Product geometry technical specification standard-oriented knowledge graph construction method

Abstract

The knowledge graph construction method for the product geometry technical specification standard is characterized by specifically comprising the following steps: step 1, data acquisition: acquiring a GPS standard text and metadata from a standard database to obtain corpora for constructing a GPS standard knowledge graph, and preprocessing the data; step 2, performing entity extraction and relation extraction based on the preprocessed GPS standard data; step 3, performing entity alignment on the extracted entities so as to reduce knowledge redundancy; 4, on the basis of the extracted entities and relationships, constructing triples of the entities, the relationships, the entities, the entities, the attributes and the attribute values; and step 5, storing the triple into a universal Neo4j graph database, carrying out visualization on the obtained GPS standard knowledge graph by using the Neo4j graph database, and carrying out GPS standard retrieval.

Description

Technical Field

[0001] The present invention belongs to the cross - field of knowledge graph technology and standardization information management, and particularly relates to a method for constructing a knowledge graph based on the Geometrical Product Specifications (GPS) standard system. Background Art

[0002] Geometrical Product Specifications (GPS) is a set of geometric technology standard systems established for all geometric products, covering the geometric features of products from macroscopic to microscopic, and involving the whole process of the entire life cycle of product development, design, manufacturing, acceptance, use, maintenance, and scrapping. It consists of many technical standards related to product geometric features and their characteristic quantities, covering standards such as the dimensions, geometric shapes and positions of workpieces, and surface topography.

[0003] With the rapid development of Industry 4.0 and intelligent manufacturing, the importance of Geometrical Product Specifications (GPS) standards in high - end manufacturing fields such as mechanical manufacturing, aerospace, and automotive manufacturing has become increasingly prominent. As the core technical standard system for quality control in modern manufacturing, GPS standards not only directly affect the entire process of product design, manufacturing, and inspection, but also are an important basic support for realizing intelligent manufacturing and digital factories.

[0004] However, the complexity and systematicness of the GPS standard system also pose huge challenges to practical applications. According to statistics, there are more than 150 currently effective GPS standards, and more than 20 new standards are under development. These standard knowledge is scattered, the content is highly professional, and there are complex association relationships such as references, substitutions, and supplements between standards and knowledge points. Although traditional standard retrieval methods (such as keyword retrieval or classification number retrieval) can achieve the query of single - standard information, they have obvious limitations: First, it is difficult to accurately locate the position and role of a specific standard in the entire GPS implementation stage; second, it is impossible to clearly present the association relationships and logical sequences between standards; third, there is a lack of systematic understanding of the overall architecture of the standard system. These problems directly lead to low standard application efficiency and insufficient ability to guide practice, seriously restricting the effective application of GPS standards in intelligent manufacturing.

[0005] By constructing a knowledge graph for product Geometrical Product Specifications (GPS) standards, it is possible to effectively integrate scattered standard knowledge. Through key technologies such as entity recognition, relationship extraction, and graph construction, the visualization expression and intelligent recommendation of complex relationships between standards can be realized. Specifically, this construction method can achieve the following goals: First, establish a multi-dimensional classification system for GPS standards, including standard types, application fields, technical characteristics, etc.; Second, construct a semantic association network between standards to clarify relationships such as citation, substitution, and supplementation between standards; Third, realize the intelligent retrieval and recommendation of standard knowledge, providing accurate standard application guidance for designers and engineers. By constructing a GPS standard knowledge graph, not only can the efficiency and accuracy of standard application be improved, but also strong standard support can be provided for intelligent manufacturing, promoting the transformation and upgrading of the manufacturing industry towards digitalization and intelligentization. Summary of the Invention

[0006] The object of the present invention is to provide a method for constructing a knowledge graph for product Geometrical Product Specifications (GPS) standards, so as to solve the difficulties in standard retrieval and application in the prior art, improve the efficiency and accuracy of designers and engineers in selecting and applying GPS standards, and provide accurate standard application guidance for designers and engineers.

[0007] A method for constructing a knowledge graph for product Geometrical Product Specifications (GPS) standards specifically includes the following steps: Step 1, data acquisition: Obtain GPS standard texts and metadata from the standard database as the corpus for constructing the GPS standard knowledge graph, and perform data preprocessing; Step 2, based on the preprocessed GPS standard data, perform entity extraction and relationship extraction; Step 3, perform entity alignment on the extracted entities to reduce knowledge redundancy; Step 4, based on the extracted entities and relationships, construct triples of entity, relationship, entity and entity, attribute, and attribute value; Step 5, store the triples in a general Neo4j graph database, use the Neo4j graph database to visualize the obtained GPS standard knowledge graph, and perform retrieval of GPS standards.

