Highway standard relationship construction, query method, equipment, media and program products

Abstract

The present invention discloses a highway standard relationship construction and query method, device, medium, and program product. The method includes: performing multidimensional semantic analysis on multimodal highway standard data to obtain multidimensional feature vectors corresponding to highway standard items, calculating multidimensional edit distances between items, obtaining a similarity matrix between highway standard item sequences, then constructing a dynamic association network based on the similarity matrix with each highway standard item as a node, clustering the nodes in the dynamic association network to obtain a cluster structure corresponding to the dynamic association network, and finally dynamically optimizing the dynamic association network and cluster structure based on the dynamic evolution characteristics corresponding to each node in the dynamic association network. The embodiment of the present invention realizes multidimensional semantic expression of multimodal highway standard data, constructs a dynamic association network through multidimensional edit distance, and realizes accurate association and dynamic update between highway standard items, thereby effectively improving highway standard data management efficiency and user query experience.

Description

Technical Field

[0001] The present invention relates to the technical field of highway data management, and in particular to highway standard relationship construction, query methods, equipment, media and program products. Background Art

[0002] Highway industry standards are an important foundation for ensuring industry standardization and promoting technological advancement. However, with the development of the highway industry, the scope of the standards system continues to expand, and the complexity of the standards content is increasing. Traditional standards management and query methods are no longer able to meet the increasingly diverse and complex needs of the industry.

[0003] The existing standard management and query methods have the following problems and defects:

[0004] 1. Existing standards management methods rely primarily on static storage and manual operations, lacking systematic, structured processing capabilities. Numerous standards exist in PDF or scanned document formats, and references between standards, version change information, and logical connections between items are not effectively explored and displayed. This requires users to read and manually compare each item to obtain the required information, which is not only time-consuming and labor-intensive, but also prone to overlooking key content or misinterpreting it.

[0005] 2. Existing standard query methods mostly rely on keyword matching and cannot support deep semantic search. This model is limited by the fact that when users' queries are complex, such as comparing differences between multiple standard entries or finding related content based on fuzzy descriptions, existing systems often struggle to provide accurate and relevant results. Furthermore, existing query systems are almost incapable of processing unstructured data such as images and formulas, significantly reducing query efficiency and user experience.

[0006] 3. The dynamic interdependence between standards is not effectively utilized. The highway industry's standards system is highly dynamic, with new standards frequently released and older standards abolished or revised. However, the existing system fails to establish a dynamic interdependence network between standards. For example, the changes between new and old versions of standards are not systematically presented, forcing users to manually organize them when applying them, which can easily affect actual use due to omissions or misunderstandings.

[0007] Therefore, it is urgent to invent a method for constructing and querying highway standard relationships to solve the problems in traditional highway industry standard management and query modes that are difficult to cope with the dynamic association of complex standard systems, deep retrieval of semantic levels, and unstructured data processing, so as to realize the dynamic construction and intelligent query of highway standard relationships, improve the efficiency of standard management and the convenience of user queries. Summary of the Invention

[0008] In view of this, embodiments of the present invention provide a highway standard relationship construction, query method, device, medium and program product, which at least partially solve the problems existing in the prior art.

[0009] Other features and advantages of the present invention will become apparent from the following detailed description, or may be learned in part by practice of the present invention.

[0010] In order to achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0011] According to a first aspect of an embodiment of the present invention, a method for establishing a highway standard relationship is provided, the method comprising:

[0012] Perform multidimensional semantic analysis on multimodal highway standard data to obtain multidimensional feature vectors corresponding to highway standard items;

[0013] Calculating the multidimensional edit distance between highway standard items based on the multidimensional feature vector to obtain a similarity matrix between highway standard item sequences;

[0014] Based on the similarity matrix between the highway standard item sequences, each highway standard item is used as a node, and the similarity between node pairs is used to establish the association edges between nodes to construct a dynamic association network;

[0015] Clustering the nodes in the dynamic association network to obtain a cluster structure corresponding to the dynamic association network;

[0016] According to the dynamic evolution characteristics of the highway standard items corresponding to each node in the dynamic association network, the dynamic association network and the cluster structure are dynamically optimized.

[0017] Furthermore, multi-dimensional semantic analysis is performed on the multimodal highway standard data to obtain multi-dimensional feature vectors corresponding to highway standard items, including:

[0018] Acquire original multimodal highway standard data and construct a multimodal highway standard data document set, wherein the multimodal highway standard data includes highway standard text, images, formulas, version information, and reference information;

[0019] Performing data structure initialization processing on the multimodal highway standard data document set to obtain a highway standard item set;

[0020] For each highway standard item in the highway standard item set;

[0021] Constructing a corresponding entry text set based on the text content in the highway standard entry, performing word segmentation processing on the entry text set based on a highway field-specific dictionary, and performing word weighting processing on the word segmentation results using a field word weight function to assign higher weights to highway field keywords to obtain a word weighted result;

[0022] Mapping the word weighted result to a vector space to obtain an entry text feature vector corresponding to the highway standard entry;

[0023] Constructing an entry image set based on the image and chart contents in the highway standard entry;

[0024] Vectorizing the images in the entry image set using a pre-trained convolutional neural network to obtain corresponding image feature vectors;

[0025] Use the image aggregation function to merge all the image feature vectors under the highway standard items to obtain the item image feature vector;

[0026] Constructing an entry formula set according to the formula contents in the highway standard entry;

[0027] Extracting symbols involved in the entry formula set and constructing a symbol set;

[0028] Establishing a symbol meaning mapping table according to the meaning of each symbol in the symbol set;

[0029] Integrating the relationship between the formula structure and the symbols into a high-dimensional vector space according to the symbol meaning mapping table and the symbol set to obtain an entry formula feature vector;

[0030] Performing vector concatenation processing using the entry text feature vector, the entry image feature vector, the entry formula feature vector, and version information and reference information corresponding to the highway standard entry to obtain a multidimensional feature vector corresponding to the highway standard entry;

[0031] A tag set is constructed for a highway standard entry according to highway standard specific attributes, wherein the highway standard specific attributes include road category, technical level, reference level, and version information.

