A recommended method and apparatus for cable design specifications

Abstract

This invention provides a method and apparatus for recommending cable design specifications. The method includes the following steps: constructing a subnet, the subnet including a laying specification subnet and several associated subnets, the laying specification subnet including specification knowledge element nodes, the specification knowledge element nodes storing cable design specifications; constructing hyperedges, the hyperedges being used to connect the subnets; constructing a hypernetwork through the subnets and the hyperedges; parsing a user task package, mapping the keywords in the user task package to the hypernetwork to become the starting hyperedges of the associated subnets; and obtaining and pushing the specification knowledge element nodes from the laying specification subnet based on the starting hyperedges. This invention, by constructing a hypernetwork and obtaining the starting hyperedges from the user task package, and then obtaining and pushing specification knowledge element nodes from the laying specification subnet, makes finding relevant specifications for cable assembly design references more convenient, accurate, and efficient.

Description

Technical Field

[0001] This invention relates to the field of information recommendation technology, and more particularly to a method and apparatus for recommending cable design specifications. Background Technology

[0002] Aircraft cabling design must meet the requirements of various design specifications. Therefore, designers need to consult relevant laying process specifications based on the design task requirements, and combine cable type, design stage, and professional knowledge to design cable routing, thereby determining the necessary knowledge such as harness branch and clamp positions and process parameters.

[0003] Aircraft cabling companies typically compile historical cabling task records into process example documents for future design reference. Designers need to refer to these documents, based on the similarity of the aircraft model and the specific task, to deduce the necessary knowledge for the current design task.

[0004] In the existing technology, most companies rely on designers to manually review relevant case studies and specification documents, resulting in low levels of reasoning intelligence and an inability to provide accurate knowledge to solve problems; moreover, it requires a lot of time and effort, leading to low design efficiency.

[0005] In the existing technology 2, the recommendation of normative knowledge elements is achieved through keyword matching. However, this method has low accuracy, and although there are many keywords in the queried knowledge elements, they may not be useful for specific application objects and application tasks. Furthermore, it lacks reasoning ability, and normative knowledge that has a significant impact on this design but does not contain keywords cannot be queried.

[0006] In the existing third technique, recommended normative knowledge elements are achieved through deep learning. The accuracy of this method depends on the number and accuracy of the cases. For some design scenarios with few cases, its recommendation accuracy is even lower than that of keyword queries. Each retrieval requires a lot of computation time, which is inefficient. Every time a new design case is accumulated, the model needs to be retrained, which is time-consuming and labor-intensive. Summary of the Invention

[0007] This invention addresses the shortcomings of existing technologies by disclosing a method for recommending cable design specifications, thereby resolving at least the problems of inaccurate query results, the inability of the retrieved specifications to effectively guide design, and the time-consuming and laborious query process in related technologies.

[0008] The technical solution provided by this invention is as follows:

[0009] In a first aspect, this invention discloses a method for recommending cable design specifications, comprising the following steps:

[0010] Construct a subnet, which includes a laying specification subnet and several related subnets. The laying specification subnet includes specification knowledge element nodes, which store cable design specifications.

[0011] Construct a hyperedge, which is used to connect the subnets;

[0012] A hypernetwork is constructed using the subnets and the hyperedges;

[0013] Parse the user task package and map the keywords in the user task package to the hypernetwork to become the starting hyperedge of the associated subnetwork;

[0014] The specification knowledge element node is obtained and pushed from the laying specification subnet according to the starting hyperedge.

[0015] This implementation method constructs a hypernet and obtains the initial hyperedge from the user task packet, and then retrieves and pushes the specification knowledge element nodes from the laying specification subnet, making it more convenient, accurate and efficient to find relevant specifications for cable assembly design.

[0016] In some implementations, the laying specification subnet model is represented as follows: Where, N K This represents the subnet of the laying specifications. Representing the aforementioned standardized knowledge element nodes; also includes the following steps:

[0017] The canonical knowledge element nodes are calculated by reconstructing the vector space model. and The degree of correlation between them, among which, and These are any two specification knowledge element nodes in the laying specification subnet.

[0018] This implementation method combines cable assembly design technology with the construction of subnets, which makes it more accurate to find relevant specifications for cable assembly design.

[0019] In some implementations, the associated subnet includes at least a laying object subnet, a laying task subnet, and a professional domain subnet; the hyperedge is formally represented as: in, Represents the nodes in the aforementioned laying specification subnet. Represents the nodes in the subnet of the laying object. This represents a node in the aforementioned laying task subnet. The nodes represent the nodes in the professional field subnet, h represents the number of nodes in the laying specification subnet, e represents the number of nodes in the laying object subnet, f represents the number of nodes in the laying task subnet, and g represents the number of nodes in the professional field subnet.

[0020] This implementation method, by combining cable assembly design process with the construction of super-edges, can make the search for relevant specifications for cable assembly design more accurate and improve the accuracy rate. It can also search for specification knowledge that has a significant impact on the design but does not contain keywords.

[0021] In some implementations, the degree of association between the subnet nodes is represented as the hyperedge association degree, where the X-layer nodes... and Y layer nodes The degree of hyperedge correlation formed is represented as The calculation formula is as follows:

[0022] in, Indicates the nodes in the subnet A subset of nodes formed by directly connected subnet nodes; Indicates passing through nodes The number of superedges between; Represents subnet nodes and The strength of the association.

[0023] This implementation method calculates the degree of association between subnet nodes, enabling keyword query methods and deep learning methods in the technical background to learn from past cases. It also solves the problems of keyword query methods lacking reasoning ability and having poor accuracy in recommended knowledge, as well as the problems of deep learning being time-consuming and laborious.