[0008] As a further preferred solution of the method for constructing a knowledge graph for product Geometrical Product Specifications (GPS) standards of the present invention, the specific steps of step 2 are as follows: Step 2.1, for GPS standard content with a specific structure, use regular expression writing to perform entity recognition; The following standards and their attributes can be extracted from GPS standard data based on rules: (1) Standard number; (2) Title; (3) Hierarchical topics: first-level, second-level, third-level; (3) Adoption method: identical, modified, not equivalent; (4) Adoption standard number; (5) Standard status: current, repealed; (6) Alternative standard number; (7) Implementation date; (8) Scope of application; (9) Normative reference documents; (10) Terms / definitions; (11) Classification: basic standard, comprehensive standard, general standard, supplementary standard; (12) Matrix coordinates: matrix rows (1-9) and columns (A-G); (13) Implementation stage: specification, assessment, verification; and perform structured storage; Step 2.11, the matrix coordinates in the standard attributes are constructed based on the GPS matrix model. The matrix rows are geometric features, including 9 items, namely 1 dimension, 2 distance, 3 shape, 4 direction, 5 position, 6 runout, 7 profile surface texture, 8 area surface texture, 9 surface defects; the matrix columns are links, including 7 items, namely A symbols and notations, B feature-feature, C feature characteristics, D conformity and non-conformity, E measurement, F measuring equipment, G calibration; Step 2.12, the implementation stage in the standard attributes is constructed based on the links in the GPS matrix model. The specification stage includes links A, B, C, the assessment stage includes link D, and the verification stage includes links E, F, G; Step 2.13, the division of hierarchical topics in the standard attributes follows a hierarchical structure from general to specific, and is constructed based on the GPS matrix model and the three-part title. The first-level topic is a macro classification, covering 9 geometric features; the second-level topic is a subdivision under the first-level topic, focusing on a specific field; the third-level topic is a specific classification, directly related to a certain standard or knowledge point; Step 2.2, for the content of the old GPS standard that does not conform to the specific structure, inherit the attribute values of its alternative standard for the missing attribute information; Step 2.3, based on the extracted entities, identify and extract the relationships between entities: adoption relationship, alternative relationship, citation relationship, hierarchical relationship, matrix relationship, application relationship, etc.

[0009] As a further preferred solution of the method for constructing a knowledge graph for product geometric technology specification (GPS) standards of the present invention, step 3 specifically includes the following steps: Step 3.1, perform data cleaning on the extracted knowledge, and then achieve the alignment of some entities; Step 3.2, for entities that are semantically consistent but not completely identical in characters, use the method of calculating cosine similarity based on the word bag model Word2vec for entity alignment.

[0010] As a further preferred solution for the method of constructing a knowledge graph for the Geometrical Product Specifications (GPS) standard of the present invention, the specific process of entity alignment based on the cosine similarity calculation of Word2vec word vectors is as follows: Step 3.21, segment the sentences in the GPS standard into characters to construct a training set, and train Word2vec word vectors; segment the entities to be represented into characters, construct a vocabulary-vector dictionary according to the vocabulary and Word2vec word vectors, obtain the vector representation of each character, add the vector representations of each character, and then average them to obtain the vector representation of the final entity; calculate the cosine similarity to perform entity alignment.

[0011] The method of constructing a knowledge graph for the Geometrical Product Specifications (GPS) standard of the present invention realizes the semantic representation of the standard and the visualization of the association relationship, enriches the logical relationships between GPS standards, such as hierarchical, matrix, application, citation relationships, etc., enhances the relevance between standards and knowledge, and helps users better understand and apply GPS standards. Brief Description of the Drawings

[0012] Figure 1 is the flowchart for constructing the GPS standard knowledge graph.

[0013] Figure 2 is the GPS standard matrix model and the division diagram of the implementation stage.

[0014] Figure 3 is the topic classification of the straightness standard.