[0032] Furthermore, the multidimensional edit distance between highway standard items is calculated based on the multidimensional feature vector to obtain a similarity matrix between highway standard item sequences, including:

[0033] generating a first highway standard item sequence and a second highway standard item sequence to be compared based on the highway standard item set;

[0034] If the first highway standard entry sequence lacks an entry to be compared corresponding to the second highway standard entry sequence, inserting the entry to be compared in the second highway standard entry sequence into a corresponding position in the first highway standard entry sequence based on a preset insertion cost function;

[0035] If the second highway standard item sequence lacks an item to be compared corresponding to the first highway standard item sequence, deleting the item to be compared without a corresponding item in the first highway standard item sequence based on a preset deletion cost function;

[0036] If there are corresponding first and second items to be compared between the first and second highway standard item sequences, then calculating, based on the multidimensional feature vectors corresponding to the first and second items to be compared, the text dimension distance, image dimension distance, formula dimension distance, and version and reference information dimension distance between the first and second items to be compared;

[0037] Obtaining an entry multidimensional edit distance based on the text dimension distance, the image dimension distance, the formula dimension distance, and the version and reference information dimension distance;

[0038] A multidimensional edit distance matrix D is generated based on the multidimensional edit distance between the first highway standard entry sequence and the second highway standard entry sequence. The elements in the multidimensional edit distance matrix D are: represents the minimum multidimensional edit distance between the first i entries in the first highway standard entry sequence and the first j entries in the second highway standard entry sequence;

[0039] The multidimensional edit distance in the multidimensional edit distance matrix is ​​converted into a similarity score within a preset range to obtain a similarity matrix between the first highway standard item sequence and the second highway standard item sequence.

[0040] Furthermore, based on the similarity matrix between the highway standard item sequences, each highway standard item is used as a node, and the similarity between node pairs is used to establish association edges between nodes to construct a dynamic association network, including:

[0041] Creating a graph structure, the graph structure including a node set and an edge set, the node set corresponding to all highway standard entries;

[0042] Based on the similarity matrix, determining whether the similarity score between the node pairs is greater than a preset similarity threshold;

[0043] If the similarity score between the node pairs is less than or equal to the preset similarity threshold, the weight of the first edge between the node pairs is zero, and no associated edge is established between the node pairs;

[0044] If the similarity score between the node pair is greater than the preset similarity threshold, obtaining a first edge weight corresponding to the node pair according to the similarity score between the node pair;

[0045] Determine whether there is version inheritance or mandatory reference between the node pairs;

[0046] If version inheritance or mandatory reference exists between the node pairs, a second edge weight is obtained according to the gravity coefficient and the first edge weight, wherein the gravity coefficient is adaptively adjusted according to the reference strength and the degree of difference between the homologous standard versions;

[0047] If there is no version inheritance or mandatory reference between the node pairs, the first edge weight is used as the second edge weight;

[0048] An associated edge is established for the node pair in the edge set, and the second edge weight is stored in a weight field of the associated edge, thereby finally obtaining a dynamic associated network with weights.

[0049] Furthermore, clustering the nodes in the dynamic association network to obtain a cluster structure corresponding to the dynamic association network includes:

[0050] All nodes in the dynamic association network are aggregated based on a hierarchical clustering algorithm to obtain at least one standard cluster, forming a cluster structure corresponding to the dynamic association network, and the average weight between member nodes in the standard cluster is greater than or equal to a preset aggregation threshold.

[0051] Furthermore, according to the dynamic evolution characteristics of the highway standard items corresponding to each node in the dynamic association network, the dynamic association network and the cluster structure are dynamically optimized, including:

[0052] Based on the dynamic association network, an optimization objective function is constructed, wherein the optimization objective function includes a graph smoothing term, an entropy regularization term, a multimodal coupling loss, and a cross-standard reference constraint term. The smoothing term is used to keep the embedding vectors of adjacent nodes similar in the multimodal space. The entropy regularization term includes the edge weight entropy within the standard cluster and an adjustable function for limiting the node state norm. The cross-standard reference constraint term is used to constrain the edge set of version inheritance or mandatory reference;

[0053] Introducing second-order or third-order entropy terms into the entropy regularization term amplifies the influence of high-confidence edges in the optimization process, allowing strong connections to be strengthened during iterations to improve the ability to express complex multimodal distributions;

[0054] Introducing cross-cluster mutual information metric between standard clusters;

[0055] Based on the optimization objective function, a global-local alternating optimization strategy is adopted to dynamically optimize the dynamic association network;

[0056] During the global optimization process, the edge weights of the associated edges remain fixed, and the nodes in the dynamic associated network are updated using variational inference. The node representations in the optimization objective function are solved based on the ADMM algorithm to update the node representations.

[0057] During the local optimization process, pulse-modify the cross-cluster association edges and the association edges with version inheritance or mandatory reference according to discrete events, wherein the discrete events include standard version releases;

[0058] Based on the optimized dynamic association network, the cluster structure is re-divided and adaptively merged to obtain an updated cluster structure.

[0059] According to a second aspect of an embodiment of the present invention, a method for querying highway standard relationships is provided, the method comprising:

[0060] Perform multi-level modeling based on query requirements;

[0061] According to the dynamic association network in the highway standard relationship construction method as described in any one of the above items, a corresponding multi-level index and fast retrieval structure is generated;

[0062] According to the query statement entered by the user, match the corresponding query node candidate set;

[0063] Outputting a query result according to the query node candidate set;

[0064] According to a third aspect of an embodiment of the present invention, there is provided a highway standard relationship management device, the device comprising: a processor and a memory;

[0065] The memory is used to store one or more program instructions;

[0066] The processor is used to run one or more program instructions to execute the steps of a highway standard relationship construction method or a highway standard relationship query method as described in any one of the above items.

[0067] According to a fourth aspect of an embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of a highway standard relationship construction method or a highway standard relationship query method as described in any one of the above items are implemented.

