Change impact analysis method combined with natural language processing

Abstract

This invention relates to the field of natural language processing technology, specifically disclosing a method for analyzing the impact scope of requirement changes combined with natural language processing. The method includes: performing natural language processing on each requirement item in a requirement document to extract causal relationships, temporal dependencies, constraint relationships, and referential relationships, obtaining a multi-dimensional semantic dependency set; constructing a multi-dimensional semantic dependency graph; calculating the impact score of each node when a change event of a target requirement item is received; identifying the requirement items corresponding to nodes whose impact scores exceed a preset impact threshold as affected requirement items, and outputting the change impact scope analysis results. This invention solves the technical problems of existing technologies, such as the lack of multi-dimensional logical dependency identification, incomplete change impact propagation analysis, and insufficient quantification of impact, caused by reliance on single semantic similarity.

Description

Technical Field

[0001] This invention relates to the field of natural language processing technology, and in particular to a method for analyzing the impact of changes in requirements in conjunction with natural language processing. Background Technology

[0002] With the continuous expansion of software engineering scale and the increasing complexity of systems, requirements management has become a core component of the entire software development lifecycle. In safety-critical fields such as aviation, finance, and healthcare, a large software system often has thousands of requirements, which frequently change during project implementation. Scope analysis of requirement changes, as a fundamental aspect of change control, directly determines the accuracy of change decisions and the controllability of project risks, and remains a significant and long-standing technical challenge in the field of software engineering.

[0003] Existing methods for analyzing the impact of change requests that incorporate natural language processing primarily identify potentially affected items by calculating the semantic similarity between request texts. Typical methods include keyword matching based on word frequency statistical models such as TF-IDF and BM25, and semantic vector similarity calculation methods based on pre-trained language models such as BERT. These methods convert the request text into semantic vectors and then use the cosine similarity of these vectors to filter out items that are semantically close to the changed request, which are then output as the scope of impact.

[0004] However, existing methods for analyzing the impact of requirement changes using natural language processing have significant drawbacks: First, they rely solely on semantic similarity for single-dimensional matching, failing to identify multi-dimensional logical dependencies such as causal relationships, temporal dependencies, constraint relationships, and referential relationships between requirement items. This leads to the systematic omission of requirement items that are "semantically dissimilar but logically strongly dependent." Second, existing methods treat the requirement set as a flat set of independent items, lacking networked modeling of dependencies between requirements. This makes it impossible to perform multi-hop propagation analysis of the impact of changes, resulting in an incomplete scope of impact identification. Third, the assessment of the degree of impact lacks quantitative basis, making it impossible to prioritize affected items according to their impact intensity, which is insufficient to support change decisions in engineering practice. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method for analyzing the impact scope of demand changes in conjunction with natural language processing, in order to solve the problems raised in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for analyzing the impact of changes in requirements using natural language processing, comprising the following steps: Natural language processing is performed on each requirement item in the requirement document. Through syntactic dependency analysis, causal relationships, temporal dependencies, constraint relationships, and referential relationships are extracted from the text of each requirement item to obtain a multi-dimensional semantic dependency set corresponding to each requirement item. Using each requirement item as a node and each dependency in the multi-dimensional semantic dependency set as a directed edge, a multi-dimensional semantic dependency graph is constructed by assigning corresponding edge type weights according to the type of each dependency. When a change event of a target requirement item is received, the node corresponding to the target requirement item is taken as the propagation start node in the multi-dimensional semantic dependency graph. The graph propagation algorithm is executed according to the edge type weight and the preset propagation depth threshold to obtain the influence degree score of each node in the multi-dimensional semantic dependency graph. The demand items corresponding to nodes whose impact score exceeds the preset impact threshold are identified as affected demand items. The affected demand items are sorted according to the impact score, and the change impact range analysis results are output.