[0015] Figure 4 is the flowchart for entity alignment based on the cosine similarity calculation of Word2vec word vectors. Detailed Embodiment

[0016] As Figure 1 shown, in this embodiment, a method for constructing a knowledge graph for the Geometrical Product Specifications (GPS) standard specifically includes the following steps: Step 1, data acquisition: Obtain the GPS standard text and metadata from the standard database as the corpus for constructing the GPS standard knowledge graph, and perform data preprocessing; Step 2, based on the preprocessed GPS standard data, perform entity extraction and relationship extraction; Step 2.1, for the GPS standard content with a specific structure, use regular expression writing to perform entity recognition; The following standards and their attributes can be extracted from GPS standard data based on rules: (1) standard number; (2) title; (3) hierarchical theme: first-level, second-level, third-level; (3) adoption method: identical, modified, not equivalent; (4) adopted standard number; (5) standard status: current, repealed; (6) alternative standard number; (7) implementation date; (8) scope of application; (9) normative reference documents; (10) terms / definitions; (11) classification: basic standard, comprehensive standard, general standard, supplementary standard; (12) matrix coordinates: matrix rows (1-9) and columns (A-G); (13) implementation stage: specification, assessment, verification; and perform structured storage; among them, the matrix coordinates in the standard attributes are constructed based on the GPS matrix model, such as Figure 2 shown in the GPS standard matrix model and the implementation stage division diagram. The matrix rows are geometric features, including 9 items, namely 1 dimension, 2 distance, 3 shape, 4 direction, 5 position, 6 runout, 7 profile surface texture, 8 area surface texture, 9 surface defects; the matrix columns are links, including 7 items, namely A symbols and notations, B feature - feature, C feature characteristics, D conformity and non - conformity, E measurement, F measuring equipment, G calibration. The implementation stage in the standard attributes is constructed based on the links in the GPS matrix model. The specification stage includes links A, B, C, the assessment stage includes link D, and the verification stage includes links E, F, G, as Figure 2 shown; the hierarchical theme division in the standard attributes follows a hierarchical structure from general to specific, and is constructed based on the GPS matrix model and the three - part title. The first - level theme is a macro - classification, covering 9 geometric features; the second - level theme is a subdivision under the first - level theme, focusing on a specific field; the third - level theme is a specific classification, directly related to a certain standard or knowledge point; such as Figure 3 shown in the theme classification with straightness as an example.

[0017] Step 2.2, for the content of the old GPS standards that do not conform to the specific structure, inherit the attribute values of its alternative standards for the missing attribute information, thus constructing an inheritance relationship between attributes; Step 3, based on the extracted entities, identify and extract the relationships between entities: adoption relationship, substitution relationship, reference relationship, hierarchical relationship, matrix relationship, application relationship, inheritance relationship, etc.

[0018] Step 4, perform entity alignment on the extracted entities to reduce knowledge redundancy; Step 4.1, perform data cleaning on the extracted knowledge, and then achieve the alignment of some entities; Step 4.2, for entities that are semantically consistent but not completely character - consistent, use the method of calculating cosine similarity based on the bag - of - words model Word2vec for entity alignment. Segment the sentences in the GPS standard into characters to construct a training set and train Word2vec word vectors; segment the entity to be represented into characters, construct a vocabulary - vector dictionary according to the vocabulary and Word2vec word vectors to obtain the vector representation of each character, add the vector representations of each character, and then average them to obtain the vector representation of the final entity; calculate the cosine similarity for entity alignment.

[0019] Figure 4 The specific process for entity alignment based on calculating cosine similarity of Word2vec word vectors is as follows: (1) Segment the sentences in the GPS standard into characters to construct a training set and train Word2vec word vectors.

[0020] (2) Represent the entity as a vector: Use the training text to construct a vocabulary and load the previously saved Word2vec word vectors. Segment the entity to be represented into characters, construct a vocabulary - vector dictionary according to the vocabulary and Word2vec word vectors to obtain the vector representation of each character, add the vector representations of each character, and then average them to obtain the vector representation of the final entity.

[0021] (3) Calculate the cosine similarity; (4) Achieve entity alignment.

[0022] Step 5, based on the extracted entities and relationships, construct triples of entity, relationship, entity - and - entity, attribute, and attribute - value. Step 6, store the triples in a general Neo4j graph database, use the Neo4j graph database to visualize the obtained GPS standard knowledge graph, and conduct retrieval of the GPS standard.

[0023] The GPS standard knowledge graph is a relational network that describes the relationships between standards and standard attributes as well as between standards. The basic description method of the standard knowledge graph. In the GPS standard knowledge graph, circular marks represent entities, and square marks represent the attributes of entities. In the GPS standard knowledge graph, entities are mainly standards. For each entity, there is a series of attributes used to describe a series of inherent attributes of the entity. For entities, the considered attributes are as follows: standard number, title, hierarchical theme, adoption method, adopted standard number, standard status, alternative standard number, implementation date, scope of application, normative reference documents, terms / definitions, classification, matrix coordinates, implementation stage.