[0068] According to a fifth aspect of an embodiment of the present invention, a computer program product is provided, which includes a computer program stored on a non-transitory computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer implements the steps of a highway standard relationship construction method or a highway standard relationship query method as described in any one of the above items.

[0069] The embodiments of the present invention provide a method, device, medium, and program product for constructing and querying highway standard relationships. The method includes: performing multidimensional semantic analysis on multimodal highway standard data to obtain multidimensional feature vectors corresponding to highway standard items; calculating the multidimensional edit distance between items based on the multidimensional feature vectors to obtain a similarity matrix between highway standard item sequences; then, based on the similarity matrix between highway standard item sequences, constructing a dynamic association network with each highway standard item as a node; clustering the nodes in the dynamic association network to obtain a cluster structure corresponding to the dynamic association network; and finally, dynamically optimizing the dynamic association network and cluster structure based on the dynamic evolution characteristics corresponding to each node in the dynamic association network. The embodiments of the present invention implement multidimensional semantic expression of multimodal highway standard data, construct a dynamic association network through multidimensional edit distance, and achieve precise association and dynamic updating between highway standard items. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0071] Figure 1 A schematic flow chart of a method for constructing highway standard relationships provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0072] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0073] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0074] Figure 1 The figure shows a flow chart of a method for constructing highway standard relationships according to an embodiment of the present invention.

[0075] like Figure 1 As shown, the highway standard relationship construction method according to an embodiment of the present invention may include step S100, step S200, step S300, step S400 and step S500.

[0076] In step S100 , multidimensional semantic analysis is performed on the multimodal highway standard data to obtain multidimensional feature vectors corresponding to highway standard items.

[0077] Specifically, the above steps include:

[0078] First, obtain the original multimodal highway standard data document and construct a multimodal highway standard data document collection ,in, Represents the nth multimodal highway standard data document in the collection. The multimodal highway standard data includes highway standard text, images, formulas, version information, and reference information.

[0079] Then, the data structure of the multimodal highway standard data document set is initialized to obtain the highway standard entry set. ,in, Represents the jth highway standard entry in the highway standard entry set. It serves as the basic unit for subsequent processing and is used to carry multimodal content such as text, images, formulas, version information, and reference information.

[0080] For each highway standard item in the highway standard item set:

[0081] According to highway standard items The text content in the corresponding entry text collection is constructed ,in, Indicates highway standard items The pth original word in .

[0082] Based on a special dictionary in the highway field, the entry text collection is segmented in combination with a word segmentation algorithm, and professional terms such as road materials, construction technology, and design criteria are given higher priority analysis.

[0083] In order to emphasize the difference between key concepts and general vocabulary in the field of highway standards, the domain word weight function is used to perform word weighting on the word segmentation results, giving highway domain keywords a higher weight and obtaining word weighted results. The domain word weight function is:

[0084] ,in, for The corresponding word vector combination, for The corresponding word weight, The greater weight given to the domain keywords, A smaller weight is given to general words.

[0085] The word weighted results are mapped to the vector space to obtain the entry text feature vector corresponding to the highway standard entry. The formula is: ,in, is a word embedding function used to map words to vector space, Append to the corresponding word vector to make the key terms in the field contribute more to the overall text features.

[0086] According to highway standard items Image and chart content in the , build entry image collection ,in, Represents the qth image in the entry image collection.

[0087] For each image in the collection , use the pre-trained convolutional neural network (CNN) model to vectorize and obtain the corresponding image feature vector .

[0088] Use image aggregation function to aggregate the same highway standard items Merge all the image feature vectors below to get the entry image feature vector ,in, It is a weighted average maximum pooling aggregation strategy used to extract comprehensive image features in highway standard entries.

[0089] The entry image feature vector generated by the embodiment of the present invention can reflect the high-dimensional image features of the visual elements of the road structure and construction process.

[0090] According to highway standard items Formula content in, build entry formula collection ,in, Represents the rth formula in the set of entry formulas.

[0091] Extract the symbols involved in the entry formula set and build a symbol set ,in, Represents the zth symbol in a symbol set, where symbols include physical quantities, units, constants, and coefficients.

[0092] A symbol meaning mapping table is established according to the meaning of each symbol in the symbol set.

[0093] According to the symbol meaning mapping table and symbol set, the symbol set is converted into a symbol sequence and a structured syntax tree for subsequent coupling with the feature vectors of text and image. The relationship between the formula structure and the symbols is then integrated into the high-dimensional vector space to obtain the entry formula feature vector. The formula is ,in, It represents the embedding and structural analysis function for symbol sequences, integrating the relationship between formula structure and symbols into a high-dimensional vector space.

[0094] Preferably, in the process of symbol analysis and formula structuring, the symbols of specific parameters of highway engineering (such as design load, pavement stiffness coefficient) can be further separated, and higher attention can be paid to these engineering quantities, and the corresponding weight coefficients can be Added to the mapping process, if the symbol For key engineering quantities in the highway field, they will be enhanced during embedding.

[0095] After completing the extraction of the three main modal features of text, image, and formula, the final multidimensional semantic vector is formed by further combining version information and reference information. The multidimensional feature vector corresponding to the highway standard item is obtained by vector splicing using the item text feature vector, item image feature vector, item formula feature vector, and version information and reference information. The formula is: ,in, Represents vector-level splicing operations, Indicates highway standard items The metadata vector corresponding to the version information and reference information.

[0096] For each highway standard entry, a label set is constructed for the highway standard entry based on the highway standard specific attributes. ,The specific attributes of the above highway standards include road ,category, technical grade, reference level and version information. ,The above tag set includes multiple types such as “highway design specifications”, “old standard version”, “mandatory clause reference” and “recommended clause reference”.

[0097] The multidimensional feature vector corresponding to all highway standard entries and label sets Stored in the database, it provides data-driven input for subsequent multi-dimensional edit distance calculation and information entropy optimization diffusion.

[0098] Through the above-mentioned multidimensional feature vector analysis process, the embodiment of the present invention completely generates a multidimensional semantic expression that combines text, visual, formula and reference information, and supplemented by highway industry labels, to achieve a three-dimensional characterization of highway standard items, providing directly callable feature data for subsequent steps.

