Knowledge graph construction method and device, storage medium and computer equipment

Abstract

The embodiment of the invention provides a knowledge graph construction method and device, a storage medium and computer equipment. The method comprises the following steps: constructing a data database related to a set field; constructing an ontology model of the knowledge graph according to the data database; performing knowledge extraction on the data database based on the deep neural network model to generate extraction information; performing knowledge fusion on the extracted information and the ontology model to generate fusion information; and carrying out visual display on the fusion information so as to construct the knowledge graph. According to the technical scheme provided by the embodiment of the invention, the data in the field of engine fault maintenance is fused and stored and displayed in a graph form, so that the high quality and intuition of the constructed knowledge graph are ensured, and the efficiency and reliability of vehicle part maintenance are further improved.

Description

[Technical field]

[0001] The present invention relates to the field of computer technology, and in particular to a method, device, storage medium and computer equipment for constructing a knowledge graph. [Background technology]

[0002] The engine is an important component of a fuel vehicle. It not only has a complex structure, but also has a high probability of failure and many types of failures. Maintenance technicians will go through a very tedious troubleshooting process when repairing engine failures.

[0003] Engine failure repair accounts for the majority of automobile maintenance work. Multi-source engine maintenance information and complex and changeable maintenance business scenarios have brought severe challenges to the efficient and reliable operation and maintenance of the engine, but the current maintenance efficiency and reliability of vehicle components are low. [Summary of the invention]

[0004] In view of this, an embodiment of the present invention provides a method, apparatus, storage medium and computer device for constructing a knowledge graph, so as to improve the efficiency and reliability of vehicle parts maintenance based on the constructed knowledge graph.

[0005] On the one hand, an embodiment of the present invention provides a method for constructing a knowledge graph, including:

[0006] Build a database of data related to the set field;

[0007] Constructing an ontology model of a knowledge graph based on the data database;

[0008] Extracting knowledge from the data database based on a deep neural network model to generate extraction information;

[0009] Performing knowledge fusion on the extracted information and the ontology model to generate fusion information;

[0010] The fusion information is visualized to construct the knowledge graph.

[0011] Optionally, constructing an ontology model of the knowledge graph according to the data database includes:

[0012] Acquire a plurality of first attribute information corresponding to the component body based on the data database;

[0013] Determine a first attribute relationship for each of the first attribute information;

[0014] The ontology model is constructed according to the first attribute information and the first attribute relationship.

[0015] Optionally, extracting knowledge from the data database based on a deep neural network model to generate extraction information includes:

[0016] Based on the BIESO method, the fault entity boundary is marked and the marking information is generated;

[0017] Based on the annotation information, knowledge extraction is performed according to the acquired entity extraction model to generate entity information and second attribute information corresponding to the component failure;

[0018] Associating the entity information with each other to generate entity relationships;

[0019] Associating the entity information with the second attribute information to generate a second attribute relationship;

[0020] The extraction information is generated according to the entity relationship and the second attribute relationship.

[0021] Optionally, before extracting knowledge based on the annotation information according to the acquired entity extraction model to generate entity information and second attribute information corresponding to the component failure, the method includes:

[0022] Based on the deep neural network model, an entity extraction model is constructed according to the fault text data in the data database. The entity extraction model includes a BERT layer, a BiLSTM layer, an Attention layer and a Softmax layer. The fault text data is encoded by the BERT layer, the BiLSTM layer generates hidden values ​​in forward and reverse directions, the Attention layer learns to obtain the distribution probability value of attention, and the entity extraction model is generated based on the Softmax layer.

[0023] Optionally, the performing knowledge fusion on the extracted information and the ontology model to generate fusion information includes:

[0024] The extracted information and the ontology model are fused with knowledge, and based on the linking technology of entity popularity, the extracted entity information is processed to generate fused information.

[0025] Optionally, the entity information includes a component failure entity, and the linking technology based on entity popularity performs data processing on the extracted entity information to generate fusion information, including:

[0026] The entity information and the ontology library corresponding to the ontology model are fused, processed, disambiguated and updated to generate fused information.

[0027] Optionally, the data database includes: component drawing data, component maintenance logs recorded by maintenance personnel, component maintenance operating instructions, component professional books and / or component failure maintenance cases published on the web page.

[0028] On the other hand, an embodiment of the present invention provides a device for constructing a knowledge graph, including:

[0029] The first construction module is used to construct a data database related to the set field;

[0030] A second construction module is used to construct an ontology model of the knowledge graph according to the data database;

[0031] A first generating module is used to extract knowledge from the data database based on a deep neural network model to generate extraction information;

[0032] A second generating module is used to perform knowledge fusion on the extracted information and the ontology model to generate fusion information;

[0033] The third construction module is used to visualize the fusion information to construct the knowledge graph.

[0034] On the other hand, an embodiment of the present invention provides a storage medium, including: the storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the above-mentioned knowledge graph construction method.

