A maritime information matching processing method and device

Abstract

The application discloses a maritime information matching processing method and device, and the method comprises the following steps: obtaining to-be-retrieved maritime text information; performing digital representation on the to-be-retrieved maritime text information to obtain a corresponding to-be-retrieved information set; performing retrieval matching processing on the to-be-retrieved information set based on a preset maritime knowledge graph database to obtain a retrieval matching result of maritime information.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of industrial data processing, text data processing and big data mining technology, and in particular to a maritime information matching processing method and device. BACKGROUND

[0002] With the rapid development of global shipping industry, maritime information presents an explosive growth trend, covering ship dynamics, port operations, AIS (Automatic Identification System) data, marine environmental factors and other multi-dimensional information. These information are mostly in the form of unstructured text (such as ship reports, port scheduling records, maritime supervision documents, etc.), and contain a large number of entities (ships, ports, AIS devices, marine elements, etc.) and complex relationships between entities (such as ship-port docking relationship, AIS device and ship binding relationship, marine element influence on ship navigation relationship, etc.), which leads to many challenges in maritime information retrieval matching.

[0003] In the prior art, maritime information retrieval mostly relies on traditional text matching methods (such as keyword matching, semantic similarity calculation based on word vectors, etc.), which has the following technical problems:

[0004] Ignoring the structured features of information: traditional methods treat text as a linear sequence, and fail to effectively capture the association between entities (i.e. graph structure features), resulting in the loss of key context information during retrieval (such as the semantic difference between "ship A docking at port B" and "port B scheduling ship A" cannot be distinguished by keyword matching).

[0005] Insufficient accuracy of entity matching: only through text similarity (such as string matching, word vector cosine similarity) to determine whether the entity matches, without considering the "neighbor relationship" of the entity (i.e. the structural features of the entity in the knowledge graph). For example, two ships with similar names may belong to different routes (associated with different ports), and traditional methods are prone to misjudgment as the same entity.

[0006] Poor robustness of edge information retrieval: the relationship between entities (edge information) often has ambiguity and diversity (such as "passing through", "docking", "passing by" all represent the association between ship and port), and the existing technology is too simple in representing edge information (such as single vector mapping), which is difficult to adapt to the matching needs of complex relationships, resulting in missed or false detection of edge information retrieval. SUMMARY

[0007] The present application mainly solves the problem of how to efficiently, completely and accurately retrieve maritime information, and discloses a maritime information matching processing method and device.

[0008] In a first aspect of the present application, a maritime information matching processing method is disclosed, comprising:

[0009] S1, obtaining maritime text information to be searched;

[0010] S2, digitally representing the maritime text information to be searched to obtain a corresponding information set to be searched;

[0011] S3, based on a preset maritime knowledge graph database, performing search matching processing on the information set to be searched to obtain a search matching result of maritime information.

[0012] The digitally representing the maritime text information to be searched to obtain a corresponding information set to be searched comprises:

[0013] S21, graph data representation of the maritime text information to be searched is performed to obtain a node information set and an edge information set;

[0014] S22, Euclidean space embedding representation of the node information set and the edge information set is performed to obtain an adjacency matrix set;

[0015] S23, the node information set, the edge information set and the adjacency matrix set are used to construct a corresponding information set to be searched.

[0016] The graph data representation of the maritime text information to be searched to obtain a node information set and an edge information set comprises:

[0017] S211, determining a factor set of graph data objects; the factor set comprises a node name set and an edge name set;

[0018] S212, based on the node name set, performing entity extraction processing on the maritime text information to be searched to obtain a node information set; the node information set comprises name information and attribute information of each node;

[0019] S213, based on the edge name set, performing relationship extraction processing on the maritime text information to be searched to obtain an edge information set; the edge information set comprises edge information corresponding to each node.

[0020] The search matching processing of the information set to be searched based on the preset maritime knowledge graph database to obtain a search matching result of maritime information comprises:

[0021] The preset maritime knowledge graph database comprises entity information and corresponding edge information; edges are connected with entities; the entity information comprises name information and corresponding adjacency matrices of entities;

[0022] S31, perform entity matching retrieval on the to-be-retrieved information set based on the preset maritime knowledge graph database to obtain an entity matching information set; the entity matching information set includes a plurality of entity information;

[0023] S32, perform edge information retrieval based on the preset maritime knowledge graph database and the entity matching information set to obtain an edge matching information set;

[0024] S33, perform reasoning processing on the entity matching information set and the edge matching information set to obtain a retrieval matching result of maritime information.