[0099] Next, in step S200 , the multidimensional edit distance between highway standard items is calculated based on the multidimensional feature vector to obtain a similarity matrix between highway standard item sequences.

[0100] Specifically, the above steps include:

[0101] Based on the multidimensional feature vectors and industry labels of the highway standard items generated in the above steps, the embodiment of the present invention constructs a multidimensional edit distance measurement mechanism that adapts to the actual application scenarios of highway standards to measure the degree of difference between different standard items.

[0102] First, based on the highway standard item set , in order to facilitate the subsequent dynamic programming process of edit distance, the first highway standard entry sequence to be compared is generated and the Second Highway Standard Entry Sequence ,in, and It can be a sequence of entries from different versions of the same standard, or a sequence of entries across different standards. Each entry in the sequence and Each corresponds to a multidimensional feature vector and , for subsequent edit distance algorithm to call.

[0103] Define the three basic operations of the edit distance algorithm (insertion, deletion, and substitution), combine the characteristics of the highway standard industry with multi-dimensional difference measurement, and conduct scenario-based modeling of operation costs. 、 and represent the insertion, deletion and replacement cost functions respectively.

[0104] If the first highway standard entry sequence Missing and Second Highway Standard Entry Sequence The corresponding entry to be compared is inserted into the second highway standard entry sequence based on the preset insertion cost function. The items to be compared in Insert the first highway standard entry into the corresponding position of the sequence. The insertion cost is defined as: ,in, represents a fixed insertion penalty, This means that insert operations are weighted based on version information or citation level. For example, if the inserted entry is a required or mandatory clause, a higher penalty value can be set.

[0105] If the second highway standard entry sequence Missing and first highway standard entry sequence For the corresponding items to be compared, based on the preset deletion cost function, the items to be compared that have no corresponding items in the first highway standard item sequence are deleted. The formula is defined as: ,in, represents the basic deletion penalty, It is used to implement a lower cost for deleting old or expired standard entries and a higher cost for deleting new version entries, so as to meet the needs of the ever-evolving scenarios of highway standards.

[0106] If there are first and second items to be compared that correspond to each other between the first highway standard item sequence and the second highway standard item sequence, then according to the multidimensional feature vectors corresponding to the first and second items to be compared and , respectively calculate the text dimension distance, image dimension distance, formula dimension distance, and version and reference information dimension distance between the first item to be compared and the second item to be compared.

[0107] The text dimension distance calculation formula is defined as: .

[0108] The image dimension distance calculation formula is: ,in, Represents the Euclidean norm of a vector, which is used to measure the difference between highway structure diagrams or construction drawings in visual space.

[0109] The formula for calculating the dimension distance is: ,in, and They respectively represent the h-th coordinate value of the corresponding symbol in the multidimensional feature vector.

[0110] The formula for calculating the distance between the version and reference information dimension is defined as: ,in, is the difference metric function for the g-th metadata attribute, It is a preset weighting coefficient for key fields, reflecting the difference in weight between the new and old versions or between cross-standard references.

[0111] The multidimensional edit distance of the entry is calculated based on the text dimension distance, image dimension distance, formula dimension distance, and version and reference information dimension distance. The multidimensional distance measurement function is: ,in, These are the weight parameters for text, image, formula, version and citation information dimensions respectively.

[0112] and The replacement cost is: ,in, By superimposing multi-dimensional difference metrics as the basic replacement penalty, we can achieve fine-grained control over multimodal content and version information.

[0113] when and When the source and version are similar, Approaching a smaller value.

[0114] Based on the above operation cost function, a multidimensional edit distance matrix D is defined according to the multidimensional edit distance between the first highway standard entry sequence and the second highway standard entry sequence. The elements in the multidimensional edit distance matrix D are: Represents the first highway standard entry sequence The first i entries and the second highway standard entry sequence The minimum multidimensional edit distance of the first j entries in the dynamic programming recursive relationship is: ,in, , that is, when both sequences are empty, the distance is zero. Indicates that accumulation can only be performed through insertion operations when this sequence is empty.

[0115] When the last item of the multidimensional edit distance matrix D After the calculation is completed, the sequence is obtained and The multidimensional edit distance value of: .

[0116] Convert the multidimensional edit distance in the multidimensional edit distance matrix into a similarity score within a preset range to obtain the first highway standard entry sequence with the second highway standard entry sequence The similarity matrix Sim between ,in, is the attenuation coefficient, which is used to convert the distance value into a similarity score in the range of (0,1]. The larger the value, the more similar the entries are.

[0117] The embodiment of the present invention realizes the acquisition of sequence under the control of multi-dimensional feature vector and scenario cost through the above dynamic programming process. and The minimum edit distance value is used to measure the degree of difference between two highway standard item sequences. The resulting similarity matrix can be directly used for subsequent network structure construction, adjacent node screening, and preliminary aggregation of standard items. This multidimensional edit distance method can effectively distinguish differences between new and old versions, minor changes, and cross-standard references, thus laying a foundation for difference measurement for highway standards management and intelligent query.

[0118] In step S300 , based on the similarity matrix between highway standard item sequences, each highway standard item is used as a node, and the similarity between node pairs is used to establish association edges between nodes to construct a dynamic association network.

[0119] Specifically, the above steps include:

[0120] First, define the graph structure , where the node set Corresponding to all highway standard entries and including nodes Related entries And the corresponding multidimensional feature vector , the edge set E is initially empty and is used to construct edge connections based on similarity thresholds and weights.

[0121] This step completes the loading of similarity information and the initial initialization of the graph, providing data input drive for the next step of establishing weighted association edges.

[0122] Then, based on the similarity matrix, we determine the node pair Is the similarity score between them greater than the preset similarity threshold, that is ,in, In the similarity matrix and The corresponding similarity score, is the preset similarity threshold.

[0123] If the node pair If the similarity score between the two nodes is less than or equal to the preset similarity threshold, the first edge weight between the node pairs is When is zero, no associated edge is established between the node pair to reduce noise connections between invalid or weakly associated nodes.