[0007] Preferably, the steps for obtaining the multi-dimensional semantic dependency set are as follows: Obtain the requirements document and preprocess it. Segment the requirements document according to the requirements number identifier, and split the requirements document into independent requirements items. Each requirements item retains its number and corresponding text content. The text of each requirement item is segmented and tagged with part-of-speech tags, and then a syntactic dependency tree is constructed based on the tagging results. Identify statements containing conditional triggering clauses in the text of each requirement item, and mark the dependency relationship between two requirement items that satisfy the conditional triggering clause as a causal relationship; Identify statements in each requirement entry text that contain timing markers, and mark the dependency relationship between two requirement entries that contain timing markers and have an execution order as a timing dependency relationship; Identify statements in each requirement item text that contain modal verbs and limiting conditions, and mark the dependency relationship of the limiting conditions on other requirement items as a constraint relationship; Identify statements in each requirement entry text that contain explicit numbering or implicit business object pointing, and mark the dependency relationship between two requirement entries that contain explicit numbering or implicit business object pointing as a reference relationship; The extracted causal relationships, temporal dependencies, constraint relationships, and reference relationships are summarized to obtain a multi-dimensional semantic dependency set corresponding to each requirement item.

[0008] Preferably, the steps for constructing the multi-dimensional semantic dependency graph are as follows: Each requirement item is mapped to a node in a multi-dimensional semantic dependency graph, and each node carries the corresponding requirement item's number and text content as node attributes. Each dependency in the multidimensional semantic dependency set is mapped to a directed edge in the multidimensional semantic dependency graph. The direction of the directed edge represents the propagation direction of the dependency, that is, from the dependency initiator node to the dependency-dependent node. Based on the type of each dependency and the degree of logical determination of each type of dependency in the propagation of the impact of demand changes, each directed edge is assigned a corresponding edge type weight. A multi-dimensional semantic dependency graph is constructed by using the nodes corresponding to each requirement item as the vertex set and the directed edges carrying edge type weights as the edge set.

[0009] Preferably, the content from the dependency initiator node to the dependency node is as follows: The directed edges of the causal relationship are from the node corresponding to the demand item of the condition triggering party to the node corresponding to the demand item of the triggered party. The directed edges of the temporal dependencies are from the node corresponding to the preceding requirement item to the node corresponding to the following requirement item. The directed edges of the constraint relationship are pointed from the node corresponding to the requirement item of the constraint issuer to the node corresponding to the requirement item of the constrained party. The directed edges of the reference relationship are those where the node corresponding to the referencer's requirement item points to the node corresponding to the referenced's requirement item.

[0010] Preferably, the steps for obtaining the influence score of each node are as follows: When a change event for a target requirement item is received, the node corresponding to the target requirement item is set as the propagation start node in the multi-dimensional semantic dependency graph, and the initial influence score of the propagation start node is set to 1.0, while the initial influence score of all other nodes in the multi-dimensional semantic dependency graph is set to 0. Calculate the criticality coefficient of each node in the multi-dimensional semantic dependency graph; Starting from the propagation initiation node, propagation proceeds outward layer by layer along the directed edges of the multidimensional semantic dependency graph. Based on the criticality coefficient of each node, the influence score of each node in the multidimensional semantic dependency graph is calculated. When there are multiple propagation paths from different directions for the i-th node, the maximum value of the influence score corresponding to all propagation paths is taken as the final influence score of the node. During the propagation process, when the influence score of a node is lower than the preset influence threshold, or when the number of directed edges traversed by the propagation reaches the preset propagation depth threshold, the propagation stops in that direction. The graph propagation algorithm is used to obtain the influence rating of each node in the multi-dimensional semantic dependency graph.

[0011] Preferably, the calculation logic for the criticality coefficient of each node is as follows: obtain the quantified value of the demand level and the quantified value of the demand type of each node corresponding to the demand item in the multi-dimensional semantic dependency graph, multiply the quantified value of the demand level and the quantified value of the demand type of each node corresponding to the demand item, and calculate the criticality coefficient of each node. The demand level quantification value refers to the numerical value assigned according to the level of the demand item in the demand hierarchy. The higher the level, the larger the corresponding demand level quantification value. The required type quantification value refers to the numerical value assigned based on the safety criticality of the type to which the required item belongs; the higher the safety criticality, the larger the required type quantification value.

