Software requirement document consistency detection method and system based on semantic analysis

Abstract

The application relates to the technical field of software detection, in particular to a software requirement document consistency detection method and system based on semantic analysis, which comprises the following steps: obtaining requirement documents, requirement modification records, version evolution records and requirement correlation data; extracting business policy change features based on semantic analysis; constructing a requirement change propagation state model; establishing a requirement change propagation chain from modification content, business policy and business constraint to an influence domain; and comparing a theoretical propagation state sequence with an actual propagation state sequence through mapping, detecting a propagation chain breakage position, determining a propagation influence domain and outputting policy synchronization repair information. The application realizes consistency detection on a requirement change propagation process, can identify implicit consistency failure problems caused by incomplete propagation in the requirement change process, and improves the accuracy and effectiveness of software requirement document consistency detection.

Description

Technical Field

[0001] This invention relates to the field of software inspection, specifically a method and system for detecting consistency in software requirements documents based on semantic analysis. Background Technology

[0002] Software requirements documents are the technical documentation that describes business functions, system behavior, and implementation constraints. In the context of agile development and continuous iteration, requirements change frequently and intertwine into complex networks. Therefore, consistency checks on requirements documents are a crucial means of ensuring software quality.

[0003] Currently, software requirements document consistency testing typically relies on technologies such as semantic analysis, natural language processing, knowledge representation, or requirements correlation analysis. These technologies parse the terminology, entity objects, contextual semantics, logical relationships, and citations within the requirements document. By constructing relationships between requirements, they compare and analyze the content of requirements across different chapters, versions, or documents to identify issues such as inconsistent terminology, semantic conflicts, logical contradictions, citation errors, and missing requirements. Some solutions also integrate version control systems to record requirement modification history and leverage requirement dependencies to assist in analyzing the impact of requirement changes on other requirements, thereby generating requirements consistency testing results and providing technical support for software development, testing, and maintenance.

[0004] The inventors of this application discovered that the aforementioned technical solutions primarily use the text content of the current version of the requirements document as the detection object, assuming that all related requirements are in the same evolutionary state. Therefore, the detection focus remains on the terminology, semantics, and logical relationships themselves, lacking a dynamic description of the requirement change propagation process. When software development enters a continuous iteration phase, what truly changes in a requirement modification is not a single piece of text, but rather its corresponding business strategies, business constraints, and system behavior rules. These changes propagate layer by layer along the relationships between requirements to multiple related requirements. In actual development, some related requirements may have already been updated synchronously, while others retain the old version content, resulting in a break in the requirement change propagation process. In layman's terms, it's like modifying a main rule; many downstream rules that depend on it should be updated together, but some have already been modified while others remain in the old state. Although on the surface every document can be read normally and seems to have no obvious errors, in reality, the update has "broken halfway through." Existing technologies lack the ability to model the propagation status, path, and integrity of requirement changes. They cannot identify the propagation process of business strategy changes spreading layer by layer along relationships, nor can they determine at which node or relationship the propagation is interrupted. Therefore, it is difficult to detect implicit consistency failures caused by incomplete requirement change propagation before explicit semantic conflicts arise. To address these issues, there is an urgent need for a software requirement document consistency detection method and system capable of establishing a requirement change propagation status model, dynamically simulating the requirement change propagation process, and detecting breaks in the propagation chain. Summary of the Invention

[0005] This application provides a software requirement document consistency detection method and system based on semantic analysis, which can transform the detection of software requirement changes from a single text modification to a propagation consistency detection process based on requirement relationships. It can identify state evolution, propagation deviation, and synchronization loss during the propagation of requirement changes, thereby improving the accuracy and completeness of software requirement document consistency detection.

[0006] To achieve the above objectives, the embodiments of this application disclose the following technical solutions:

[0007] On the one hand, this solution discloses a software requirements document consistency detection method based on semantic analysis. By acquiring relevant data from the requirements document, it identifies the characteristics of business strategy changes and establishes a requirements change propagation state model based on requirements relationships. This model enables requirements changes to propagate from modified content to the affected domain according to these relationships. Furthermore, by generating theoretical and actual propagation state sequences and comparing them, the consistency of requirements change propagation is detected.

[0008] This solution establishes a correlation between the requirement change propagation process and the actual requirement modification status, enabling software requirement consistency detection to expand from static text comparison to requirement change propagation process analysis, thereby improving the ability to detect missing requirement synchronization.

[0009] Furthermore, by analyzing the requirement reference relationships, business process relationships, interface call relationships, test case relationships, and historical version synchronization records in the requirement association data, a propagation potential field model is established, and the propagation status between requirement nodes is adjusted according to changes in the propagation potential field.

[0010] By introducing a propagation potential field model, the correlation strength and historical synchronization behavior between different demand nodes can be used to calculate the demand change propagation state, thereby reflecting the differences in demand change propagation capabilities under different correlation relationships.

[0011] Furthermore, by combining the remaining propagation potential energy to determine the theoretical propagation state and analyzing the state changes between continuous propagation nodes, the location of the propagation chain break is determined based on the deviation between the actual response state and the theoretical propagation state.

[0012] By analyzing the state changes of continuously propagating nodes, we can avoid making judgments based solely on deviations from the state of a single node, thereby improving the reliability of identifying anomalies in the propagation of demand changes.

[0013] Furthermore, a synchronous twin evolution model is established in the digital space between the theoretical propagation trajectory and the actual evolution trajectory, and the propagation state model is updated according to the differences between the theoretical trajectory and the actual trajectory to meet the needs of change.

[0014] By utilizing the dynamic synchronization between the theoretical propagation process and the actual evolution process, the requirement change propagation model can be continuously revised, enabling the propagation consistency detection to adapt to the continuous iteration process of software projects.

[0015] Furthermore, by constructing parameters for changes in the propagation environment and combining them with time-varying deviation vectors, the uncertainty perturbations and state transition relationships in the synchronous twin evolution model are adjusted.

[0016] By introducing factors related to changes in the project iteration environment, the demand propagation state model can be dynamically adjusted according to the actual project evolution, thereby improving its detection and adaptability during long-term operation.

[0017] Furthermore, the current effective propagation path is identified based on the version branch merging relationship and the requirement substitution relationship, and the theoretical propagation path of abnormal interruption is reconstructed and restored.

[0018] By correcting the changes in propagation paths that occur during collaborative software development, misjudgments caused by version branch changes or collaboration processes are reduced, thereby improving the accuracy of propagation consistency detection results.

[0019] Furthermore, by analyzing the archiving markers of associated nodes, proactive version replacement instructions, and actual propagation status change trends, we can identify state deviations that are part of normal business evolution.

[0020] By filtering out changes in propagation status caused by legitimate business evolution, the impact of non-abnormal modification behavior on the propagation chain break detection results is reduced.

[0021] Furthermore, by analyzing the scope of affected demand chapters based on the location of the propagation chain break, and generating a strategy omission and correction differential compensation template based on the difference between the theoretical propagation state sequence and the actual propagation state sequence.

[0022] By further converting the results of propagation chain break detection into policy synchronization repair information, an information closed loop is achieved from demand anomaly identification to repair guidance.

[0023] Furthermore, the system detects propagation and repair behaviors through the software development collaboration platform interface and performs pre-emptive interception control if the corresponding repair is not completed.

[0024] By linking the results of requirement consistency checks with the software development collaboration process, the linkage between requirement synchronization status checks and subsequent development process control is achieved, thereby improving the closed-loop nature of requirement change management.

[0025] On the other hand, this solution discloses a software requirement document consistency detection method and system based on semantic analysis, which forms a complete detection system through a data acquisition and identification module, a requirement change propagation state modeling module, a propagation potential field calculation module, a propagation state deduction module, a propagation state adaptive update module, a propagation consistency detection module, a propagation path reconstruction module, and a propagation influence domain analysis module.

[0026] This technical solution establishes a demand change propagation status analysis link, enabling demand modification information to be propagated in conjunction with demand relationships. It also achieves demand consistency detection based on the difference between the theoretical propagation status and the actual propagation status, thereby solving the problem that traditional detection methods are unable to identify anomalies in demand change propagation.