[0124] If the node pair If the similarity score between the two nodes is greater than the preset similarity threshold, the first edge weight corresponding to the node pair is obtained based on the similarity score between the node pairs. .

[0125] Then, the version, reference, and inheritance relationship of highway standards are further integrated to determine whether there is version inheritance or mandatory reference between node pairs.

[0126] If there is version inheritance or mandatory reference between node pairs, then according to the gravity coefficient and the first edge weight , get the second edge weight , the calculation formula is: , where the gravitational coefficient Adaptive adjustment is made based on the citation intensity and the difference between the versions of the homologous standard. If two entries belong to different versions of the same standard and there are only minor revisions, The larger the value; if the two items are cross-standard references, then The value is relatively small.

[0127] If there is no version inheritance or mandatory reference between the node pairs, the first edge weight is directly used as the second edge weight.

[0128] According to the edge weight, establish an associated edge in the edge set for the node pair , and the second edge weight Stored in associated edges The weight field of is used to complete the weighted connection of the graph structure G, and finally obtain a dynamic association network with weights.

[0129] Next, in step S400 , the nodes in the dynamic association network are clustered to obtain a cluster structure corresponding to the dynamic association network.

[0130] Specifically, the above steps include:

[0131] In order to identify closely connected highway standard items in the dynamic association network, an aggregation threshold is set. , which is used to determine whether there is sufficient association strength between nodes.

[0132] If the node pair The edge weights between , they are considered as candidate connections within the same aggregation community.

[0133] During the aggregation process, all nodes in the dynamic association network are divided into communities based on the hierarchical clustering algorithm: ,in, Represents the kth aggregate community, in which the average weight between member nodes is greater than or equal to the preset aggregation threshold .

[0134] For the The node set indicates that its entries have high similarity in multimodal features and version reference associations, and can be regarded as a standard cluster. For example, adjacent version entries of the same source standard often form a cluster, and entries referenced across standards can also enter the same cluster when the edge weight is high enough.

[0135] Each standard cluster The output is the preliminary aggregation result, including the set of nodes in the cluster and its edge weight , get the cluster structure and output all communities List of included highway standard items, number of nodes and their version reference distribution.

[0136] The dynamic association network and cluster information are stored in the database.

[0137] The embodiment of the present invention divides highly similar communities among highway standard items through the above steps. Subsequent steps will continue to perform deep reasoning based on the information summary optimization diffusion model to further refine the semantic associations and dynamic evolution attributes between items.

[0138] Finally, in step S500 , the dynamic association network and the cluster structure are dynamically optimized according to the dynamic evolution characteristics of the highway standard items corresponding to each node in the dynamic association network.

[0139] Specifically, the above steps include:

[0140] Through the previous step, we get a dynamic association network , where each edge With initial weight , which is obtained by integrating multimodal feature information such as version, reference, image, formula, etc. of highway standards, as well as preliminary aggregation standard clusters , used to identify a set of entries with high similarity, homology, or related versions.

[0141] However, in actual scenarios, the associations between highway standard items are not always static. The following problems exist: the dynamic evolution of multimodal features at different time periods, such as construction drawing updates, formula parameter changes, and the addition or abolition of mandatory references; and the high-dimensional coupling of cross-standard reference logic, which is difficult to characterize using only traditional graph traversal or simple entropy regularization.

[0142] Based on this, an embodiment of the present invention provides the following multimodal entropy-driven diffusion correction model, which combines network differential equations with information entropy regularization to achieve dual iterative optimization of network structure and cluster division.

[0143] For node sets , where each node Corresponding highway standard items .

[0144] To optimize both the topology and multimodal semantics simultaneously, we introduce the following four comprehensive objectives:

[0145] Graph smoothing term: , Represents nodes that are continuously updated during the iteration process The graph smoothing term encourages the embedding vectors of adjacent nodes to remain similar in the multimodal space.

[0146] Entropy regularization term: , the first part of the formula is the edge weight entropy within the cluster, is an adjustable function ( or ), used to limit the norm of node states, is the regularization term.

[0147] Multimodal coupling loss: ,in, is a vector addition function that directly adds all vectors together. is the current node representation in the Cain space.

[0148] Cross-standard reference constraints: For node pairs with mandatory references or version inheritance , additional constraints ,in, represents a mandatory reference or version inheritance edge set, The number of high-weight edges for forced reference or version inheritance, A custom version difference mapping function.

[0149] According to the above four items, the overall optimization objective function is obtained: , where x is the node representation and W is the edge weight.

[0150] In order to capture the dynamic correlation evolution between highway standard items, a network differential equation is proposed to model the diffusion process of node states. Different from the traditional discrete iteration, the continuous time idea is adopted to characterize the propagation and attenuation of multimodal characteristics.

[0151] First, the discrete Upgrade to continuous status, , and make it satisfy the following network diffusion equation in vector form: ,in, is the Laplace operator, It is an externally driven item used to simulate mandatory references, version evolution, or external information injection.

[0152] For edge weights Introducing the time dimension, we get , ,in, represents the gradient of the regularization term with respect to the edge weight, It is a pulse increment based on version inheritance or reference events. is the learning rate constant. Based on this formula, the edge weights are allowed to be dynamically updated according to the regularization and reference features during the diffusion process.

[0153] Since highway standards are often updated based on version numbers or years, differential equations can be combined with discrete events: when the system is at a certain time When the standard version release point is reached, the edge weight pulse update and cluster structure recalculation are triggered. Based on this, the present invention can capture the complex evolution process of "continuous diffusion + discrete jump".

[0154] On the basis of the above entropy regularization term, it is further expanded to high-order entropy and cross-cluster mutual information to characterize multimodal coupling and cross-standard information flow. In order to improve the ability to express complex multimodal distributions, the second-order or third-order entropy term is introduced: , incorporate this formula into the total entropy regularization term In the optimization process, the influence of high-confidence edges is amplified, so that strong connections are strengthened in the iteration.

