A dynamic analysis method of cross-cultural discourse network

Abstract

The application provides a dynamic analysis method of a cross-cultural discourse network, comprising the following steps: obtaining source cultural metaphor expression adaptation rewriting data of a bridge node through a discourse network analysis platform, and identifying a replacement identification sequence of a nested pragmatic premise; separating the source text pragmatic premise and the replacement pragmatic premise according to the replacement identification sequence of the nested pragmatic premise, determining the depth level of the source text pragmatic premise and the replacement pragmatic premise, and generating a stripping replacement level matrix; extracting a downstream node backtracking verification behavior log from the stripping replacement level matrix, and statistically verifying the frequency and depth of tracking through the source text pragmatic premise and the replacement pragmatic premise, and constructing a verification intensity distribution diagram; constructing a distortion propagation risk diagram according to a propagation range of a fracture effect, analyzing the distortion propagation risk diagram, identifying a section diffusion transformation node, locating a potential blocking path starting point, and establishing a transformation node set.

Description

Technical Field

[0001] This invention relates to the field of information technology, and in particular to a dynamic analysis method for cross-cultural discourse networks. Background Technology

[0002] In cross-cultural information dissemination, information is forwarded step by step through multiple intermediate nodes, forming a discourse dissemination chain with a network topology. Among them, the bridging nodes located at the boundaries of different cultural circles play a transitive role in transforming the source cultural discourse into a form acceptable to the target culture. In reality, this can be manifested as cross-cultural media accounts, institutional translators, or content redistribution platforms, undertaking the core pragmatic transformation function. This transformation involves not only language translation but also the rewriting of the source cultural metaphorical expressions and their underlying pragmatic premises.

[0003] Current research generally suggests that the deeper the rewriting at bridging nodes, the stronger the acceptance within the target cultural sphere. However, this assessment overlooks the fact that downstream nodes are not passive recipients but actively engage in backtracking and verification. When downstream nodes trace the pragmatic relevance of the rewritten content to the source text, deeper rewriting implies a more thorough replacement of the original pragmatic premises. This makes it more difficult for downstream nodes to establish a credible mapping between the rewritten content and the source cultural context, leading them to question its authenticity and actively interrupt its dissemination. This indicates that the positive relationship between rewriting depth and the disruption effect can be reversed with the intervention of backtracking and verification.

[0004] The disruption effect spreads along the propagation chain as segmented discourse disconnects, the scope of which is limited by the pragmatic tolerance boundaries of the target cultural sphere and the magnitude of chain-like semantic shifts. However, existing methods lack the ability to quantitatively capture the number of pragmatic premises stripped, the frequency of retrospective verification, and the critical conditions for disruption diffusion, making it difficult to achieve targeted control over the disruption path.

[0005] Therefore, how to construct a dynamic monitoring mechanism for pragmatic premise stripping and retrospective verification behavior in the discourse network platform, identify the nodes of break-through diffusion and transformation, and achieve targeted intervention in the propagation path has become the core problem that this study urgently needs to solve. Summary of the Invention

[0006] This invention provides a dynamic analysis method for cross-cultural discourse networks. The method includes: acquiring source culture metaphor expression adaptation and rewriting data of bridging nodes through a discourse network analysis platform, and identifying replacement identifier sequences of nested pragmatic premises. Based on the substitution identifier sequence of nested pragmatic premises, the source text pragmatic premises and the substitution pragmatic premises are separated, the depth level of the source text pragmatic premises and the substitution pragmatic premises is determined, and a substitution level matrix is ​​generated. The downstream node backtracking verification behavior logs are extracted from the stripping and replacement hierarchy matrix. The verification frequency and tracing depth are statistically analyzed by the pragmatic premises of the source text and the pragmatic premises of the replacement text, and a verification intensity distribution map is constructed. Identify high-frequency questioning regions in the verification intensity distribution map, analyze the semantic distortion accumulation path from bridging nodes to downstream nodes in the high-frequency questioning regions, set the pragmatic tolerance boundary of the target cultural circle, and evaluate the propagation range of the fracture effect by combining the semantic distortion accumulation path and the pragmatic tolerance boundary of the target cultural circle. Based on the propagation range of the fracture effect, a distortion propagation risk map is constructed. The distortion propagation risk map is analyzed to identify segmental diffusion transformation nodes, locate the starting point of potential blocking paths, and establish a set of transformation nodes. Based on the set of transformation nodes, the rewriting depth of upstream bridging nodes is traced, the fracture propagation path is marked and high-risk propagation chains are isolated, pragmatic adaptation feedback units are deployed to send tolerance boundary signals back to the bridging nodes, and an optimized rewriting depth vector containing backtracking verification suppression markers is generated.

[0007] Furthermore, the process of obtaining source cultural metaphorical expression adaptation and rewriting data of bridging nodes through a discourse network analysis platform, and identifying replacement identifier sequences of nested pragmatic premises, includes: The source cultural metaphor expression adaptation and rewriting data of bridging nodes are obtained through a discourse network analysis platform. The comparison records of source text metaphor fragments and target text adaptation fragments are extracted from the rewriting data. The comparison records contain the pragmatic premises of the source text and their corresponding target adaptation content. The nesting level of each pragmatic premise is marked according to the nesting and inclusion relationship of the pragmatic premises of the source text in the comparison records, and a distribution map of pragmatic premises containing nesting levels is obtained. For each nested level in the pragmatic premise distribution map, the word differences between the pragmatic premise of the source text and the target adapted content are compared one by one. If there is a word in the target adapted content that is semantically inconsistent with the pragmatic premise of the source text, the position of the word is marked as a replacement site. The replacement sites are arranged according to the order of their appearance in the discourse propagation chain to obtain the replacement identifier sequence of the nested pragmatic premises.

[0008] Furthermore, the step of separating the source text pragmatic premises and the alternative pragmatic premises based on the substitution identifier sequence of nested pragmatic premises, determining the depth level of the source text pragmatic premises and the alternative pragmatic premises, and generating a stripping substitution level matrix includes: Based on the substitution identifier sequence of nested pragmatic premises, the source text pragmatic premise to be replaced and the replaced pragmatic premise are extracted from each substitution point. Pairing records of source text pragmatic premises and replaced pragmatic premises are established according to the correspondence of substitution points to obtain a premise pairing set. For each pairing record in the premise pairing set, the nesting position of the pragmatic premise of the source text in the original discourse structure is traced, and the number of other pragmatic premises wrapped by the pragmatic premise of the source text is counted as the source text depth level. At the same time, the nesting position of the alternative pragmatic premise in the target adapted discourse structure is traced, and the number of other adapted contents wrapped by the alternative pragmatic premise is counted as the replacement depth level, so as to obtain the bidirectional depth level annotation of each pairing record. The numerical difference between the source text depth level and the replacement depth level is extracted from the bidirectional depth level annotation. If the source text depth level is higher than the replacement depth level, the paired record is marked as a deep stripping type. If the source text depth level is lower than or equal to the replacement depth level, the paired record is marked as a shallow replacement type, thus forming a level difference record containing the stripping type marker. Based on the hierarchical difference records, a two-dimensional matrix is ​​constructed with the source text depth level as the row index and the replacement depth level as the column index. The span of each paired record is filled into the corresponding row and column intersection position to form a stripping replacement level matrix covering all replacement nodes.

[0009] Furthermore, the generation of the stripping and replacement hierarchy matrix further includes: From the separated source text pragmatic premises and alternative pragmatic premises, extract the source text depth level and target adaptation level corresponding to each alternative node. The source text depth level is the nesting depth value of the source text pragmatic premise in the source cultural discourse system, and the target adaptation level is the nesting depth value of the alternative pragmatic premise in the target cultural discourse system, so as to obtain the bidirectional hierarchical annotation record of each alternative node. For each replacement node in the bidirectional hierarchical annotation record, the cross-layer rewriting direction is determined by comparing the numerical values ​​of the source text depth level and the target adaptation level. If the source text depth level is higher than the target adaptation level, the rewriting direction is to transform to a shallower level; if the source text depth level is lower than the target adaptation level, the rewriting direction is to transform to a deeper level. At the same time, the difference between the two level values ​​is used as the level crossing range, resulting in a node feature set that includes the rewriting direction and the crossing range. Based on the source text depth level and target adaptation level of all replacement nodes in the node feature set, a two-dimensional matrix is ​​constructed with the source text depth level as the row dimension and the target adaptation level as the column dimension. The span of each replacement node is filled into the corresponding row and column intersection position. The position where the span exceeds the preset normal adaptation threshold is marked as an abnormal cross-layer region, forming a stripping replacement level matrix covering all replacement nodes.

[0010] Furthermore, the step of extracting downstream node backtracking verification behavior logs from the stripping and replacement hierarchy matrix, and constructing a verification intensity distribution map by statistically analyzing the verification frequency and tracing depth of the source text pragmatic premises and the replaced pragmatic premises, includes: Extract the downstream node backtracking verification behavior logs corresponding to each replacement node from the stripping and replacement hierarchy matrix. The behavior logs record the operation records of the downstream node initiating a query to the source text for each replacement node and comparing it with the rewritten content, thereby obtaining a set of backtracking verification records associated with each row and column position in the matrix. For each record in the backtracking verification record set, the number of queries initiated by downstream nodes is counted based on the pairing relationship between the pragmatic premises of the source text and the alternative pragmatic premises, which is used as the verification frequency. At the same time, the number of propagation chain nodes traversed by the downstream node when backtracking from the current position to the source text is counted as the tracing depth, thus obtaining the node verification feature containing the verification frequency and the tracing depth. Based on the node verification characteristics, using the row and column positions of the stripping and replacement hierarchy matrix as coordinate references, the verification frequency and tracing depth of each replacement node are mapped to the corresponding coordinate positions. The verification frequency and tracing depth are multiplied to obtain the verification intensity value, which is then filled into each coordinate position to construct a verification intensity distribution map.