[0027] This invention addresses the shortcomings of existing software requirement document consistency detection methods, which primarily focus on semantic consistency and struggle to identify implicit consistency failures caused by incomplete synchronous propagation during requirement changes. It proposes a semantic analysis-based method and system for software requirement document consistency detection. This method acquires requirement documents, requirement modification records, version evolution records, and related requirement data. Through semantic analysis, it extracts the business strategy change characteristics corresponding to the source requirement modifications and constructs a requirement change propagation state model based on the dependencies between requirements. This model establishes a hierarchical requirement change propagation chain, from modified content, business strategies, and business constraints to the impact domain. Furthermore, it maps and compares theoretical and actual propagation state sequences to determine the locations of propagation chain breaks and outputs software requirement consistency detection results containing policy synchronization and repair information based on the propagation impact domain. By expanding the consistency detection object from static requirement text to the dynamic requirement change propagation process, this method enables the detection of the synchronous integrity of requirement changes and can identify potential consistency failures before obvious semantic conflicts arise in the requirement documents. In short, when a requirement in a software project is modified, this solution not only determines whether the modified text is correct, but also whether the modification has affected all related requirements. This avoids situations where "one part is changed, but related functions, interfaces, or test content remain the old version," thus improving the consistency detection capability in the software requirement management process. Attached Figure Description

[0028] Figure 1 This is a flowchart of the overall method for software requirements document consistency detection according to an embodiment of the present invention;

[0029] Figure 2 This is a schematic diagram of the demand change propagation state model and propagation potential field construction in an embodiment of the present invention;

[0030] Figure 3 This is a flowchart illustrating the refined determination of the propagation chain break location according to an embodiment of the present invention.

[0031] Figure 4 This is a flowchart of the digital twin adaptive correction and propagation path reconstruction process according to an embodiment of the present invention;

[0032] Figure 5 This is a flowchart of the consistency detection result output and closed-loop interception control in an embodiment of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In the following description, numerous specific details are set forth to provide a comprehensive understanding of the present invention. The present invention may be practiced without some or all of these specific details. In other instances, well-known processes have not been described in detail to avoid unnecessarily obscuring the present invention.

[0034] When used in conjunction with the terms "comprising," "method comprising," or similar language in this specification and appended claims, the singular forms "a," "some," and "the" include plural references unless the context clearly indicates otherwise. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0035] Terminology definition:

[0036] The requirement change propagation state model is a digital state machine model used to describe the dynamic propagation process of software requirement changes in a requirement association topology network. It transforms the association relationships between static requirement texts into a dynamic topology mapping system with time evolution capabilities and state transition relationships, and is used to uniformly describe the propagation state and evolution relationship between each requirement node during the requirement change propagation process.

[0037] The propagation potential field model is a state transmission resistance and propagation energy field model built on the correlation strength of demand nodes and historical synchronization behavior. It is used to quantify the propagation attenuation characteristics, propagation capability changes, and activation constraint capabilities on downstream demand nodes during the demand propagation process of business strategy changes, providing propagation constraints for propagation state updates.

[0038] The demand propagation digital twin model is a two-way mirror closed-loop model that synchronously maps the theoretical propagation trajectory and the actual propagation trajectory in the digital space. It is used to continuously obtain the propagation deviation between the two and dynamically correct the propagation state transition parameters in combination with the project iteration environment.

[0039] The propagation path reconstruction model is a propagation path correction model that combines version branch evolution relationship, demand substitution relationship and multi-person concurrent collaboration topology. It is used to identify and restore false propagation interruptions caused by multi-source collaboration and filter false propagation break signals caused by non-real demand changes.

[0040] The strategy omission differential compensation template is a structured propagation repair template automatically generated based on the propagation difference between the theoretical propagation state and the actual propagation state. It is used to guide the revision of the chapters of requirements that have not been modified synchronously, and serves as the data basis for subsequent strategy synchronous repair information generation, software development collaboration platform interception and control, and requirement work order generation.

[0041] Example 1

[0042] This embodiment provides a software requirements document consistency detection method based on semantic analysis. By establishing a requirement change propagation state model, the requirement modification process is extended from static text consistency analysis to dynamic propagation process analysis. It combines a propagation potential field model to quantify propagation capability, a requirement propagation digital twin model to adaptively correct the propagation state, and a propagation path reconstruction model to restore effective propagation paths. This enables dynamic propagation trajectory deduction of requirement changes, precise location of propagation chain breaks, and closed-loop control for propagation repair. The method includes the following steps:

[0043] Step 1: Obtain the target project's requirements document, requirement modification history, version evolution history, and requirement-related data, and extract the characteristics of business strategy changes.

[0044] In this embodiment, the system first establishes a software requirements data acquisition environment. Through open data interfaces, it calls data interfaces from a requirements management system (such as Jira, requirements specifications, and user story documents), a version control system (such as GitLab commit logs, branch records, and version snapshots), and a test management system to acquire requirements documents, requirement modification records, version evolution records, and requirement-related data generated during the target project's development process. A unified data index is then established according to the requirement number. Subsequently, the system parses the requirement-related data, extracting requirement reference relationships, business process relationships, interface call relationships, test case relationships, and historical version synchronization records. These are then integrated to form a requirement-related topology, which is further used to construct a corresponding multi-source-related topology matrix.

[0045] Based on semantic representation learning, context dependency analysis, and business entity recognition, the requirement text is preprocessed to complete text segmentation, lexical analysis, and entity recognition. This identifies business objects, actions, constraints, and objectives within the requirement, extracting the core business rule changes before and after the requirement modification. For example, when the requirement text is changed from "all users log in using SMS verification codes" to "high-risk accounts log in using multi-factor authentication," the system recognizes that the business strategy has changed from a unified authentication strategy to a risk-based authentication strategy, thereby establishing corresponding semantic tags and forming a set of business strategy change features. These business strategy change features, along with the requirement number, version number, and associated requirement number, are encapsulated to form requirement propagation source data, serving as a standardized data input interface package for subsequent propagation state modeling and potential initialization.

[0046] Step 2: Construct a demand change propagation state model, a propagation potential field model, and a propagation state update mechanism.

[0047] After obtaining the source data for requirement propagation, this embodiment analyzes the requirement number, version number, associated requirement number, and business strategy change characteristics contained in the source data. It further combines this data with requirement association data to construct a requirement change propagation state model, which describes the propagation pattern of requirement modifications throughout the entire requirement network. This embodiment no longer treats the relationships between requirements as ordinary reference relationships, but rather abstracts the requirement association network into a dynamic propagation network with propagation capabilities. Each requirement node represents a requirement object, and each association edge represents a business propagation relationship between requirements. Different association edges correspond to different propagation capabilities.

[0048] First, based on the requirement association data, the reference relationships, business process relationships, interface call relationships, test relationships, and historical version synchronization records between requirements are analyzed, and a unified requirement association topology is established. Each requirement node stores the current business strategy change status, business constraint synchronization status, related requirement response status, and version evolution status for that requirement, while nodes maintain propagation connections through association topology edges. Subsequently, a requirement change propagation state model is established within the entire requirement association topology, enabling business strategy changes to drive changes in business constraint synchronization status, related requirement response status, and version evolution status step-by-step according to business rules, thus forming a propagation chain that propagates layer by layer from modified content, business strategy, business constraints to the affected domain.

[0049] Based on requirement-related data, a requirement change propagation state model and a propagation potential field model are constructed, abstracting the requirement-related network into a dynamic topology network with different propagation capabilities. The system statistically analyzes the reference relationships between requirement nodes, business process relationships, interface call relationships, test case relationships, and historical version synchronization records to establish association weights, and generates the propagation potential field distribution in the propagation network based on the association weights.

[0050] To describe the propagation capability of business strategy changes in the propagation network, this embodiment further converts the characteristics of business strategy changes into initial propagation potential energy. Initial propagation potential energy can be understood as the propagation capability carried by the requirement modification, which is continuously constrained by the propagation potential field during propagation and decays with changes in propagation distance and relationships. During propagation, at each associated node, the system recalculates the current propagation potential energy based on the propagation decay function in the propagation potential field model, and continues to pass the calculated remaining propagation potential energy to the next associated node.

[0051] In this embodiment, the propagation of business strategy change characteristics in the demand association network does not decay at a fixed rate, but is simultaneously influenced by association weights, propagation path lengths, and propagation resistance between different demand nodes. Association weights reflect the degree of business dependence between two demand nodes, propagation path lengths reflect the number of intermediate demand nodes through which the business strategy change passes, and propagation resistance reflects the difficulty of synchronizing demand modifications across different R&D stages. Therefore, the propagation process of business strategy change can be abstracted as a potential energy decay process constrained by association weights.

[0052] Specifically, the initial propagation potential energy formed by the transformation of business strategy change characteristics is used as the input of the propagation source. When the propagation potential energy is transmitted from the current demand node to the next associated demand node along the demand change propagation chain, it is constrained by the associated weight of the current associated topology edge and the propagation resistance, and the propagation potential energy gradually decreases. Based on the mapping relationship between the energy maintenance and attenuation law during the continuous transmission of propagation potential energy, a recursive equation for propagation potential energy is constructed:

[0053] ;

[0054] This leads to the remaining propagation potential energy received by the j-th associated demand node during the demand change propagation process:

[0055] ;

[0056] The first equation describes the propagation potential energy decay process along a single propagation edge, while the second equation describes the change in business strategy along the complete propagation path. The cumulative decay process during continuous propagation across multiple nodes.