[0035] On the other hand, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions, wherein the program instructions, when loaded and executed by the processor, implement the steps of the above-mentioned knowledge graph construction method.

[0036] In the technical solution of the method for constructing a knowledge graph provided by the embodiment of the present invention, a data database related to a set field is constructed; an ontology model of the knowledge graph is constructed based on the data database; knowledge is extracted from the data database based on a deep neural network model to generate extracted information; knowledge fusion is performed on the extracted information and the ontology model to generate fused information; and the fused information is visualized to construct a knowledge graph. In the technical solution provided by the embodiment of the present invention, data in the field of engine fault maintenance are integrated and stored and displayed in the form of a graph, which ensures the high quality and intuitiveness of the knowledge graph, thereby improving the efficiency and reliability of vehicle component maintenance.

Brief Description of the Drawings

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0038] Figure 1A flowchart of a method for constructing a knowledge graph provided by one embodiment of the present invention;

[0039] Figure 2 A flowchart of an ontology model for constructing a knowledge graph provided by one embodiment of the present invention;

[0040] Figure 3 A schematic diagram of a classification system of an ontology model provided by an embodiment of the present invention;

[0041] Figure 4 A partial modeling diagram of an ontology model provided by an embodiment of the present invention;

[0042] Figure 5 A flowchart of extracting knowledge from a data database and generating extracted information is provided in one embodiment of the present invention;

[0043] Figure 6 A schematic diagram of the structure of an entity extraction model provided by one embodiment of the present invention;

[0044] Figure 7 A schematic diagram of a knowledge graph provided by an embodiment of the present invention;

[0045] Figure 8 A schematic diagram of the structure of a knowledge graph construction device provided by one embodiment of the present invention;

[0046] Fig. 9 A schematic diagram of a computer device provided by an embodiment of the present invention. [Specific implementation method]

[0047] In order to better understand the technical solution of the present invention, the embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0048] It should be clear that the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings.

[0050] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0051] The engine is an important part of a fuel vehicle. It is not only complex in structure, but also has a high probability of failure and many types of failures. When repairing an engine failure, maintenance technicians will have to go through a very tedious troubleshooting process. If the engine failure repair process can be intelligently analyzed and decided, the cause of the failure can be quickly and accurately determined, and effective maintenance measures can be taken, the efficiency of engine failure repair can be improved, economic losses can be reduced, and especially for maintenance technicians with less experience, auxiliary analysis and decision-making can effectively guide the maintenance process and save a lot of time. At present, most domestic and foreign scholars have conducted research on engine fault diagnosis from a model-based and data-driven perspective. In the model-based fault diagnosis process, it is difficult to integrate the domain knowledge of the engine failure cause, making it difficult to establish a physical mathematical model. The knowledge graph is a data structure expressed by a graph, which can describe the many-to-many complex coupling relationship between the fault phenomenon and the fault cause, and then analyze the fault cause, trace the source, and provide fault repair guidance through rule knowledge reasoning.

[0052] Knowledge graph, known as knowledge domain visualization or knowledge domain mapping map in the library and information industry, is a series of various graphics that show the development process and structural relationship of knowledge. It uses visualization technology to describe knowledge resources and their carriers, and mine, analyze, construct, draw and display knowledge and their mutual connections. Knowledge graph is a modern theory that combines the theories and methods of applied mathematics, graphics, information visualization technology, information science and other disciplines with methods such as metrology citation analysis and co-occurrence analysis, and uses visualized graphs to vividly display the core structure, development history, frontier fields and overall knowledge architecture of the discipline to achieve the purpose of multidisciplinary integration. Knowledge graph can provide practical and valuable reference for disciplinary research.

[0053] As an intelligent technology with very comprehensive semantic relationships and structured expressions, knowledge graph is the foundation of artificial intelligence and plays a vital role in interpretable artificial intelligence. It can intuitively and effectively display entities and their relationships in a structured manner in massive amounts of information. If knowledge graph technology is combined with the field of engine fault maintenance, it will effectively solve the problems of fault finding, troubleshooting, and maintenance difficulties in the field of engine maintenance, time-consuming and labor-intensive maintenance information integration, and low maintenance efficiency. It can condense fault maintenance information and provide maintenance technicians with timely and accurate maintenance guidance.

[0054] The multi-source engine maintenance information and the complex and ever-changing maintenance business scenarios have brought severe challenges to the efficient and reliable operation and maintenance of the engine. The current engine maintenance solution lacks unified data modeling at the knowledge level, making it difficult to effectively process historical maintenance data. The maintenance of engine failures occupies the main maintenance work of automobile maintenance. In order to reduce the number of unnecessary disassembly and assembly times for maintaining engine failures and improve the economy and practicality of various parts in the car, an accurate and efficient method is needed to achieve convenient, fast and maintenance of automobile failures. Knowledge graph technology can structure the engine failure information modeling, make intelligent analysis and decision-making on the engine failure maintenance process, quickly and accurately determine the cause of the failure and take effective maintenance measures, improve the efficiency of engine failure maintenance, reduce economic losses, and make the engine work in the best condition. Especially for less experienced maintenance technicians, the use of knowledge graphs to assist in analysis and decision-making will effectively guide the maintenance process and save a lot of troubleshooting time, which is very meaningful for ensuring the normal and safe operation of automobile engines.