[0012] Preferably, the content of the output change impact range analysis results is as follows: The influence score of each node in the multi-dimensional semantic dependency graph is compared with the preset influence threshold, and the demand items corresponding to the nodes whose influence scores exceed the preset influence threshold are identified as affected demand items. The affected demand items are sorted in descending order of their impact rating to obtain a sorted list of affected demand items. Based on the sorted list of affected requirement items, generate and output the results of the change impact scope analysis.

[0013] As described above, the method for analyzing the impact scope of demand changes combined with natural language processing provided by this invention has at least the following beneficial effects: This invention provides a method for analyzing the impact of demand changes by combining natural language processing. Through syntactic dependency analysis, it simultaneously extracts four types of multi-dimensional semantic dependencies—causal, temporal, constraint, and referential—from the text of demand items. This overcomes the shortcomings of existing technologies that rely solely on semantic similarity for single-dimensional matching, leading to the systematic omission of strong logical dependencies and improving the completeness of dependency identification among demand items. Based on this, a multi-dimensional semantic dependency graph is constructed using directed edges of these multi-dimensional semantic dependencies. An edge type weight and path type combination correction coefficient are introduced to design a graph propagation algorithm, enabling networked, multi-level propagation analysis of the impact of changes. This solves the problem of existing technologies treating demand sets as flat sets of items and failing to identify multi-level indirect impacts. Furthermore, by introducing a node criticality coefficient into the impact score calculation, the quantitative values ​​of the demand level and demand type of each demand item are included in the impact score calculation. This allows the impact score to reflect the actual contribution of the demand item's own attributes to the impact of the change. The affected demand items are then ranked and output based on the impact score, solving the problem of existing technologies lacking quantitative basis for impact assessment and failing to support priority ranking in change decisions. Attached Figure Description

[0014] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0015] Figure 1 This is a flowchart illustrating the method for analyzing the impact of demand changes in conjunction with natural language processing, as described in this invention. Detailed Implementation

[0016] 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 some embodiments of the present invention, and not all embodiments. 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.

[0017] Please see Figure 1 As shown, this invention provides a method for analyzing the impact scope of demand changes by incorporating natural language processing, comprising the following steps: S1. Perform natural language processing on each requirement item in the requirement document. Through syntactic dependency analysis, extract causal relationships, temporal dependencies, constraint relationships, and referential relationships from the text of each requirement item to obtain a set of multi-dimensional semantic dependencies corresponding to each requirement item. In this embodiment, the steps for obtaining the multi-dimensional semantic dependency set are as follows: The requirements document is obtained and preprocessed. It is segmented according to the requirement number identifier, and the requirements document is split into independent requirement items. Each requirement item retains its number and corresponding text content. The purpose of the preprocessing is to transform the unstructured requirements document into structured data in units of items, so as to provide input for subsequent syntactic dependency analysis.

[0018] The text of each requirement item is segmented and tagged with parts of speech. Then, a syntactic dependency tree is constructed based on the tagged results. The syntactic dependency tree is rooted with the predicate verb as the root node and directed edges with the grammatical dominance relationships between words. It expresses the syntactic structure of the requirement item text. Syntactic dependency analysis can reveal the grammatical relationships between the subject, predicate, object and modifier in the requirement text, which is the basis for extracting various semantic dependencies in the future.

[0019] Identify statements containing conditional triggering clauses in the text of each requirement item, and mark the dependency relationship between two requirement items that satisfy the conditional triggering clause as a causal relationship; the conditional triggering clause includes at least one of "if...then...", "when......", and "if...then...". It should be specifically noted that the causal relationship reflects the logical dependency between the requirement items that "preconditions trigger subsequent behaviors". For example, there is a causal relationship between the requirement item "if a user fails to log in more than 5 times in a row, the system should lock the account" and the requirement item "the account lock status should be recorded in the security audit log". When the former changes, the latter may be affected.

[0020] Identify statements containing timing markers in the text of each requirement item, and mark the dependency relationship between two requirement items that contain timing markers and have an execution order as a timing dependency relationship; the timing markers include at least one of "after", "before", "only after completion", and "at the same time". It should be specifically noted that the time-series dependency reflects the logical dependency of the "execution order constraint" between the requirement items. For example, there is a time-series dependency between the requirement item "data transmission can only be performed after data encryption is completed" and the requirement item "the system should support AES-256 data encryption". Changes in encryption requirements will propagate to transmission requirements along the time-series dependency.