[0011] Furthermore, the identification of high-frequency questioning regions in the verification intensity distribution map, analysis of the semantic distortion accumulation path from bridging nodes to downstream nodes in the high-frequency questioning regions, setting the pragmatic tolerance boundary of the target cultural sphere, and assessing the propagation range of the fracture effect by combining the semantic distortion accumulation path and the pragmatic tolerance boundary of the target cultural sphere, including: High-frequency questioning areas are identified from the verification intensity distribution map. The verification intensity value at each coordinate position is compared with a preset questioning frequency threshold. If the verification intensity value at a certain coordinate position exceeds the questioning frequency threshold, the coordinate position and its adjacent consecutive positions exceeding the threshold are defined as high-frequency questioning areas, thus obtaining a set of high-frequency questioning areas. For each high-frequency questioning region in the set of high-frequency questioning regions, locate the bridging node and downstream node corresponding to the region, track the changes in semantic content segment by segment along the propagation chain from the bridging node to the downstream node, and accumulate the semantic distortion accumulation path from the bridging node to the downstream node according to the degree of semantic difference between the rewritten content and the source text in each propagation segment. Based on the tolerance threshold and tolerance gap type pre-established in the target cultural circle, a pragmatic tolerance boundary is set. The pragmatic tolerance boundary is superimposed and compared with the semantic distortion accumulation path. If the cumulative distortion value of a certain segment in the semantic distortion accumulation path exceeds the tolerance threshold of the pragmatic tolerance boundary, the segment is marked as an over-limit path segment, and a set of over-limit path segments is obtained. For each of the out-of-limit path segments in the set of out-of-limit path segments, the number of downstream nodes affected by the segment and the coverage of the propagation layers are counted along the propagation chain downstream to assess the actual propagation range of the fracture effect.

[0012] Furthermore, the assessment of the propagation range of the fracture effect also includes: Extract the tolerance threshold and tolerance gap type of each target cultural circle from the established pragmatic tolerance boundary. The tolerance threshold is the upper limit of semantic offset acceptance set in advance for the circle, and the tolerance gap type is the rejection mark of the circle for specific category pragmatic premise substitution, so as to obtain the tolerance feature record of each target cultural circle. Based on the tolerance feature record, the distortion magnitude and distortion nature of each propagation segment in the semantic distortion accumulation path are read. The distortion magnitude is the degree of deviation between the semantics of the source text and the semantics of the rewritten content within the segment, and the distortion nature is the pragmatic premise substitution category involved in the segment. If the distortion magnitude of a segment exceeds the tolerance threshold of the corresponding layer, or the distortion nature of the segment belongs to the tolerance gap type of the corresponding layer, then the segment is marked as an over-limit segment and its path node position is recorded to obtain the association set of over-limit segments and path nodes. For each overlimited segment in the associated set, the speech propagation chain is traversed downstream. The number of downstream nodes that receive the distorted content of the overlimited segment is counted as the diffusion coverage breadth. The number of propagation chain layers that the distorted content traverses when it is transmitted downstream is counted as the distortion transmission level. The actual propagation range of the break effect is determined.

[0013] Furthermore, the process of constructing a distortion propagation risk map based on the propagation range of the fracture effect, analyzing the distortion propagation risk map, identifying segmental diffusion transformation nodes, locating the starting point of potential blocking paths, and establishing a set of transformation nodes includes: Based on the actual propagation range of the fracture effect, and using the topological structure of the discourse propagation chain as a basis, each over-limit segment and its diffusion coverage and distortion transmission level are marked at the corresponding position of the propagation chain. An edge set is constructed based on the propagation connection relationship between nodes, and the product of the diffusion coverage and distortion transmission level of each node is used as the node weight to obtain the distortion propagation risk map. For each node in the distortion propagation risk graph, based on the number of branches and the length of the distribution path that the node distributes downstream after receiving upstream distorted content, if the number of branches exceeds a preset branch threshold and the length of the distribution path exceeds a preset path threshold, then the node is marked as a diffusion conversion node, and a set of diffusion conversion nodes is obtained. For each diffusion conversion node in the set of diffusion conversion nodes, backtrack upstream along the propagation chain to the nearest overlimited section, mark the path node where the overlimited section is located as the potential blocking path start point, and obtain the blocking path start point record; Based on the record of the starting point of the blocking path and the set of diffusion and transformation nodes, each diffusion and transformation node is associated and paired with its corresponding starting point of the blocking path. The nodes are arranged in order of their position in the propagation chain to establish a set of transformation nodes that includes the node position, the starting point of the blocking path and the propagation path to which they belong.

[0014] Furthermore, the establishment of the conversion node set also includes: The distortion degree and the preconditions for triggering propagation of each segment of the diffusion and transformation node are retrieved from the distortion propagation risk map. The distortion degree is the semantic deviation value between the distorted content received and forwarded by the node and the source text. The preconditions are the number of paths and the frequency of receiving distorted content from upstream by the node. The propagation triggering characteristic record of each diffusion and transformation node is obtained. For each diffusion transformation node in the propagation trigger feature record, identify the connection position of the node with the upstream node and the connection position with the downstream node in the propagation chain. If the connection between the node and the upstream node is a single path and there are multiple branch connections with the downstream node, then mark the node as a candidate blocking node and record the position of its upstream and downstream connection breakpoints to obtain a set of candidate blocking nodes. For each candidate blocking node in the candidate blocking node set, its blocking capability is evaluated based on the number of distorted content sources lost by the downstream branches after the node is cut off, and its intervention feasibility is evaluated based on the number of path layers of the node from the bridging node in the propagation chain. The blocking capability and intervention feasibility are integrated into the attribute record of the node to establish a set of conversion nodes that includes the node location, blocking capability and the path to which it belongs.

[0015] Furthermore, the process of tracing the rewriting depth of upstream bridging nodes based on the set of transformation nodes, marking the fracture propagation path and isolating high-risk propagation chains, deploying pragmatic adaptation feedback units to send tolerance boundary signals back to the bridging nodes, and generating an optimized rewriting depth vector containing backtracking verification suppression markers includes: Based on the set of transformation nodes, the bridging node positions corresponding to each transformation node are traced upstream along the propagation chain. The hierarchical span when the bridging node performs a replacement operation on the pragmatic premise of the source text is extracted from the rewriting record of the bridging node as the rewriting depth. At the same time, the propagation path from the bridging node to each transformation node is marked as a break diffusion path. If the number of downstream nodes covered by the break diffusion path exceeds a preset risk threshold, the propagation chain to which the path belongs is marked as a high-risk propagation chain and its downstream propagation connection is interrupted, resulting in a set of isolated high-risk propagation chains. Read the tolerance threshold and tolerance gap type of the target cultural circle corresponding to the bridging node in the high-risk propagation chain set, encapsulate the tolerance threshold and tolerance gap type into a tolerance boundary signal and send it back to the bridging node to obtain the tolerance boundary reception record of the bridging node. Based on the tolerance boundary reception record, the rewrite depth of each bridging node is compared with the received tolerance boundary signal. If the rewrite depth exceeds the tolerance threshold, a backtracking verification suppression flag is added at the rewrite depth value position. The rewrite depth values ​​of each bridging node and the suppression flag are integrated into a vector form to generate an optimized rewrite depth vector containing the backtracking verification suppression flag.

[0016] The technical solutions provided by the embodiments of the present invention may include the following beneficial effects: This invention discloses a dynamic analysis method for cross-cultural discourse networks, proposing a systematic solution to the propagation risks of semantic distortion and fragmentation effects in cross-cultural communication. By analyzing the source cultural metaphorical expressions and adaptation rewriting data of bridging nodes, this invention identifies substitution sequences of nested pragmatic premises, constructs a matrix to strip substitution levels, accurately locates the cumulative paths of semantic distortion and high-frequency questioning areas, and assesses the propagation range of fragmentation effects by combining the pragmatic tolerance boundaries of the target cultural sphere. This generates a distortion propagation risk map, identifies segmented diffusion transformation nodes, and establishes a set of blocking paths. By tracing back the upstream rewriting depth, pragmatic adaptation feedback units are deployed to send optimization signals back to bridging nodes, ultimately generating a rewriting depth vector containing suppression markers. The core innovation of this invention lies in effectively isolating high-risk propagation chains through dynamic analysis and feedback mechanisms, reducing the risk of semantic distortion in cross-cultural communication, improving the accuracy and cultural tolerance of discourse adaptation, and providing scientific support for cross-cultural communication. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a dynamic analysis method for cross-cultural discourse networks according to the present invention.

[0018] Figure 2 This is a schematic diagram illustrating a dynamic analysis method for cross-cultural discourse networks according to the present invention.

[0019] Figure 3 This is another schematic diagram of a dynamic analysis method for cross-cultural discourse networks according to the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0021] like Figures 1-3 This embodiment of a dynamic analysis method for cross-cultural discourse networks may specifically include: Step S101: Obtain source cultural metaphor expression adaptation and rewriting data of bridging nodes through the discourse network analysis platform, and identify the replacement identifier sequence of nested pragmatic premises.