[0055] In response to the various problems faced by the above-mentioned automobile engine fault maintenance, an embodiment of the present invention proposes an application framework of an automobile engine auxiliary maintenance decision-making knowledge graph including a source data layer, a graph construction layer, a fault retrieval layer, and a reasoning decision layer. First, the multi-source information such as the fault phenomenon, cause analysis, maintenance steps, and maintenance process specifications in the engine fault maintenance process are integrated into an expressible, reasonable, and controllable structured data network, namely, a knowledge graph. As a semantic network that can associate everything, knowledge graph technology can better utilize the accumulated fault maintenance experience data in the current fault maintenance process, provide experience guidance for the current engine fault maintenance, improve the efficiency and accuracy of engine maintenance, and realize the guidance and intelligent decision-making processing of fault maintenance using knowledge graphs.

[0056] An embodiment of the present invention provides a method for constructing a knowledge graph. Figure 1 A flowchart of a method for constructing a knowledge graph provided by an embodiment of the present invention is shown in FIG. Figure 1 As shown, the method includes:

[0057] Step 102: Build a data database related to the set field.

[0058] In the embodiment of the present invention, each step is performed by a computer device. For example, the computer device includes a computer, a server, or a tablet computer.

[0059] In the embodiment of the present invention, the set domain may include a component failure maintenance domain, for example, the component failure maintenance domain is an engine failure maintenance domain. In the embodiment of the present invention, the engine failure maintenance domain is taken as an example for description.

[0060] In an embodiment of the present invention, the data database may include unstructured data and semi-structured data related to engine fault maintenance. For example, the data database includes: component drawing data, component maintenance logs recorded by maintenance personnel, component maintenance operation instructions, component professional books and / or component fault maintenance cases published on the web page, where the component may be an engine.

[0061] Step 104: Construct an ontology model of the knowledge graph based on the data database.

[0062] In this step, the ontology concept of the knowledge graph can be constructed, and the maintenance data-related knowledge can be extracted from the engine fault maintenance data database through manual extraction, and the ontology model can be constructed in combination with the engine fault maintenance expert knowledge.

[0063] Figure 2 A flowchart of an ontology model for constructing a knowledge graph is provided in one embodiment of the present invention, such as Figure 2 As shown, step 104 includes:

[0064] Step 1042: Acquire a plurality of first attribute information corresponding to the component body based on the database.

[0065] In the embodiment of the present invention, the first attribute information may include fault phenomenon, car brand, engine model, etc.

[0066] Step 1044: determine a first attribute relationship for each first attribute information.

[0067] For example, the relationship between an engine and a country is manufacturing, and the relationship between an engine and a car is assembly.

[0068] Step 1046: construct an ontology model according to the first attribute information and the first attribute relationship.

[0069] Specifically, the ontology concept of the knowledge graph can be constructed, and the seven-step method can be used to construct the ontology concepts related to engine fault maintenance, and the knowledge graph model layer can be constructed to construct the ontology model of the knowledge graph. The specific steps are as follows.

[0070] When determining the process of building the ontology model, not only the engine maintenance characteristics should be considered, but also the scope and composition of the aforementioned data database should be considered, and the appearance of invalid ontologies should be avoided as much as possible. There are no filled instances of concept classes in the data database. At the same time, the actual application scenarios based on the upper layer of the ontology model should be considered, and concepts with low learning value or low user attention should be removed. Among them, the knowledge graph can be logically divided into two levels: the data layer and the pattern layer. The pattern layer is above the data layer and is refined knowledge. It usually refers to the ontology model of the knowledge graph, which defines the types, attributes and scopes of concepts and relationships. For example: the engine can be used as the ontology information of the pattern layer, and the crankshaft, as a component of the engine, can be used as the entity information of the data layer. Invalid ontologies can include repeated ontologies such as: "the heart of a fuel vehicle" and "engine", which refer to the same concept. If the concept of "engine" is used as the ontology, then "the heart of a fuel vehicle" is an invalid ontology.

[0071] The process of constructing the ontology model in the engine maintenance field is further explained:

[0072] (1) Clarify the domain to which the ontology model belongs

[0073] Because of the complexity of the composition of the engine structure, which determines its diverse fault types, it is difficult to make the constructed ontology model cover the entire field of engine maintenance and the quality of the ontology is difficult to guarantee. Therefore, the embodiment of the present invention focuses on constructing the fields of common engine failure phenomena, maintenance methods and maintenance tools. In order to obtain a high-quality ontology model, it is necessary to comprehensively consider the actual application scenarios of the ontology, the characteristics of the engine maintenance field, the quality and scale of the data database, etc. to determine the scope of the ontology model. Through this step, the concepts of the main classes in the ontology model can be determined. For example, in addition to the failure phenomenon, it should also include closely related car brands, engine models, etc.