[0057] in, This represents the remaining propagation potential energy obtained by the j-th associated demand node, which is used to characterize the impact of business strategy changes on the demand node after propagation. It is calculated by the demand change propagation state model.

[0058] This represents the propagation potential energy carried by the i-th demand node when it propagates to downstream nodes, which originates from the transformation result of business strategy change characteristics;

[0059] The initial propagation potential energy generated by the modification of the source requirement is first quantified based on the business strategy change characteristics obtained in Step 1. Specifically, the number of changes in business objects, business rules, and business constraints are counted separately, and the impact of each change is calculated in conjunction with the semantic impact range of the corresponding modification content. Then, the initial propagation potential energy corresponding to the source requirement is generated according to a weighted fusion method. The number of changes in business objects, business rules, and business constraints can all be automatically obtained from the semantic analysis results of the requirement document, and the semantic impact range is calculated based on the number of related requirement nodes corresponding to the modification content.

[0060] This represents the association weight between requirement node i and requirement node j, used to describe the propagation association strength between the two requirement nodes. It is obtained through requirement association data, including the number of references between requirement chapters, the number of business process associations, the number of interface call relationships, the number of test case associations, and the number of successful historical version synchronizations, and is obtained after normalization processing.

[0061] This represents the propagation path length between demand node i and demand node j, and is used to represent the number of propagation levels through which business strategy changes pass. It is obtained by path search through the associated topology in the demand change propagation state model.

[0062] This represents the propagation path length attenuation coefficient, used to describe the degree of influence of path length on propagation capability under different types of demand relationships. It is obtained by statistically analyzing the synchronization success rate corresponding to different propagation distances in the historical version evolution record.

[0063] This represents the propagation resistance between requirement node i and requirement node j, taking into account the number of version branches between currently related requirements, the number of multi-person collaboration conflicts, and the historical average synchronization latency. The number of version branches is obtained from the version control system, the number of multi-person collaboration conflicts is obtained from version conflict records, and the historical average synchronization latency is calculated based on the requirement synchronization timestamp. The system performs normalization on the above indicators and merges them according to preset weights to generate the propagation resistance.

[0064] This represents the propagation resistance adjustment coefficient, used to adjust the proportion of the influence of propagation resistance on the propagation potential energy attenuation process. It is adaptively adjusted according to the changes in propagation environment parameters. When the changes in propagation environment parameters increase, the influence of propagation resistance on propagation potential energy attenuation is increased; when the changes in propagation environment parameters decrease, the influence of propagation resistance is reduced accordingly.

[0065] This represents the set of demand change propagation paths from the source demand node to the target demand node, which is obtained by searching for effective propagation paths in the demand association topology.

[0066] The model described above incorporates requirement association weights into the propagation potential decay process, preventing requirement changes from adopting a uniform diffusion approach. Instead, it dynamically adjusts propagation capabilities based on the actual requirement association network. When there is a strong business dependency between two requirement nodes, their association weight is high, and the propagation potential decay is small. When there is only a weak association between requirement nodes, or when propagation occurs indirectly through multiple version nodes, the propagation potential continuously decreases, thus better aligning with the actual pattern of requirement changes spreading layer by layer during software development.

[0067] After updating the propagation potential, the system dynamically updates the state transition relationships of the corresponding nodes in the demand change propagation state model based on the remaining propagation potential. When a node receives a high propagation potential, it indicates that the current demand has a high propagation capability, and its business strategy changes can continue to propagate to downstream demands. When a node receives a low propagation potential, it indicates that the propagation capability has significantly weakened, and the probability of subsequent node propagation decreases accordingly. Through this method, different demand nodes can form differentiated propagation states, rather than using a uniform propagation approach.

[0068] To ensure that the propagation state model can reflect the actual R&D process, this embodiment further continuously corrects the transition relationship between propagation states based on historical version synchronization records, so that the correlation relationship with a high long-term synchronization frequency maintains a high propagation capability, while the demand relationship with a long-term synchronization failure or a weak correlation gradually reduces the propagation capability.

[0069] Step 3: Generate a theoretical propagation state sequence based on the demand change propagation state model, and perform adaptive correction by combining it with the demand propagation digital twin model.

[0070] In this embodiment, after establishing the requirement change propagation state model, the requirement change propagation state model, propagation potential field model, propagation chain structure, and state transition relationships corresponding to each associated requirement node, output from step 2, are used as input to dynamically deduce the propagation process of business strategy change characteristics throughout the entire requirement change propagation chain, thereby obtaining a theoretical propagation state sequence that reflects the theoretical propagation process of requirement changes. Simultaneously, semantic analysis is performed again on the current version of the requirement document for the target project to obtain the actual modification characteristics corresponding to each associated requirement node, constructing an actual propagation state sequence. Furthermore, the requirement propagation digital twin model is used to continuously synchronize the theoretical propagation trajectory and the actual evolution trajectory, enabling the requirement change propagation state model to continuously and adaptively correct itself as the project evolves, providing a dynamically updated theoretical reference for subsequent propagation chain break detection.

[0071] In practice, the business strategy change state output in step 2 is first used as the propagation starting state, and the state transition relationship in the requirement change propagation state model is used as the propagation constraint. Forward deduction is then performed sequentially on each associated requirement node in the propagation chain. For each associated requirement node, the system reads the propagation potential field strength, remaining propagation potential energy, and state transition relationship corresponding to that node. Based on the current node's propagation state, the theoretical propagation state of the next node is determined. This allows the business strategy change state to propagate according to the business strategy change state, business constraint synchronization state, associated requirement response state, and version evolution state, ultimately forming a theoretical propagation state sequence covering the entire requirement association network.

[0072] To ensure that the theoretical propagation state sequence can truly reflect the continuous evolution of software requirements, during the propagation simulation, after each node state update is completed, the new propagation state is fed back to the requirement change propagation state model. This ensures that the next node state simulation is always based on the latest propagation state, rather than using a one-time static calculation method to complete the entire propagation process. This forms a dynamic propagation process that is recursively pushed from node to node and updated layer by layer.

[0073] Preferably, during state deduction, each requirement node can be abstracted into a propagation state unit. Each propagation state unit includes at least the current propagation state, remaining propagation capacity, propagation direction, and propagation history information. Subsequently, according to the topology of the requirement change propagation chain, a topological sort is performed on all propagation state units to ensure that the propagation state is always updated according to the business dependency direction, avoiding repeated calculations of the propagation state due to circular dependencies between nodes.

[0074] To describe the above-described propagation state update process, a propagation state recursion model can be established here in this embodiment.

[0075] On the other hand, to obtain the true evolution status of software requirements, this embodiment simultaneously performs semantic analysis on the current version of the requirements document. Specifically, firstly, all requirement chapters of the current version of the target project are obtained, and semantic parsing is performed on the requirement chapters to identify business objects, business actions, business rules, interface definitions, and constraints in the chapters. Combined with the business strategy change characteristics obtained in step 1, it is determined whether each requirement node has responded to the current business strategy change.

[0076] Furthermore, for each associated requirement node, the corresponding current version modification content, historical version modification records, and requirement modification records are extracted, and combined with the version evolution record to determine whether the node has completed business strategy synchronization. When it is detected that a requirement node has made corresponding modifications according to the current business strategy, the node is marked as responsive; if the requirement node still retains the historical business strategy content, it is marked as unresponsive; if only partial modifications have been completed, it is marked as partially responsive. Subsequently, according to the node order corresponding to the theoretical propagation state sequence, the response states of all nodes are combined to form the actual propagation state sequence.

[0077] To avoid the inability to directly correspond the theoretical propagation status with the actual propagation status due to different release times of different versions, this embodiment further establishes a digital twin model of demand propagation.

[0078] Further establish a digital twin model of demand propagation, and establish a synchronous evolution model in digital space of the theoretical propagation trajectory composed of the theoretical propagation state sequence and the actual evolution trajectory composed of the actual propagation state sequence.

[0079] In practical implementation, the digital twin model first constructs a theoretical propagation trajectory based on the theoretical propagation state sequence, and then constructs an actual evolution trajectory based on the actual propagation state sequence. A one-to-one correspondence is established between the two propagation trajectories according to the requirement node identifier, version number, propagation level, and propagation time. As the project iterates continuously, new versions of the requirement documents are continuously added to the digital twin model, the actual evolution trajectory is continuously updated, and the theoretical propagation trajectory is synchronously recalculated based on the propagation state model, thus achieving continuous synchronization between the theoretical space and the real space.