[0024] In some implementations, parsing the user task packet and mapping the keywords in the user task packet to the hypernet to become the starting hyperedge of the associated subnet specifically includes:

[0025] Entity recognition is performed on the user task package, the entity includes a first entity in the domain dictionary and a second entity in the task package, and the user task package contains several keywords representing laying semantic information;

[0026] Calculate the similarity between the first entity and the second entity;

[0027] Determine whether the similarity between the first entity and the second entity exceeds a preset value;

[0028] When the similarity between the first entity and the second entity exceeds a preset value, the first entity in the task package is replaced with the second entity in the domain dictionary;

[0029] When it is determined that the similarity between the first entity and the second entity does not exceed a preset value, the entity in the task package is replaced with the entity with the highest similarity in the domain dictionary;

[0030] The keywords in the user task package are mapped to the hypernet to become the starting hyperedge, which is a hyperedge connecting the laying object subnet, the laying task subnet, and the professional domain subnet.

[0031] This implementation method addresses the common characteristics of the laying requirements section in the task package and the cable laying process specifications and process examples for complex cables. By using the nested named entity method to parse the laying process design task package, it maps it to the hypernet to become the starting hyperedge. This resolves the design task into a hyperedge that only involves the first three layers of subnets, making it more convenient, accurate, and efficient to find relevant specifications for cable assembly design.

[0032] In some implementations, the step of obtaining and pushing the specification knowledge element node from the laying specification subnet based on the starting hyperedge specifically includes:

[0033] Based on the supernet Bayesian algorithm and the initial hyperedge, calculate the probability that the specification knowledge element node in the laying specification subnet is recommended;

[0034] The recommended normative knowledge element nodes are sorted according to the probability.

[0035] The standardized knowledge element nodes are pushed and sorted.

[0036] This implementation method calculates the probability of the obtained standard knowledge element nodes, sorts and recommends nodes according to the calculated probabilities, making the search for relevant standards for cable assembly design references more accurate.

[0037] In some implementations, calculating the probability that the specification knowledge element node in the laying specification subnet is recommended based on the supernet Bayesian algorithm and the initial hyperedge specifically includes:

[0038] choose Adjacent nodes in their respective subnets in and Represents the nodes in the subnet of the laying object. and This represents a node in the aforementioned laying task subnet. and Represents a node in the subnet of the aforementioned professional field;

[0039] calculate and The degree of correlation between them and The degree of correlation between them and The degree of correlation between them;

[0040] Calculate according to the formula The corresponding total correlation ω R The formula is as follows:

[0041]

[0042] In the formula, R is the feasible region. Indicating the subnet of the laying object The degree of correlation between nodes Indicating the laying task subnet The degree of correlation between nodes Indicating the professional field subnet The degree of association between nodes;

[0043] Traverse all adjacent nodes and calculate the association probability between the hyperedge of the laying specification subnet node and the laying object subnet, the laying task subnet, and the professional domain subnet. The value of is given by the following formula:

[0044]

[0045] In the formula, Indicates the process The number of hyperedges, Indicates the process The number of superedges of a node.

[0046] Calculate the initial hyperedge probability of nodes in the laying object subnet, the laying task subnet, and the professional domain subnet. The value of is given by the following formula:

[0047]

[0048] In the formula, Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of superedges of a node.

[0049] Calculate the probability of the occurrence of a hyperedge containing nodes of the laying specification subnet according to the formula. The value of is given by the following formula:

[0050]

[0051] In the formula, Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. yes The node is a neighboring node in the standard layer.

[0052] Computational Recommendation Standard Knowledge Elements probability The formula is as follows:

[0053]

[0054] This implementation method calculates the recommendation probability based on the supernet Bayesian algorithm, making it more convenient, accurate, and efficient to find relevant specifications for cable assembly design.

[0055] In some implementations, new cable design specifications are added to the hypernet by adding new nodes and hyperedges.

[0056] This implementation addresses the problem of low efficiency in deep learning, where each retrieval requires significant computation time, by adding new cable design specifications to the hypernetwork through new nodes and hyperedges. Simultaneously, it makes the recommendation of specification knowledge elements more accurate and faster.

[0057] According to a second aspect of the present invention, a cable design specification recommendation device is disclosed, comprising:

[0058] A construction module is used to construct subnets, which include a laying specification subnet and several related subnets. The laying specification subnet includes specification knowledge element nodes, which store cable design specifications.

[0059] The construction module is also used to construct a hyperedge, which is used to connect the subnet;

[0060] The construction module is also used to construct a hypernetwork through the subnet and the hyperedge;

[0061] The parsing module is used to parse the user task package and map the keywords in the user task package to the hypernet to become the starting hyperedge of the associated subnet;

[0062] The push module is used to obtain and push the specification knowledge element node from the laying specification subnet according to the starting hyperedge.

[0063] In some implementations, a cable design specification recommendation device further includes an addition module for adding new cable design specifications to the hypernet by adding new nodes and hyperedges.

[0064] Compared with the prior art, the present invention has at least the following beneficial effects:

[0065] 1. By constructing a hypernet and parsing the initial hyperedge obtained from the user task package, the specification knowledge element nodes are obtained and pushed from the laying specification subnet, making it more convenient, accurate and efficient to find relevant specifications for cable assembly design references;

[0066] 2. By combining cable assembly design process with the construction of super-edge, it is possible to make the search for relevant specifications for cable assembly design more accurate and improve the accuracy rate. It is also possible to search for specification knowledge that has a significant impact on the design but does not contain keywords.

[0067] 3. By calculating the degree of association between subnet nodes, the keyword query method and deep learning method in the technical background can learn from past cases, while solving the problems of keyword query method lacking reasoning ability and poor recommendation knowledge accuracy, and the problem of deep learning being time-consuming and laborious.