[0074] (2) Ontology model analysis

[0075] Based on the database (expert opinions, search for engine maintenance related information, and seek the organizational structure of engine maintenance resources on other platforms), learn and analyze the characteristics of the engine maintenance field, and determine the specific ontology composition based on the determined ontology model scope, classify the concepts and construct hierarchical relationships and detailed concept class attributes and relationships. Ontology model analysis should also consider the practical application of the ontology model and cover the field of engine fault maintenance. For example, in the engine maintenance semantic search system, the answers to users' engine fault maintenance questions should be met as much as possible.

[0076] (3) Reuse existing ontology models in this field

[0077] Investigate the existing engine maintenance ontology models and consider their reusability. Since ontology models are highly general and abstract, rebuilding the ontology model in this field will result in high costs, and the constructed ontology model will not be guaranteed due to insufficient professional knowledge. Based on this, before building the ontology model in this field, it is necessary to extensively query relevant data resources in this field or investigate existing third-party data systems, and fully analyze these ontology models.

[0078] (4) List the first attribute information (entity elements) in the engine maintenance field

[0079] Under the guidance and cooperation of experts in the field of engine maintenance, the concepts, relationships and attributes related to the engine fault maintenance entity are summarized. The main purpose of this step is to provide "raw materials" for the subsequent process. In this process, it is only necessary to list the expected elements as much as possible, and there is no need to classify the above elements without errors, conflicts and complete clarity. In the embodiment of the present invention, more specific concepts of each system such as the lubrication system, fuel supply system, crank-connecting rod mechanism, cooling system, and valve mechanism are listed for the engine equipment related elements. For example, the crank-connecting rod mechanism has multiple elements such as the crankshaft body group, flywheel group, and piston connecting rod group. With these concept groupings, it will be easier to locate the location of the engine failure.

[0080] (5) Define relationships and attributes.

[0081] This step mainly defines relationships and attributes for each type of the first attribute information (ontology elements) mentioned above. Relationships are used to define the relationship between different concepts, such as the relationship between engine and country is manufacturing, and the relationship between engine and car is assembly. Attributes are used to describe the inherent characteristics of concepts, such as the model and type of the engine, the type of cooling system, etc.

[0082] (6) Ontology model evaluation

[0083] Ontology model evaluation is a process of comprehensively evaluating various factors that affect the quality of ontology. Ontology model evaluation is carried out through expert inspection and other methods. The content includes confirming whether the goals set in the ontology model analysis phase meet the requirements and verifying whether the constructed ontology model is correct. If the evaluation fails, improvements need to be made or the concept classes and core attributes need to be redefined.

[0084] (7) Preliminary construction and iterative upgrade of the ontology model

[0085] After the above steps, the initial construction of the ontology model can be completed. However, with the development of science and technology, new faults or new engine components will appear in the field of engine fault maintenance, which will be accompanied by new maintenance methods. New knowledge can be updated and iterated to always keep the timeliness of engine fault maintenance.

[0086] In the embodiment of the present invention, the popular and mature ontology construction tool protégé software can be selected to realize the visualization of the engine maintenance ontology. protégé is an ontology editing and modeling tool developed using Java language, and can also be used as an open source knowledge editor.

[0087] The protégé software has the concepts of hypernyms and hyponyms that reflect the relationship of inclusion and being included between words. For example, "engine maintenance" is the hypernym of "cause of engine failure", "cause of engine failure" is the hypernym of "damage to parts", and "engine maintenance" is also the hypernym of "damage to parts". "OWL:Thing" is the highest level hypernym and the default top-level class; therefore, in the embodiment of the present invention, the class of "engine maintenance" is first constructed and used as the hyponym of the top-level class "OWL:Thing"; according to the actual purpose of the engine maintenance ontology, the hierarchical relationship of concepts in the field of engine failure maintenance is constructed. Figure 3 A schematic diagram of a classification system of an ontology model provided for one embodiment of the present invention, through comprehensive organization of engine maintenance-related knowledge and ontology model construction theory, combined with a large amount of maintenance data, triple data is generated according to engine maintenance knowledge ontology-related concepts and attribute relationships, which are divided into general triples and triples specific to the engine maintenance field.

[0088] like Figure 3 As shown, for example, the general triple has three relations: {maintenance personnel, employed in, maintenance enterprise}, {maintenance enterprise, name is, unit name}. The triples specific to engine maintenance have the following relations: {engine equipment, including, crank-connecting rod mechanism}, {fault phenomenon, type is, poor starting}, {fault cause, is, equipment damage}, {maintenance method, including, maintenance tools}, {fault phenomenon, cause is, fault cause}, {fault cause, solution is, maintenance method}, {maintenance method, use, maintenance tools}.