[0021] Identify statements in each requirement item text that contain modal verbs and limiting conditions, and mark the dependency relationship of the limiting conditions on other requirement items as a constraint relationship; the modal verbs include at least one of "must not", "must", "shall satisfy", and "not exceed"; It should be specifically noted that the constraints reflect the logical dependency of "rule constraints" between requirement items. For example, the requirement item "system response time shall not exceed 200 milliseconds" constrains all functional requirement items related to system response. When this constraint changes, all functional requirement items bound by it are affected.

[0022] Identify statements in each requirement entry text that contain explicit numbering or implicit business object pointing, and mark the dependency relationship between two requirement entries that contain explicit numbering or implicit business object pointing as a reference relationship; Explicit numbering refers to the situation where the number of another requirement item directly appears in the requirement item text, such as "see REQ-0021"; implicit business object referencing refers to the situation where two requirement items involve the same business object (such as "user account" or "order record") but are not explicitly referenced by number. Implicit reference relationships are identified by comparing the business object terms in the text of each requirement item. Reference relationships are the most direct logical connection between requirement items, and changes to the referenced requirement item must be evaluated simultaneously with the referrer.

[0023] The extracted causal relationships, temporal dependencies, constraint relationships, and reference relationships are summarized to obtain a multi-dimensional semantic dependency set corresponding to each requirement item.

[0024] S2. Using each requirement item as a node and each dependency in the multi-dimensional semantic dependency set as a directed edge, assign corresponding edge type weights according to the type of each dependency to construct a multi-dimensional semantic dependency graph. In this embodiment, the steps for constructing the multi-dimensional semantic dependency graph are as follows: Each requirement item is mapped to a node in a multi-dimensional semantic dependency graph, and each node carries the corresponding requirement item's number and text content as node attributes. It should be noted that there is a one-to-one correspondence between nodes and requirement items. The number of requirement items in the requirement document corresponds to the number of nodes in the multi-dimensional semantic dependency graph, thus ensuring that the graph structure fully covers the requirement set.

[0025] Each dependency in the multidimensional semantic dependency set is mapped to a directed edge in the multidimensional semantic dependency graph. The direction of the directed edge represents the propagation direction of the dependency, that is, from the dependency initiator node to the dependency-dependent node. In this embodiment, the content pointed from the dependency initiator node to the dependency node is as follows: The directed edge of the causal relationship points from the node corresponding to the demand item of the triggering party to the node corresponding to the demand item of the triggered party, indicating that when the triggering party changes, the impact propagates along the directed edge to the triggered party. The directed edges of the temporal dependencies are from the node corresponding to the preceding requirement item to the node corresponding to the following requirement item. The directed edges of the constraint relationship are pointed from the node corresponding to the requirement item of the constraint issuer to the node corresponding to the requirement item of the constrained party. The directed edges of the reference relationship are those where the node corresponding to the referencer's requirement item points to the node corresponding to the referenced's requirement item.

[0026] Based on the type of each dependency and the degree of logical determination of each dependency in the propagation of the impact of demand changes, each directed edge is assigned a corresponding edge type weight. Specifically, the various dependencies are sorted from strong to weak in terms of logical determination. The stronger the degree of logical determination of a dependency type, the higher the edge type weight, and the weaker the degree of logical determination of a dependency type, the lower the edge type weight.

[0027] It should be specifically noted that causal relationships express the mandatory logic of "preconditions triggering subsequent behaviors," and have the strongest determinative effect on the propagation of changes; constraint relationships express the mandatory logic of "rule constraints," and have the second strongest determinative effect on the propagation of changes; temporal dependencies express the associative logic of "execution order constraints," and have the next strongest determinative effect on the propagation of changes; and reference relationships express the associative logic of "object pointing," and have the weakest determinative effect on the propagation of changes. Accordingly, the edge type weights corresponding to each type of dependency relationship, from highest to lowest, are causal relationships, constraint relationships, temporal dependencies, and reference relationships.