[0068] 4. By adding new cable design specifications to the hypernetwork through new nodes and hyperedges, the problem of low efficiency caused by the large amount of computation time required for each retrieval in deep learning is solved, while making the recommendation of specification knowledge elements more accurate and faster. Attached Figure Description

[0069] The preferred embodiments will now be described in a clear and easy-to-understand manner, with reference to the accompanying drawings, to further explain the above-mentioned characteristics, technical features, advantages, and implementation methods of this solution.

[0070] Figure 1 This is a flowchart illustrating an embodiment of the cable design specification recommendation method provided by the present invention;

[0071] Figure 2 This is a flowchart illustrating another embodiment of the cable design specification recommendation method provided by the present invention;

[0072] Figure 3 This is a flowchart illustrating another embodiment of the cable design specification recommendation method provided by the present invention;

[0073] Figure 4 This is a schematic diagram of an embodiment of the cable design specification recommendation device provided by the present invention;

[0074] Figure 5 This is a schematic diagram of another embodiment of the cable design specification recommendation device provided by the present invention;

[0075] Figure 6 This is a schematic diagram of another embodiment of the cable design specification recommendation device provided by the present invention;

[0076] Figure 7This is a schematic diagram of another embodiment of the cable design specification recommendation device provided by the present invention;

[0077] Figure 8 This is a schematic diagram of a hypernetwork according to an embodiment of the present invention;

[0078] Figure 9 This is a flowchart of a knowledge push process for cable laying tasks according to an embodiment of the present invention;

[0079] Figure 10 This is a flowchart of a task package parsing process according to an embodiment of the present invention.

[0080] Explanation of icon numbers:

[0081] The module includes a construction module 10, a parsing module 20, a push module 30, an addition module 40, an entity recognition module 21, a calculation module 22, a judgment module 23, a replacement module 24, a mapping module 25, a calculation module 31, a sorting module 32, and a push sub-module 33. Detailed Implementation

[0082] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings and other implementation methods can be obtained based on these drawings without any creative effort.

[0083] To keep the drawings concise, only the parts relevant to the invention are shown schematically in each figure, and they do not represent the actual structure of the product. Furthermore, for ease of understanding, in some figures, only one of components with the same structure or function is shown schematically, or only one is labeled. In this document, "one" can mean not only "only one" but also "more than one".

[0084] Standardized knowledge element: In an enterprise knowledge base, the smallest independent knowledge unit that can be freely accessed, organized, retrieved, and utilized from standardized documents is called a standardized knowledge element, which is generally a paragraph in a standardized document.

[0085] Subnet: A subnet is a network that has no nesting relationship.

[0086] Hypernet: A hypernet is a heterogeneous network composed of heterogeneous nodes and heterogeneous relationships. In this paper, it is a network composed of multiple nested networks.

[0087] Hyperedge: A hyperedge is a link connecting subnets, connecting them together through subnet nodes.

[0088] Object subnetting: An object subnet is a complex network composed of cable laying objects and their attributes. The object subnetting model is represented as N. O =(V O E O ), where V O It is a node, E O It refers to the relationships between subnet nodes of the laying object. This refers to the objects being laid, including cables, connectors, and auxiliary materials. Represents object properties, in Establishing connections between different laying objects; the greater the correlation, the greater the likelihood that a certain laying standard applies to the two laying objects. This establishes a connection between the laying object and its attributes, with a correlation value of 1. Implement association relationships between attributes.

[0089] Relationship Corresponding correlation calculate:

[0090] Two types are defined in the ontology model of the laying object. Relationships, composed_of, and effect, assumptions and If these two relationships exist, then and The formula for calculating the correlation between them is:

[0091]

[0092]

[0093] in, Indicates that the laying objects appear simultaneously. and Case studies of laying tasks, This indicates that there are simultaneously laying objects. and And cases where the relationship between objects is composed_of.

[0094] Relationship Corresponding correlation calculate:

[0095] The size reflects the likelihood that two attributes will appear in the same document knowledge element, if object attributes are set. and attributes If they both appear in the knowledge elements of a specification document, the degree of correlation between them can be calculated using the Jaccard coefficient:

[0096]

[0097] In the formula For all contained nodes The set of normative knowledge elements, For all contained nodes The set of normative knowledge elements.

[0098] Object subnet association path (Path) O For subnet nodes of objects that do not have a direct relationship: and Assuming there exists a path in the object subnet that can achieve... and The path is defined as an associated path: The set of direct associations corresponding to this set is: The correlation degree corresponding to this path is calculated as follows:

[0099]

[0100] Assuming the object subnet has two nodes and If there are n related paths, the path with the highest correlation score is taken as the correlation score between the two nodes, as shown in the following formula:

[0101]

[0102] In one embodiment, refer to the accompanying drawings. Figure 1 And attached diagram Figure 8 The cable design specification recommendation method provided by this invention includes the following steps:

[0103] S110, Construct a subnet. The subnet includes a laying specification subnet and several related subnets. The laying specification subnet includes specification knowledge element nodes, which store cable design specifications.

[0104] Specifically, a laying specification subnet and several related subnets are constructed. There must be at least one related subnet. In actual use, the subnets can be expanded or deleted according to the needs, but there must be a subnet containing specification knowledge elements, that is, a laying specification subnet.

[0105] Taking the construction of a laying specification subnet as an example, the laying specification subnet is represented as N. K =(V K V K To represent the knowledge elements of the laying specifications, it is necessary to calculate the correlation degree W between the knowledge elements of each document. KThis application uses a semantically weighted similarity calculation method to measure the relevance between different knowledge texts. For two standardized representations of standardized knowledge elements... and The formula for calculating the text similarity between the two is as follows:

[0106]

[0107] Furthermore, the specification knowledge element nodes also store cable design case information.