[0025] The entity matching retrieval on the to-be-retrieved information set based on the preset maritime knowledge graph database to obtain an entity matching information set includes:

[0026] S311, respectively perform text similarity calculation on the name information of each entity in the preset maritime knowledge graph database and the name information of each node in the node information set in the to-be-retrieved information set to obtain a corresponding name text similarity value;

[0027] S312, extract entities and nodes with all name text similarity values greater than a preset first threshold value; and construct node pairs by using the extracted entities and nodes of each name text similarity value;

[0028] S313, perform matching degree calculation on the adjacency matrix of each node pair to obtain a matching degree value;

[0029] S314, construct an entity matching information set by using the entity information of the entities in the node pairs with all matching degree values greater than a preset second threshold value.

[0030] The matching degree calculation on the adjacency matrix of each node pair to obtain a matching degree value includes:

[0031] S3131, for each node pair, obtain the adjacency matrix corresponding to the entity of the node pair and the adjacency matrix corresponding to the node;

[0032] S3132, perform element-by-element subtraction on the adjacency matrix corresponding to the entity and the adjacency matrix corresponding to the node to obtain a first difference matrix;

[0033] S3133, perform eigenvalue decomposition on the first difference matrix to obtain an eigenvalue set;

[0034] S3134, calculate the average value γ0 of the eigenvalue set;

[0035] S3135, calculate the rank value z of the first difference matrix;

[0036] S3136, the average value of the feature value set and the rank value of the first difference matrix are statistically calculated to obtain a first matching value;

[0037] S3137, the adjacent matrix corresponding to the entity and the adjacent matrix corresponding to the node are divided item by item to obtain a second difference matrix;

[0038] S3138, the second difference matrix is calculated to obtain a second matching value;

[0039] S3139, the first matching value and the second matching value are fused to obtain a matching degree value.

[0040] The expression of the statistical calculation is:

[0041] p1=z×sin(t 1-α (n)),

[0042]

[0043] Wherein, n is the degree of freedom value, alpha is the proportion value, gamma max is the maximum eigenvalue of the eigenvalue set, t 1-α (n) is the 1-alpha quantile of t distribution with n degrees of freedom, and p1 is the first matching value;

[0044] The expression of the fusion calculation is:

[0045] p=D n (p1+p2),

[0046] Wherein, p is the matching degree value, p2 is the second matching value, and D n is the n-order Weibull function.

[0047] In a second aspect of the embodiment of the application, a maritime information matching processing device is disclosed, which comprises:

[0048] A memory storing executable program codes;

[0049] A processor coupled with the memory;

[0050] The processor calls the executable program codes stored in the memory to execute the maritime information matching processing method.

[0051] In a third aspect of the embodiment of the application, a computer storage medium is disclosed, which stores computer instructions, and when the computer instructions are called by a computer, the maritime information matching processing method is executed.

[0052] The fourth aspect of the embodiment of the present application discloses an information data processing terminal, which is used for realizing the matching processing method of maritime information.

[0053] The present application has the following advantages:

[0054] The structured features of the reserved information improve the retrieval integrity: by converting the to-be-retrieved maritime text into a node information set and an edge information set (graph data representation), and further embedding into an adjacency matrix, the association between entities and entities is completely retained, solving the problem of losing structural information in traditional text matching, and making the retrieval result more consistent with the complex semantics of maritime information.

[0055] Multi-dimensional entity matching improves matching accuracy: in the entity matching process, candidate entities are first screened through text similarity, and then the matching degree is calculated based on the structural features (eigenvalue, rank value, etc.) of the adjacency matrix, considering the "name semantics" and "structural context" of the entity, effectively distinguishing entities with similar names but different association relationships, and reducing the false matching rate.

[0056] Robust edge information retrieval enhances relationship matching accuracy: through edge vector fusion center calculation (comprehensive arithmetic mean, geometric mean, median vector and distance distribution characteristics), a more representative center vector is obtained; combined with the difference calculation model constructed by the trigonometric function, the subtle differences of the edge vector are amplified, the matching ability for fuzzy relationships and diverse expressions is improved, and the missed detection and false detection are reduced.

[0057] Fusion reasoning mechanism improves the reliability of retrieval results: the combination of path reasoning, rule reasoning and vector reasoning verifies the matching results from the aspects of structural association, explicit logic and semantic similarity, makes up for the limitations of single reasoning method, and makes the retrieval results more comprehensive and reliable.