[0022] The source cultural metaphorical expression adaptation and rewriting data of bridging nodes are obtained through a discourse network analysis platform. From this rewriting data, comparison records of source text metaphorical fragments and target text adaptation fragments are extracted. These comparison records contain the pragmatic premises of the source text and their corresponding target adaptation content. Based on the nesting and inclusion relationships of the source text pragmatic premises in the comparison records, the nesting levels of each pragmatic premise are marked, resulting in a pragmatic premise distribution map containing nesting levels. For each nesting level in the pragmatic premise distribution map, the word differences between the source text pragmatic premises and the target adaptation content are compared one by one. If a word in the target adaptation content is semantically inconsistent with the pragmatic premise of the source text, its location is marked as a replacement site. These replacement sites are arranged according to their order of appearance in the discourse propagation chain, resulting in a sequence of replacement identifiers for nested pragmatic premises.

[0023] In cross-cultural discourse communication scenarios, bridging nodes play a crucial role in transforming metaphorical expressions from the source culture into forms acceptable to the target culture. A discourse network analysis system built using the Neo4j graph database records the entire process of bridging nodes adapting and rewriting source text. The system takes as input metaphorical fragments from the source text and target cultural corpora. Its working principle involves extracting metaphorical features using a BERT model and applying a rule-based transformation algorithm to generate adapted fragments. The output consists of rewritten data containing the original expressions of the metaphorical fragments from the source text and the adapted target text fragments transformed by the bridging nodes. The construction process includes data collection, node modeling, and algorithm integration, ensuring reproducibility with a similarity threshold set to 0.8.

[0024] Specifically, the pragmatic premises in the comparative records have a nested inclusion relationship, that is, a certain pragmatic premise may be based on another pragmatic premise.

[0025] In one embodiment, a list of pragmatic premises is taken as input, and the semantic dependencies between premises are analyzed using a dependency parsing algorithm to construct a directed dependency graph, where nodes represent premises and edges represent dependency directions. If the validity of premise A depends on premise B, an edge from A to B is added, and the dependency depth is calculated: the root node has a depth of 0, and the child node depth is increased by 1. If the depth is greater than 1, it is determined to be a nested level, and the dependent premise is marked as a deeper level, for example, depth 2 is marked as level 2. This layer-by-layer labeling forms a pragmatic premise distribution graph. Based on the pragmatic premise distribution graph, for each nested level, the pragmatic premises of the source text are compared with the target adapted content. If the semantic direction of a term in the target adapted content is inconsistent with the pragmatic premise of the source text, the position of that term is marked as a replacement site. According to the temporal order of the replacement sites in the propagation chain, a sequence of replacement identifiers for nested pragmatic premises is obtained.

[0026] Step S102: Based on the substitution identifier sequence of nested pragmatic premises, the source text pragmatic premises and the substitution pragmatic premises are separated, the depth level of the source text pragmatic premises and the substitution pragmatic premises is determined, and a stripping substitution level matrix is ​​generated.

[0027] Based on the substitution identifier sequence of nested pragmatic premises, the source text pragmatic premise to be replaced and the substituted pragmatic premise are extracted from each substitution point. Pairing records of source text pragmatic premises and substituted pragmatic premises are established according to the correspondence of substitution points, resulting in a premise pairing set. For each pairing record in the premise pairing set, the nesting position of the source text pragmatic premise in the original discourse structure is traced, and the number of other pragmatic premises wrapped around the original pragmatic premise is counted as the source text depth level. Simultaneously, the nesting position of the substituted pragmatic premise in the target adapted discourse structure is traced, and the number of other adapted contents wrapped around the substituted pragmatic premise is counted as the substitution depth level, obtaining a bidirectional depth level annotation for each pairing record. The numerical difference between the source text depth level and the substitution depth level is extracted from the bidirectional depth level annotation. If the source text depth level is higher than the substitution depth level, the pairing record is marked as a deep stripping type; if the source text depth level is lower than or equal to the substitution depth level, the pairing record is marked as a shallow substitution type, forming a level difference record containing stripping type markers. Based on the hierarchical difference records, a two-dimensional matrix is ​​constructed with the source text depth level as the row index and the replacement depth level as the column index. The span of each paired record is filled into the corresponding row and column intersection position. The span calculation formula is D=|SR|, where S is the source text depth level and R is the replacement depth level. Positions with spans exceeding a preset threshold of 3 are marked as abnormal cross-layer regions. At the same time, a peeling type mark is superimposed to form a peeling replacement level matrix covering all replacement nodes.

[0028] In cross-cultural discourse communication scenarios, the nested pragmatic premise substitution identifier sequence records all substitution site information for the bridging node's adaptation and rewriting of the source text. When extracting pairing records from the substitution identifier sequence, for each substitution site, the corresponding source text pragmatic premise and substitution pragmatic premise are read. The source text pragmatic premise refers to the expression unit in the original discourse that carries the meaning of a specific cultural context, and the substitution pragmatic premise refers to the expression unit after the bridging node rewrites it to adapt to the target culture. The two form a one-to-one pairing relationship and converge into a premise pairing set.

[0029] Specifically, the establishment of the premise pairing set depends on the sequential traversal of the substitution sites.

[0030] In one possible implementation, the depth level is determined by tracing the nesting position of pragmatic premises within the discourse structure. The nesting position refers to the hierarchical relationship in which a pragmatic premise is wrapped by other pragmatic premises within the overall discourse structure; the more outer pragmatic premises, the deeper the nesting level. By tracing the nesting position of the replacement pragmatic premise within the target matching discourse structure and counting the number of matching contents wrapped by it, the replacement depth level is obtained. By annotating both the source text depth level and the replacement depth level for each paired record, a bidirectional depth level annotation is formed.

[0031] It should be noted that the bidirectional depth-level annotation reflects the degree of influence of the bridging node rewriting operation on the nested pragmatic premise structure. When a deeply nested pragmatic premise in the source text is rewritten as a shallowly nested alternative pragmatic premise in the target discourse, it means that the deep cultural context of the original discourse has been stripped away during the rewriting process. Furthermore, paired records are classified and labeled based on the numerical difference between the source text depth level and the alternative depth level. If the source text depth level is higher than the alternative depth level, it indicates that the rewriting process transforms the deeply nested pragmatic premise of the source text into a shallow expression; such paired records are labeled as deep stripping type. If the source text depth level is lower than or equal to the alternative depth level, it indicates that the rewriting process maintains or deepens the nesting level; such paired records are labeled as shallow substitution type.

[0032] In one embodiment, the source text depth level of a certain paired record is level three and the replacement depth level is level one, with a difference of two. Since the source text depth level is higher than the replacement depth level, it is marked as a deep stripping type.

[0033] Understandably, deeply stripped pairings hold special significance in cross-cultural communication. Such rewriting operations often involve a fundamental transformation of the core context of the source culture, making it easier for downstream nodes to detect semantic breaks between the source text and the rewritten content during backtracking verification. When constructing a two-dimensional matrix based on hierarchical difference records containing stripping type markers, the row index is formed by the value range of the source text's depth level, and the column index is formed by the value range of the replacement depth level. For each pairing record, its row position is determined based on its source text depth level, its column position is determined based on its replacement depth level, and the stripping type marker of that pairing record is filled into the row-column intersection.

[0034] For example, if a paired record has a source text depth level of three, a replacement depth level of one, and a stripping type of deep stripping, then a deep stripping marker is filled in the third row and first column of the matrix. After traversing all paired records and completing the filling, a stripping and replacement level matrix is ​​obtained. This stripping and replacement level matrix visually presents the distribution characteristics of bridging node rewriting operations across different depth levels. The regions in the matrix with concentrated deep stripping markers correspond to rewriting operations with a high degree of pragmatic premise stripping.

[0035] Extract the original cultural affiliation level and target adaptation level corresponding to each substitution from the separated source text pragmatic premises and alternative pragmatic premises. Identify the cross-level rewriting direction and level crossing range of each substitution node. Analyze the substitution operations and their depth distribution where the level gap exceeds the normal adaptation range, and form a stripped substitution level matrix covering all substitution nodes.

[0036] From the separated source text pragmatic premises and alternative pragmatic premises, the source text depth level and target adaptation level corresponding to each replacement node are extracted. The source text depth level is the nesting depth value of the source text pragmatic premise in the source cultural discourse system, and the target adaptation level is the nesting depth value of the alternative pragmatic premise in the target cultural discourse system, thus obtaining a bidirectional hierarchical annotation record for each replacement node. For each replacement node in the bidirectional hierarchical annotation record, the cross-layer rewriting direction is determined by comparing the values ​​of the source text depth level and the target adaptation level. If the source text depth level is higher than the target adaptation level, the rewriting direction is a shallower transformation; if the source text depth level is lower than the target adaptation level, the rewriting direction is a deeper transformation. Simultaneously, the difference between the two level values ​​is used as the level leap range, resulting in a node feature set containing the rewriting direction and leap range. Based on the source text depth level and target adaptation level of all replacement nodes in the node feature set, a two-dimensional matrix is ​​constructed with the source text depth level as the row dimension and the target adaptation level as the column dimension. The span of each replacement node is filled into the corresponding row and column intersection position. The position where the span exceeds the preset normal adaptation threshold is marked as an abnormal cross-layer region. The threshold is set to 3. The calculation process is to input the historical layer difference dataset and use the formula D=μ+σ, where D is the threshold, μ is the mean, and σ is the standard deviation. The output threshold is used for marking, forming a stripping replacement layer matrix covering all replacement nodes. This matrix is ​​then input into downstream nodes to analyze the cross-cultural adaptation distribution characteristics and optimize the replacement strategy.