[0108] Furthermore, a four-layer subnet can be constructed, with several related subnets including the laying object subnet, the laying task subnet, and the professional domain subnet.

[0109] S120, constructs a superedge, which is used to connect subnets;

[0110] Specifically, a hyperedge (HE) is a link connecting subnets. It connects subnets together through subnet nodes. Taking the above four-layer subnet as an example, the nodes inside the dashed lines of the ellipse belong to one case. Therefore, the edges connecting these nodes (all the dashed lines enclosed by the ellipse) constitute a direct hyperedge. A hyperedge can be formally represented as: in, Represents the nodes in the subnet of the laying specifications. Represents the nodes in the subnet of the laying object. This represents a node in the subnet for the laying task. The nodes represent the professional domain subnet, h represents the number of nodes in the laying specification subnet, e represents the number of nodes in the laying object subnet, f represents the number of nodes in the laying task subnet, and g represents the number of nodes in the professional domain subnet.

[0111] S130 constructs a hypernetwork through subnets and hyperedges;

[0112] Specifically, through V O V T V F V K The super-edge formed by these elements connects the laying object subnet, the laying task subnet, the professional domain subnet, and the laying specification subnet, thus forming a super network.

[0113] S140, parse the user task packet and map the keywords in the user task packet to the hypernet to become the starting hyperedge of the associated subnet;

[0114] Specifically, the actual cable laying process design task is allocated in the form of task packages. A task package is a semi-structured text containing a large number of keywords that represent the semantic information of the laying process. A task package can be formally expressed as:

[0115] Package=<Object,Task,Feild>

[0116] Here, "Object" represents the set of vocabulary related to laying objects, which are generally cable components or cable components and their attributes. The more detailed the object, the more accurate the final knowledge obtained. "Task" represents the set of vocabulary related to laying tasks. "Feld" represents the set of vocabulary related to the professional field. The user task package is parsed, and the keywords in the user task package are mapped to a hypernet to become the starting hyperedge of the related subnet. In other words, the design task is parsed into a hyperedge that only involves the first three layers of subnets.

[0117] S150: Obtain and push specification knowledge element nodes from the laying specification subnet based on the starting superedge.

[0118] This embodiment constructs a hypernet and parses the initial hyperedge obtained from the user task package to obtain and push specification knowledge element nodes from the laying specification subnet, making it more convenient, accurate and efficient to find relevant specifications for cable assembly design references.

[0119] In one embodiment, the laying specification subnet model is represented as follows: Where, N K Represents the standard subnetting installation. Represents a standardized knowledge element node; it also includes steps:

[0120] Computation of normative knowledge element nodes by reconstructing the vector space model and The degree of correlation between them, among which, and These are any two specification knowledge element nodes in the laying specification subnet.

[0121] Specifically, the calculation steps are as follows:

[0122] Step 1: Calculate the set and Sim(T) is the feature similarity between each feature item in the dataset. pi ,T qj Construct the vector space matrix A:

[0123]

[0124] Step 2: Select the maximum value in A, max(Sim(T) pi ,T qj If max(Sim(T)) pi ,T qj If the value is greater than the preset value δ, then the characteristic item T will be... pi and T qj Merge into feature term TT r and TT rIt is added as an element to the new set TT, and simultaneously removed from the set. and Remove feature term T from the middle pi and T qj ;

[0125] Step 3: Repeat steps 1 and 2 above until A is empty or max(Sim(T) is empty. pi ,T qj ))<δ;

[0126] Step 4: Set and The remaining feature terms, together with the empty element, form feature term classes, which are then added to TT to obtain a new feature term set TT = {tt1, tt2, ..., tt}. r}, consider TT as an r-dimensional coordinate system, where each dimension represents a feature class;

[0127] Step 5: In the r-dimensional coordinate system, based on the new feature term set TT, obtain... and The feature vectors are respectively The similarity calculation formula based on the reconstructed vector space model is as follows:

[0128]

[0129] In the formula, 1≤k≤r; α k ω is the correction factor. pk ω qk For feature term class tt k In the text and The weights in the equation.

[0130] This embodiment combines cable assembly design technology with subnet construction, which makes it more accurate to find relevant specifications for cable assembly design.

[0131] In one embodiment, based on the above embodiments, the degree of association between subnet nodes is represented as the hyperedge association degree, where X-layer nodes... and Y layer nodes The degree of hyperedge correlation formed is represented by E. XY The calculation formula is as follows:

[0132] in, Indicates nodes in a subnet A subset of nodes formed by directly connected subnet nodes; Indicates passing through nodes The number of superedges between; Represents subnet nodes and The strength of the association.

[0133] Specifically, the hyperedge correlation degree between the laying object subnet and the laying task subnet. Indicates the subnet node of the laying object and laying task subnet nodes The degree of correlation between them is calculated using the following formula:

[0134]

[0135] in Indicates the nodes in the cable subnet A subset of nodes formed by directly connected object subnet nodes; Indicates passing through nodes The number of superedges between; Represents cable subnet nodes and The strength of the association.

[0136] A higher correlation value indicates a higher correlation between the laying object and the laying task, and also reflects the importance of the specification knowledge containing the laying object to the laying task.

[0137] This embodiment calculates the degree of association between subnet nodes, enabling keyword query methods and deep learning methods in the technical background to learn from past cases. It also solves the problems of keyword query methods lacking reasoning ability and having poor accuracy in recommended knowledge, as well as the problems of deep learning being time-consuming and laborious.

[0138] In one embodiment, refer to the accompanying drawings. Figure 2 Based on the above embodiments, step S140 specifically includes:

[0139] S210, perform entity recognition on the user task package. The entities include the first entity in the domain dictionary and the second entity in the task package. The user task package contains several keywords that represent laying semantic information.