[0058] Adapt to the characteristics of maritime information, strong practicability: special graph representation and matching strategy are designed for entity types (ships, ports, AIS devices, etc.) and relationship types (docking, scheduling, influence, etc.) specific to the maritime field, which is more consistent with the maritime business scenario and can be directly applied to maritime supervision, ship scheduling, risk warning and other practical businesses. BRIEF DESCRIPTION OF DRAWINGS

[0059] Figure 1 The flowchart of the implementation of the method of the present application. DETAILED DESCRIPTION

[0060] In order to better understand the content of the present application, an embodiment is given here.

[0061] Figure 1 The flowchart of the implementation of the method of the present application.

[0062] In a first aspect, the present application discloses a maritime information matching processing method, comprising:

[0063] S1, obtaining a to-be-searched maritime text information;

[0064] S2, performing digital representation on the to-be-searched maritime text information to obtain a corresponding to-be-searched information set;

[0065] S3, performing search matching processing on the to-be-searched information set based on a preset maritime knowledge graph database to obtain a search matching result of maritime information.

[0066] The digital representation on the to-be-searched maritime text information to obtain the corresponding to-be-searched information set comprises:

[0067] S21, performing graph data representation on the to-be-searched maritime text information to obtain a node information set and an edge information set;

[0068] S22, performing Euclidean space embedding representation on the node information set and the edge information set to obtain an adjacency matrix set;

[0069] S23, constructing the corresponding to-be-searched information set by using the node information set, the edge information set and the adjacency matrix set.

[0070] The graph data representation on the to-be-searched maritime text information to obtain the node information set and the edge information set comprises:

[0071] S211, determining a factor set of a graph data object; the factor set comprises a node name set and an edge name set;

[0072] S212, performing entity extraction processing on the to-be-searched maritime text information based on the node name set to obtain a node information set; the node information set comprises name information and attribute information of each node;

[0073] S213, performing relation extraction processing on the to-be-searched maritime text information based on the edge name set to obtain an edge information set; the edge information set comprises edge information corresponding to each node;

[0074] The node name set comprises a ship entity name, a port entity name, an AIS entity name and a marine factor entity name;

[0075] The edge name set comprises relations between various entities;

[0076] The Euclidean space embedding representation on the node information set and the edge information set to obtain the adjacency matrix set comprises

[0077] Structurally parsing the node information set and the edge information set to obtain an adjacency matrix and an adjacency list of each node;

[0078] Determining a graph structure feature to be retained; the graph structure feature to be retained is a distance to an entity less than a preset distance threshold or a similarity value of an edge related to a node less than a preset similarity value threshold;

[0079] Processing the adjacency matrix and the adjacency list of each node based on the graph structure feature to be retained to obtain a processed adjacency matrix and a processed adjacency list;

[0080] Constructing an adjacency matrix set by using all the processed adjacency matrices.

[0081] The processing of the adjacency matrix and the adjacency list of each node based on the graph structure feature to be retained can be implemented by using a Word2Vec function of a gensim library, a node2vec library (compatible with a DeepWalk mode) or the like;

[0082] The searching and matching processing of the to-be-searched information set based on the preset maritime knowledge graph database to obtain a searching and matching result of maritime information includes:

[0083] The preset maritime knowledge graph database includes entity information and corresponding edge information; an edge is connected with an entity; the entity information includes name information of the entity and a corresponding adjacency matrix;

[0084] S31, performing entity matching searching on the to-be-searched information set based on the preset maritime knowledge graph database to obtain an entity matching information set; the entity matching information set includes a plurality of entity information;

[0085] S32, performing edge information searching based on the preset maritime knowledge graph database and the entity matching information set to obtain an edge matching information set; the edge matching information set includes a set of edge information corresponding to each entity matching information;

[0086] S33, performing reasoning processing on the entity matching information set and the edge matching information set to obtain the searching and matching result of maritime information.

[0087] The reasoning processing can be implemented by using path reasoning based on a graph structure (a multi-hop association method), deductive reasoning based on a rule (an explicit logic derivation method) or vector reasoning based on embedding (a semantic similarity completion method).