[0037] In the process of cross-cultural discourse communication, the extraction of the original cultural affiliation level and the target adaptation level is the foundation for identifying the rewriting behavior of bridging nodes. The original cultural affiliation level refers to the nesting depth of the pragmatic premise of the source text within the source cultural discourse system, where it is surrounded by other pragmatic premises. The larger the value, the more core the pragmatic premise is in the cultural context.

[0038] Specifically, the target adaptation hierarchy employs a nested depth-of-field calculation approach, statistically analyzing the position of the replacement pragmatic premise within the target cultural discourse system. The calculation process is as follows: The input discourse system is a tree structure. A Depth-First Search (DFS) algorithm is used to traverse from the root node. The depth *d* is defined as the path length from the root to that node; for the root node, *d*=0, and for child nodes, *d*=1. For example, if premise A is a second-level node under the root, then *d*=2. By sequentially reading the source text pragmatic premise and the replacement pragmatic premise corresponding to each replacement node, their *d* values ​​are determined, forming a bidirectional hierarchical annotation record.

[0039] In one embodiment, the direction of cross-level rewriting is determined based on the numerical relationship between the original cultural affiliation level and the target adaptation level. If the original cultural affiliation level is higher than the target adaptation level, it indicates that the bridging node rewrites the deeply nested pragmatic premises in the source culture into shallow expressions in the target culture, and the rewriting direction is determined to be a shallow transformation; conversely, if the original cultural affiliation level is lower than the target adaptation level, the rewriting direction is determined to be a deep transformation. The extent of the level crossing is determined by the difference between the values ​​of the two levels.

[0040] It should be noted that the preset standard adaptation threshold is set based on the pragmatic tolerance characteristics of the target cultural sphere. The specific process is as follows: 100 pragmatic samples are collected from the target cultural sphere, and the tolerance score T for each sample is calculated using the formula D=(∑T). i ) / N, where T i Let N be the score of the i-th sample, and N be the number of samples (100). The average tolerance D is used as a threshold, ranging from 0 to 10. If D is less than 5, the threshold is 3; if D is between 5 and 7, the threshold is 4; if D is greater than 7, the threshold is 5. When the level crossing exceeds this threshold, the corresponding matrix position is marked as an abnormal cross-level region. Further, the stripping and replacement level matrix is ​​constructed with the original cultural affiliation level as the row dimension and the target adaptation level as the column dimension. For all replacement nodes in the node feature set, their row and column intersection positions in the matrix are located based on the bidirectional level values ​​of each node. The level crossing range is filled into this position, and abnormal cross-level regions are visually marked in the matrix, thus completing the construction of a stripping and replacement level matrix covering all replacement nodes.

[0041] Step S103: Extract the downstream node backtracking verification behavior log from the stripping and replacement hierarchy matrix, and construct a verification intensity distribution map by statistically analyzing the verification frequency and tracing depth of the source text pragmatic premise and the replacement pragmatic premise.

[0042] Extract the downstream node backtracking verification behavior logs corresponding to each replacement node from the stripping and replacement hierarchy matrix. These logs record the operations of downstream nodes initiating queries against the source text for each replacement node and comparing them with the rewritten content, thus obtaining a set of backtracking verification records associated with each row and column position in the matrix. For each record in the backtracking verification record set, the number of times downstream nodes initiate queries for that pairing is counted based on the pairing relationship between the pragmatic premise of the source text and the pragmatic premise of the replacement, serving as the verification frequency. Simultaneously, the number of propagation chain nodes traversed by the downstream node when backtracking from its current position towards the source text is counted as the tracing depth, resulting in node verification features containing verification frequency and tracing depth. Based on these node verification features, using the row and column positions of the stripping and replacement hierarchy matrix as coordinate references, map the verification frequency and tracing depth of each replacement node to their corresponding coordinate positions. Multiply the verification frequency by the tracing depth to obtain the verification intensity value and fill it into each coordinate position to construct a verification intensity distribution map.

[0043] In the process of cross-cultural discourse dissemination, the backtracking and verification behavior of downstream nodes is a key indicator for judging the stability of the dissemination chain. The backtracking and verification behavior log records all the operational traces of downstream nodes actively initiating queries towards the source text and comparing them with the rewritten content after receiving the rewritten content from the bridging node.

[0044] Specifically, the stripping and replacement hierarchy matrix is ​​defined as an M-row, N-column two-dimensional array, where M represents the number of replacement levels and N represents the number of nodes in each level, used to represent the replacement relationships between nodes. The matrix construction process is as follows: the input is a list of node identifiers and hierarchy relationship data; Dijkstra's algorithm is used to calculate the shortest replacement path; the output is a matrix filled with replacement node identifiers. For example, when the number of levels is 3 and the number of nodes is 4, the matrix elements are as follows: A[1,2]=node X. When extracting behavior logs from this matrix, based on the replacement node identifiers corresponding to each row and column position in the matrix, the historical records of the replacement node being queried by downstream nodes are obtained, forming a backtracking verification record set that corresponds one-to-one with the matrix coordinates. Subsequently, this set is used to count the query frequency. If the frequency exceeds the threshold of 0.5, a verification alarm is triggered to ensure that the output of the previous step is correlated with the subsequent statistics.

[0045] In one embodiment, the verification frequency is calculated for each pairing of pragmatic premises in the source text and their replacement pragmatic premises. A pragmatic premise refers to an implicit pragmatic assumption in the text, such as the source text assuming the reader has prior knowledge. The pairing rule is as follows: input the source text and the replacement text, compare their respective premise vectors using a cosine similarity algorithm, and pairings are formed if the similarity exceeds 0.8. When the same pair is queried by multiple downstream nodes, or queried multiple times by the same downstream node, the cumulative number of queries is used as the verification frequency for that pair. A higher verification frequency indicates a stronger willingness among downstream nodes to verify the rewritten content of the pair.

[0046] It should be noted that the tracing depth reflects the distance that a downstream node travels through the propagation chain when tracing back to the source text. The specific calculation process is as follows: Input the propagation chain graph, start from the downstream node, use the breadth-first search algorithm to traverse to the source node, and output the depth D=N-1, where D is the depth and N is the total number of nodes in the path.

[0047] For example, if the link is node A-node B-source node C, then N=3 and D=2.

[0048] In one possible implementation, the tracing depth is determined by counting the number of intermediate propagation nodes that a downstream node passes through when tracing back from its location towards the source text.

[0049] For example, if a downstream node is located at the fifth layer of the propagation chain and its backtracking query directly points to the bridging node at the second layer, then the number of intermediate nodes traversed is two, and the tracing depth is denoted as two. Further, the verification intensity distribution map is constructed using the row and column structure of the stripped replacement hierarchy matrix as the coordinate reference. The stripping operation involves inputting the replacement hierarchy matrix M, extracting row vectors as x-axis coordinates and column vectors as y-axis coordinates using a matrix factorization algorithm such as SVD, and outputting the coordinate reference. For each replacement node, its verification frequency is multiplied by the tracing depth to obtain the verification intensity value. This value reflects both the activity level of the query and the penetration distance of the backtracking. The verification frequency is derived from system log data, and a counting algorithm is used to count the number of times each node is queried by downstream nodes. For example, if node A is queried 5 times, its frequency is 5. The verification intensity value is then filled into the corresponding matrix coordinate positions, thus forming a verification intensity distribution map covering all replacement nodes. This distribution map visually presents the intensity characteristics of each replacement node's backtracking verification by downstream nodes.

[0050] Step S104: Identify high-frequency questioning regions in the verification intensity distribution map, analyze the semantic distortion accumulation path from bridging nodes to downstream nodes in the high-frequency questioning regions, set the pragmatic tolerance boundary of the target cultural circle, and evaluate the propagation range of the fracture effect by combining the semantic distortion accumulation path and the pragmatic tolerance boundary of the target cultural circle.

[0051] High-frequency questioning regions are identified from the verification intensity distribution map. The verification intensity values ​​at each coordinate position are compared with a preset questioning frequency threshold. If the verification intensity value at a certain coordinate position exceeds the questioning frequency threshold, that coordinate position and its adjacent consecutive exceeding-threshold positions are designated as high-frequency questioning regions, thus obtaining a set of high-frequency questioning regions. For each high-frequency questioning region in the set, the bridging node and downstream node positions corresponding to that region are located. The semantic content changes are tracked segment by segment along the propagation chain from the bridging node to the downstream node. The semantic difference between the rewritten content and the source text within each propagation segment is accumulated segment by segment to obtain the semantic distortion accumulation path from the bridging node to the downstream node. A pragmatic tolerance boundary is set based on a pre-established tolerance threshold and tolerance gap type for the target cultural sphere. The pragmatic tolerance boundary is superimposed and compared with the semantic distortion accumulation path. If the accumulated distortion value of a segment in the semantic distortion accumulation path exceeds the tolerance threshold of the pragmatic tolerance boundary, that segment is marked as an over-limit path segment, thus obtaining a set of over-limit path segments. For each of the out-of-limit path segments in the set of out-of-limit path segments, the number of downstream nodes affected by the segment and the coverage of the propagation layers are counted along the propagation chain downstream. The number of downstream nodes and the coverage of the propagation layers of all out-of-limit path segments are summarized to assess the actual propagation range of the fracture effect.

