complex and the search result is determined by multiple knowledge factors, simply returning the final text content is insufficient to meet the user's requirement for reliable results. Therefore, after determining the search result, this application can further trace back along the associated paths in the associated knowledge structure to obtain the knowledge relationships that support the formation of the search result.

[0181] Preferably, the association path may include rule semantic relationships, definition semantic relationships, process semantic relationships, case semantic relationships, and auxiliary semantic relationships. For example, when a search result corresponds to the content of a certain rule, this application can further obtain the object definition information corresponding to the rule to explain the scope of the object in the rule; obtain the corresponding process information to explain the execution process of the rule; obtain the corresponding case information to explain the application of the rule in a specific context; and obtain the corresponding auxiliary information to explain the supplementary basis involved in the rule judgment process.

[0182] For example, when a user queries whether a certain business operation complies with regulations, this application determines a certain rule as the search result. Furthermore, this application can backtrack through the associated knowledge structure to discover that the rule applies to the current query request because the query object meets a certain defined scope, the current context matches the conditions in a certain case, and certain supplementary explanations further support this judgment. Thus, users can not only obtain the search results but also understand the knowledge formation process corresponding to those results.

[0183] Furthermore, the tracing process in this application does not simply return the text location of the search result, but rather reconstructs the knowledge association path in the formation process of the search result based on the associated knowledge structure. This approach improves the transparency of the search process, enabling query results in complex rule scenarios to have more sufficient explanatory basis.

[0184] Through the above methods, this application forms a complete processing chain from rule document content form recognition, differential processing, knowledge association construction, retrieval matching to result tracing, enabling different content forms in rule documents to continuously participate in the knowledge organization and application process, thereby improving the stability and reliability of rule document processing.

[0185] In this embodiment, a complete rule document processing scenario is used to illustrate the collaborative processing process among the rule form, definition form, process form, case form, and auxiliary form in this application.

[0186] In a business management scenario, there exists a rule document outlining the procedures for handling business matters. This rule document includes definitions of business objects, processing requirements, execution processes, typical cases, and related explanatory content. Different sections of the document serve different knowledge functions. If traditional text retrieval methods are used, it typically only allows finding document fragments containing the same or similar words based on user-input keywords, failing to understand the relationships between different sections.

[0187] For example, the rule document includes the following: The first part specifies the requirements for processing business applications, clarifying that "applicants may submit applications when they meet conditions such as complete materials and compliant identity." The second part explains the specific scope of "applicants," such as which individuals, organizations, or business entities fall within the eligible scope. The third part describes the processing flow, including application submission, material review, condition confirmation, and result feedback. The fourth part describes a real-world case, such as an applicant failing to provide complete materials due to special circumstances, but supplementing them according to the specific situation and ultimately completing the business processing. The fifth part explains the background for the formulation of this business rule and the basis for handling special circumstances.

[0188] Regarding the aforementioned rule document content, this application first identifies multiple document content units based on document structure, text expression, and contextual relationships, and then performs morphological pattern recognition on each document content unit. Specifically, the first part, the processing requirements, contains clearly defined constraints, applicable conditions, and behavioral requirements, and is therefore identified as a rule form. The second part, the scope description, defines the meaning of the business object and is therefore identified as a definition form. The third part, the processing flow, reflects the execution relationship between multiple processing steps and is therefore identified as a flow form. The fourth part, the actual processing cases, includes specific subjects, actions, occurrence conditions, and processing results, and is therefore identified as a case form. The fifth part, the explanatory content, although not directly specifying processing requirements, explains the reasons for the rule's applicability and is therefore identified as an auxiliary form.

[0189] Furthermore, this application employs corresponding processing methods based on different form patterns. For rule-based content, this application extracts the relationships between the applicant, material conditions, processing actions, and restrictive requirements to form semantic relationships. For example, it determines the constraint relationship between "applicant," "meeting the material completeness condition," and "submitting the application," enabling the rule content to serve as a basis for subsequent search judgments. For definition-based content, this application extracts the relationship between the business object and its defined scope. For example, it associates the conceptual scope of "applicant" with processing rules, allowing subsequent queries to determine the applicable rules based on the actual scope of the object. For process-based content, this application extracts the process relationship between application submission, review and confirmation, and result feedback, enabling the search process to not only return processing requirements but also provide the corresponding execution process. For case-based content, this application focuses on analyzing the contextual factors in the case. For example, it identifies the specific status of the applicant, the reasons for missing materials, supplementary conditions, and the final processing result in the case, and determines the correlation between these factors and the application of the rules.