[0088] The entity matching searching of the to-be-searched information set based on the preset maritime knowledge graph database to obtain the entity matching information set includes:

[0089] S311, respectively perform text similarity calculation on name information of each entity in the preset maritime knowledge graph database and name information of each node in the node information set in the to-be-retrieved information set, to obtain a corresponding name text similarity value;

[0090] S312, extract entities and nodes with all name text similarity values greater than a preset first threshold value; and construct a node pair by using each entity and node with a name text similarity value extracted.

[0091] S313, perform matching degree calculation on an adjacency matrix of each node pair to obtain a matching degree value.

[0092] S314, construct an entity matching information set by using entity information of entities in a node pair with all matching degree values greater than a preset second threshold value.

[0093] The matching degree calculation on the adjacency matrix of each node pair to obtain the matching degree value comprises:

[0094] S3131, for each node pair, obtain an adjacency matrix corresponding to an entity of the node pair and an adjacency matrix corresponding to a node.

[0095] S3132, perform item-by-item subtraction on matrix elements of the adjacency matrix corresponding to the entity and the adjacency matrix corresponding to the node to obtain a first difference matrix.

[0096] S3133, perform eigenvalue decomposition on the first difference matrix to obtain a characteristic value set.

[0097] S3134, calculate an average value γ0 of the characteristic value set.

[0098] S3135, calculate a rank value z of the first difference matrix.

[0099] S3136, perform statistical calculation on the average value of the characteristic value set and the rank value of the first difference matrix to obtain a first matching value.

[0100] S3137, perform item-by-item division on matrix elements of the adjacency matrix corresponding to the entity and the adjacency matrix corresponding to the node to obtain a second difference matrix.

[0101] S3138, perform second matching value calculation on the second difference matrix to obtain a second matching value.

[0102] S3139, perform fusion calculation on the first matching value and the second matching value to obtain the matching degree value.

[0103] The statistical calculation expression is:

[0104] p1 = z * sin(t 1-α (n)),

[0105]

[0106] wherein n is a degree of freedom value, a is a proportion value, g max is a maximum eigenvalue of the eigenvalue set, t 1-α (n) is a 1-a quantile of a t-distribution with a degree of freedom n, and p1 is a first matching value.

[0107] The statistical calculation integrates the "overall eigenvalue difference" (eigenvalue distribution) and "structural complexity" (rank value) of the matrix into the matching value calculation by introducing the eigenvalue mean, maximum eigenvalue, matrix rank value, and t-distribution quantile. The introduction of the sine function can nonlinearly amplify significant differences, and the t-distribution quantile dynamically adapts to the fluctuation degree of the eigenvalue, so that p1 can more accurately reflect the structural similarity of the adjacency matrix, especially for the identification of subtle differences in the "neighbor relationship" of maritime entities.

[0108] The expression of the fusion calculation is:

[0109] p = D n (p1 + p2),

[0110] wherein p is a matching degree value, p2 is a second matching value, D n is an n-order Weber function, and n is determined by the degree of freedom value.

[0111] The expression of the fusion calculation uses an n-order Weber function (D n ) to fuse the first matching value (p1, reflecting structural differences) and the second matching value (p2, reflecting element proportion differences), and the nonlinear characteristics of the Weber function can dynamically adjust the weight according to the degree of freedom (n), realizing the adaptive balance of "structural features" and "element features". Compared with simple weighted summation, this method can better adapt to the diversity of maritime entity relationships (such as scenes where some entity relationships are dense and some are sparse), and improve the robustness of matching degree calculation.

[0112] The second matching value calculation on the second difference matrix includes:

[0113] Calculating the norm values of the second difference matrix to obtain the l2 norm a1 and the Frobenius norm a2;

[0114] Performing CEEMDAN transformation on each row vector of the second difference matrix to obtain corresponding transformed vectors;

[0115] Performing matching fusion calculation on all transformed vectors to obtain the second matching value.

[0116] The expression for the matching and fusion calculation is:

[0117]

[0118] Where, φ maxi and φ 0i ...

[0119] The matching fusion calculation combines the l2 norm and Frobenius norm to characterize the overall proportional features of the matrix. It decomposes the row vectors using CEEMDAN transform and extracts extrema and mean values ​​to capture subtle differences in the proportions of local elements. The combination of the arctangent function and the sine function can compress outlier interference while amplifying effective differences, enabling p2 to accurately reflect the proportional similarity between adjacent matrix elements. This is suitable for the association matching of marine entity attributes (such as ship tonnage and port throughput).