[0080] To quantify the difference between the theoretical propagation trajectory and the actual evolution trajectory, this embodiment further calculates the time-varying deviation vector.

[0081] In this embodiment, the theoretical propagation trajectory originates from the theoretical propagation state sequence derived through continuous deduction of the requirement change propagation state model, while the actual evolution trajectory originates from the actual propagation state sequence formed by parsing the current version of the requirement document. Due to version iterations, collaborative modifications by multiple users, and asynchronous requirement propagation during software development, the difference between the theoretical propagation trajectory and the actual evolution trajectory is usually not merely a difference in node states, but also involves both propagation time and propagation level offsets. If the model is corrected based solely on a single state difference, it is difficult to accurately reflect the true error of the propagation model during dynamic evolution.

[0082] Therefore, this embodiment maps the differences in node states, propagation time, and propagation levels between the theoretical propagation trajectory and the actual evolution trajectory to a unified propagation deviation space, and constructs a time-varying deviation vector, enabling propagation deviations from different sources to be fed back to the demand propagation digital twin model in a unified form. Based on the mapping relationship between the theoretical propagation trajectory and the actual evolution trajectory, a time-varying deviation vector calculation equation is constructed, resulting in:

[0083] ;

[0084] in, ;

[0085] The time-varying deviation vector corresponding to time t is used to comprehensively characterize the overall deviation between the theoretical propagation trajectory and the actual evolution trajectory. It is calculated by the demand propagation digital twin model and used as a feedback quantity to update the demand change propagation state model.

[0086] N represents the number of related demand nodes participating in the current propagation analysis, which is obtained statistically based on the propagation chain structure in the demand change propagation state model.

[0087] The propagation influence weight of the i-th associated demand node describes its importance in the overall propagation process. It is calculated based on the association weight, propagation potential field strength, and remaining propagation potential energy in the demand association topology established in step 2, and then normalized. The node influence weight is calculated based on the node's out-degree, the average association weight, and the historical synchronization frequency. Specifically, the node's out-degree is obtained statistically from the association topology, the average association weight is calculated based on the propagation potential field model, and the historical synchronization frequency is obtained statistically from historical version synchronization records. These are then normalized to generate the node influence weight.

[0088] This represents the difference in node state between the theoretical propagation state and the actual propagation state of the i-th associated demand node at time t. It is calculated based on the propagation state of the corresponding node in the theoretical propagation state sequence and the actual propagation state sequence.

[0089] The normalized baseline value represents the difference in the propagation state of the i-th associated demand node. It is obtained from the historical propagation state statistics and is used to eliminate the difference in the units of different propagation states.

[0090] This represents the difference in propagation time between the theoretical propagation time and the actual propagation time of the i-th related requirement node, which is calculated based on the version submission time, requirement modification time and version release time in the version evolution record.

[0091] This represents the normalized baseline value for the difference in propagation time, which is obtained statistically based on the historical version evolution record of the current project.

[0092] This represents the difference in propagation level between the theoretical and actual propagation levels of the i-th associated demand node, which is calculated based on the propagation chain depth and propagation mapping relationship in the demand change propagation state model.

[0093] This represents the normalized baseline value for the propagation hierarchy difference, which is determined by the maximum propagation hierarchy in the current demand association network.

[0094] These represent the node state difference weight, propagation time difference weight, and propagation level difference weight, respectively, which are used to adjust the proportion of the influence of different types of propagation deviation on the time-varying deviation vector. They are calculated based on historical propagation deviation data using an adaptive normalization method, and satisfy the condition that the sum of the three is 1.

[0095] The deviation direction vector of the i-th associated demand node in the propagation deviation space is used to represent the propagation direction corresponding to the propagation deviation. It is generated based on the propagation direction relationship between the theoretical propagation trajectory and the actual evolution trajectory. Each demand node corresponds to a propagation direction vector, which is generated based on the propagation direction in the demand change propagation chain. Its components correspond to the business strategy propagation direction, the business constraint propagation direction, and the associated demand propagation direction, respectively. After normalization, the node direction vector is formed.

[0096] The above calculations yield a time-varying deviation vector between the theoretical propagation trajectory and the actual evolution trajectory during the current demand change propagation process. The digital twin model continuously corrects the state transition relationships in the demand change propagation state model based on this time-varying deviation vector, ensuring that the theoretical propagation trajectory increasingly approximates the actual evolution trajectory, thereby improving the accuracy and stability of subsequent propagation chain break detection.

[0097] The above calculations yield a time-varying deviation vector that reflects the predictive capability of the current propagation model. This vector not only reflects whether a node has deviated, but also the location, direction, and duration of the deviation, providing feedback for adaptive correction of the propagation model.

[0098] Furthermore, this embodiment constructs propagation environment change parameters based on the target project's operating environment. Specifically, it obtains the target project's current iteration cycle, requirement modification frequency, number of version branches, and historical propagation deviation data. The project iteration cycle reflects the speed of requirement updates; the requirement modification frequency reflects the activity level of requirement changes; the number of version branches reflects the complexity of multi-person collaborative development; and the historical propagation deviation data reflects the accumulation of historical propagation prediction errors. Subsequently, the above data is uniformly converted into propagation environment change parameters.

[0099] The parameters related to changes in the propagation environment do not directly participate in the propagation state calculation. Instead, they serve as the basis for correcting environmental disturbances in the digital twin model, dynamically updating the internal parameters of the propagation model. When the project enters a high-frequency demand iteration phase, the parameters related to changes in the propagation environment can improve the sensitivity of propagation state updates; when the project enters a stable maintenance phase, the magnitude of propagation state updates is reduced, ensuring that the propagation model always remains consistent with the actual R&D pace.

[0100] To enable continuous learning of the propagation model, this embodiment further adjusts the environmental disturbances in the digital twin model based on the parameters of the propagation environment change, and updates the demand change propagation state model together with the time-varying deviation vector.

[0101] The system re-derives the theoretical propagation state sequence based on the updated state transition relationship and inputs it into the twin model for comparison again, thereby achieving continuous cyclical correction of the model.

[0102] Step 4: Establish propagation mapping relationship, determine the location of propagation chain break, and perform propagation path reconstruction and pseudo-propagation break signal filtering.

[0103] Before comparing theoretical and actual sequences, a propagation mapping matrix is ​​first established, which includes node mapping, time mapping, and propagation hierarchy mapping. Specifically, node mapping is established based on the unique identifier of the requirement and the interface object; version time mapping is established based on the version evolution record; and propagation hierarchy mapping is established based on the propagation chain depth, thereby generating a unified mapping matrix as the basis for state comparison.

[0104] Preferably, in this embodiment, a unified propagation mapping matrix can be constructed based on the above three mapping relationships.

[0105] After the propagation mapping relationship is established, a mapping comparison is performed between the theoretical propagation state sequence and the actual propagation state sequence.

[0106] Specifically, the system reads the theoretical propagation state node by node, starting from the source node, according to the propagation chain direction. Simultaneously, it reads the corresponding actual propagation state based on the propagation mapping relationship. For each propagation node, it compares the theoretical and actual propagation states to see if they are consistent in terms of propagation state, propagation time, and propagation level. When all three dimensions are consistent, the node is considered to be propagating normally; if any dimension deviates, the system records the propagation deviation information for the current node and continues the analysis down to the next downstream node.

[0107] To avoid misjudgment caused by accidental changes in local nodes, this embodiment does not directly determine the breakage of the propagation chain based on the deviation of a single node, but further calculates the deviation degree of the propagation state.

[0108] The deviation of the propagation state takes into account the degree of difference in propagation state, propagation duration, changes in propagation level, and continuity of propagation path, so that the deviation can more realistically reflect the degree of propagation distortion in the process of demand propagation.

[0109] Since the theoretical propagation state sequence and the actual propagation state sequence have established a unified propagation mapping relationship through node mapping, time mapping, and propagation level mapping, there is a one-to-one correspondence between each propagation node. However, simply comparing the differences in node states cannot reflect the true degree of propagation distortion caused by the combined effects of propagation potential field attenuation, propagation level accumulation, and time delay during the propagation process. Therefore, this embodiment further introduces a propagation consistency mapping constraint, using the remaining propagation potential energy obtained in step 2 as the propagation capability constraint, and the theoretical propagation trajectory and actual evolution trajectory generated in step 3 as the propagation reference. A propagation deviation function is constructed through node state deviation, propagation time deviation, and propagation level deviation, so that the deviation degree of different propagation nodes can be uniformly mapped to the same evaluation space.