[0028] In one specific embodiment, the edge type weight for causal relationships is 0.9, the edge type weight for constraint relationships is 0.8, the edge type weight for temporal dependencies is 0.7, and the edge type weight for reference relationships is 0.6. The edge type weights are preferred values ​​in this embodiment. In practical applications, the weights of each edge type can be adjusted by changing historical data according to the needs of specific project domains.

[0029] A multi-dimensional semantic dependency graph is constructed by using the nodes corresponding to each requirement item as the vertex set and the directed edges carrying edge type weights as the edge set.

[0030] The multi-dimensional semantic dependency graph fully preserves the four types of logical dependencies in the demand set—causal, temporal, constraint, and referential—as well as their propagation direction and strength information, providing structured input for subsequent calculation of the scope of change impact based on graph propagation algorithms.

[0031] S3. When a change event of the target requirement item is received, the node corresponding to the target requirement item is taken as the propagation start node in the multi-dimensional semantic dependency graph. The graph propagation algorithm is executed according to the edge type weight and the preset propagation depth threshold to obtain the influence degree score of each node in the multi-dimensional semantic dependency graph. In this embodiment, the steps for obtaining the influence score of each node are as follows: When a change event for a target requirement item is received, the node corresponding to the target requirement item is set as the propagation start node in the multi-dimensional semantic dependency graph, and the initial influence score of the propagation start node is set to 1.0. The initial influence scores of all other nodes in the multi-dimensional semantic dependency graph are set to 0. The meaning of the initial setting is: when the change event occurs, the target requirement item itself is completely affected, and the other nodes have not yet been affected. The subsequent influence propagates outward layer by layer along the directed edge direction.

[0032] Calculate the criticality coefficient of each node in the multi-dimensional semantic dependency graph; the criticality coefficient of each node is determined by the quantitative value of the requirement level and the quantitative value of the requirement type of each requirement item. In this embodiment, the calculation logic of the criticality coefficient of each node is as follows: obtain the demand level quantification value and demand type quantification value of the corresponding demand item of each node in the multi-dimensional semantic dependency graph, multiply the demand level quantification value and demand type quantification value of the corresponding demand item of each node, and calculate the criticality coefficient of each node. The demand level quantification value refers to the numerical value assigned according to the level of the demand item in the demand hierarchy. The higher the level, the larger the corresponding demand level quantification value. The required type quantification value refers to a numerical value assigned based on the safety criticality of the type to which the required item belongs. The higher the safety criticality, the larger the required type quantification value. Both the required level quantification value and the required type quantification value are pre-determined by those skilled in the art based on the specific project's requirement specification document and do not require automatic acquisition using existing technologies.

[0033] Specifically, the requirement hierarchy is divided into three levels: user requirements, system requirements, and software requirements; and the requirement types are divided into five types: security requirements, functional requirements, interface requirements, performance requirements, and constraint requirements. In a specific embodiment, the quantified values ​​for the requirement hierarchy corresponding to user requirements, system requirements, and software requirements are 1.2, 1.0, and 0.8, respectively; the quantified values ​​for the requirement types corresponding to security requirements, functional requirements, interface requirements, performance requirements, and constraint requirements are 1.0, 0.8, 0.7, 0.6, and 0.5, respectively; the criticality coefficient of each node is the product of the quantified value of the requirement hierarchy and the quantified value of the requirement type for the corresponding requirement item. The above values ​​are preferred values ​​in this embodiment, and in practical applications, each quantified value can be adjusted according to the historical data of requirement changes in the specific project domain.

[0034] It should be specifically noted that the purpose of introducing the criticality coefficient of each node is: under the same propagation path, the higher the criticality of the requirement item node, the stronger its response to the impact of the change, and its impact score should be amplified accordingly, so that the impact score can reflect the actual contribution of the requirement item's own attributes to the impact of the change.