[0120] The CEEMDAN transform is a fully adaptive empirical mode decomposition transform of noise sets;

[0121] The step of retrieving edge information based on the preset maritime knowledge graph database and entity matching information set to obtain the edge matching information set includes:

[0122] For each edge information in the edge information set of the information to be retrieved, perform text vector conversion processing to obtain the corresponding edge vector;

[0123] The center vector is obtained by merging all edge vectors;

[0124] For each entity information in the entity matching information set, the edge vector corresponding to the edge information is compared with the center vector to calculate the corresponding difference value.

[0125] An edge matching information set is constructed by using all edge information in the entity matching information set whose difference value is less than a preset third threshold value.

[0126] The expression calculated by the fusion center is:

[0127] Calculate the arithmetic mean vector for all edge vectors to obtain the arithmetic mean vector and the Euclidean distance between the arithmetic mean vector and all edge vectors;

[0128] Calculate the geometric mean vector for all edge vectors to obtain the geometric mean vector and the Euclidean distance between the geometric mean vector and all edge vectors;

[0129] Calculate the median vector for all edge vectors to obtain the median vector and the Euclidean distance between the median vector and all edge vectors;

[0130] perform fusion center calculation on the arithmetic mean vector, the geometric mean vector and the median vector to obtain a center vector;

[0131] The expression of the fusion center calculation is:

[0132]

[0133] wherein z j is the jth element of the center vector, a1 j , a2 j and a3 j are the jth elements of the arithmetic mean vector, the geometric mean vector and the median vector respectively, μ1, μ2 and μ3 are the mean values of the Euclidean distances of the arithmetic mean vector, the geometric mean vector and the median vector from all edge vectors respectively, and θ1, θ2 and θ3 are the variances of the Euclidean distances of the arithmetic mean vector, the geometric mean vector and the median vector from all edge vectors respectively.

[0134] The fusion center calculation integrates the statistical characteristics of the arithmetic mean (a1), the geometric mean (a2) and the median (a3) vectors, converts the ratio of the mean value (μ) and the variance (θ) into a weight through the inverse sine function, and dynamically balances the representativeness of the three kinds of mean values (such as the mean weight of the distance distribution set is higher). This method overcomes the problem that a single mean vector is sensitive to abnormal edge vectors, so that the center vector (z_j) can better represent the overall characteristics of the edge information to be searched, and the stability of the edge information search is improved.

[0135] The expression of the vector difference calculation is:

[0136]

[0137] wherein cq is the difference value, τ j is the jth element of the edge vector of the edge information corresponding to the entity information, z j is the jth element of the center vector, and M is the total number of elements of the edge vector of the edge information.

[0138] The vector difference calculation amplifies the relative difference (such as the proportional difference and the absolute difference) of the edge vector elements through the combination of the sine function and the tangent function, and is particularly sensitive to the subtle difference close to zero. Compared with the traditional Euclidean distance, this method can more accurately capture the semantic difference (such as the subtle difference of the vectors of “short-term stop” and “long-term stop”) of the maritime edge information, and improves the accuracy of edge matching.

[0139] The edge vector of the edge information corresponding to the entity information is obtained by performing text vector conversion on the edge information;

[0140] The edge matching information set comprises a set of edge information corresponding to each entity matching information.

[0141] The first threshold value can be 0.8, the second threshold value can be 0.9 or 1.2, and the third threshold value can be 1.5.

[0142] The maritime knowledge graph database is obtained by representing maritime knowledge sets as knowledge graphs.

[0143] The graph structure-based storage is based on graph theory, represents entities as nodes, and represents relationships between entities as edges, and both nodes and edges can define attributes. This storage method can intuitively reflect the association between entities in the knowledge graph, has good performance advantages for complex relationship queries, supports horizontal and vertical expansion, and can handle large-scale data. Graph databases usually use SPARQL as a query language, and Neo4j, Dgraph, etc. are common graph databases.

[0144] The representation method of the relationship vector includes a translation hypothesis (such as TransE) method, a complex space modeling (such as RotatE) method, and a pre-trained language model fusion (such as KG-BERT) method.

[0145] The to-be-retrieved maritime text information can be input by a user from a related computer input device.

[0146] In a second aspect, the present application discloses a maritime information matching processing device, which comprises:

[0147] A memory storing executable program codes;

[0148] A processor coupled with the memory;

[0149] The processor calls the executable program codes stored in the memory to execute the maritime information matching processing method.

[0150] In a third aspect, the present application discloses a computer storage medium storing computer instructions, which are called by a computer to execute the maritime information matching processing method.