[0190] In this process, the case examples are not simply examples of the rules, but rather serve to illustrate how the rules apply in specific circumstances. For instance, a general rule may require complete documentation before processing, but in a case where special circumstances led to completion after supplementary processing, this demonstrates that special conditions can influence the determination of rule applicability. Therefore, this application, through case-based processing, can uncover applicable conditions beyond the textual rules.

[0191] Regarding the supplementary content, this application analyzes its relationship with the rule content and case content. For example, the background of rule formulation may explain why a certain processing requirement needs to be set, and the special handling instructions may explain why an exception can be applied to a case. Although the above content is not the direct basis for enforcement, it can help understand the boundaries of the rules, thus forming a supplementary semantic relationship.

[0192] Furthermore, this application generates a related knowledge structure based on the aforementioned semantic relationships of different forms. In this related knowledge structure, rule-type content serves as the core constraint node, establishing object-limiting relationships with definition-type content, execution-related relationships with process-type content, applicability-related relationships with case-type content, and support-related relationships with auxiliary-type content. Through these relationships, a complete knowledge chain can be formed from "rule requirements" to "applicable conditions" and then to "actual cases."

[0193] When a user submits an inquiry, such as: "Can this service be processed if an applicant's materials are incomplete but there are special circumstances?", This application first analyzes the semantic features of the query request to determine that the query object is a business processing application and the query behavior is processing a business. It further analyzes the contextual features of the query to determine that the current query involves conditions such as incomplete materials and special circumstances. Subsequently, this application matches the above query features with the retrieval adaptation unit. If only traditional keyword retrieval is used, it may only return the rule content of "processing is only possible after materials are complete," ignoring the crucial factor of special circumstances. However, when this application matches based on associated knowledge structures, it can further associate case-related content to discover cases where incomplete materials were processed but were completed after special handling; simultaneously, it associates auxiliary content to obtain explanatory evidence supporting this special handling method; finally, it combines the rule-related content to determine search results that better fit the current context.

[0194] Through the above comprehensive processing, this application realizes the transformation from rule text to knowledge association structure, and then to contextualized retrieval results, enabling different content forms in the rule document to participate in the judgment process.

[0195] As can be seen from the above embodiments, this application does not simply classify or search rule documents using keywords, but rather differentiates them based on the knowledge roles played by different content, creating an interconnected knowledge structure from rules, definitions, processes, cases, and supporting content. Specifically, case studies address how rules apply to specific contexts, while supporting content addresses how non-core content supports rule understanding; both collectively improve the accuracy and completeness of the rule document retrieval process.

[0196] In summary, this application addresses the issue of different content in rule documents playing different knowledge representation roles. Instead of treating rule documents as a collection of information composed of uniform text content, it analyzes the semantic attribution of different content across multiple candidate form modes, such as rule form, definition form, process form, case form, and auxiliary form, based on the structural features, textual expression features, and contextual association features corresponding to the document content units. This analysis determines the target form mode corresponding to each document content unit. Through this method, this application can identify the actual knowledge role played by different content in rule documents, avoiding the problem of indistinguishable content roles caused by uniform processing based solely on surface text features in traditional methods.

[0197] Furthermore, this application adopts a differentiated processing strategy for different morphological modes, and performs corresponding content morphological semantic representation on different document content units, so that different types of content form morphological semantic relationships that match their expressive functions. Specifically, regarding rule-based content, this application can extract the relationships between rule-constrained objects, constraints, and execution requirements, transforming rule content from simple textual descriptions into semantic relationships with applicable scope. Regarding definition-based content, this application can determine the association between conceptual objects and defined scopes, ensuring accurate understanding of the meaning of objects within the rules. Regarding process-based content, this application can identify the executing subject, operational actions, execution sequence, and state change relationships, enabling the rule execution process to be expressed in a structured manner. Regarding case-based content, this application can further analyze contextual factors such as the subject's state, behavioral process, occurrence conditions, environmental factors, and result state within the case, ensuring that case content is not merely rule explanation material but can reflect the applicability of rules under specific application conditions. Regarding auxiliary content, this application can identify auxiliary information that, while not directly forming rule constraints, conceptual definitions, process steps, or case descriptions, can influence rule understanding and association construction, enabling non-subjective content in rule documents to also play a supporting knowledge role.

[0198] Furthermore, this application determines the knowledge associations between different content forms by integrating the morphological semantic relationships corresponding to different document content units, and generates an associated knowledge structure corresponding to the rule document. Through this associated knowledge structure, the discrete text content arranged in chapter order in traditional rule documents can be transformed into an associated structure organized based on knowledge interaction relationships, effectively connecting rule content, definition content, process content, case content, and auxiliary content. For example, rule content can be associated with corresponding definition content to clarify the scope of application, with corresponding process content to determine the execution process, with corresponding case content to determine the applicable method in a specific context, and with corresponding auxiliary content to obtain supplementary evidence affecting the judgment. Therefore, this application improves the completeness of the knowledge organization of rule documents and avoids the problem of insufficient rule understanding caused by the fragmentation of different types of content in existing methods.