[0035] Starting from the propagation initiation node, propagation proceeds outward layer by layer along the directed edges of the multidimensional semantic dependency graph. Based on the criticality coefficient of each node, the influence score of each node in the multidimensional semantic dependency graph is calculated. The calculation formula is as follows: ,in, S i The influence of the i-th node in the multi-dimensional semantic dependency graph is scored, where i is the node number in the multi-dimensional semantic dependency graph. S 0 The initial impact level of the propagation starting node is scored, with a value of 1.0. d i This represents the number of directed edges traversed from the starting node to the i-th node. This represents the edge type weight corresponding to the k-th directed edge in the propagation path, where k is the index of the directed edge in the propagation path, ranging from 1 to... d i , It is the product of the edge type weights of all directed edges along the propagation path. The path type combination correction coefficient is determined by the dependency relationship type combination of each directed edge on the propagation path. It is used to correct the overall influence propagation intensity of paths with different dependency relationship type combinations. λ is the propagation attenuation factor, λ∈(0,1), used to characterize the law that the influence score decreases with the increase of propagation level. In a specific embodiment, λ is taken as 0.8. The node criticality coefficient of the i-th node is determined by the product of the demand level quantification value and the demand type quantification value of the corresponding demand item.

[0036] It should be further noted that when the i-th node has multiple propagation paths from different directions, the path type combination correction coefficients corresponding to different propagation paths may be different; this applies when calculating the influence score of the i-th node. S i When, in the formula The path type combination correction coefficient corresponding to the propagation path that maximizes the influence score of this node.

[0037] It should be specifically noted that the path type combination correction coefficient The path type combination correction coefficient is determined based on the combination of dependency types of each directed edge along the propagation path. The value can be adjusted by changing historical data according to the needs of specific project areas; In one specific embodiment, when all directed edges on the propagation path are causal or constraint relationships... A value of 1.2 indicates that a propagation path consisting of directed edges of strongly dependent types has an amplifying effect on the propagation of influence; when the directed edges on the propagation path contain a mixed combination of temporal dependencies and reference relationships, A value of 0.9 indicates that a propagation path composed of a mixture of directed edges of weak dependency types has a diminishing effect on the propagation of influence; for other combinations of types, The value is 1.0.

[0038] It should be specifically explained that the purpose of introducing the path type combination correction coefficient is as follows: Existing graph propagation algorithms only consider the weight of a single directed edge, while ignoring the corrective effect of the combination of dependency types of multiple directed edges on the propagation path on the overall propagation intensity. When the propagation path is composed of directed edges of strong dependency types, the overall propagation intensity should be appropriately amplified; when the propagation path is composed of a mixture of directed edges of weak dependency types, the overall propagation intensity should be appropriately attenuated. The above correction mechanism enables the influence degree score to more accurately reflect the comprehensive role of multiple dependency types in the actual propagation path.

[0039] When there are multiple propagation paths from different directions for the i-th node, the maximum value of the influence score corresponding to all propagation paths is taken as the final influence score of the node. During the propagation process, when the influence score of a node is lower than the preset influence threshold, or when the number of directed edges traversed by the propagation reaches the preset propagation depth threshold, the propagation stops in that direction. In one specific embodiment, the preset propagation depth threshold is set to 4, meaning that the propagation influence extends to nodes within 4 directed edges from the propagation starting node.

[0040] The graph propagation algorithm is used to obtain the influence rating of each node in the multi-dimensional semantic dependency graph.

[0041] S4. Identify the demand items corresponding to nodes whose impact score exceeds the preset impact threshold as affected demand items, sort the affected demand items according to the impact score, and output the change impact range analysis results.