[0140] Specifically, the nested named entity method is used to parse the laying process design task package. The identified entities differ from the entity nodes in the constructed laying process specification knowledge hypernetwork, making a complete match impossible. Therefore, entity linking technology is needed to link the entities in the task package with those in the graph. Combined with the constructed evaluation object dictionary, the BERT model is used to convert the identified laying entities into vector forms with semantic information.

[0141] S220, Calculate the similarity between the first entity and the second entity;

[0142] Specifically, similarity is calculated using Euclidean distance; the closer the semantics, the smaller the corresponding Euclidean distance.

[0143] S230, determine whether the similarity between the first entity and the second entity exceeds a preset value;

[0144] S240, when the similarity between the first entity and the second entity exceeds a preset value, the first entity in the task package is replaced with the second entity in the domain dictionary;

[0145] S250, when it is determined that the similarity between the first entity and the second entity does not exceed the preset value, the entity in the task package is replaced with the entity with the highest similarity in the domain dictionary;

[0146] Specifically, by combining the similarity results between entities in the domain dictionary and entities in the task package, it is determined whether two entities are equal. If the similarity exceeds a preset value, the entity in the task package is directly replaced with the entity in the domain dictionary; if they are not equal, the entity with the highest similarity in the domain dictionary is selected for replacement.

[0147] S260 maps keywords in the user task package to a hypernet to become the starting hyperedge. The starting hyperedge is the hyperedge that connects the laying object subnet, the laying task subnet, and the professional domain subnet.

[0148] Specifically, based on the entity concept in the hypernetwork, task packages in natural language form are mapped to the knowledge hypernetwork, formally represented as the initial edge HE. TP :

[0149]

[0150] In the formula, V i O For nodes in the object subnet, It is a process subnet node. It is a subnet node of the domain.

[0151] This embodiment takes into account that the laying requirements in the task package and the complex cable laying process specifications and process examples have the same characteristics. By using the nested named entity method to parse the laying process design task package, and then mapping it to the hypernet as the starting hyperedge, the design task is parsed into a hyperedge that only involves the first three layers of subnets, making it more convenient, accurate and efficient to find relevant specifications for cable assembly design.

[0152] In one embodiment, refer to the accompanying drawings. Figure 10 Parsing a user task package includes the following steps:

[0153] Select a task package;

[0154] Entity recognition based on improved machine reading comprehension algorithms;

[0155] Calculate the semantic similarity between identified entities and ontology elements;

[0156] Determine if the similarity exceeds a threshold;

[0157] When the similarity exceeds the threshold, the ontology element is replaced with the identified entity.

[0158] If the similarity does not exceed the threshold, replace the ontology element with a similar entity.

[0159] Specifically, based on the constructed evaluation object dictionary, the BERT model is used to convert the identified laying entities into vector forms with semantic information. Then, Euclidean distance is used to calculate the similarity between the two; the closer the two are semantically, the smaller the corresponding Euclidean distance. Next, the similarity results between entities in the domain dictionary and entities in the task package are combined to determine whether the two entities are equal. If the similarity exceeds a threshold, the entity in the task package is directly replaced with an entity in the domain dictionary; if they are not equal, the entity with the highest similarity in the domain dictionary is selected for replacement.

[0160] In one embodiment, refer to the accompanying drawings. Figure 3 And attached diagram Figure 9 S150 specifically includes:

[0161] S310, Based on the supernet Bayesian algorithm and the initial hyperedge, calculate the probability that the standard knowledge element node in the laying standard subnet is recommended;

[0162] S320, sort the recommended normative knowledge element nodes according to probability;

[0163] S330, push and sort the standardized knowledge element nodes.

[0164] Specifically, the dashed box represents the hypernet model, the solid line represents the initial hyperedge of the design task parsing, and the ellipse represents the standardized knowledge elements with high recommendation probabilities calculated according to the formula. After the task package is parsed into the initial hyperedge, the probability of each node in the standardized subnet being recommended under the influence of the initial hyperedge is calculated based on the hypernet Bayesian algorithm; then, the recommended knowledge elements are sorted according to the probability; finally, the deployed knowledge elements are pushed to the process designers.

[0165] This embodiment calculates the probability of the obtained standard knowledge element nodes, sorts and recommends nodes according to the calculated probabilities, making the search for relevant standards for cable assembly design references more accurate.

[0166] In one embodiment, calculating the probability that a specification knowledge element node in the laying specification subnet is recommended, based on the supernet Bayesian algorithm and the initial hyperedge, specifically includes:

[0167] choose Adjacent nodes in their respective subnets in and Represents the nodes in the subnet of the laying object. and This represents a node in the subnet for the laying task. and Represents nodes in a subnet of a specific professional field;

[0168] calculate and The degree of correlation between them and The degree of correlation between them and The degree of correlation between them;

[0169] Assuming the target subnet has two nodes and If there are n related paths, the path with the highest correlation score is taken as the correlation score between the two nodes, as shown in the following formula:

[0170]

[0171] Similarly, calculation and The formula for the degree of correlation between them is:

[0172]

[0173] calculate and The formula for the degree of correlation between them is:

[0174]

[0175] Calculate according to the formula The corresponding total correlation ω R The formula is as follows:

[0176]

[0177] In the formula, R is the feasible region. Indicates the subnet of the laying object The degree of correlation between nodes Indicates the laying task subnet The degree of correlation between nodes Indicating professional field subnets The degree of association between nodes;

[0178] Traverse all adjacent nodes and calculate the association probability between the hyperedges of the laying specification subnet nodes and the laying object subnet, the laying task subnet, and the professional domain subnet. The value of is given by the following formula:

[0179]

[0180] In the formula, Indicates the process The number of hyperedges, Indicates the process The number of superedges of a node.