[0110] Furthermore, considering that nodes with higher propagation potential energy theoretically have a higher probability of successful propagation, deviations from the same state should have a greater anomaly weight in regions with high propagation potential energy; while regions with lower propagation potential energy are allowed a certain degree of propagation attenuation. Based on the above propagation constraints, a propagation deviation equation is constructed, leading to:

[0111] ;

[0112] When the propagation chain contains consecutive propagation nodes, to avoid misjudgment caused by accidental deviation of a single node, this embodiment further constructs a cumulative deviation function for the propagation chain:

[0113] ;

[0114] Among them, when the cumulative propagation deviation corresponding to the continuous propagation node continuously exceeds the propagation consistency threshold, and at the same time meets the downward trend of propagation status and the absence of version response, it serves as an important basis for determining the location of subsequent propagation chain break.

[0115] The propagation consistency threshold is determined based on the statistical distribution of propagation deviation of historical normal projects. Specifically, the propagation deviation of each propagation node during the historical normal demand propagation process is statistically analyzed, and the percentile value of the corresponding propagation deviation distribution is calculated. The corresponding percentile value is used as the propagation consistency threshold; or, based on historical propagation deviation samples, a propagation consistency threshold that can simultaneously take into account the false alarm rate and the false negative rate is obtained through cross-validation.

[0116] In the formula, This represents the propagation deviation of the i-th propagation node, used to quantify the degree of inconsistency between the theoretical propagation state and the actual propagation state. It is calculated from the propagation mapping relationship.

[0117] This represents the theoretical propagation state corresponding to the i-th propagation node in the theoretical propagation state sequence, which is generated by forward derivation of the demand change propagation state model in step 3.

[0118] This indicates the actual propagation state of the corresponding propagation node in the actual propagation state sequence, which is obtained through semantic analysis of the current version of the requirement document, requirement modification records, and version evolution records.

[0119] This represents a stabilizing compensation factor to prevent the denominator from being zero; a very small positive number can be used to achieve numerical stability. The theoretical propagation time corresponding to the theoretical propagation node is obtained from the theoretical propagation trajectory record; This indicates the actual propagation time corresponding to the actual propagation node, which is obtained by parsing the version submission time, requirement modification time, and version release time in the version control system. This indicates the standard propagation cycle corresponding to the current target project, which is obtained from project iteration cycle statistics. This represents the theoretical propagation level, which is calculated by changing the propagation chain topology depth according to requirements. This indicates the actual propagation level, obtained by parsing the propagation path based on the requirements of the current version. This represents the maximum propagation level in the current propagation chain, obtained from demand-related topology analysis; This represents the remaining propagation potential energy of the i-th propagation node after the propagation potential energy decay calculation in step 2, which is calculated by the propagation potential field model. This represents the initial propagation potential energy of the propagation source node, obtained by transforming the characteristics of changes in business strategies. This represents the propagation path continuity coefficient, calculated based on the consistency of propagation direction, the continuity of propagation hierarchy, and the recovery status of propagation state between the current node and its adjacent nodes. Specifically, the consistency of propagation direction is calculated based on the angle between propagation direction vectors; the continuity of propagation hierarchy is obtained statistically based on the propagation hierarchy mapping relationship; and the recovery status of propagation state is determined by whether subsequent nodes have re-reached the theoretical propagation state. These represent the impact weights corresponding to deviations in propagation state, propagation time, and propagation level, respectively, which can be obtained through statistical optimization based on historical demand propagation data. This represents the propagation potential energy enhancement coefficient, used to adjust the amplification effect of propagation potential energy on deviation. This represents the propagation continuity correction coefficient, used to enhance the impact of propagation path continuity on deviation calculation; , All parameters are adaptively updated based on the propagation environment change parameters output by the digital twin model. The larger the propagation environment change parameters, the greater the impact of environmental disturbances on the deviation calculation; the smaller the propagation environment change parameters, the smaller the impact of environmental disturbances.

[0120] This represents the cumulative deviation of the propagation chain, used to evaluate the overall consistency of continuous propagation paths; N represents the number of continuous propagation nodes participating in the cumulative calculation, which is automatically determined based on the continuous segments of the propagation chain. The importance weight of the i-th propagation node is calculated based on the requirements reference relationship, business process relationship, interface call relationship, test case relationship and historical version synchronization record.

[0121] After calculating the propagation deviation, the system obtains the propagation deviation information corresponding to each propagation node in the propagation chain, but it still cannot immediately determine the location of the propagation chain break. This is because there are many temporary modifications during the software development process, such as a person in charge not yet submitting a requirement update, or a version still being under development. In these cases, the propagation deviation is only a transient propagation deviation, not a true propagation chain break.

[0122] Therefore, this embodiment further analyzes the trend of changes in the propagation state along the demand change propagation chain.

[0123] In the specific implementation process, multiple propagation nodes are continuously read in the order of the propagation chain, and the changes in propagation state between adjacent propagation nodes are continuously analyzed. The changes in propagation state include not only changes in the numerical value of the propagation state, but also the downward trend of the propagation state, changes in the propagation interval, and the recovery status of the propagation state.

[0124] Among them, the downward trend of the propagation status is used to describe whether the propagation status continues to decline; the change in the propagation interval is used to reflect whether the update of the propagation status has stopped significantly; and the recovery status of the propagation status is used to determine whether subsequent nodes have resumed normal propagation.

[0125] When the propagation state deviates only at a single node and subsequent nodes resume normal propagation, it is considered a local propagation fluctuation. When the propagation state continues to decline at multiple nodes and no propagation recovery occurs, it is considered that the propagation anomaly is continuously spreading along the propagation chain.

[0126] Furthermore, in this embodiment, the theoretical propagation state of the corresponding propagation node is determined based on the remaining propagation potential energy obtained in step 2, and the actual response state of the current node is read.

[0127] Then it is determined whether the actual response state of the continuous propagation node is lower than the theoretical propagation state determined by the remaining propagation potential energy.

[0128] Analyzing the actual response state of consecutive nodes along the propagation chain, if the actual response state is lower than the theoretical state determined by the remaining potential energy, and the downward trend satisfies a preset propagation anomaly pattern that includes a combination of features such as continuous decline, recovery failure, and time lag, and the version evolution record shows that no corresponding version response has been formed, it is determined that a chain propagation state anomaly has occurred, and the associated topological edge that first satisfies the above conditions is determined as the propagation chain break position. In this embodiment, the preset propagation anomaly pattern is generated statistically based on historical demand propagation records. First, historical propagation chain break samples and normal propagation samples are labeled, and the number of nodes with continuous propagation state decline, the number of nodes with propagation state recovery, and the proportion of propagation time lag are counted to construct a propagation anomaly feature vector; then, an anomaly pattern library is established based on the frequency of occurrence of each propagation anomaly feature vector in historical samples. When the number of nodes with continuous propagation state decline reaches or exceeds the average continuous decline length of historical anomaly samples, the propagation recovery state never recovers to the theoretical propagation state, and the proportion of propagation time lag exceeds the statistical range corresponding to historical normal propagation samples, it is determined that the preset propagation anomaly pattern is satisfied.

[0129] Before detecting deviations in the actual propagation state, this embodiment first obtains the current version branch topology, multi-user concurrent collaboration impedance, and historical version activity indicators of each node's documents. It then identifies the current effective propagation path by combining the version branch merging relationship and the demand substitution relationship, and performs propagation path reconstruction accordingly.

[0130] Multi-user concurrent collaboration impedance refers to the communication delays, code merge conflict risks, and synchronization resistance to requirement changes that arise when multiple developers simultaneously modify multiple dependent requirement chapters or code branches in an agile development and continuous integration environment. In practice, multi-user concurrent collaboration impedance is quantified by combining the following statistical indicators: Concurrent Modification Density: the number of independent developers simultaneously editing or submitting code to the target requirement node and its associated downstream nodes within a preset time window; Branch Intertwining Degree: the number of code branches under development corresponding to the target requirement node; Historical Conflict Frequency: the cumulative number of code merge conflicts or requirement version overlaps occurring in the module containing the associated node in the historical version evolution record. The higher the value of the above statistical indicators, the higher the resistance and complexity of multi-user collaboration, and the greater the corresponding value of multi-user concurrent collaboration impedance.

[0131] When a theoretical propagation path is detected to be interrupted due to multi-person collaboration, the propagation is not immediately identified as abnormal. Instead, the propagation path reconstruction model is used to search for a new propagation path.

[0132] During the process of reconstructing the propagation path, the system re-finds the propagation path based on the version merging relationship and reconnects the propagation nodes based on the demand substitution relationship, so that the originally interrupted propagation path can be re-established in a new propagation direction.

[0133] To re-establish a successful propagation path, the corresponding reconstructed theoretical propagation state sequence is regenerated.

[0134] Subsequently, the original theoretical propagation state sequence is replaced with the reconstructed theoretical propagation state sequence, and the propagation mapping relationship is re-established and the propagation deviation analysis is re-executed to ensure that subsequent consistency detection is always based on the current real propagation path.