[0151] In a fourth aspect, the present application discloses an information data processing terminal for implementing the maritime information matching processing method.

[0152] The above merely illustrates the embodiments of the present application but should not be taken as limitations. Various modifications and variations can be made to the present application based on the skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the scope of claims of the present application.

Claims

1. A method for matching and processing maritime information, characterized in that, include: S1, obtain the maritime text information to be retrieved; S2, digitally represent the maritime text information to be retrieved to obtain the corresponding set of information to be retrieved; S3, based on a preset maritime knowledge graph database, performs retrieval and matching processing on the set of information to be retrieved, and obtains retrieval and matching results for maritime information, including: The preset maritime knowledge graph database includes entity information and corresponding edge information; edges are connected to entities; the entity information includes the entity's name information and the corresponding adjacency matrix; S31, based on a preset maritime knowledge graph database, perform entity matching retrieval on the set of information to be retrieved to obtain an entity matching information set; the entity matching information set includes several entity information, including: S311, For the name information of each entity in the preset maritime knowledge graph database, and the name information of each node in the node information set in the information set to be retrieved, respectively, text similarity calculation is performed to obtain the corresponding name text similarity value; S312, extract all entities and nodes whose name text similarity value is greater than a preset first threshold value; construct node pairs using the entities and nodes with each extracted name text similarity value; S313, calculate the matching degree for the adjacency matrix of each node pair to obtain the matching degree value, including: S3131, for each pair of nodes, obtain the adjacency matrix of the entity corresponding to the node pair and the adjacency matrix of the node. S3132, Subtract each element from the adjacency matrix corresponding to the entity and the adjacency matrix corresponding to the node to obtain the first difference matrix; S3133, Perform eigenvalue decomposition on the first difference matrix to obtain an eigenvalue set; S3134, calculate the average value of the eigenvalue set. ; S3135, calculate the rank value z of the first difference matrix; S3136, Statistical calculation is performed on the average value of the feature set and the rank value of the first difference matrix to obtain the first matching value; S3137, divide the adjacency matrix of the entity and the adjacency matrix of the node element by element to obtain the second difference matrix; S3138, Calculate the second matching value on the second difference matrix to obtain the second matching value; S3139, perform a fusion calculation on the first matching value and the second matching value to obtain a matching degree value; S314, using the entity information of entities in all node pairs with matching degree values ​​greater than the preset second threshold value, construct an entity matching information set; S32, based on the preset maritime knowledge graph database and entity matching information set, perform edge information retrieval to obtain the edge matching information set; S33, perform reasoning processing on the entity matching information set and the edge matching information set to obtain the retrieval matching results of maritime information.

2. The maritime information matching and processing method as described in claim 1, characterized in that, The step of digitally representing the maritime text information to be retrieved to obtain the corresponding set of information to be retrieved includes: S21, The maritime text information to be retrieved is represented by graph data to obtain a set of node information and a set of edge information; S22, Euclidean space embedding representation is performed on the node information set and edge information set to obtain the adjacency matrix set; S23, using the node information set, edge information set, and adjacency matrix set, construct the corresponding set of information to be retrieved.

3. The maritime information matching and processing method as described in claim 2, characterized in that, The process of representing the maritime text information to be retrieved using graph data to obtain a set of node information and a set of edge information includes: S211, Determine the feature set of the graph data object; the feature set includes a node name set and an edge name set; S212, based on the node name set, perform entity extraction processing on the maritime text information to be retrieved to obtain a node information set; the node information set includes the name information and attribute information of each node; S213, Based on the set of edge names, perform relation extraction processing on the maritime text information to be retrieved to obtain a set of edge information; the set of edge information includes the edge information corresponding to each node.

4. The maritime information matching and processing method as described in claim 1, characterized in that, The expression for the statistical calculation is: ， , , Where n is the number of degrees of freedom. This is a proportional value. The largest eigenvalue in the set of eigenvalues. For a t-distribution with n degrees of freedom, 1- Quantiles, p1 is the first matching value; The expression for the fusion calculation is: ， Where p is the matching degree value, and p2 is the second matching value. It is an nth-order Weber function.

5. A maritime information matching and processing device, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the maritime information matching processing method as described in any one of claims 1 to 4.

6. A computer-storable medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked by the computer, are used to execute the maritime information matching and processing method as described in any one of claims 1 to 4.

Images