[0181] Calculate the initial hyperedge probability of nodes in the laying object subnet, laying task subnet, and professional domain subnet. The value of is given by the following formula:

[0182]

[0183] In the formula, Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of superedges of a node.

[0184] Calculate the probability of the occurrence of a hyperedge containing nodes of the laying specification subnet according to the formula. The value of is given by the following formula:

[0185]

[0186] In the formula, Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. yes The node is a neighboring node in the standard layer.

[0187] Computational Recommendation Standard Knowledge Elements probability The formula is as follows:

[0188]

[0189] This embodiment calculates the recommendation probability based on the supernet Bayesian algorithm, making it more convenient, accurate, and efficient to find relevant specifications for cable assembly design.

[0190] In one embodiment, new cable design specifications are added to the hypernet by adding nodes and hyperedges.

[0191] Specifically, when new cases are accumulated, it is not necessary to retrain the entire model as in deep learning; simply adding nodes and hyperedges to the hypernetwork is sufficient.

[0192] This embodiment addresses the problem of low efficiency in deep learning, where each retrieval requires significant computation time, by adding new cable design specifications to the hypernetwork through new nodes and hyperedges. This also makes the recommendation of specification knowledge elements more accurate and faster.

[0193] In one embodiment, based on the same technical concept, refer to the accompanying drawings. Figure 4 The cable design specification recommendation device provided in this application includes: a construction module 10, a parsing module 20, and a push judgment module 30.

[0194] Module 10 is used to construct subnets, which include a laying specification subnet and several related subnets. The laying specification subnet includes specification knowledge element nodes, which store cable design specifications.

[0195] Specifically, module 10 constructs a laying specification subnet and several related subnets. There must be at least one related subnet. In actual use, the subnets can be expanded or deleted according to the needs, but there must be a subnet containing specification knowledge elements, that is, a laying specification subnet.

[0196] Taking the construction of a laying specification subnet as an example, the laying specification subnet is represented as N. K =(V K V K To represent the knowledge elements of the laying specifications, it is necessary to calculate the correlation degree W between the knowledge elements of each document. K This application uses a semantically weighted similarity calculation method to measure the relevance between different knowledge texts. For two standardized representations of standardized knowledge elements... and The formula for calculating the text similarity between the two is as follows:

[0197]

[0198] Furthermore, the specification knowledge element nodes also store cable design case information.

[0199] Furthermore, a four-layer subnet can be constructed, with several related subnets including the laying object subnet, the laying task subnet, and the professional domain subnet.

[0200] Module 10 is also used to construct hyperedges, which are used to connect subnets;

[0201] Specifically, a hyperedge is a link connecting subnets. It connects subnets together through subnet nodes. Taking the above four-layer subnet as an example, the hyperedge can be formally represented as: in, Represents the nodes in the subnet of the laying specifications. Represents the nodes in the subnet of the laying object. This represents a node in the subnet for the laying task. The nodes represent the professional domain subnet, h represents the number of nodes in the laying specification subnet, e represents the number of nodes in the laying object subnet, f represents the number of nodes in the laying task subnet, and g represents the number of nodes in the professional domain subnet.

[0202] Module 10 is also used to construct hypernetworks using subnets and hyperedges;

[0203] Specifically, through V O V T V F V K The super-edge formed by these elements connects the laying object subnet, the laying task subnet, the professional domain subnet, and the laying specification subnet, thus forming a super network.

[0204] Parsing module 20 is used to parse user task packets and map the keywords in the user task packets to the hypernet to become the starting hyperedge of the associated subnet;

[0205] Specifically, the actual cable laying process design task is allocated in the form of task packages. A task package is a semi-structured text containing a large number of keywords that represent the semantic information of the laying process. A task package can be formally expressed as:

[0206] Package=<Object,Task,Feild>

[0207] Here, Object is the set of vocabulary related to laying objects. Laying objects are generally cable parts or cable parts and their attributes. The more detailed the object, the more accurate the knowledge obtained. Task is the set of vocabulary related to laying tasks. Field is the set of vocabulary related to the professional field. Parsing module 20 parses the user task package and maps the keywords in the user task package to the hypernet to become the starting hyperedge of the related subnet. That is, it parses the design task into a hyperedge that only involves the first three layers of subnets.

[0208] The push module 30 retrieves and pushes specification knowledge element nodes from the laying specification subnet based on the starting superedge.

[0209] In this embodiment, the construction module 10 constructs a hypernet and the parsing module 20 parses the starting hyperedge obtained from the user task package. The push module 30 obtains and pushes the specification knowledge element nodes from the laying specification subnet, making it more convenient, accurate and efficient to find relevant specifications for cable assembly design references.

[0210] In one embodiment, refer to the accompanying drawings. Figure 5 The cable design specification recommendation device provided in this application also includes: an addition module 40 for adding new cable design specifications to the hypernet by adding new nodes and hyperedges.

[0211] Specifically, when new cases are accumulated, it is not necessary to retrain the entire model as in deep learning. Instead, simply add nodes and hyperedges to the hypernetwork by adding module 40.

[0212] This embodiment adds new cable design specifications to the hypernetwork by adding new nodes and hyperedges in module 40, which solves the problem that deep learning requires a lot of time to compute and is inefficient for each retrieval. At the same time, it makes the recommendation of specification knowledge elements more accurate and faster.

[0213] In one embodiment, refer to the accompanying drawings. Figure 6 The cable design specification recommendation device provided in this application includes a parsing module 20 comprising: an entity recognition module 21, a calculation module 22, a judgment module 23, a replacement module 24, and a mapping module 25.

[0214] Entity recognition module 21 is used to perform entity recognition on user task package. The entities include the first entity in the domain dictionary and the second entity in the task package. The user task package contains several keywords that represent laying semantic information.