[0155] For several standard clusters formed during the initial aggregation In actual scenarios, some cross-cluster entries may also have important reference relationships. Therefore, we introduce cross-cluster mutual information measurement and control the information flow between clusters: ,in, is the dot product of the node representation, if If the value is too large, it indicates that there is a potential need to merge the two clusters. If it is too small, its edge weight will be further reduced to maintain the difference between clusters.

[0156] Since the optimization objective function Also includes node representation and edge weights It is difficult to directly solve the high-order entropy terms and cross-cluster mutual information constraints. Therefore, the embodiment of the present invention adopts a global-local alternating optimization strategy to dynamically optimize the dynamic association network.

[0157] During global optimization, the edge weights of the associated edges Fixed, using variational inference decomposition to represent nodes in dynamic association networks Update, and then optimize the objective function based on the ADMM algorithm in Indicates solving.

[0158] During the local optimization process, cross-cluster association edges and association edges with version inheritance or mandatory references are pulsed and modified according to discrete events (standard version releases): if a reference is abolished in this version, the corresponding edge weight decays to zero; if a new mandatory reference is added, the corresponding edge weight is increased.

[0159] Continue to optimize until dynamic convergence occurs and cross-cluster mutual information is adjusted to a reasonable range.

[0160] The node representation and edge weights in the optimized dynamic association network have changed, resulting in the original cluster No longer reasonable.

[0161] Based on this, according to the updated node representation and edge weight W, recalculate the association between any two nodes, ,in, is a decreasing function ( ), which is used to fuse multimodal spatial distance with graph edge weight to obtain a new edge metric.

[0162] Using the updated edge weights Perform community detection to obtain the cluster set after secondary division .

[0163] If there is a cluster and Cross-cluster mutual information If the number of nodes is greater than the set threshold or exceeds expectations, they will be merged into a larger cluster; if there are signs of division within a cluster, it will be split to ensure that the network maintains a reasonable cluster structure during version iterations and strong reference changes.

[0164] Furthermore, the updated dynamic association network is output, where the edge set keeps the node pairs unchanged but the weights become , the weight is iteratively modified under the joint action of high-order entropy regularization, cross-cluster mutual information and differential equation diffusion model.

[0165] Output the cluster set after secondary partitioning and compare it with the initial aggregated By comparing and analyzing the reference edges of the merged or split items and the versions that mutated in the process, an evolution graph or snapshot sequence can be drawn to trace the dynamic evolution of highway standards over time or version numbers.

[0166] In a highway standard management system, the multimodal entropy-driven diffusion correction model proposed in an embodiment of the present invention can be run periodically or event-triggered. Once a major standard update is detected (such as the implementation of a new version, mandatory reference rule changes), the model is restarted to keep the network and cluster structure up to date.

[0167] In addition, an embodiment of the present invention further provides a method for querying highway standard relationships, which specifically includes:

[0168] In highway standards management and application scenarios, users often need to perform cross-standard, cross-version, and cross-modal queries on a large number of standard entries, such as "finding the latest version of a construction drawing reference" and "retrieving historical evolution records related to a formula parameter."

[0169] Through the dynamic association network (including node representation, edge weight, version information and aggregate cluster structure) in the highway standard relationship construction method as described above, an embodiment of the present invention provides a highway standard relationship query method, thereby achieving more efficient, more accurate and more comprehensive retrieval on the basis of the above.

[0170] First, multi-level modeling was conducted based on query requirements. To meet the query needs of different users, queries were divided into the following types: Keyword queries: searching for keywords or professional terms within highway standard texts; Reference path queries: exploring the inherited or mandatory reference paths of an item across different versions or standards; Modal combination queries: complex searches based on multimodal elements such as text, images, and formulas. "Scenario-based semantic templates" were defined for different user groups, including engineers, designers, and regulatory agencies, such as construction drawing verification, parameter verification, and compliance approval. These templates help the system automatically match scenario-based rules when interpreting query intent.

[0171] Incorporating version and timeliness information into the query semantic model allows users to filter or limit query results to specific years or editions.

[0172] In the dynamic association network of the highway standard relationship construction method described above, each node (item) has multimodal information such as text features, image features, formula features, and reference relationships, and each associated edge records cross-version or cross-standard weight information. To meet the needs of massive standard item retrieval, hierarchical indexes are established for nodes and edges, including: First-level index (text / semantic): The multimodal vector of the node is converted into Clustering is performed to speed up similar retrieval; Secondary index (reference edge): For mandatory references and high-weight references in the association network, separate inverted indexes or graph search indexes are established to quickly find reference paths.

[0173] Furthermore, because highway standards evolve over multiple versions, a version index tree is created for each version number and cross-referenced with the node and edge indices. When a query request arrives, the relevant entries and edges within the specified version range are located based on this "version index tree + node / edge indices."

[0174] The system obtains user-entered query statements, such as "Find mandatory references related to a certain construction drawing in the latest specifications," and performs natural language parsing on the query statement, including word segmentation, part-of-speech tagging, and named entity recognition (which can identify standard names, version numbers, construction scenario keywords, etc.), and matches them with professional vocabulary and multimodal terminology libraries. The system can also vectorize the query statement based on a pre-trained model (such as BERT and GPT) to obtain a query vector, and then calculate the similarity with the highway standard item node vector to find the query node candidate set that best matches the query intent.

[0175] Preferably, based on the "scenario-based semantic template" defined in the above steps, the intent of the parsed query is inferred: if a "version verification" scenario is detected, the retrieval rules of the version index tree and the mandatory reference edge are used; if an "image verification" scenario is detected, the image feature index is additionally called, etc.

[0176] Based on the query node candidate set, a comprehensive ranking is performed based on the following factors: semantic cosine similarity with the query vector q; the influence of the node in the associated network (derived from the above-mentioned aggregation metrics, etc.); and weighted support for the result entries by edge weights (citation strength).

[0177] Form the final ranking score , outputting several high-scoring query results. These results include not only text but also construction drawings and supporting formulas. The system integrates the matched images, formulas, and text summaries for display. For example, the most relevant images or formulas are displayed directly in the result list, and the associated paths (across standards and versions) of the items are visualized in the interactive interface.