[0135] To describe the propagation path reconstruction process, a propagation path optimization model can be further established.

[0136] After the propagation path reconstruction is completed, this embodiment continues to perform pseudo-propagation break signal filtering.

[0137] Since some demand nodes have been permanently archived or replaced by new demands, there may still be a deviation between the theoretical propagation state and the actual propagation state. However, this deviation is a normal business evolution and not a real propagation anomaly.

[0138] Therefore, this embodiment reads the archive markers of associated nodes in the version control system and the active version replacement instructions.

[0139] Subsequently, the propagation trend of the actual propagation state sequence was decomposed.

[0140] Specifically, the propagation state sequence is divided into two components: long-term evolution trend and short-term modification fluctuation.

[0141] Long-term evolution trends mainly reflect the direction of long-term business evolution in terms of demand; short-term modification fluctuations mainly reflect local modifications made in the current version.

[0142] Preferably, this embodiment can use a multi-scale time series decomposition method to complete the propagation trend separation.

[0143] After completing the propagation trend decomposition, when a propagation deviation occurs at the target node, the first step is to determine whether the long-term evolution trend signal is consistent with the archive marker or the active version replacement instruction.

[0144] If they match, it indicates that the current deviation in propagation is a legitimate business evolution, rather than a propagation failure.

[0145] At this point, a filtering operator is generated by the propagation path reconstruction model to strip the propagation deviation state of the corresponding target node and mark the node as a legitimate change node so that it no longer participates in the propagation chain breakage determination.

[0146] If the long-term evolution trend does not match the archiving marker and the active version replacement instruction, the current propagation deviation state is retained and continues to participate in the final propagation consistency analysis as a candidate position for the propagation chain break.

[0147] Step 5: Determine the propagation impact domain based on the location of the propagation chain break, and output the software requirement consistency detection result containing strategy synchronization repair information based on the propagation impact domain.

[0148] In this embodiment, after detecting the break point in the propagation chain during the requirement change propagation process, the corresponding propagation impact domain is further determined based on the break point, and strategy synchronization repair information is generated based on the propagation impact domain to form a closed-loop detection result for the consistency status of the software requirement document. The propagation impact domain describes the scope of related requirement nodes that have not completed synchronized propagation after a propagation chain break, including the affected requirement chapters, related business processes, related interfaces, and related test cases.

[0149] Specifically, firstly, the associated topological edge information corresponding to the location of the propagation chain break is obtained, and the upstream and downstream node information connected to the location of the propagation chain break in the requirement change propagation state model is read. Based on the requirement reference relationship, business process relationship, interface call relationship, test case relationship, and historical version synchronization record, the diffusion analysis of the associated requirement nodes downstream of the location of the propagation chain break is performed.

[0150] In one implementation, the location where the propagation chain breaks is taken as the initial propagation node, and the process is traversed downstream along the state transition relationship in the requirement change propagation state model. When it is detected that there are requirement reference relationships, interface call relationships, business process relationships, or test relationships and historical version synchronization records among the associated requirement nodes, the corresponding associated requirement node is added to the candidate affected node set, and it is determined whether it is in an affected state based on the propagation state information corresponding to each associated requirement node.

[0151] The propagation status information is generated by the aforementioned requirement change propagation status model, including business strategy change status, business constraint synchronization status, related requirement response status, and version evolution status. When the related requirement response status of a related requirement node does not reach the theoretical propagation status, or the version evolution status does not produce a corresponding version response, the related requirement node is determined to be an affected node after the propagation chain is broken.

[0152] Furthermore, the boundaries of the propagation influence domain are determined based on the topological connections between multiple affected nodes. In this embodiment, the propagation influence domain is not simply determined based on the text section range, but rather on the state propagation range during the requirement change propagation process. That is, for the same requirement section, if it does not have a valid propagation association with the break point in the propagation chain, it is not included in the propagation influence domain; while requirement sections that are geographically distant but have business process or interface call associations are included in the propagation influence domain.

[0153] Preferably, when calculating the boundary of the propagation influence domain, a comprehensive judgment can be made based on the propagation path length, association weight, and degree of deviation of the propagation state between related demand nodes.

[0154] Because the propagation impact does not spread evenly to all associated demand nodes after a propagation chain break, but is constrained by the demand association topology, the decay of the propagation potential field, and the degree of deviation from propagation consistency, determining the propagation impact domain solely based on the propagation path length or node adjacency relationship can easily lead to the misinclusion of associated nodes with weak propagation capabilities within the repair scope, or the omission of distant demand nodes affected by highly associated propagation.

[0155] Based on this, this embodiment takes the location of the propagation chain break as the propagation source point, and combines the remaining propagation potential energy calculated in step 2, the propagation deviation obtained in step 4, and the demand-related topology to establish a propagation influence domain diffusion boundary function, so that the propagation influence can be dynamically diffused according to the actual propagation capacity, and the final set of affected demand nodes is determined based on the comprehensive influence intensity.

[0156] Furthermore, based on the topological mapping relationship between the location of the propagation chain break and the associated demand nodes, a propagation influence domain diffusion equation is constructed, leading to:

[0157] ;

[0158] Furthermore, to determine the boundary of the propagation influence domain, a propagation influence determination function is defined:

[0159] ;

[0160] Specifically, if the propagation impact intensity of the associated demand node is not lower than the propagation impact threshold, then the associated demand node is included in the propagation impact domain; otherwise, it is determined that it has not been effectively affected by the current propagation chain break and will not participate in the generation of subsequent strategy synchronization repair information.

[0161] By calculating the diffusion boundary of the propagation impact domain as described above, the scope of propagation impact no longer depends on a fixed chapter range, but is dynamically determined based on demand correlation, propagation capability, and degree of propagation distortion. This improves the accuracy of propagation impact domain identification and reduces the scope of invalid repairs.

[0162] In the formula, This represents the propagation impact intensity corresponding to the j-th associated demand node, used to measure the comprehensive propagation impact of the propagation chain break on this demand node, and is calculated jointly by the propagation potential field model, the propagation deviation analysis results, and the demand association topology. This represents the remaining propagation potential energy after propagation to the j-th associated demand node, which is obtained from the propagation potential energy attenuation calculation result in step 2. This represents the initial propagation potential energy corresponding to the propagation source node, which is obtained by transforming the characteristics of business strategy changes. This represents the propagation path length from the break point of the propagation chain to the j-th associated demand node, which is obtained by counting the number of associated topology edges traversed in the demand association topology; This represents the propagation path attenuation coefficient, which describes the natural attenuation of propagation capability as the propagation path increases. It is calculated based on historical version synchronization records and propagation success rate statistics. The overall association weight of the nodes is calculated by statistically analyzing the weights of all associated topological edges pointing to the target node; The propagation deviation degree corresponding to the j-th associated demand node is obtained from the deviation degree calculation result of the theoretical propagation state sequence and the actual propagation state sequence in step 4, and is used to reflect the degree of propagation distortion of the current node. This represents the propagation impact domain, which is the set of related demand nodes that are ultimately determined to be affected by the break in the propagation chain; This represents the j-th associated demand node, which is uniquely determined by the demand node identifier in the demand change propagation state model. This represents the threshold for determining the propagation impact, used to distinguish whether something belongs to the propagation impact domain. It is obtained based on statistical analysis of historical propagation impact domain samples. Specifically, statistical analysis is performed on the propagation impact intensity corresponding to historical propagation chain break events, the statistical distribution of the propagation impact values ​​of affected nodes is calculated, and the propagation impact determination threshold is determined based on the statistical quantile corresponding to the coverage rate of the impact range, so that the propagation impact domain can cover the vast majority of nodes with real affected needs.

[0163] After determining the scope of the propagation impact, a strategy omission compensation template is further generated for chapters with unsynchronized modification requirements.

[0164] Specifically, the theoretical propagation state sequence and the actual propagation state sequence corresponding to the break point of the propagation chain are obtained. The propagation differences between the two state sequences after the break point are analyzed. Propagation node information that exists in the theoretical state but is missing in the actual state is extracted. Based on the business strategy change state, business constraint synchronization state and related demand response state corresponding to the missing nodes, a strategy omission and modification differential compensation template is generated.

[0165] The strategy omission differential compensation template is used to describe the requirements that have not been modified synchronously due to the break in the propagation chain. It includes at least the chapter identifier of the requirement to be supplemented, the content of related business strategy changes, the business constraint information that needs to be adjusted synchronously, the related interface adjustment information, and the related test case adjustment information.

[0166] In one specific implementation, the target node state in the theoretical propagation state sequence is used as the baseline state, and the corresponding node state in the actual propagation state sequence is used as the current state. The missing propagation part of the requirement change is obtained through state difference analysis.