[0042] In this embodiment, the output change impact range analysis results are as follows: The influence score of each node in the multi-dimensional semantic dependency graph is compared with a preset influence threshold. The demand items corresponding to the nodes whose influence scores exceed the preset influence threshold are identified as affected demand items. The preset influence threshold is used to filter nodes with extremely low influence, so as to avoid including demand items with only a very weak dependency relationship with the target demand item in the scope of influence, thereby controlling the accuracy of the change influence scope analysis results. In one specific embodiment, the preset impact threshold is set to 0.1, meaning that the requirement items corresponding to nodes with an impact score lower than 0.1 are not included in the affected requirement items. The above value is the preferred value in this embodiment. In practical applications, the preset impact threshold can be adjusted according to the specific project's requirements for the accuracy of change impact analysis. It should be specifically noted that the preset impact threshold and the preset propagation depth threshold together form the termination control mechanism of the graph propagation algorithm; the preset propagation depth threshold limits the spread of the impact range from the perspective of propagation level, while the preset impact threshold filters out nodes with low impact from the perspective of impact intensity. The two work together to ensure that the analysis results of the change impact range are balanced between completeness and accuracy. The affected requirements are sorted in descending order of their impact scores to obtain a sorted list of affected requirements. The purpose of this sorting is to make the results of the change impact scope analysis intuitively reflect the relative degree of impact of each affected requirement on the change, and to provide a basis for requirement change reviewers to prioritize the requirements with high impact. Based on the sorted list of affected requirement items, generate and output the change impact scope analysis results; the change impact scope analysis results include the target requirement item number and change event description, the number and text content of each affected requirement item, the impact degree score corresponding to each affected requirement item, the propagation path between each affected requirement item and the target requirement item, and the dependency relationship type of each directed edge on the path. It should be specifically noted that the propagation path between each affected requirement item and the target requirement item, as well as the dependency type of each directed edge on the path, are retained in the change impact scope analysis results. The purpose is to enable requirement change reviewers to trace the specific dependency path of each affected requirement item into the impact scope, thereby understanding the logical basis for the impact propagation and supporting change decisions. The above propagation path information comes directly from the execution process record of the S3 graph propagation algorithm and does not require additional calculation. In this embodiment, it should be specifically noted that the above formulas are all dimensionless calculations. The formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0043] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0044] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for analyzing the impact of changes in requirements using natural language processing, characterized in that: Includes the following steps: Natural language processing is performed on each requirement item in the requirement document. Through syntactic dependency analysis, causal relationships, temporal dependencies, constraint relationships, and referential relationships are extracted from the text of each requirement item to obtain a multi-dimensional semantic dependency set corresponding to each requirement item. Using each requirement item as a node and each dependency in the multi-dimensional semantic dependency set as a directed edge, assign corresponding edge type weights according to the type of each dependency, and map each dependency in the multi-dimensional semantic dependency set to a directed edge in the multi-dimensional semantic dependency graph. The direction of the directed edge represents the propagation direction of the dependency, that is, from the dependency initiator node to the dependent node, thus constructing a multi-dimensional semantic dependency graph. When a change event of the target requirement item is received, the node corresponding to the target requirement item is taken as the propagation start node in the multi-dimensional semantic dependency graph. The graph propagation algorithm is executed according to the edge type weight and the preset propagation depth threshold, and the keyness coefficient of each node in the multi-dimensional semantic dependency graph is calculated to obtain the influence degree score of each node in the multi-dimensional semantic dependency graph. The demand items corresponding to nodes whose impact score exceeds the preset impact threshold are identified as affected demand items. The affected demand items are sorted according to the impact score, and the change impact range analysis results are output.

2. The method for analyzing the impact scope of demand changes combined with natural language processing according to claim 1, characterized in that: The steps for obtaining the multi-dimensional semantic dependency set are as follows: Obtain the requirements document and preprocess it. Segment the requirements document according to the requirements number identifier, and split the requirements document into independent requirements items. Each requirements item retains its number and corresponding text content. The text of each requirement item is segmented and tagged with part-of-speech tags, and then a syntactic dependency tree is constructed based on the tagging results. Identify statements containing conditional triggering clauses in the text of each requirement item, and mark the dependency relationship between two requirement items that satisfy the conditional triggering clause as a causal relationship; Identify statements in each requirement entry text that contain timing markers, and mark the dependency relationship between two requirement entries that contain timing markers and have an execution order as a timing dependency relationship; Identify statements in each requirement item text that contain modal verbs and limiting conditions, and mark the dependency relationship of the limiting conditions on other requirement items as a constraint relationship; Identify statements in each requirement entry text that contain explicit numbering or implicit business object pointing, and mark the dependency relationship between two requirement entries that contain explicit numbering or implicit business object pointing as a reference relationship; The extracted causal relationships, temporal dependencies, constraint relationships, and reference relationships are summarized to obtain a multi-dimensional semantic dependency set corresponding to each requirement item.