[0178] It should be noted that if the user submits a "cross-version reference tracing" query, the call to the node Related high-weight edges , recursively draw the reference path in the visual interface; if the edge weight is weak or there is a conflict mark, the user will be reminded through color or warning labels.

[0179] In addition, an embodiment of the present invention also provides a highway standard relationship management device, which includes: a processor and a memory; the memory is used to store one or more program instructions; the processor is used to run one or more program instructions to execute the steps of a highway standard relationship construction method or a highway standard relationship query method as described above.

[0180] In addition, an embodiment of the present invention further provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the steps of a highway standard relationship construction method or a highway standard relationship query method as described above.

[0181] In addition, an embodiment of the present invention further provides a computer program product, which includes computer program instructions. When the computer program instructions are executed by a processor, the steps of the above-mentioned highway standard relationship construction method or highway standard relationship query method are implemented.

[0182] The highway standard relationship construction and query methods, devices, media, and program products proposed in the embodiments of the present invention can partially address the existing issues of insufficient multimodal information processing capabilities, inaccurate difference measurement, and poor dynamic adaptability. Using a multidimensional semantic expression model, the multimodal content of highway standard documents, including text, images, and formulas, is parsed to generate a unified multidimensional semantic feature vector. The influence of domain weights is also taken into account, giving higher weight to key terms, road design drawings, engineering formulas, and other information during feature extraction, thereby improving the comprehensiveness of standard item parsing.

[0183] To measure the differences between highway standard items, an embodiment of the present invention proposes a multi-dimensional edit distance calculation method based on scenario optimization. This method uses a dynamic programming algorithm to combine multimodal feature differences in text, images, formulas, and version reference information for measurement. In particular, the definition of a scenario-based operation cost function enables insertion, deletion, and substitution operations to flexibly adapt to the actual characteristics of highway standards. For example, for the addition or abolition of mandatory references, the dynamic adjustment of the weight parameters in the operation cost function can accurately reflect the semantic differences between standard items caused by version changes.

[0184] In the process of constructing a dynamic association network, the embodiment of the present invention completes the construction of a dynamic association network of highway standard items clustering and weighting by combining the similarity matrix with the multi-dimensional edit distance results. Through the dynamic aggregation algorithm, it can not only quickly identify the high similarity between standard items, but also comprehensively process multimodal features and version reference information. It can cope with complex cross-version and cross-standard reference scenarios, and at the same time support real-time updates of the dynamic evolution of the item relationship network. Combined with the high-order entropy regularization optimization model and cross-cluster mutual information constraints, the rationality of item clustering is further strengthened, making the semantic association between items more accurate.

[0185] Furthermore, the present invention constructs a multimodal entropy-driven diffusion model, based on the diffusion process of network differential equations, to capture the dynamic evolution of associations between highway standard items. This model not only diffuses and corrects semantic information between items but also dynamically adjusts the structure and edge weights of the association network by incorporating external drivers such as mandatory references and version inheritance. Based on this method combining continuous diffusion with discrete transitions, embodiments of the present invention can respond in real time to standard item version updates and reference relationship changes, ensuring the timeliness of the item relationship network compared to existing technologies.

[0186] In the embodiments of the present invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules within the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the aforementioned method. The storage medium can be a memory, for example, volatile memory or non-volatile memory, or can include both volatile and non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM).The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory. Those skilled in the art will appreciate that in one or more of the above examples, the functions described in the present invention can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media includes any medium that facilitates the transmission of computer programs from one place to another. The storage medium can be any available medium that can be accessed by a general or special-purpose computer. Although the present invention has been described in detail above using general instructions and specific embodiments, it will be apparent to those skilled in the art that modifications or improvements can be made based on the present invention. Therefore, these modifications or improvements made without departing from the spirit of the present invention are within the scope of protection claimed in the present invention.

[0187] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Those skilled in the art can make some simple modifications, equivalent changes or modifications based on the technical content disclosed above, which all fall within the scope of protection of the present invention.

Claims

1. A method for constructing highway standard relationships, characterized in that: The method comprises: Perform multidimensional semantic analysis on multimodal highway standard data to obtain multidimensional feature vectors corresponding to highway standard items; Based on the multidimensional feature vector, the multidimensional edit distance between the highway standard items is calculated to obtain a similarity matrix between the highway standard item sequences, including: generating a first highway standard item sequence and a second highway standard item sequence to be compared based on the highway standard item set; If the first highway standard entry sequence lacks an entry to be compared corresponding to the second highway standard entry sequence, inserting the entry to be compared in the second highway standard entry sequence into a corresponding position in the first highway standard entry sequence based on a preset insertion cost function; If the second highway standard item sequence lacks an item to be compared corresponding to the first highway standard item sequence, deleting the item to be compared without a corresponding item in the first highway standard item sequence based on a preset deletion cost function; If there are corresponding first and second items to be compared between the first and second highway standard item sequences, then calculating, based on the multidimensional feature vectors corresponding to the first and second items to be compared, the text dimension distance, image dimension distance, formula dimension distance, and version and reference information dimension distance between the first and second items to be compared; Obtaining an entry multidimensional edit distance based on the text dimension distance, the image dimension distance, the formula dimension distance, and the version and reference information dimension distance; A multidimensional edit distance matrix D is generated based on the multidimensional edit distance between the first highway standard entry sequence and the second highway standard entry sequence. The elements in the multidimensional edit distance matrix D are: represents the minimum multidimensional edit distance between the first i entries in the first highway standard entry sequence and the first j entries in the second highway standard entry sequence; Converting the multidimensional edit distance in the multidimensional edit distance matrix into a similarity score within a preset range to obtain a similarity matrix between the first highway standard item sequence and the second highway standard item sequence; Based on the similarity matrix between the highway standard item sequences, each highway standard item is used as a node, and the similarity between node pairs is used to establish association edges between nodes to construct a dynamic association network, including: Creating a graph structure, the graph structure including a node set and an edge set, the node set corresponding to all highway standard entries; Based on the similarity matrix, determining whether the similarity score between the node pairs is greater than a preset similarity threshold; If the similarity score between the node pairs is less than or equal to the preset similarity threshold, the weight of the first edge between the node pairs is zero, and no associated edge is established between the node pairs; If the similarity score between the node pair is greater than the preset similarity threshold, obtaining a first edge weight corresponding to the node pair according to the similarity score between the node pair; Determine whether there is version inheritance or mandatory reference between the node pairs; If version inheritance or mandatory reference exists between the node pairs, a second edge weight is obtained according to the gravity coefficient and the first edge weight, wherein the gravity coefficient is adaptively adjusted according to the reference strength and the degree of difference between the homologous standard versions; If there is no version inheritance or mandatory reference between the node pairs, the first edge weight is used as the second edge weight; Establishing an association edge for the node pair in the edge set, storing the second edge weight in a weight field of the association edge, and finally obtaining a dynamic association network with weights; Clustering the nodes in the dynamic association network to obtain a cluster structure corresponding to the dynamic association network; According to the dynamic evolution characteristics of the highway standard items corresponding to each node in the dynamic association network, the dynamic association network and the cluster structure are dynamically optimized.