[0089] In the above described triple, the engine includes two major mechanisms and five major systems (the two major mechanisms of the engine are the crank-connecting rod mechanism and the valve mechanism, and the five major systems of the engine include the fuel supply system, cooling system, lubrication system, ignition system and starting system.), and further includes more specific parts: crankcase, piston ring, valve group, cylinder liner, cylinder gasket, piston pin retainer, connecting rod cover, cylinder head, crankshaft, piston, sprocket, positioning lock, connecting rod bolt, piston pin, O-ring, connecting rod, connecting rod bearing, flat top piston, convex top piston, camshaft and other parts that make up the engine. Specific engine parts also have an engine maintenance triple relationship, such as: {engine, including, crank-connecting rod mechanism}, {crank-connecting rod mechanism, including, crank}, {crankshaft, material, 45 steel}, {engine, parts, crankcase}, {connecting rod bearing, composed of, piston connecting rod}, {piston, type, flat top piston}, {piston pin retainer, installed on, piston}.

[0090] Through the above triple information and combined with the existing specific data, the class (Class), property (Property, also called Data property) and relationship (Relation, also called Object property) in the field of engine maintenance can be obtained. Table 1 is a table of classes, properties and relationships in the field of engine maintenance, as shown in Table 1 below.

[0091] Table 1

[0092]

[0093]

[0094] Protégé can be used to model the ontology concepts of the five main categories and 22 subcategories in the field of engine fault maintenance. The five main categories include: engine equipment, engine fault phenomenon, fault cause, maintenance method and maintenance enterprise. Each main category contains several subcategories. Taking the engine as an example, its subcategories can be divided into fuel supply system, lubrication system, crank-connecting rod mechanism, cooling system, ignition system, starting system and valve mechanism.

[0095] After the ontology model evaluation, the ontology model of the engine maintenance field was initially completed. Figure 4 A partial modeling diagram of an ontology model provided by an embodiment of the present invention, Figure 4 The visualization effect of the engine domain ontology model built in protégé is shown below. Figure 4As shown, it includes first attribute information and first attribute relationship, and the connecting line between the first attribute information represents the first attribute relationship, for example, {maintenance personnel, works in, maintenance enterprise}, {maintenance enterprise, name is, unit name}, {engine equipment, includes, crank-connecting rod mechanism}, {fault phenomenon, type is, poor starting}, {fault cause is, equipment damage}, {maintenance method, includes, maintenance tools}, {fault phenomenon, cause is, fault cause}, {maintenance method, uses, maintenance tools}.

[0096] Step 106: Extract knowledge from the data database based on the deep neural network model to generate extraction information.

[0097] In an embodiment of the present invention, knowledge extraction can be performed on a data database in the field of engine fault maintenance, including the extraction of entities of engine fault maintenance, entity attributes, and relationships between entities, so as to complete the construction of a data layer of a knowledge graph.

[0098] Figure 5 A flowchart of extracting knowledge from a data database and generating extracted information is provided in accordance with an embodiment of the present invention. Figure 5 As shown, step 106 includes:

[0099] Step 1062: Annotate the set fault entity boundary based on the BIESO method to generate annotation information.

[0100] In the embodiment of the present invention, the BIESO method shown in Table 2 can be used to mark the fault entity boundary of the engine. The types of entities refer to four categories: "engine and components", "fault manifestation", "fault cause" and "maintenance method", and the symbol set can be "ENGP", "FAU", "CAU" and "SOL", and their meanings are shown in Table 3.

[0101] Table 2

[0102]

[0103] Table 3

[0104]

[0105] Step 1064: construct an entity extraction model based on the fault text data in the data database based on the deep neural network model.

[0106] In an embodiment of the present invention, the deep neural network model includes BERT-BiLSTM-Att. BERT-BiLSTM-Att is a natural language processing model based on deep learning, which includes a pre-trained model (BERT), a bidirectional long short-term memory network (BiLSTM) and an attention mechanism (Attention).

[0107] In the embodiment of the present invention, the entity extraction model can realize the entity and attribute extraction model of engine fault maintenance, and can use unstructured data to build a knowledge graph of engine fault maintenance to provide basic data corpus. Among them, entity information refers to things that exist independently and have specific meanings in the real world, which can be people, places, organizations, products, events, etc. For example, various engine parts, maintenance tool names, maintenance personnel names, etc. can all be used as entity information.

[0108] In the embodiment of the present invention, Figure 6 A schematic diagram of the structure of an entity extraction model provided by an embodiment of the present invention is shown in FIG. Figure 6 As shown in the figure, the entity extraction model includes a BERT layer (encoder), a BiLSTM layer (layer), an Attention layer and a Softmax layer. The fault text data xi=[x1,x2,...xt,...xn]T is encoded by the BERT layer, the BiLSTM layer generates hidden values ​​hi=[h1,h2,...ht,...hn]T in the forward and reverse directions, and the Attention layer learns to obtain the distribution probability value of attention ai=[a1,a2,...at,...an]T. The key idea is to compare the output of the hidden layer of the calculation model and the alignment relationship between the representation vector in the entire text, and generate an entity extraction model based on the Softmax layer.