[0215] Specifically, the entity recognition module 21 uses the nested named entity method to parse the laying process design task package. The entities obtained after recognition differ from the entity nodes in the constructed laying process specification knowledge hypernetwork, and cannot be completely matched. Therefore, entity linking technology is needed to link the entities in the task package with the entities in the graph. Combined with the constructed evaluation object dictionary, the BERT model is used to convert the identified laying entities into vector forms with semantic information.

[0216] The calculation module 22 is used to calculate the similarity between the first entity and the second entity;

[0217] Specifically, the calculation module 22 uses Euclidean distance to calculate similarity; the closer the semantics, the smaller the corresponding Euclidean distance.

[0218] The judgment module 23 is used to determine whether the similarity between the first entity and the second entity exceeds a preset value;

[0219] Replacement module 24 is used to replace the first entity in the task package with the second entity in the domain dictionary when the similarity between the first entity and the second entity exceeds a preset value.

[0220] The replacement module 24 is also used to replace the entity in the task package with the entity with the highest similarity in the domain dictionary when it is determined that the similarity between the first entity and the second entity does not exceed a preset value.

[0221] Specifically, based on the similarity results between entities in the domain dictionary and entities in the task package, the judgment module 23 determines whether two entities are equal. If the similarity exceeds a preset value, the replacement module 24 directly replaces the entity in the task package with the entity in the domain dictionary; if they are not equal, the replacement module 24 selects the entity with the highest similarity in the domain dictionary for replacement.

[0222] The mapping module 25 is used to map keywords in the user task package to the hypernet as the starting hyperedge. The starting hyperedge is the hyperedge that connects the laying object subnet, the laying task subnet, and the professional domain subnet.

[0223] Specifically, based on the entity concept in the hypernetwork, the mapping module 25 maps the task package in natural language form to the knowledge hypernetwork, formally represented as the starting edge HE. TP :

[0224]

[0225] In the formula, V i O For nodes in the object subnet, It is a process subnet node. It is a subnet node of the domain.

[0226] This embodiment takes into account that the laying requirements in the task package and the complex cable laying process specifications and process examples have the same characteristics. By using the nested named entity method to parse the laying process design task package, and then mapping it to the hypernet as the starting hyperedge, the design task is parsed into a hyperedge that only involves the first three layers of subnets, making it more convenient, accurate and efficient to find relevant specifications for cable assembly design.

[0227] In one embodiment, refer to the accompanying drawings. Figure 7 The push module 30 includes: a calculation module 31, a sorting module 32, and a push submodule 33.

[0228] Calculation module 31 calculates the probability that the standard knowledge element node in the laying standard subnet is recommended based on the supernet Bayesian algorithm and the initial hyperedge;

[0229] The sorting module 32 sorts the recommended normative knowledge element nodes according to probability;

[0230] Sub-module 33 pushes standardized knowledge element nodes after sorting.

[0231] Specifically, after the task package is parsed into an initial hyperedge by the parsing module 20, the calculation module 31 calculates the probability of each node in the laying specification subnet being recommended under the influence of the initial hyperedge using the hypernet Bayesian algorithm; then the sorting module 32 sorts the recommended knowledge elements according to the probability; finally, the push submodule 33 pushes the laying knowledge elements to the process designers.

[0232] In this embodiment, the probability of the obtained standard knowledge element nodes is calculated by the calculation module 31, the sorting module 32 sorts them according to the probability calculated by the calculation module 31, and the push sub-module 33 recommends them according to the standard knowledge element nodes sorted by the sorting module 32, so that the search for relevant standards for cable assembly design reference is more accurate.

[0233] In one embodiment, this application also discloses an electronic device, including a memory and a processor. The memory is used to store a running program, and the processor is used to execute the running program stored in the memory to implement the operations performed by the cable design specification recommended method of the above embodiments. The processor may be a CPU, a controller, a microcontroller, a microprocessor, or other data processing chip.

[0234] In one embodiment, the present invention provides a computer-readable storage medium storing a control program, which, when executed by a processor, is used to implement the cable design specification recommendation method as described above. The technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various method embodiments of the present invention. The computer-readable storage medium includes various media capable of carrying computer program code, such as USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), and random access memory (RAM).

[0235] Since the cable design specification recommendation program adopts all the technical solutions of all the aforementioned embodiments when it is executed by the processor, it has at least all the beneficial effects brought about by all the technical solutions of all the aforementioned embodiments, which will not be repeated here.

[0236] It should be noted that the above embodiments can be freely combined as needed. The above description is only a preferred embodiment of the present invention. It should be pointed out that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A cable design specification recommendation method characterized by comprising: Including the following steps: Construct a subnet, which includes a laying specification subnet and several related subnets. The laying specification subnet includes specification knowledge element nodes, which store cable design specifications. The related subnets include at least a laying object subnet, a laying task subnet, and a professional domain subnet. Construct a hyperedge, which is used to connect the subnets; A hypernetwork is constructed using the subnets and the hyperedges; Parse the user task package and map the keywords in the user task package to the hypernetwork to become the starting hyperedge of the associated subnetwork; The specification knowledge element node is obtained from the laying specification subnet based on the starting hyperedge and pushed accordingly; The step of parsing the user task package and mapping the keywords in the user task package to the hypernetwork to become the starting hyperedge of the associated subnet includes: performing entity recognition on the user task package, wherein the entity includes a first entity in the domain dictionary and a second entity in the task package, and the user task package contains several keywords representing laying semantic information; calculating the similarity between the first entity and the second entity; determining whether the similarity between the first entity and the second entity exceeds a preset value; when the similarity between the first entity and the second entity exceeds the preset value, replacing the second entity in the task package with the first entity in the domain dictionary; when the similarity between the first entity and the second entity does not exceed the preset value, replacing the entity in the task package with the entity with the highest similarity in the domain dictionary; and mapping the keywords in the user task package to the hypernetwork to become the starting hyperedge, wherein the starting hyperedge is a hyperedge connecting the laying object subnet, the laying task subnet, and the professional domain subnet.