2. A highway standard relationship construction method according to claim 1, characterized in that: Multi-dimensional semantic analysis is performed on the multimodal highway standard data to obtain the multi-dimensional feature vectors corresponding to the highway standard items, including: Acquire original multimodal highway standard data and construct a multimodal highway standard data document set, wherein the multimodal highway standard data includes highway standard text, images, formulas, version information, and reference information; Performing data structure initialization processing on the multimodal highway standard data document set to obtain a highway standard item set; For each highway standard item in the highway standard item set; Constructing a corresponding entry text set based on the text content in the highway standard entry, performing word segmentation processing on the entry text set based on a highway field-specific dictionary, and performing word weighting processing on the word segmentation results using a field word weight function to assign higher weights to highway field keywords to obtain a word weighted result; Mapping the word weighted result to a vector space to obtain an entry text feature vector corresponding to the highway standard entry; Constructing an entry image set based on the image and chart contents in the highway standard entry; Vectorizing the images in the entry image set using a pre-trained convolutional neural network to obtain corresponding image feature vectors; Use the image aggregation function to merge all the image feature vectors under the highway standard items to obtain the item image feature vector; Constructing an entry formula set according to the formula contents in the highway standard entry; Extracting symbols involved in the entry formula set and constructing a symbol set; Establishing a symbol meaning mapping table according to the meaning of each symbol in the symbol set; Integrating the relationship between the formula structure and the symbols into a high-dimensional vector space according to the symbol meaning mapping table and the symbol set to obtain an entry formula feature vector; Performing vector concatenation processing using the entry text feature vector, the entry image feature vector, the entry formula feature vector, and version information and reference information corresponding to the highway standard entry to obtain a multidimensional feature vector corresponding to the highway standard entry; A tag set is constructed for a highway standard entry according to highway standard specific attributes, wherein the highway standard specific attributes include road category, technical level, reference level, and version information.

3. A highway standard relationship construction method according to claim 1, characterized in that: Clustering the nodes in the dynamic association network to obtain a cluster structure corresponding to the dynamic association network includes: All nodes in the dynamic association network are aggregated based on a hierarchical clustering algorithm to obtain at least one standard cluster, forming a cluster structure corresponding to the dynamic association network, and the average weight between member nodes in the standard cluster is greater than or equal to a preset aggregation threshold.

4. A highway standard relationship construction method according to claim 1, characterized in that: According to the dynamic evolution characteristics of the highway standard items corresponding to each node in the dynamic association network, the dynamic association network and the cluster structure are dynamically optimized, including: Based on the dynamic association network, an optimization objective function is constructed, wherein the optimization objective function includes a graph smoothing term, an entropy regularization term, a multimodal coupling loss, and a cross-standard reference constraint term. The smoothing term is used to keep the embedding vectors of adjacent nodes similar in the multimodal space. The entropy regularization term includes the edge weight entropy within the standard cluster and an adjustable function for limiting the node state norm. The cross-standard reference constraint term is used to constrain the edge set of version inheritance or mandatory reference; Introducing second-order or third-order entropy terms into the entropy regularization term amplifies the influence of high-confidence edges in the optimization process, allowing strong connections to be strengthened during iterations to improve the ability to express complex multimodal distributions; Introducing cross-cluster mutual information metric between standard clusters; Based on the optimization objective function, a global-local alternating optimization strategy is adopted to dynamically optimize the dynamic association network; During the global optimization process, the edge weights of the associated edges remain fixed, and the nodes in the dynamic associated network are updated using variational inference. The node representations in the optimization objective function are solved based on the ADMM algorithm to update the node representations. During the local optimization process, pulse-modify the cross-cluster association edges and the association edges with version inheritance or mandatory reference according to discrete events, wherein the discrete events include standard version releases; Based on the optimized dynamic association network, the cluster structure is re-divided and adaptively merged to obtain an updated cluster structure.

5. A highway standard relationship query method, characterized in that: The method comprises: Perform multi-level modeling based on query requirements; According to the dynamic association network in the highway standard relationship construction method according to any one of claims 1 to 4, a corresponding multi-level index and fast retrieval structure is generated; According to the query statement entered by the user, match the corresponding query node candidate set; Outputting a query result based on the query node candidate set.

6. A highway standard relationship management device, characterized in that: The device includes: a processor and a memory; The memory is used to store one or more program instructions; The processor is used to run one or more program instructions to execute the steps of the highway standard relationship construction method according to any one of claims 1 to 4 or the highway standard relationship query method according to claim 5.

7. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a highway standard relationship construction method according to any one of claims 1 to 4 or a highway standard relationship query method according to claim 5.

8. A computer program product, characterized in that The computer program product includes computer program instructions, which, when executed by a processor, implement the steps of a highway standard relationship construction method according to any one of claims 1 to 4 or a highway standard relationship query method according to claim 5.

Images