[0052] In the process of cross-cultural discourse dissemination, the identification of high-frequency questioning regions is based on the verification intensity numerical characteristics of each coordinate position in the verification intensity distribution map. The verification intensity distribution map uses the row and column structure of the stripped replacement hierarchy matrix as the coordinate reference, with each coordinate position corresponding to the verification intensity value of a replacement node. The process of stripping the replacement hierarchy matrix is ​​as follows: Input the original replacement hierarchy matrix M, which is n rows and m columns in size, where the element values ​​are the hierarchy values ​​of the replacement nodes; use a traversal algorithm to scan the matrix row by row, identify and retain replacement nodes with hierarchy values ​​greater than 0, and remove non-replacement nodes with hierarchy values ​​of 0; output the stripped matrix structure, which only contains the row and column coordinates of the replacement nodes, as the reference for the distribution map. When the verification intensity value at a certain coordinate position exceeds a preset questioning frequency threshold, it indicates that the replacement node is frequently back-verified by downstream nodes, posing a high risk of propagation breakage. This coordinate position and its adjacent consecutive positions exceeding the threshold are collectively designated as high-frequency questioning regions, forming a set of high-frequency questioning regions.

[0053] Specifically, the semantic distortion accumulation path is traced for each high-frequency questioning region in the set of high-frequency questioning regions. The tracing process starts from the bridging node and proceeds downstream along the propagation chain segment by segment, comparing the semantic differences between the rewritten content and the source text within each propagation segment. The degree of semantic difference is determined by comparing the pragmatic premises of the propagated content with those of the source text. If the pragmatic premises in the propagated content are completely different from those in the source text, the semantic difference of that segment is high; if the pragmatic premises in the propagated content are partially consistent with those in the source text, the semantic difference of that segment is low. The semantic difference of each segment is accumulated segment by segment according to the propagation order, forming the semantic distortion accumulation path from the bridging node to the downstream node.

[0054] In one possible implementation, the semantic distortion accumulation path exhibits an increasing characteristic, meaning that the further the propagation segment is from the bridging node, the higher its cumulative distortion value.

[0055] It should be noted that the pragmatic tolerance boundary is set based on the tolerance threshold and tolerance gap type pre-established for the target cultural sphere. The tolerance threshold is the maximum acceptable upper limit for the degree of semantic deviation within that sphere, and the tolerance gap type is the rejection indicator for specific categories of pragmatic premise substitution within that sphere. When comparing the pragmatic tolerance boundary with the semantic distortion accumulation path, the distortion value of each segment of the accumulation path is compared with the tolerance threshold segment by segment under the same coordinate system.

[0056] In one embodiment, if the cumulative distortion value of a segment in the semantic distortion accumulation path is level four and the tolerance threshold is level three, then the segment exceeds the pragmatic tolerance boundary by one level and is marked as an over-limit path segment. If the pragmatic premise substitution category involved in a segment happens to belong to the tolerance gap type, the segment is also marked as an over-limit path segment even if its cumulative distortion value does not exceed the tolerance threshold. By traversing all segments in the semantic distortion accumulation path, a set of over-limit path segments is formed. Further, the assessment of the propagation range of the break effect is carried out for each over-limit path segment in the set of over-limit path segments. The assessment process counts two indicators downstream along the propagation chain: one is the number of downstream nodes affected by the over-limit path segment, that is, how many downstream nodes receive the distorted content of the segment; the other is the coverage of the propagation layers, that is, how many layers of the propagation chain depth the distorted content of the segment traverses when it is transmitted downstream.

[0057] Understandably, the number of downstream nodes reflects the breadth of the horizontal influence of the over-limit path segment, while the coverage of the propagation layers reflects its vertical penetration depth. By summarizing the number of downstream nodes and the coverage of the propagation layers for all over-limit path segments, the actual propagation range of the fracture effect is determined. This propagation range identifies the overall area in the cross-cultural discourse propagation chain where a propagation fracture may occur due to the accumulation of semantic distortion exceeding the pragmatic tolerance boundary.

[0058] Extract the tolerance thresholds and tolerance gap types of each target cultural sphere from the established pragmatic tolerance boundaries. Compare the distortion amplitude and distortion nature of each segment in the semantic distortion accumulation path to identify the distortion segments that exceed the tolerance thresholds and their corresponding path nodes. Evaluate the diffusion coverage and distortion transmission level of each over-limit segment to downstream nodes to determine the actual propagation range of the fracture effect.

[0059] The tolerance threshold and tolerance gap type of each target cultural sphere are extracted from the pre-defined pragmatic tolerance boundaries. The tolerance threshold is the upper limit of semantic deviation acceptance for that sphere, and the tolerance gap type is the rejection indicator of a specific category of pragmatic premise substitution for that sphere. This yields a tolerance feature record for each target cultural sphere. Based on the tolerance feature record, the distortion magnitude and distortion nature of each propagation segment in the semantic distortion accumulation path are read. The distortion magnitude is the degree of deviation between the semantics of the source text and the semantics of the rewritten content within that segment, and the distortion nature is the category of pragmatic premise substitution involved in that segment. If the distortion magnitude of a segment exceeds the tolerance threshold of the corresponding sphere, or if the distortion nature of the segment belongs to the tolerance gap type of the corresponding sphere, then the segment is marked as an over-limit segment, and the position of its corresponding path node is recorded, resulting in an association set of over-limit segments and path nodes. For each overlimited segment in the associated set, the speech propagation chain is traversed downstream. The number of downstream nodes that receive the distorted content of the overlimited segment is counted as the diffusion coverage breadth. The number of propagation chain layers that the distorted content traverses when it is transmitted downstream is counted as the distortion transmission level. The actual propagation range of the break effect is determined based on the diffusion coverage breadth and the distortion transmission level.

[0060] In the process of cross-cultural discourse communication, the pragmatic tolerance boundary of the target cultural sphere determines the limit of its acceptance of rewritten content from foreign discourse. The tolerance threshold refers to the upper limit of acceptance set by the target cultural sphere for the degree of semantic deviation. When the semantic deviation between the rewritten content and the source text exceeds this upper limit, the communication nodes within the target cultural sphere tend to question or interrupt the communication.

[0061] For example, Western cultural circles typically have a lower tolerance threshold for discourses on attribution of responsibility. Collective attribution refers to assigning responsibility to a group, while individual attribution refers to assigning responsibility to an individual. Pragmatic premises refer to the implicit assumptions underlying the discourse. When the pragmatic premise of collective attribution in the source text is rewritten as individual attribution, the text is first input into the BERT model to identify the premises, outputting premise vectors A and B. The magnitude of the rewriting is calculated using the cosine similarity formula sim(A,B)=A·B / (|A||B|), where A and B are vectors. If the similarity is below 0.8, it exceeds the acceptance limit of that cultural circle, and downstream nodes will distrust the rewritten content.

[0062] Specifically, the tolerance gap type identifies the natural rejection of specific categories of pragmatic premise replacement by the target cultural circle. The identification method is to statistically analyze historical communication data for categories with a rejection rate greater than 80% after replacement, and the judgment rule is that if the replacement involves this category, it is marked as a gap.

[0063] In one embodiment, a target cultural circle exhibits a tolerance gap for pragmatic premise substitutions involving apology methods. Regardless of the magnitude of the substitution, any substitution operation involving this category will trigger a backtracking verification behavior at the circle's nodes. This behavior involves the node tracing back from the current content to the source text, including extracting the source semantic vector, calculating the similarity, and outputting the verification result. If the similarity is less than 0.9, propagation is rejected. Based on the tolerance feature record, i.e., the preset rejection rate database, each segment of the semantic distortion accumulation path, i.e., each link in the propagation chain from the source to the target, is compared one by one. The distortion magnitude is determined by the formula D equals one minus the cosine similarity between the source text semantic vector S and the rewritten content semantic vector R, where S and R are extracted using the BERT model. A D greater than 0.2 is considered to exceed the tolerance threshold. The distortion nature is determined based on the pragmatic premise substitution category involved in the segment. The category classification standard is divided into three categories according to pragmatic function: politeness, emotion, and instruction. The determination process involves inputting the replacement text into the classifier model and outputting the category. If the distortion amplitude of a certain segment exceeds the tolerance threshold, or the distortion nature belongs to the tolerance gap type, then the segment is marked as an over-limit segment.

[0064] It should be noted that the identification of out-of-limit sections is accompanied by the recording of the positions of the path nodes to which they belong. The position of the path node is defined as the node ID and coordinates in the propagation chain graph. For example, if the node ID is N1 and the coordinates are (0,0), a hash table data structure is used to form an association set of out-of-limit sections and path nodes. The recording process involves mapping the starting node ID of each out-of-limit section to the section identifier when traversing the propagation chain. Furthermore, for each out-of-limit section in the association set, the BFS algorithm is used to traverse downstream along the propagation chain. The input is the propagation chain graph and the starting node, and the output is the number of downstream nodes as the diffusion coverage breadth, for example, a breadth of 10. The maximum path depth traversed when the distorted content is transmitted downstream is used as the distortion propagation level, for example, a level of 5. The two are combined to determine the actual propagation range of the fracture effect. The weight calculation formula is W=B×0.6+L×0.4, where B is the breadth and L is the level.

[0065] Step S105: Construct a distortion propagation risk map based on the propagation range of the fracture effect, analyze the distortion propagation risk map, identify segmental diffusion transformation nodes, locate the starting point of potential blocking paths, and establish a set of transformation nodes.