[0024] Entity nodes of a certain GPS standard can be constructed through entity and attribute information. After constructing the nodes, the relationships between standards and standards, and between attributes and attributes are further constructed. Entity relationships include standard adoption relationships, substitution relationships, reference relationships, hierarchical relationships, matrix relationships, application relationships, and inheritance relationships.

[0025] The above specific implementation cases are used to explain the present invention and are only the preferred embodiments of the present invention, rather than limiting the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and scope of the claims of the present invention fall within the protection scope of the present invention.

Claims

1. A method for constructing a knowledge graph for product geometric technology specification standards, characterized in that: The method specifically includes the following steps: Step 1, data acquisition: Obtain GPS standard texts and metadata from a standard database as the corpus for constructing a GPS standard knowledge graph, and perform data preprocessing. Step 2, based on the preprocessed GPS standard data, perform entity extraction and relationship extraction. Step 3, perform entity alignment on the extracted entities to reduce knowledge redundancy. Step 4, based on the extracted entities and relationships, construct triples of entity, relationship, entity and entity, attribute, and attribute value. Step 5, store the triples in a general Neo4j graph database, use the Neo4j graph database to visualize the obtained GPS standard knowledge graph, and perform GPS standard retrieval.

2. The knowledge graph construction method according to claim 1, wherein: Step 2.1, for GPS standard content with a specific structure, use regular expressions to perform entity recognition. Based on the rules, the following standards and their attributes can be extracted from GPS standard data: (1) standard number; (2) title; (3) hierarchical theme: first level, second level, third level (3) adoption method: identical, modified, not equivalent; (4) adopted standard number; (5) standard status: current, abolished; (6) alternative standard number; (7) implementation date; (8) scope of application; (9) referenced normative documents; (10) terms / definitions; (11) classification: basic standard, comprehensive standard, general standard, supplementary standard; (12) matrix coordinates: matrix rows (1-9) and columns (A-G); (13) implementation stage: specification, assessment, verification. And perform structured storage. Step 2.11, the matrix coordinates in the standard attributes are constructed based on the GPS matrix model. The matrix rows are geometric features, including 9 items, namely 1 dimension, 2 distance, 3 shape, 4 direction, 5 position, 6 runout, 7 profile surface texture, 8 area surface texture, 9 surface defects; the matrix columns are links, including 7 items, namely A symbols and notations, B element to element, C element features, D conformity and non-conformity, E measurement, F measuring equipment, G calibration. Step 2.12, the implementation stage in the standard attributes is constructed based on the links in the GPS matrix model. The specification stage includes links A, B, C, the assessment stage includes link D, and the verification stage includes links E, F, G. Step 2.13, the hierarchical theme division in the standard attributes follows a hierarchical structure from general to specific, and is constructed based on the GPS matrix model and a three-part title. The first-level theme is a macro classification, covering 9 geometric features. The second-level theme is a subdivision under the first-level theme, focusing on a specific field; the third-level theme is a specific classification, directly related to a certain standard or knowledge point. Step 2.2, for GPS old standard content that does not conform to the specific structure, inherit the attribute values of its alternative standard for the missing attribute information. Step 2.3, based on the extracted entities, identify and extract the relationships between entities: adoption relationship, substitution relationship, reference relationship, hierarchical relationship, matrix relationship, application relationship, etc.

3. The knowledge graph construction method according to claim 1, wherein: The specific steps of step 3 are as follows: Step 3.1, perform data cleaning on the extracted knowledge, and then achieve the alignment of some entities. Step 3.2: For entities that are semantically consistent but not completely character - consistent, use the method of calculating cosine similarity based on the Word2vec bag - of - words model for entity alignment.

4. The knowledge graph construction method according to claim 3, wherein: The specific process of calculating cosine similarity based on Word2vec word vectors for entity alignment is as follows: Step 3.21: Segment the sentences in the GPS standard into characters to construct a training set and train Word2vec word vectors; segment the entities to be represented into characters, construct a vocabulary - vector dictionary according to the vocabulary and Word2vec word vectors, obtain the vector representation of each character, add the vector representations of each character, and then average them to obtain the vector representation of the final entity; calculate the cosine similarity for entity alignment.

Images