[0167] Furthermore, based on the content of the requirement document corresponding to the unsynchronized propagation node, semantic analysis methods are used to extract the missing business semantics and generate the corresponding strategy omission and differential compensation template.

[0168] For example, when the source requirement is modified to "change the login authentication strategy from SMS verification code authentication to multi-factor authentication", the theoretical propagation state sequence may include the identity authentication requirement section, the verification code service requirement section, the access control requirement section, the exception handling requirement section, and the security testing requirement section. If the actual propagation state sequence only includes the update of the identity authentication requirement section, while the verification code service requirement section and the test case section have not changed synchronously, then the policy omission compensation template records the verification code service interface adjustment requirement, the security testing process adjustment requirement, and the exception handling rule synchronization requirement.

[0169] After generating the strategy omission differential compensation template, the strategy omission differential compensation template and the affected requirement chapter scope are encapsulated to generate a requirement consistency detection result containing a list of requirements to be revised.

[0170] In this embodiment, the requirement consistency detection result includes at least the location of the propagation chain break, the propagation impact domain, and policy synchronization repair information. The policy synchronization repair information includes a policy omission differential compensation template and the affected requirement chapter range. Specifically, the propagation chain break location identifies the position where a state mismatch occurs during the requirement change propagation process, the propagation impact domain describes the affected range, and the policy omission differential compensation template guides the subsequent requirement synchronization revision process.

[0171] Furthermore, after outputting the strategy to synchronize the repair information, a closed-loop interception and control process is established with the software development collaboration platform.

[0172] Specifically, the strategy omission and differential compensation template is converted into a merge control instruction and sent to the software development collaboration platform interface. The software development collaboration platform interface is used to interact with the version control system, requirements management system, and task management system to obtain the target code branch status, requirement change task status, and revision execution status.

[0173] When the software development collaboration platform interface receives the merge control instruction, it reads the revision record corresponding to the target code branch or requirement change task and determines whether there is any propagation fix behavior targeting the scope of the requirement chapter within the propagation impact domain.

[0174] The propagation and remediation actions include revising requirements documents, adjusting interface code, updating test cases, and adjusting business processes.

[0175] If the target code branch or requirement change task has completed the corresponding propagation and repair behavior, the subsequent software development process is allowed to continue; if the corresponding propagation and repair behavior is not detected, a task blocking state is generated, and the target code branch merging operation or task flow advancement process is intercepted and locked in advance through the software development collaboration platform interface.

[0176] Through the above process, the requirement consistency detection is expanded from a single detection process into a closed-loop process that includes requirement change identification, propagation status analysis, fault location, impact domain determination, and repair execution verification.

[0177] Example 2

[0178] In this embodiment, a software requirements document consistency detection system based on semantic analysis adopts a layered data processing architecture. Each module forms a complete data processing pipeline according to the requirement change propagation process. The system is deployed on a software development platform and can establish data communication connections with the requirements management platform, version control system, test management platform, and software development collaboration platform through network interfaces. The system comprises several modules: a data acquisition and identification module, serving as the data entry point, responsible for uniformly acquiring requirement documents, requirement modification records, version evolution records, and requirement-related data, and outputting business strategy change characteristics and actual propagation state sequences; a requirement change propagation state modeling module, establishing a requirement change propagation state model and a requirement change propagation chain based on business strategy change characteristics; a propagation potential field calculation module, further constructing a propagation potential field model and calculating propagation intensity; a propagation state deduction module, generating a theoretical propagation state sequence based on propagation intensity; a propagation state adaptive update module, establishing a synchronous twin evolution model between the theoretical and actual propagation trajectories and continuously updating the requirement change propagation state model; a propagation consistency detection module, establishing a propagation mapping relationship using the theoretical and actual propagation state sequences and detecting propagation chain break locations; a propagation path reconstruction module, reconstructing and restoring abnormal propagation paths and filtering pseudo-propagation break signals; and a propagation influence domain analysis module, determining the propagation influence domain based on the propagation chain break location, generating software requirement consistency detection results, and distributing strategy synchronization repair information through the software development collaboration platform interface, thus forming a closed-loop processing architecture encompassing requirement change identification, propagation deduction, anomaly location, repair guidance, and execution verification.

[0179] In this embodiment, the data acquisition and identification module receives requirement documents and requirement modification records from the requirement management platform, version evolution records from the version control system, and requirement-related data from the test management platform via standard data interfaces. It then uses semantic analysis to identify business strategy change characteristics and generates an actual propagation state sequence based on the current version of the requirement document. Subsequently, the business strategy change characteristics and requirement-related data are synchronously input into the requirement change propagation state modeling module. This module establishes a requirement change propagation state model that includes business strategy change states, business constraint synchronization states, related requirement response states, and version evolution states, and constructs a requirement change propagation chain that propagates step-by-step from modified content to the influence domain. The propagation potential field calculation module continues to receive the requirement change propagation state model and requirement-related data. Based on requirement reference relationships, business process relationships, interface call relationships, test case relationships, and historical version synchronization records, it establishes a propagation potential field model, calculates the state propagation strength and propagation attenuation results corresponding to each propagation path, and uses the propagation strength as a unified propagation constraint for subsequent propagation deductions. Through the continuous processing of these three modules, the original requirement data is gradually transformed into propagation state data that can describe the propagation laws of business strategies, providing a unified data foundation and propagation constraints for subsequent theoretical propagation state deductions.

[0180] In this embodiment, the propagation state deduction module receives the state propagation intensity output by the propagation potential field calculation module and the state transition relationship in the demand change propagation state model. It then performs forward deduction on each associated demand node in the propagation chain to generate a theoretical propagation state sequence. This theoretical propagation state sequence, along with the actual propagation state sequence output by the data acquisition and identification module, is input into the propagation state adaptive update module. This module establishes a synchronous twin evolution model in digital space corresponding to the theoretical and actual propagation trajectories. Based on project iteration cycles, demand modification frequency, number of version branches, and historical propagation deviations, it constructs propagation environment change parameters to continuously correct the demand change propagation state model, ensuring that the theoretical propagation state sequence remains synchronized with the actual R&D process. The updated theoretical propagation state sequence is further sent to the propagation consistency detection module. This module establishes a propagation mapping relationship consisting of node mapping, time mapping, and propagation hierarchy mapping. It then performs a node-by-node comparison between the theoretical and actual propagation state sequences, calculates the propagation deviation, and locates the break in the demand propagation chain. Through this continuous data stream, the theoretical propagation model evolves from static deduction to dynamic learning, improving the accuracy of demand propagation prediction and the ability to identify propagation anomalies.

[0181] In this embodiment, the propagation path reconstruction module receives the propagation chain break location output by the propagation consistency detection module, and simultaneously acquires the current version's branch merging relationship, requirement replacement relationship, and multi-person collaboration information. It identifies propagation path changes caused by version evolution or collaborative development, reconstructs an effective propagation path, and generates a corrected theoretical propagation state sequence. Simultaneously, it filters false propagation break signals using archiving markers and proactive version replacement commands, ensuring high reliability of the propagation anomaly location results. Subsequently, the propagation impact domain analysis module receives the propagation chain break location and the corrected theoretical propagation state sequence, performs propagation impact domain analysis on downstream related requirement nodes, generates a propagation impact domain, strategy synchronization repair information, and a list of requirements to be revised, and distributes this information to the requirement management system, version control system, and task management system via the software development collaboration platform interface. When it detects that the requirement corresponding to the propagation impact domain has not yet completed propagation repair, the software development collaboration platform interface executes branch merging pre-interception, thus forming a linkage mechanism between propagation anomaly detection and propagation repair execution, enabling requirement consistency detection results to directly drive closed-loop control of the R&D process.

[0182] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation methods of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A software requirements document consistency detection method based on semantic analysis, characterized in that, Includes the following steps: Step 1: Obtain the target project's requirement documents, requirement modification records, version evolution records, and requirement-related data; use semantic analysis to extract the business strategy change characteristics corresponding to the source requirement modification content. Step 2: Based on the characteristics of business strategy changes and the correlation and dependency relationships between different requirements in the requirement-related data, construct a requirement change propagation state model and establish a requirement change propagation chain that transmits requirement changes layer by layer from modified content, business strategy, business constraints to the domain of influence. The requirement change propagation state model includes at least the business strategy change state, business constraint synchronization state, related requirement response state, and version evolution state, and establishes a state transition relationship that transmits the business strategy change state to the business constraint synchronization state, related requirement response state, and version evolution state level by level. Step 3: Run the requirement change propagation state model to forward extrapolate the transmission process of business strategy change characteristics in the requirement change propagation chain, so that the business strategy change characteristics propagate layer by layer along the requirement change propagation process, generating the corresponding theoretical propagation state sequence. At the same time, based on semantic analysis of the current version of the requirement document, obtain the actual modification characteristics of each related node under the current version, and construct the corresponding actual propagation state sequence. Step 4: Establish a propagation mapping relationship that includes node mapping, time mapping, and propagation level mapping. Map and compare the theoretical propagation state sequence with the actual propagation state sequence, calculate the deviation of the actual propagation state sequence from the theoretical propagation state sequence, and determine the location of the propagation chain break in the demand change propagation process based on the deviation. Step 5: Determine the propagation impact domain based on the location of the propagation chain break, and output the software requirement consistency detection result containing strategy synchronization repair information based on the propagation impact domain.