[0066] Based on the actual propagation range of the fracture effect and the topological structure of the discourse propagation chain, each over-limit segment, its diffusion coverage breadth, and distortion transmission level are marked at the corresponding position in the propagation chain. An edge set is constructed based on the propagation connection relationship between nodes, and the product of the diffusion coverage breadth and distortion transmission level of each node is used as the node weight to obtain a distortion propagation risk map. For each node in the distortion propagation risk map, based on the number of branches and the length of the distribution path after the node receives upstream distorted content, if the number of branches exceeds a preset branch threshold and the length of the distribution path exceeds a preset path threshold, the node is marked as a diffusion conversion node, and a diffusion conversion node set is obtained. For each diffusion conversion node in the diffusion conversion node set, the propagation chain is traced upstream to the nearest over-limit segment, and the path node position where the over-limit segment is located is marked as the potential blocking path start point, obtaining a blocking path start point record. Based on the blocking path start point record and the diffusion conversion node set, each diffusion conversion node is associated and paired with its corresponding blocking path start point, and arranged according to the position order of the diffusion conversion nodes in the propagation chain to establish a conversion node set containing the node position, blocking start point, and its propagation path.

[0067] In the process of cross-cultural discourse communication, the construction of a distortion communication risk graph is based on the actual scope of the discontinuity effect and the topological structure of the discourse communication chain. This topological structure uses each communication node as a vertex and the information transmission relationships between nodes as edges, forming a directed network. The specific construction process involves inputting discourse communication log data, including node identifiers and transmission records. First, the vertex set V is extracted as all unique communication nodes. Second, the edge set E is constructed as ordered pairs from the source node to the target node. Finally, the edge weight w is calculated. ij =log(1+f ij ), where w ij Let f be the edge weight from node i to j. ij Let i be the information transmission frequency from i to j. The frequency is obtained from log statistics. The final output is a weighted directed network graph.

[0068] Specifically, the node weights are determined based on the product of the diffusion coverage breadth and the distortion propagation level.

[0069] For example, if the diffusion coverage of a certain over-limit section is five downstream nodes and the distortion propagation level is three layers deep, then the weight value of that node is fifteen. The higher the weight value, the wider the influence range and the greater the penetration depth of that node in the distortion propagation process. By labeling all relevant nodes in the propagation chain with weights and establishing edge set relationships, a complete distortion propagation risk map is formed.

[0070] In a possible implementation, the identification of diffusion transformation nodes is carried out for each node in the distortion propagation risk map. The diffusion transformation nodes refer to the nodes that serve as the hub for distributing distorted content in the propagation chain. Such nodes receive the distorted content transmitted from the upstream and forward it to multiple downstream branches, featuring the characteristic of amplifying the scope of distortion propagation. Different from candidate blocking nodes, the latter focus on potential blocking value rather than the already occurred amplification. The identification process is determined based on two indicators: one is the number of branches distributed by the node to the downstream, and the other is the length of the propagation chain covered by the distribution path. Calculate the influence score IF = the number of branches × the path length. If IF > 10, it is marked as a diffusion transformation node. If 5 < IF ≤ 10, it is marked as a candidate blocking node, where the number of branches refers to the number of downstream branches, and the path length refers to the length of the longest downstream chain.

[0071] It should be noted that the preset branch threshold and the preset path threshold are set through the following process. First, input the propagation network graph G of the target cultural circle, including the number of nodes N and the number of edges E. Calculate the network density D = E / (N × (N - 1) / 2). Then, the branch threshold B = 0.5 × D + 0.1, and the path threshold P = 2 × D + 1. Different circles are set according to the difference in D values.

[0072] For example, for a circle with D = 0.3, B = 0.25, and P = 1.6; for a circle with D = 0.6, B = 0.4, and P = 2.2. Further, for each diffusion transformation node in the set of diffusion transformation nodes, locate the starting point of the blocking path. The location process traces back upstream along the propagation chain to trace the source path of the distorted content received by the diffusion transformation node until reaching the location of the nearest over-limit section. The nearest over-limit section refers to the over-limit section with the fewest number of nodes passed between it and the diffusion transformation node on the propagation chain. Mark the path node location where the over-limit section is located as the potential starting point of the blocking path, and this starting point is the upstream cut-in position for implementing propagation intervention for this diffusion transformation node.

[0073] In one embodiment, a certain diffusion transformation node is located on the seventh layer of the propagation chain. When tracing back upstream, it passes through the nodes on the sixth layer and the fifth layer in sequence. When an over-limit section mark is found at the position of the node on the fourth layer, the node on the fourth layer is recorded as the potential starting point of the blocking path corresponding to this diffusion transformation node.

[0074] Understandably, locating the starting point of the blocking path aims to identify the last controllable position before distorted content enters the diffusion and conversion nodes, facilitating targeted intervention in the propagation path at that location. Based on the record of the blocking path starting point and the set of diffusion and conversion nodes, the conversion node set is constructed by associating and pairing the two. For each diffusion and conversion node, its position information in the propagation chain, the corresponding blocking path starting point position, and the propagation path identifier to which the node belongs are integrated and recorded. All diffusion and conversion nodes are arranged according to their position order in the propagation chain, forming an ordered set of conversion nodes. This set visually presents the correspondence between each diffusion and conversion node and its upstream blocking starting point, as well as its distribution characteristics within the overall propagation network.

[0075] The distortion degree and preconditions for triggering propagation of each segment's diffusion transformation node are retrieved from the distorted propagation risk map. The locations of nodes with blocking potential and their upstream and downstream connection breaks in the propagation chain are identified. The ability of each candidate blocking node to cut off the segment's diffusion chain and its actual intervention feasibility are analyzed. A set of transformation nodes including node location, blocking ability, and the path to which they belong is established.

[0076] The distortion degree and preconditions for triggering propagation of each segment's diffusion conversion node are retrieved from the distortion propagation risk map. The distortion degree is the semantic deviation between the distorted content received and forwarded by the node and the source text. The preconditions are the number of paths and frequency of receiving distorted content from upstream. Propagation triggering characteristic records for each diffusion conversion node are obtained. For each diffusion conversion node in the propagation triggering characteristic record, its connection position with upstream and downstream nodes in the propagation chain is identified. If the connection between the node and the upstream node is a single path and there are multiple branch connections with downstream nodes, the node is marked as a candidate blocking node, and its upstream and downstream connection breakpoints are recorded to obtain a set of candidate blocking nodes. For each candidate blocking node in the candidate blocking node set, its cutting capability is assessed based on the number of downstream branches that lose their source of distorted content after the node is cut off. Its intervention feasibility is assessed based on the number of path layers from the bridging node in the propagation chain. The cutting capability and intervention feasibility are integrated into the node's attribute record to establish a set of conversion nodes including node location, blocking capability, and the path it belongs to.

[0077] In the process of cross-cultural discourse dissemination, extracting the characteristic information of diffusion and transformation nodes from the distortion dissemination risk map is the basis for identifying candidate blocking nodes. The degree of distortion refers to the semantic deviation between the distorted content received and forwarded by the node and the source text. The larger the deviation value, the more significant the difference between the content disseminated by the node and the original discourse.

[0078] Specifically, candidate blocking node identification is based on connection structure features and a comprehensive score for each node is calculated in conjunction with preconditions. First, the PageRank algorithm is used to obtain a structure score S. Then, T = 0.6 × N + 0.4 × F is calculated, where N is the number of paths through which the node receives distorted content from upstream, and F is the reception frequency. If T is greater than 3, the node is selected as a candidate blocking node.

[0079] In one embodiment, candidate blocking nodes are identified based on the connection structure characteristics of the nodes in the propagation chain. If a diffusion conversion node has only a single connection path with an upstream node, but multiple branch connections exist between the node and downstream nodes, then the node is marked as a candidate blocking node. Such nodes are characterized by a single entry point for receiving distorted content, but dispersed exit points for distributing distorted content. Cutting off the single connection between the node and the upstream node can block the sources of distorted content from multiple downstream branches.

[0080] It should be noted that recording the locations of the upstream and downstream connection breakpoints provides coordinate data for locating subsequent blocking paths. Furthermore, the assessment of the cutting capability is determined based on the amount of distorted content lost from downstream branches after the candidate blocking node is cut off; the calculation formula is C=∑(D i ), where C is the cutting capability value, D i To determine the number of sources lost by the i-th downstream branch, count D by traversing the downstream branch tree structure. i For example, if node A is cut off, downstream branch 1 loses 2 sources and downstream branch 2 loses 3 sources, then C=5. The larger the C value, the stronger the cutting-off capability of the node. The assessment of intervention feasibility is determined based on the minimum path layer from the node to the bridging node in the propagation chain. The calculation process uses Dijkstra's algorithm. The input is the propagation chain graph structure and the set of bridging nodes. The output is the shortest path layer L from each candidate node to the nearest bridging node. For example, the path layer L from node B to the nearest bridging node is 2. The smaller the L value, the closer the intervention position is to the source of the propagation chain, and the higher the intervention feasibility. The cutting-off capability and intervention feasibility are integrated into the attribute records of each candidate blocking node, and arranged according to the position of the node in the propagation chain to form a set of transformation nodes that includes the node position, blocking capability, and the path to which it belongs.

[0081] Step S106: Based on the set of transformation nodes, trace the rewriting depth of the upstream bridging nodes, mark the fracture propagation path and isolate high-risk propagation chains, deploy pragmatic adaptation feedback units to send tolerance boundary signals back to the bridging nodes, and generate an optimized rewriting depth vector containing backtracking verification suppression markers.