[0109] Step 1066: perform knowledge extraction based on the annotation information and the acquired entity extraction model to generate entity information and second attribute information corresponding to the component failure.

[0110] In the embodiment of the present invention, since the data in the entity extraction model are not connected and have relatively isolated characteristics, the relationship corresponding to the entity information must be extracted, and the various entity information is related to each other. The relationship can be defined according to the lexical properties they have. The common entity or attribute part-of-speech relationships in the engine maintenance corpus are shown in Table 4 below.

[0111] Table 4

[0112]

[0113]

[0114] In the embodiment of the present invention, it is necessary to filter out redundant relationships. In the process, minor relationships may be treated as major relationships, making the "network" of the graph chaotic. Therefore, the selected minor relationships need to be excluded and not considered. For example, D is a subset of F, and D and F have the relationship that D is a subset of E and E is a subset of F. Then the relationship that D is a subset of F can be filtered out.

[0115] Step 1068: Associate the entity information with each other to generate entity relationships.

[0116] Step 1070: Correlate the entity information with the second attribute information to generate a second attribute relationship.

[0117] Step 1072: Generate extraction information based on the entity relationship and the second attribute relationship.

[0118] Based on the extraction of the above entity information and the second attribute information, the relationship between entities or entities and attributes is extracted from the fault maintenance corpus, and the entity relationship and the second attribute relationship are linked through the associated semantic relationship to form a network knowledge base, that is, to extract information. When constructing the knowledge graph of engine fault maintenance, the engine equipment and fault name can be used as nodes, and the basic information of the fault phenomenon and engine equipment can be used as the second attribute information, such as the number and size of the parts, the engine fault condition, the mileage at the time of the fault, the cause of the fault and the solution, etc. as the nodes of the graph, and the relationship between the attributes (second attribute relationship), such as "the reason is", "the maintenance method is", "the precautions are", "the next step", "installed in", "included", etc., is regarded as the connection between the nodes. Based on the storage method of the knowledge graph triples, a comprehensive and easy-to-use network knowledge base for automobile engine fault maintenance can be constructed.

[0119] Step 108: Perform knowledge fusion on the extracted information and the ontology model to generate fusion information.

[0120] Specifically, the extracted information and the ontology model are fused, and based on the linking technology of entity attributes and entity popularity, the extracted entity information is processed to generate fused information.

[0121] In the embodiment of the present invention, entity information includes component failure entities. In order to build a high-quality maintenance knowledge base and eliminate the multivariate, fuzzy and erroneous knowledge after extraction, the entity information and the ontology library corresponding to the ontology model can be fused, processed, disambiguated and updated through knowledge merging and entity linking to generate fused information to solve the problems of fuzzy relationships between extracted entities and lack of logic. The specific fusion process is as follows:

[0122] Due to the complexity of engine failure, the extracted fault knowledge is fragmented and heterogeneous, and auxiliary maintenance decision-making requires the use of complex and organized knowledge. Based on this, it is necessary to merge the multi-source information extracted from different texts, merge the extracted information with the data in the ontology library, eliminate the contradictory entities and repeated knowledge in the process of maintenance knowledge fusion, merge the engine maintenance knowledge through the ontology mapping rule method, and obtain the merged result with the help of the engine failure concept ontology similarity calculation. Among them, the ontology library is constructed in the process of building the above ontology model, and the data in the ontology library is a related concept layer, such as: (maintenance tools-----engine failure), and the middle line represents the relationship between the maintenance tools and the engine (engine failure requires the help of maintenance tools for maintenance operations).

[0123] In the embodiment of the present invention, there are two methods for knowledge merging:

[0124] 1. Entity linking: Entity linking is the task of linking the entity information mentioned in the text with the corresponding entity in its knowledge base, which is to solve the ambiguity problem between entity information. In detail, given a mesh knowledge base rich in a series of entities and a corpus with annotated mentions, the goal of the entity linking task is to match each mention to the entity information corresponding to it in the knowledge base. If there is no entity information corresponding to the mention in the mesh knowledge base, it is considered that the mention cannot be linked to the current mesh knowledge base and is marked as NIL. For example, "vehicle controller" and "vehicle control unit" in the mesh knowledge base are identified and linked as the same entity information.

[0125] 2. Knowledge merging: In the knowledge graph, information from different data sources may lead to the generation of duplicate entity information. Duplicate entity merging aims to identify and merge these entity information. Method: Rule-based: Merge entities based on predefined rules, such as the same name and attribute values. Machine learning: Use training data to learn the entity merging pattern and automatically identify and merge duplicate entities. For example, if two entity information "IBM Corporation" and "International Business Machines Corporation" have the same address and founding year, they can be merged into the same entity information.

[0126] Step 110: Visualize the fused information to construct a knowledge graph.