[0199] Furthermore, this application generates retrieval adaptation units based on associated knowledge structures, enabling the retrieval object to move beyond the original text fragments in the rule document and instead construct adaptation units that reflect the complete meaning of the rules based on the knowledge relationships formed between different content forms. When responding to a retrieval request, this application not only matches based on the semantic features of the query but also comprehensively judges the query context features combined with case association information, rule association information, and auxiliary association information in the retrieval adaptation units. For example, when a user queries whether a certain behavior meets the rule requirements under specific conditions, this application can comprehensively match the specific contextual factors in the case form, the constraints in the rule form, and the association evidence in the auxiliary form, rather than simply returning general rule content based on keywords. This improves the matching degree between the retrieval results and actual application scenarios.

[0200] Furthermore, this application enhances the adaptability of the rule document retrieval process to complex application scenarios by introducing case-based and auxiliary formats. The case-based format incorporates contextual factors from specific events into the rule judgment process, enabling abstract rules to determine their scope of application based on actual conditions. The auxiliary format retains and utilizes supporting explanatory content from the rule document, allowing information easily overlooked in traditional processing to participate in knowledge association construction. Through the synergistic effect of these two formats with the rule-based, definition-based, and process-based formats, this application achieves a more complete expression of rule knowledge.

[0201] Furthermore, this application achieves closed-loop optimization of the rule document processing process through a complete processing flow, from rule document content parsing, morphological pattern recognition, differentiated processing, knowledge association construction to retrieval matching. Compared with retrieval methods that rely solely on text similarity or keyword matching, this application can fully utilize the relationships between different contents within the rule document, enabling the retrieval results to not only include textual information related to the query content, but also to provide associated definitional information, process information, case information, and auxiliary information that influence the judgment, thereby improving the accuracy, completeness, and application adaptability of rule document processing.

[0202] Please see Figure 2 This application also provides a content-format-aware rule-based document parsing and extraction device, comprising: The document feature acquisition module is used to acquire multiple document content units in a rule document and extract the structural features, text expression features, and contextual features corresponding to each document content unit. The morphological pattern determination module is used to analyze the semantic belonging degree of each document content unit in multiple candidate morphological patterns based on the structural features, the text expression features and the context association features, determine the morphological belonging strength of each document content unit, and determine the target morphological pattern of the document content unit according to the morphological belonging strength. The differentiation processing module is used to perform content morphological semantic representation on the document content unit according to the differentiation processing strategy corresponding to the target morphological pattern, and obtain the corresponding morphological semantic relationship. The knowledge association construction module is used to perform fusion analysis on the semantic association features between different content forms based on the morphological semantic relationships corresponding to different document content units, determine the knowledge association relationships corresponding to different content forms, and generate the associated knowledge structure corresponding to the rule document. The retrieval processing module is used to generate a retrieval adaptation unit based on the associated knowledge structure, and in response to the retrieval request, match the retrieval adaptation unit with the query semantic features and query context features corresponding to the retrieval request, and extract the corresponding retrieval results.

[0203] 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 content-form awareness rule document parsing and extraction steps as described above in this application.

[0204] 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 content-form-aware rule document parsing and extraction method described above in this application.

[0205] 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 content-format-aware rule-based document parsing and extraction method, characterized in that, include: Obtain multiple document content units from the rule document, and extract the structural features, textual expression features, and contextual association features corresponding to each document content unit; Based on the structural features, text expression features, and context association features, the semantic attribution degree of each document content unit in multiple candidate morphological modes is analyzed to determine the morphological attribution strength of each document content unit, and the target morphological mode corresponding to the document content unit is determined according to the morphological attribution strength. Based on the differentiated processing strategy corresponding to the target morphological pattern, the document content unit is subjected to content morphological semantic representation to obtain the corresponding morphological semantic relationship; Based on the morphological semantic relationships corresponding to different document content units, the semantic association features between different content forms are fused and analyzed to determine the knowledge association relationships corresponding to different content forms and generate the associated knowledge structure corresponding to the rule document. Based on the aforementioned knowledge structure, a retrieval adaptation unit is generated. In response to a retrieval request, the retrieval adaptation unit is matched with the query semantic features and query context features corresponding to the retrieval request, and the corresponding retrieval results are extracted.