2. The method for recommending cable design specifications according to claim 1, characterized in that, The laying specification subnet model is represented as NK = ( ), where NK represents the laying specification subnet, Representing the aforementioned standardized knowledge element nodes; also includes the following steps: calculating the correlation degree between the normative knowledge element nodes by reconstructing a vector space model and wherein, and are any two normative knowledge element nodes in the laying normative subnetwork respectively.

3. The cable design specification recommendation method according to claim 1, characterized in that, The hyperedge is formally represented as: HE = { , ,..., , ,..., , ,..., , ,..., },in, Represents the nodes in the aforementioned laying specification subnet. Represents the nodes in the subnet of the laying object. This represents a node in the aforementioned laying task subnet. The nodes represent the nodes in the professional field subnet, h represents the number of nodes in the laying specification subnet, e represents the number of nodes in the laying object subnet, f represents the number of nodes in the laying task subnet, and g represents the number of nodes in the professional field subnet.

4. The cable design specification recommendation method according to claim 1, characterized in that, The degree of association between the subnet nodes is represented by the hyperedge association degree, where the X-layer nodes and Y layer nodes The degree of hyperedge correlation formed is represented as ( , The calculation formula is as follows: ( , ) = D ( , )+ ; in, Indicates the nodes in the subnet A subset of nodes formed by directly connected subnet nodes; D ( , () indicates the point passed through , The number of superedges between; Represents subnet nodes and The strength of the association.

5. The method for recommending cable design specifications according to claim 1, characterized in that, The step of obtaining and pushing the specification knowledge element node from the laying specification subnet according to the starting hyperedge specifically includes: Based on the supernet Bayesian algorithm and the initial hyperedge, calculate the probability that the specification knowledge element node in the laying specification subnet is recommended; The recommended normative knowledge element nodes are sorted according to the probability. The standardized knowledge element nodes are pushed and sorted.

6. The method for recommending cable design specifications according to claim 5, characterized in that, The step of calculating the probability of the recommended specification knowledge element node in the laying specification subnet based on the supernet Bayesian algorithm and the initial hyperedge specifically includes: choose , , Adjacent nodes in their respective subnets , , ,in and Represents the nodes in the subnet of the laying object. and This represents a node in the aforementioned laying task subnet. and Represents a node in the subnet of the aforementioned professional field; calculate and The degree of correlation between them and The degree of correlation between them and The degree of correlation between them; Calculate according to the formula , , Corresponding total correlation The formula is as follows: ＝ * ； In the formula, R is the feasible region, R = { , , | , }, Indicates the nodes in the subnet A subset of nodes formed by directly connected subnet nodes. Indicates the nodes in the subnet A subset of nodes formed by directly connected subnet nodes. Indicates the nodes in the subnet A subset of nodes formed by directly connected subnet nodes. Indicating the subnet of the laying object The degree of correlation between nodes Indicating the laying task subnet The degree of correlation between nodes Indicating the professional field subnet The degree of association between nodes; Traverse all adjacent nodes and calculate the association probability between the hyperedge of the laying specification subnet node and the laying object subnet, the laying task subnet, and the professional domain subnet. The value of is given by the following formula: ， In the formula, D ( , , , () indicates the process , , , The number of hyperedges, D( () indicates the process The number of superedges of a node; Calculate the initial hyperedge probability of nodes in the laying object subnet, the laying task subnet, and the professional domain subnet. The value of is given by the following formula: ， In the formula, Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. Indicates the process The number of superedges of a node; Calculate the probability of the occurrence of a hyperedge containing nodes of the laying specification subnet according to the formula. The value of is given by the following formula: ， In the formula, Indicates the process The number of hyperedges of a node. Indicates the process The number of hyperedges of a node. yes The node is a neighboring node in the laying specification layer; Computational Recommendation Standard Knowledge Elements probability The formula is as follows: 。 7. The method for recommending cable design specifications according to claim 2, characterized in that, The new cable design specifications are added to the hypernet by adding new nodes and hyperedges.

8. A cable design specification recommendation device, characterized in that, include: A construction module is used to construct subnets, which include a laying specification subnet and several related subnets. The laying specification subnet includes specification knowledge element nodes, which store cable design specifications. The related subnets include at least a laying object subnet, a laying task subnet, and a professional domain subnet. The construction module is also used to construct a hyperedge, which is used to connect the subnet; The construction module is also used to construct a hypernetwork through the subnet and the hyperedge; The parsing module is used to parse the user task package and map the keywords in the user task package to the hypernet to become the starting hyperedge of the associated subnet; The push module is used to obtain and push the specification knowledge element node from the laying specification subnet according to the starting hyperedge; The parsing module is further configured to: perform entity recognition on the user task package, wherein the entities include a first entity in the domain dictionary and a second entity in the task package, and the user task package contains several keywords representing laying semantic information; calculate the similarity between the first entity and the second entity; determine whether the similarity between the first entity and the second entity exceeds a preset value; when the similarity between the first entity and the second entity exceeds the preset value, replace the second entity in the task package with the first entity in the domain dictionary; when the similarity between the first entity and the second entity does not exceed the preset value, replace the entity in the task package with the entity with the highest similarity in the domain dictionary; and map the keywords in the user task package to the hypernetwork to become the starting hyperedge, wherein the starting hyperedge is a hyperedge connecting the laying object subnet, the laying task subnet, and the professional domain subnet.

9. The cable design specification recommendation device according to claim 8, characterized in that, Also includes: An additional module is added to the hypernet to add new cable design specifications by adding new nodes and hyperedges.

Images