[0082] Based on the set of transformation nodes, the bridging node positions corresponding to each transformation node are traced upstream along the propagation chain. The rewriting depth is extracted from the rewriting records of each bridging node when it performs a substitution operation on the pragmatic premises of the source text. Simultaneously, the propagation path from the bridging node to each transformation node is marked as a break-through path. If the number of downstream nodes covered by the break-through path exceeds a preset risk threshold, the propagation chain to which the path belongs is marked as a high-risk propagation chain, and its downstream propagation connection is interrupted, resulting in an isolated set of high-risk propagation chains. For each bridging node in the set of high-risk propagation chains, the tolerance threshold and tolerance gap type of the target cultural sphere corresponding to the bridging node are read. The tolerance threshold and tolerance gap type are encapsulated into a tolerance boundary signal and transmitted back to the bridging node, obtaining the tolerance boundary reception record of the bridging node. Based on the tolerance boundary reception record, the rewrite depth of each bridging node is compared with the received tolerance boundary signal. If the rewrite depth exceeds the tolerance threshold, a backtracking verification suppression flag is added at the rewrite depth value position. The rewrite depth values ​​of each bridging node and the suppression flag are integrated into a vector form to generate an optimized rewrite depth vector containing the backtracking verification suppression flag.

[0083] In the process of cross-cultural discourse dissemination, tracing upstream to bridging nodes based on the set of transformation nodes is a key step in achieving targeted intervention in the dissemination path. The tracing process traverses backward along the dissemination chain, starting from the position of each transformation node, passing through intermediate dissemination nodes in sequence, until reaching the bridging node that performs the source text rewriting operation.

[0084] For example, when a certain transformation node is located at the sixth layer of the propagation chain, tracing upstream may involve passing through the fifth and fourth layers of nodes in sequence, and finally locating the bridging node at the second layer.

[0085] Specifically, the rewriting depth is determined by extracting the hierarchical span from the rewriting records of bridging nodes. The bridging node rewriting records refer to the logs of pragmatic premise substitution operations in the source text, including the original cultural affiliation level and the target adaptation level for each node. The extraction process is as follows: input the source text log, traverse the records using the analytic hierarchy process (AHP), calculate the hierarchical span D = |PT|, where P is the original hierarchical value (e.g., 3), T is the target hierarchical value (e.g., 5), and the average value of D (e.g., 2) is output, which is the numerical expression of the rewriting depth.

[0086] In one embodiment, a broken propagation path refers to a propagation path generated from a bridging node using a depth-first search algorithm, and a transformation node refers to a node in the path where information has been mutated. The path generation process takes bridging nodes and a network graph as input and outputs a list of covered downstream nodes; the number of downstream nodes is calculated by counting the terminal nodes of the path. If the number of downstream nodes covered by a broken propagation path exceeds a preset risk threshold of 10, the propagation chain to which this path belongs is marked as a high-risk propagation chain, and propagation is interrupted by disconnecting downstream connections, forming an isolated set of high-risk propagation chains. This isolation operation aims to prevent distorted content from continuing to spread downstream.

[0087] It should be noted that the feedback of the tolerance boundary signal is performed on each bridging node in the high-risk propagation chain set. The tolerance boundary signal is generated by the target cultural sphere, which refers to a specific social group such as an Asian youth community. The tolerance threshold T is calculated by averaging historical data, e.g., T=0.5 represents the upper limit of acceptance. The tolerance gap type G is divided into two categories: high G=1 and low G=0. The signal generation process inputs group feedback data and outputs it in JSON format, e.g., {t:0.5, g:1}. The feedback mechanism uses a RESTful API call. After receiving the signal, the bridging node forms a tolerance boundary reception record, including key-value pair details of the timestamp, threshold T, and type G. Furthermore, the construction of the optimized rewrite depth vector is based on comparing the tolerance boundary reception record with the rewrite depth D of each bridging node, and the difference is calculated using the Euclidean distance algorithm. The vector dimension is defined as the number of bridging nodes N, e.g., N=5. For each bridging node, if its rewrite depth D exceeds the tolerance threshold T, e.g., D>0.5, a backtracking verification suppression flag M=1 is generated; otherwise, M=0. This flag indicates that the bridging node should reduce the rewrite magnitude of the corresponding dimension in subsequent rewrite operations, e.g., multiply the magnitude by 0.8. The rewrite depth values ​​D of each bridging node and the suppression flag M are integrated into a vector form V=[D1,M1;D2,M2;……;DN,MN], for example, when N=2, V=[0.6,1;0.4,0], forming an optimized rewrite depth vector containing the backtracking verification suppression flag.

[0088] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, the personal information processing rules are clearly informed through signs / information, and authorization is obtained through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0089] The above are only some preferred embodiments of the present invention, but the present invention is not limited thereto, and many improvements and modifications can be made. Any improvements and modifications made based on the basic principles of the present invention should be considered to fall within the protection scope of the present invention.

Claims

1. A dynamic analysis method for cross-cultural discourse networks, characterized in that, The method includes: obtaining source culture metaphor expression adaptation and rewriting data of bridging nodes through a discourse network analysis platform, and identifying replacement identifier sequences of nested pragmatic premises; Based on the substitution identifier sequence of nested pragmatic premises, the source text pragmatic premises and the substitution pragmatic premises are separated, the depth level of the source text pragmatic premises and the substitution pragmatic premises is determined, and a substitution level matrix is ​​generated. The downstream node backtracking verification behavior logs are extracted from the stripping and replacement hierarchy matrix. The verification frequency and tracing depth are statistically analyzed by the pragmatic premises of the source text and the pragmatic premises of the replacement text, and a verification intensity distribution map is constructed. Identify high-frequency questioning regions in the verification intensity distribution map, analyze the semantic distortion accumulation path from bridging nodes to downstream nodes in the high-frequency questioning regions, set the pragmatic tolerance boundary of the target cultural circle, and evaluate the propagation range of the fracture effect by combining the semantic distortion accumulation path and the pragmatic tolerance boundary of the target cultural circle. Based on the propagation range of the fracture effect, a distortion propagation risk map is constructed. The distortion propagation risk map is analyzed to identify segmental diffusion transformation nodes, locate the starting point of potential blocking paths, and establish a set of transformation nodes. Based on the set of transformation nodes, the rewriting depth of upstream bridging nodes is traced, the fracture propagation path is marked and high-risk propagation chains are isolated, pragmatic adaptation feedback units are deployed to send tolerance boundary signals back to the bridging nodes, and an optimized rewriting depth vector containing backtracking verification suppression markers is generated.

2. The method according to claim 1, characterized in that, The process of obtaining source cultural metaphorical expression adaptation and rewriting data of bridging nodes through a discourse network analysis platform, and identifying replacement identifier sequences of nested pragmatic premises, includes: The source cultural metaphor expression adaptation and rewriting data of bridging nodes are obtained through a discourse network analysis platform. The comparison records of source text metaphor fragments and target text adaptation fragments are extracted from the rewriting data. The comparison records contain the pragmatic premises of the source text and their corresponding target adaptation content. The nesting level of each pragmatic premise is marked according to the nesting and inclusion relationship of the pragmatic premises of the source text in the comparison records, and a distribution map of pragmatic premises containing nesting levels is obtained. For each nested level in the pragmatic premise distribution map, the word differences between the pragmatic premise of the source text and the target adapted content are compared one by one. If there is a word in the target adapted content that is semantically inconsistent with the pragmatic premise of the source text, the position of the word is marked as a replacement site. The replacement sites are arranged according to the order of their appearance in the discourse propagation chain to obtain the replacement identifier sequence of the nested pragmatic premises.

3. The method according to claim 1, characterized in that, The process of separating the source text pragmatic premises and the alternative pragmatic premises based on the substitution identifier sequence of nested pragmatic premises, determining the depth level of the source text pragmatic premises and the alternative pragmatic premises, and generating a stripping substitution level matrix includes: Based on the substitution identifier sequence of nested pragmatic premises, the source text pragmatic premise to be replaced and the replaced pragmatic premise are extracted from each substitution point. Pairing records of source text pragmatic premises and replaced pragmatic premises are established according to the correspondence of substitution points to obtain a premise pairing set. For each pairing record in the premise pairing set, the nesting position of the pragmatic premise of the source text in the original discourse structure is traced, and the number of other pragmatic premises wrapped by the pragmatic premise of the source text is counted as the source text depth level. At the same time, the nesting position of the alternative pragmatic premise in the target adapted discourse structure is traced, and the number of other adapted contents wrapped by the alternative pragmatic premise is counted as the replacement depth level, so as to obtain the bidirectional depth level annotation of each pairing record. The numerical difference between the source text depth level and the replacement depth level is extracted from the bidirectional depth level annotation. If the source text depth level is higher than the replacement depth level, the paired record is marked as a deep stripping type. If the source text depth level is lower than or equal to the replacement depth level, the paired record is marked as a shallow replacement type, thus forming a level difference record containing the stripping type marker. Based on the hierarchical difference records, a two-dimensional matrix is ​​constructed with the source text depth level as the row index and the replacement depth level as the column index. The span of each paired record is filled into the corresponding row and column intersection position to form a stripping replacement level matrix covering all replacement nodes.