2. The software requirements document consistency detection method based on semantic analysis according to claim 1, characterized in that, The requirement association data includes the reference relationships between requirement chapters, the association relationships between business processes, the association relationships between API calls, the association relationships between test cases, and historical version synchronization records; Step 2 involves constructing the requirement change propagation state model, which includes: Parse the requirement association data, extract the requirement reference relationship, business process association relationship, interface call association relationship, test case association relationship and historical version synchronization record between different requirements, and map them into the association weight of the association topology edge. Based on the association weight, construct the propagation potential field distribution to form a propagation potential field model based on the requirement association strength and historical synchronization behavior. The propagation potential field model is used to characterize the change in propagation strength caused by the association constraint when business strategy changes propagate between different requirement nodes. The characteristics of business strategy changes are converted into initial propagation potential energy. Using the propagation decay function in the propagation potential field model, the propagation potential energy is adaptively decayed in the propagation chain according to the association weight of the associated topology edge, and the remaining propagation potential energy allocated to each downstream node in the propagation chain is obtained. The state transition relationships of each associated demand node in the demand change propagation state model are updated based on the remaining propagation potential energy to reflect the differences in propagation intensity between different demand nodes.

3. The software requirements document consistency detection method based on semantic analysis according to claim 2, characterized in that, The process of determining the location of the break in the propagation chain during the demand change propagation process based on the degree of deviation includes: Based on the remaining propagation potential energy, determine the theoretical propagation state of each associated demand node; Analyze the downward trend of propagation status, changes in propagation intervals, and state recovery status between consecutive propagation nodes along the demand change propagation chain. Determine whether the actual response state of the continuously propagating node is lower than the theoretical propagation state determined by the remaining propagation potential energy; When the actual response state of a continuous propagation node is lower than the theoretical propagation state and the downward trend of the propagation state meets the preset propagation anomaly mode, and no corresponding version response is formed, it is determined that the chain propagation state has stopped abnormally during the demand change propagation process, and the corresponding associated topological edge is determined as the location of the propagation chain break.

4. The software requirements document consistency detection method based on semantic analysis according to claim 1, characterized in that, In steps 3 and 4, a demand propagation digital twin model is introduced for adaptive parameter correction: Establish a synchronous twin evolution model between the theoretical propagation trajectory composed of theoretical propagation state sequences and the actual evolution trajectory composed of actual propagation state sequences in the digital space; The time-varying deviation vector is calculated by comparing the theoretical propagation trajectory with the actual evolution trajectory in real time. The time-varying deviation vector is used to update the demand change propagation state model, correcting the propagation mapping relationship between the theoretical propagation trajectory and the actual evolution trajectory, as well as the state transition relationship in the demand change propagation state model.

5. The software requirements document consistency detection method based on semantic analysis according to claim 4, characterized in that, The process of updating the demand change propagation state model using the time-varying deviation vector includes: Obtain the target project's current iteration cycle, requirement modification frequency, number of version branches, and historical propagation deviation data to construct parameters for changes in the propagation environment. The uncertainties in the synchronous twin evolution model are adaptively adjusted by using parameters of change in the propagation environment, and the state transition relationships and evolution parameters in the demand change propagation state model are dynamically corrected by combining time-varying deviation vectors. The state deduction is re-executed based on the corrected evolution parameters to update the theoretical propagation state sequence.

6. The software requirements document consistency detection method based on semantic analysis according to claim 1, characterized in that, Step 4, before detecting deviations from the actual propagation state, also includes the following propagation path reconstruction steps: Obtain the current version branch topology, multi-user concurrent collaboration impedance, and activity metrics of each node's documents in historical versions; identify the current effective propagation path based on version branch merging relationships and requirement replacement relationships; and construct a propagation path reconstruction model. When an interruption is detected in the inferred propagation path due to multi-source cooperation, the abnormal interruption path in the theoretical propagation state sequence is reconstructed and restored using the propagation path reconstruction model to generate a reconstructed theoretical propagation state sequence. The reconstructed theoretical propagation state sequence is used to replace the theoretical propagation state sequence, serving as the comparison basis in the propagation consistency determination process.

7. The software requirements document consistency detection method based on semantic analysis according to claim 6, characterized in that, After generating the reconstructed theoretical propagation state sequence, the process also includes a filtering step for spurious propagation break signals: Extract the archiving markers and active version replacement instructions of the associated nodes in the version control system, decompose the long-term evolution trend and short-term modification fluctuation of the actual propagation state sequence, and extract the long-term evolution trend signal and short-term modification fluctuation signal. When a deviation of the target node is detected, if the long-term evolution trend signal matches the archive marker or the active version replacement instruction, the deviation is determined to be a legitimate change, and the filtering operator constructed by the propagation path reconstruction model is invoked to strip the deviation state of the target node in order to eliminate the false propagation break signal.

8. The software requirements document consistency detection method based on semantic analysis according to claim 1, characterized in that, Step 5, which outputs the requirement consistency detection results and strategy synchronization repair information, includes: Calculate the diffusion boundary of downstream related demand nodes at the break point of the demand change propagation chain to determine the scope of affected demand chapters; Based on the difference between the theoretical propagation state sequence and the actual propagation state sequence at the break point, a strategy omission compensation template is automatically generated for chapters with unsynchronized modification requirements. The strategy omission differential compensation template and the affected requirement chapter scope are encapsulated into strategy synchronization repair information, and a requirement consistency detection result containing a list of requirements to be revised is generated.

9. The software requirements document consistency detection method based on semantic analysis according to claim 8, characterized in that, After outputting the policy synchronization and repair information, the following closed-loop interception control steps are also included: Send the strategy omission and differential compensation template as a merge control command to the software development collaboration platform interface; The software development collaboration platform interface is used to detect whether the target code branch or requirement change task contains propagation and fix behaviors targeting the affected requirement section scope; If no corresponding propagation and repair behavior is detected, the software development collaboration platform interface will intercept and lock the branch merging operation or task flow advancement of the target code branch in advance.

10. A software requirements document consistency detection system based on semantic analysis, characterized in that, include: The data acquisition and identification module is used to acquire the target project's requirement documents, requirement modification records, version evolution records, and requirement-related data. Based on semantic analysis, it processes the requirement modification records and requirement documents to identify the business strategy change characteristics corresponding to the source requirement modification and construct the actual propagation state sequence. The requirement change propagation state modeling module is used to construct a requirement change propagation state model based on the characteristics of business strategy changes and requirement-related data. It establishes a requirement change propagation chain from modified content, business strategy, business constraints to the domain of influence. The requirement change propagation state model includes at least the business strategy change state, business constraint synchronization state, related requirement response state, and version evolution state, and establishes a state transition relationship from the business strategy change state to other states in a step-by-step manner. The propagation potential field calculation module is used to construct a propagation potential field model based on the demand correlation strength and historical synchronization behavior, and to calculate the state propagation strength and attenuation during the theoretical evolution process based on the propagation potential field model. The propagation state deduction module is used to perform forward deduction by changing the propagation state model in combination with the state propagation intensity and operational requirements, and generate a theoretical propagation state sequence. The propagation state adaptive update module is used to establish a synchronous twin evolution model of theoretical and actual propagation trajectories, and to build a propagation state model based on the propagation environment change parameter update requirements constructed based on the project iteration cycle and historical propagation deviation data. The propagation consistency detection module is used to receive the actual propagation state sequence and the theoretical propagation state sequence, establish a propagation mapping relationship including node mapping, time mapping and propagation level mapping, and calculate the actual deviation to detect the break position of the demand change propagation chain. The propagation path reconstruction module is used to identify the current valid propagation path based on the version branch merging relationship and requirement substitution relationship of the current version, correct and reconstruct the theoretical propagation path that is interrupted due to multi-source collaboration anomalies, and filter out false propagation break signals. The propagation impact domain analysis module is used to determine the corresponding propagation impact domain based on the break point of the propagation chain of requirement changes, generate software requirement consistency detection results containing strategy synchronization and repair information, and distribute them to the software development collaboration platform interface.

Images