[0127] In the embodiment of the present invention, the Neo4j graph database can be selected to complete the storage of engine maintenance knowledge. Compared with other relational databases, the selected Neo4j graph database has more complete functions, Cypher query language and relatively standard management mechanism, and it is easier to connect to the Web development framework. In the graph database, even a large-scale knowledge graph can efficiently realize the query of nodes and relationships.

[0128] The environment that the system depends on is shown in Table 5 below. When configuring Java JDK, add the newly added variable JAVA_HOME, whose value is the path where the JDK is installed; add a new system variable CLASSPATH and set it to ".; %JAVA_HOME%\lib\dt.jar; %JAVA_HOME%\lib\tools.jar;" and add a value "%JAVA_HOME%\bin; %JAVA_HOME%\jre\bin;" to the system variable PATH to complete the system environment configuration.

[0129] Table 5

[0130]

[0131] The method of importing engine failure maintenance data into Neo4j is mainly to use CSV format files for import.

[0132] The specific steps to import batch CSV files into Neo4j are as follows:

[0133] The relevant data, attributes and relationships during the disassembly of the engine and the assembly of the parts after maintenance are saved in different Excel files. It should be noted that an id can be set for each entity and relationship.

[0134] Save the Excel file as a .CSV file.

[0135] Put the CSV file obtained in the previous step in the import directory of the newly created library in Neo4j. You can open the import directory under OpenFolder of the newly created library.

[0136] Open all CSV files in the import directory separately with Notepad and save them as .CSV files. At this time, the encoding format must be modified to UTF-8 format, otherwise garbled characters will appear when imported directly.

[0137] The relevant Cypher language for importing CSV files is as follows:

[0138] --Statements to import engine maintenance entities and their attributes:

[0139] LOAD CSV with headers from "file: / entity-01.csv" as line with linecreate(:jiedian01

[0140] {

[0141] id:line.id,

[0142] name:line.name,

[0143] attribute:line.attribute,

[0144] label:line.label

[0145] });

[0146] -- Statement to import engine maintenance relationship table:

[0147] LOAD CSV WITH HEADERS FROM"file: / / / rela-01.csv"AS line

[0148] match(from:jiedian01{id:line.from_id}),(to:jiedian02{id:line.to_id})

[0149] merge(from)-[r:relationship{P1:line.P1,P2:line.P2,relationship:line.relationship}]->(to)

[0150] According to the above statements, all generated CSV files can be imported into Neo4j, thus completing the construction and visualization of the preliminary knowledge graph of engine fault maintenance.

[0151] Figure 7 A schematic diagram of a knowledge graph provided by an embodiment of the present invention is shown as follows: Figure 7 As shown, for example, in this knowledge graph, "basic process" is connected to "tools", "basic process" is connected to "lubrication", "basic process" is connected to "knocking", "tools" is connected to "electrician pliers", and "tools" is connected to "cross socket".

[0152] In the technical solution provided by the embodiment of the present invention, a data database related to a set field is constructed; an ontology model of a knowledge graph is constructed based on the data database; knowledge is extracted from the data database based on a deep neural network model to generate extracted information; knowledge fusion is performed on the extracted information and the ontology model to generate fused information; the fused information is visualized to construct a knowledge graph. In the technical solution provided by the embodiment of the present invention, data in the field of engine fault maintenance are integrated and stored and displayed in the form of a graph, which ensures the high quality and intuitiveness of the knowledge graph, thereby improving the efficiency and reliability of vehicle component maintenance.

[0153] In the technical solution provided by the embodiment of the present invention, knowledge extraction of engine fault maintenance is performed based on the knowledge graph in the field of engine fault maintenance and the BERT-BiLSTM-Att deep neural network model. The model has the characteristics of autonomous learning information and is free from dependence on professional knowledge.

[0154] In the technical solution provided by the embodiment of the present invention, the knowledge graph in the field of engine fault maintenance is based on a linking technology that combines entity attributes and entity popularity. Through knowledge merging and entity linking, the entity information and the ontology library corresponding to the ontology model can be merged, processed, disambiguated and updated, which has a better linking effect, effectively eliminates redundant and erroneous knowledge, and ensures the high quality of constructing the knowledge graph.

[0155] In the technical solution provided by the embodiment of the present invention, the knowledge graph based on the field of engine fault maintenance can condense the unstructured data of the engine into a structured data display, including graph visualization and relationship visualization, which can display data quickly and intuitively, reduce data search time, and can be further applied to the fields of intelligent search and intelligent question and answer to improve the intelligent use of data.

[0156] In the technical solution provided by the embodiment of the present invention, a method for managing, integrating and storing maintenance data in the field of engine fault maintenance is provided based on the knowledge graph in the field of engine fault maintenance, which shows the relationship between engine fault maintenance knowledge more clearly than traditional maintenance data. The data in the field of engine fault maintenance is integrated and stored and displayed in the form of a graph. Knowledge reasoning can be performed based on the graph through relevant algorithms to obtain further value in the field of engine fault maintenance and give full play to the role of data.