4. The method according to claim 3, characterized in that, The generation of the stripping and replacement hierarchy matrix also includes: From the separated source text pragmatic premises and alternative pragmatic premises, extract the source text depth level and target adaptation level corresponding to each alternative node. The source text depth level is the nesting depth value of the source text pragmatic premise in the source cultural discourse system, and the target adaptation level is the nesting depth value of the alternative pragmatic premise in the target cultural discourse system, so as to obtain the bidirectional hierarchical annotation record of each alternative node. For each replacement node in the bidirectional hierarchical annotation record, the cross-layer rewriting direction is determined by comparing the numerical values ​​of the source text depth level and the target adaptation level. If the source text depth level is higher than the target adaptation level, the rewriting direction is to transform to a shallower level; if the source text depth level is lower than the target adaptation level, the rewriting direction is to transform to a deeper level. At the same time, the difference between the two level values ​​is used as the level crossing range, resulting in a node feature set that includes the rewriting direction and the crossing range. Based on the source text depth level and target adaptation level of all replacement nodes in the node feature set, a two-dimensional matrix is ​​constructed with the source text depth level as the row dimension and the target adaptation level as the column dimension. The span of each replacement node is filled into the corresponding row and column intersection position. The position where the span exceeds the preset normal adaptation threshold is marked as an abnormal cross-layer region, forming a stripping replacement level matrix covering all replacement nodes.

5. The method according to claim 1, characterized in that, The process of extracting downstream node backtracking verification behavior logs from the stripping and replacement hierarchy matrix, and constructing a verification intensity distribution map by statistically analyzing the verification frequency and tracing depth using the pragmatic premises of the source text and the replaced pragmatic premises, includes: Extract the downstream node backtracking verification behavior logs corresponding to each replacement node from the stripping and replacement hierarchy matrix. The behavior logs record the operation records of the downstream node initiating a query to the source text for each replacement node and comparing it with the rewritten content, thereby obtaining a set of backtracking verification records associated with each row and column position in the matrix. For each record in the backtracking verification record set, the number of queries initiated by downstream nodes is counted based on the pairing relationship between the pragmatic premises of the source text and the alternative pragmatic premises, which is used as the verification frequency. At the same time, the number of propagation chain nodes traversed by the downstream node when backtracking from the current position to the source text is counted as the tracing depth, thus obtaining the node verification feature containing the verification frequency and the tracing depth. Based on the node verification characteristics, using the row and column positions of the stripping and replacement hierarchy matrix as coordinate references, the verification frequency and tracing depth of each replacement node are mapped to the corresponding coordinate positions. The verification frequency and tracing depth are multiplied to obtain the verification intensity value, which is then filled into each coordinate position to construct a verification intensity distribution map.

6. The method according to claim 1, characterized in that, The high-frequency questioning regions in the identification and verification intensity distribution map are analyzed. The semantic distortion accumulation path from bridging nodes to downstream nodes within these high-frequency questioning regions is analyzed. A pragmatic tolerance boundary for the target cultural sphere is defined. The propagation range of the disruption effect is assessed by combining the semantic distortion accumulation path and the pragmatic tolerance boundary of the target cultural sphere, including: High-frequency questioning areas are identified from the verification intensity distribution map. The verification intensity value at each coordinate position is compared with a preset questioning frequency threshold. If the verification intensity value at a certain coordinate position exceeds the questioning frequency threshold, the coordinate position and its adjacent consecutive positions exceeding the threshold are defined as high-frequency questioning areas, thus obtaining a set of high-frequency questioning areas. For each high-frequency questioning region in the set of high-frequency questioning regions, locate the bridging node and downstream node corresponding to the region, track the changes in semantic content segment by segment along the propagation chain from the bridging node to the downstream node, and accumulate the semantic distortion accumulation path from the bridging node to the downstream node according to the degree of semantic difference between the rewritten content and the source text in each propagation segment. Based on the tolerance threshold and tolerance gap type pre-established in the target cultural circle, a pragmatic tolerance boundary is set. The pragmatic tolerance boundary is superimposed and compared with the semantic distortion accumulation path. If the cumulative distortion value of a certain segment in the semantic distortion accumulation path exceeds the tolerance threshold of the pragmatic tolerance boundary, the segment is marked as an over-limit path segment, and a set of over-limit path segments is obtained. For each of the out-of-limit path segments in the set of out-of-limit path segments, the number of downstream nodes affected by the segment and the coverage of the propagation layers are counted along the propagation chain downstream to assess the actual propagation range of the fracture effect.

7. The method according to claim 6, characterized in that, The assessment of the propagation range of the fracture effect also includes: Extract the tolerance threshold and tolerance gap type of each target cultural circle from the established pragmatic tolerance boundary. The tolerance threshold is the upper limit of semantic offset acceptance set in advance for the circle, and the tolerance gap type is the rejection mark of the circle for specific category pragmatic premise substitution, so as to obtain the tolerance feature record of each target cultural circle. Based on the tolerance feature record, the distortion magnitude and distortion nature of each propagation segment in the semantic distortion accumulation path are read. The distortion magnitude is the degree of deviation between the semantics of the source text and the semantics of the rewritten content within the segment, and the distortion nature is the pragmatic premise substitution category involved in the segment. If the distortion magnitude of a segment exceeds the tolerance threshold of the corresponding layer, or the distortion nature of the segment belongs to the tolerance gap type of the corresponding layer, then the segment is marked as an over-limit segment and its path node position is recorded to obtain the association set of over-limit segments and path nodes. For each overlimited segment in the associated set, the speech propagation chain is traversed downstream. The number of downstream nodes that receive the distorted content of the overlimited segment is counted as the diffusion coverage breadth. The number of propagation chain layers that the distorted content traverses when it is transmitted downstream is counted as the distortion transmission level. The actual propagation range of the break effect is determined.

8. The method according to claim 1, characterized in that, The process involves constructing a distorted propagation risk map based on the propagation range of the fracture effect, analyzing the distorted propagation risk map, identifying segmental diffusion transformation nodes, locating the starting points of potential blocking paths, and establishing a set of transformation nodes, including: Based on the actual propagation range of the fracture effect, and using the topological structure of the discourse propagation chain as a basis, each over-limit segment and its diffusion coverage and distortion transmission level are marked at the corresponding position of the propagation chain. An edge set is constructed based on the propagation connection relationship between nodes, and the product of the diffusion coverage and distortion transmission level of each node is used as the node weight to obtain the distortion propagation risk map. For each node in the distortion propagation risk graph, based on the number of branches and the length of the distribution path that the node distributes downstream after receiving upstream distorted content, if the number of branches exceeds a preset branch threshold and the length of the distribution path exceeds a preset path threshold, then the node is marked as a diffusion conversion node, and a set of diffusion conversion nodes is obtained. For each diffusion conversion node in the set of diffusion conversion nodes, backtrack upstream along the propagation chain to the nearest overlimited section, mark the path node where the overlimited section is located as the potential blocking path start point, and obtain the blocking path start point record; Based on the record of the starting point of the blocking path and the set of diffusion and transformation nodes, each diffusion and transformation node is associated and paired with its corresponding starting point of the blocking path. The nodes are arranged in order of their position in the propagation chain to establish a set of transformation nodes that includes the node position, the starting point of the blocking path and the propagation path to which they belong.

9. The method according to claim 8, characterized in that, The establishment of the conversion node set also includes: The distortion degree and the preconditions for triggering propagation of each segment of the diffusion and transformation node are retrieved from the distortion propagation risk map. The distortion degree is the semantic deviation value between the distorted content received and forwarded by the node and the source text. The preconditions are the number of paths and the frequency of receiving distorted content from upstream by the node. The propagation triggering characteristic record of each diffusion and transformation node is obtained. For each diffusion transformation node in the propagation trigger feature record, identify the connection position of the node with the upstream node and the connection position with the downstream node in the propagation chain. If the connection between the node and the upstream node is a single path and there are multiple branch connections with the downstream node, then mark the node as a candidate blocking node and record the position of its upstream and downstream connection breakpoints to obtain a set of candidate blocking nodes. For each candidate blocking node in the candidate blocking node set, its blocking capability is evaluated based on the number of distorted content sources lost by the downstream branches after the node is cut off, and its intervention feasibility is evaluated based on the number of path layers of the node from the bridging node in the propagation chain. The blocking capability and intervention feasibility are integrated into the attribute record of the node to establish a set of conversion nodes that includes the node location, blocking capability and the path to which it belongs.

10. The method according to claim 1, characterized in that, The process of tracing the rewriting depth of upstream bridging nodes based on the set of transformation nodes, marking the fracture propagation path and isolating high-risk propagation chains, deploying pragmatic adaptation feedback units to send tolerance boundary signals back to the bridging nodes, and generating an optimized rewriting depth vector containing backtracking verification suppression markers includes: Based on the set of transformation nodes, the bridging node positions corresponding to each transformation node are traced upstream along the propagation chain. The hierarchical span when the bridging node performs a replacement operation on the pragmatic premise of the source text is extracted from the rewriting record of the bridging node as the rewriting depth. At the same time, the propagation path from the bridging node to each transformation node is marked as a break diffusion path. If the number of downstream nodes covered by the break diffusion path exceeds a preset risk threshold, the propagation chain to which the path belongs is marked as a high-risk propagation chain and its downstream propagation connection is interrupted, resulting in a set of isolated high-risk propagation chains. Read the tolerance threshold and tolerance gap type of the target cultural circle corresponding to the bridging node in the high-risk propagation chain set, encapsulate the tolerance threshold and tolerance gap type into a tolerance boundary signal and send it back to the bridging node to obtain the tolerance boundary reception record of the bridging node. Based on the tolerance boundary reception record, the rewrite depth of each bridging node is compared with the received tolerance boundary signal. If the rewrite depth exceeds the tolerance threshold, a backtracking verification suppression flag is added at the rewrite depth value position. The rewrite depth values ​​of each bridging node and the suppression flag are integrated into a vector form to generate an optimized rewrite depth vector containing the backtracking verification suppression flag.

Images