3. The method for analyzing the impact scope of demand changes combined with natural language processing according to claim 1, characterized in that: The steps for constructing a multi-dimensional semantic dependency graph are as follows: Each requirement item is mapped to a node in a multi-dimensional semantic dependency graph, and each node carries the corresponding requirement item's number and text content as node attributes. Each dependency in the multidimensional semantic dependency set is mapped to a directed edge in the multidimensional semantic dependency graph. The direction of the directed edge represents the propagation direction of the dependency, that is, from the dependency initiator node to the dependency-dependent node. Based on the type of each dependency and the degree of logical determination of each type of dependency in the propagation of the impact of demand changes, each directed edge is assigned a corresponding edge type weight. A multi-dimensional semantic dependency graph is constructed by using the nodes corresponding to each requirement item as the vertex set and the directed edges carrying edge type weights as the edge set.

4. The method for analyzing the impact scope of demand changes combined with natural language processing according to claim 3, characterized in that: The content pointed from the initiating node to the dependent node is as follows: The directed edges of the causal relationship are from the node corresponding to the demand item of the condition triggering party to the node corresponding to the demand item of the triggered party. The directed edges of the temporal dependencies are from the node corresponding to the preceding requirement item to the node corresponding to the following requirement item. The directed edges of the constraint relationship are pointed from the node corresponding to the requirement item of the constraint issuer to the node corresponding to the requirement item of the constrained party. The directed edges of the reference relationship are those where the node corresponding to the referencer's requirement item points to the node corresponding to the referenced's requirement item.

5. The method for analyzing the impact scope of demand changes combined with natural language processing according to claim 1, characterized in that: The steps for obtaining the influence rating of each node are as follows: When a change event for a target requirement item is received, the node corresponding to the target requirement item is set as the propagation start node in the multi-dimensional semantic dependency graph, and the initial influence score of the propagation start node is set to 1.0, while the initial influence score of all other nodes in the multi-dimensional semantic dependency graph is set to 0. Calculate the criticality coefficient of each node in the multi-dimensional semantic dependency graph; Starting from the propagation initiation node, propagation proceeds outward layer by layer along the directed edges of the multidimensional semantic dependency graph. Based on the criticality coefficient of each node, the influence score of each node in the multidimensional semantic dependency graph is calculated. When there are multiple propagation paths from different directions for the i-th node, the maximum value of the influence score corresponding to all propagation paths is taken as the final influence score of the node. During the propagation process, when the influence score of a node is lower than the preset influence threshold, or when the number of directed edges traversed by the propagation reaches the preset propagation depth threshold, the propagation stops in that direction. The graph propagation algorithm is used to obtain the influence rating of each node in the multi-dimensional semantic dependency graph.

6. The method for analyzing the impact scope of demand changes combined with natural language processing according to claim 5, characterized in that: The calculation logic for the criticality coefficient of each node is as follows: obtain the demand level quantification value and demand type quantification value of the corresponding demand item of each node in the multi-dimensional semantic dependency graph, multiply the demand level quantification value and demand type quantification value of the corresponding demand item of each node, and calculate the criticality coefficient of each node. The demand level quantification value refers to the numerical value assigned according to the level of the demand item in the demand hierarchy. The higher the level, the larger the corresponding demand level quantification value. The required type quantification value refers to the numerical value assigned based on the safety criticality of the type to which the required item belongs; the higher the safety criticality, the larger the required type quantification value.

7. The method for analyzing the impact scope of demand changes combined with natural language processing according to claim 1, characterized in that: The results of the analysis of the scope of impact of the output change are as follows: The influence score of each node in the multi-dimensional semantic dependency graph is compared with the preset influence threshold, and the demand items corresponding to the nodes whose influence scores exceed the preset influence threshold are identified as affected demand items. The affected demand items are sorted in descending order of their impact rating to obtain a sorted list of affected demand items. Based on the sorted list of affected requirement items, generate and output the results of the change impact scope analysis.

Images