2. The method according to claim 1, characterized in that, The multiple document content units in the obtained rule document include: Based on the chapter structure, clause relationships, and text context relationships in the rule document, the content of the rule document is analyzed to identify multiple candidate semantic fragments; For each candidate semantic fragment, based on its scope of application in the rule document and its degree of association with other semantic fragments, a document content unit is determined to carry the same content form expression.

3. The method according to claim 1, characterized in that, The extraction of structural features, textual expression features, and contextual features corresponding to each document content unit includes: Based on the organizational position, clause level, and referencing relationship of the document content unit in the rule document, determine the structural features used to reflect the role of the content; Based on the restrictive expressions, definitional expressions, procedural expressions, and contextual expressions in the document content units, determine the textual expression features that reflect the semantic function of the content; Based on the interpretive, constraint, and applicability relationships between the document content units and other content units, contextual association features that reflect the content association effect are determined.

4. The method according to claim 1, characterized in that, Determining the morphological attribution strength corresponding to each document content unit includes: Based on the content expression rules corresponding to each candidate morphological pattern, the structural features, text expression features and contextual association features corresponding to the document content unit are matched and analyzed to determine the morphological matching result between the document content unit and each candidate morphological pattern. Based on the morphological matching results, further analyze the semantic role, association scope, and knowledge expression purpose of the document content unit in the rule document, and determine the morphological attribution strength used to characterize the degree to which the document content unit undertakes the corresponding content morphological function.

5. The method according to claim 1, characterized in that, Determining the target morphological pattern corresponding to the document content unit based on the morphological attribution strength includes: When the document content unit corresponds to multiple candidate morphological patterns, multiple candidate semantic interpretation paths are determined according to the morphological attribution strength corresponding to each candidate morphological pattern. For each of the candidate semantic interpretation paths, analyze the corresponding knowledge association results and their impact on the subsequent retrieval process; Based on the association effects corresponding to different candidate semantic interpretation paths, the target morphological pattern corresponding to the document content unit is determined, and subsequent processing is performed according to the processing path corresponding to the target morphological pattern.

6. The method according to claim 1, characterized in that, The step of processing the document content unit according to the differentiated processing path corresponding to the target form pattern includes: When the target form mode is a rule form mode, the restriction objects, restriction behaviors, triggering conditions and exception conditions in the document content unit are analyzed for correlation to determine the constraint range corresponding to the rule. The applicable boundaries of the rule are determined based on the reference and applicability relationships between the constraint scope and other related content in the rule document; Based on the applicable boundaries, rule retrieval association conditions are established so that the corresponding rule content can be matched according to the target object and applicable conditions during the subsequent retrieval process.

7. The method according to claim 1, characterized in that, The step of processing the document content unit according to the differentiated processing path corresponding to the target form pattern includes: When the target morphological pattern is a defined morphological pattern, the concept name, explanation, limiting conditions and associated objects in the document content unit are parsed to determine the semantic limitation scope of the corresponding concept; Based on the semantic definition and the reference relationship between the concept and other content in the rule document that involves the concept, determine the scope of applicability of the concept in different rule scenarios; Based on the aforementioned scope of application, the object matching conditions in the retrieval process are modified to avoid erroneous retrieval results caused by the expansion of the concept scope.

8. A content-format-aware rule-based document parsing and extraction device, characterized in that, include: The document feature acquisition module is used to acquire multiple document content units in a rule document and extract the structural features, text expression features, and contextual features corresponding to each document content unit. The morphological pattern determination module is used to analyze the semantic belonging degree of each document content unit in multiple candidate morphological patterns based on the structural features, the text expression features and the context association features, determine the morphological belonging strength of each document content unit, and determine the target morphological pattern of the document content unit according to the morphological belonging strength. The differentiation processing module is used to perform content morphological semantic representation on the document content unit according to the differentiation processing strategy corresponding to the target morphological pattern, and obtain the corresponding morphological semantic relationship. The knowledge association construction module is used to perform fusion analysis on the semantic association features between different content forms based on the morphological semantic relationships corresponding to different document content units, determine the knowledge association relationships corresponding to different content forms, and generate the associated knowledge structure corresponding to the rule document. The retrieval processing module is used to generate a retrieval adaptation unit based on the associated knowledge structure, and in response to the retrieval request, match the retrieval adaptation unit with the query semantic features and query context features corresponding to the retrieval request, and extract the corresponding retrieval results.

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 content-form-aware rule-based document parsing and extraction method as described 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, which, when executed by a processor, implements the steps of the content-form-aware rule-based document parsing and extraction method as described in any one of claims 1 to 7.

Images