[0157] An embodiment of the present invention provides a device for constructing a knowledge graph. Figure 8 A schematic diagram of a knowledge graph construction device provided by an embodiment of the present invention is shown in FIG. Figure 8 As shown, the device includes: a first building module 11, a second building module 12, a first generating module 13, a second generating module 14 and a third building module 15.

[0158] The first construction module 11 is used to construct a data database related to a set field.

[0159] The second construction module 12 is used to construct an ontology model of the knowledge graph based on the data database.

[0160] The first generating module 13 is used to extract knowledge from the data database based on a deep neural network model to generate extraction information.

[0161] The second generating module 14 is used to perform knowledge fusion on the extracted information and the ontology model to generate fusion information.

[0162] The third construction module 15 is used to visualize the fusion information to construct the knowledge graph.

[0163] In the embodiment of the present invention, the second construction module 12 is specifically used to obtain multiple first attribute information corresponding to the component ontology based on the data database; determine a first attribute relationship for each first attribute information; and construct the ontology model according to the first attribute information and the first attribute relationship.

[0164] In an embodiment of the present invention, the first generation module 13 is specifically used to label the set fault entity boundary based on the BIESO method to generate labeling information; based on the labeling information, perform knowledge extraction according to the acquired entity extraction model to generate entity information and second attribute information corresponding to the component fault; associate the entity information with each other to generate an entity relationship; associate the entity information with the second attribute information to generate a second attribute relationship; and generate the extraction information according to the entity relationship and the second attribute relationship.

[0165] In an embodiment of the present invention, the first generation module 13 is specifically used to construct an entity extraction model based on the deep neural network model according to the fault text data in the data database, and the entity extraction model includes a BERT layer, a BiLSTM layer, an Attention layer and a Softmax layer. The fault text data is encoded by the BERT layer, the BiLSTM layer generates hidden values ​​in the forward and reverse directions, the Attention layer learns the distribution probability value of attention, and generates the entity extraction model based on the Softmax layer.

[0166] In the embodiment of the present invention, the second generating module 14 is specifically used to perform knowledge fusion on the extracted information and the ontology model, and perform data processing on the extracted entity information based on the linking technology of entity popularity to generate fusion information.

[0167] In the embodiment of the present invention, the entity information includes a component failure entity, and the second generating module 14 is specifically used to fuse, process, disambiguate and update the entity information with an ontology library corresponding to the ontology model to generate fused information.

[0168] In the embodiment of the present invention, the data database includes: component drawing data, component maintenance logs recorded by maintenance personnel, component maintenance operation instructions, component professional books and / or component failure maintenance cases published on the web page.

[0169] In the technical solution provided by the embodiment of the present invention, a data database related to a set field is constructed; an ontology model of a knowledge graph is constructed based on the data database; knowledge is extracted from the data database based on a deep neural network model to generate extracted information; knowledge fusion is performed on the extracted information and the ontology model to generate fused information; the fused information is visualized to construct a knowledge graph. In the technical solution provided by the embodiment of the present invention, data in the field of engine fault maintenance are integrated and stored and displayed in the form of a graph, which ensures the high quality and intuitiveness of the knowledge graph, thereby improving the efficiency and reliability of vehicle component maintenance.

[0170] The knowledge graph construction device provided in this embodiment can be used to implement the above Figure 1 For the method of constructing the knowledge graph in , the specific description can be found in the embodiment of the method of constructing the knowledge graph mentioned above, and will not be repeated here.

[0171] An embodiment of the present invention provides a storage medium, which includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the steps of the embodiment of the above-mentioned knowledge graph construction method. For a specific description, please refer to the embodiment of the above-mentioned knowledge graph construction method.

[0172] An embodiment of the present invention provides a computer device, including a memory and a processor, the memory is used to store information including program instructions, the processor is used to control the execution of the program instructions, and the program instructions, when loaded and executed by the processor, implement the steps of the embodiment of the above-mentioned knowledge graph construction method. For a specific description, please refer to the embodiment of the above-mentioned knowledge graph construction method.

[0173] Fig. 9 A schematic diagram of a computer device provided by an embodiment of the present invention. Fig. 9As shown, the computer device 20 of this embodiment includes: a processor 21, a memory 22, and a computer program 23 stored in the memory 22 and executable on the processor 21. When the computer program 23 is executed by the processor 21, the construction method applied to the knowledge graph in the embodiment is implemented. To avoid repetition, it is not described one by one here. Alternatively, when the computer program is executed by the processor 21, the functions of each model / unit in the construction device applied to the knowledge graph in the embodiment are implemented. To avoid repetition, it is not described one by one here.

[0174] The computer device 20 includes, but is not limited to, a processor 21 and a memory 22. Those skilled in the art will appreciate that Fig. 9 It is only an example of the computer device 20 and does not constitute a limitation of the computer device 20. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the computer device may also include input and output devices, network access devices, buses, etc.

[0175] The processor 21 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0176] The memory 22 may be an internal storage unit of the computer device 20, such as a hard disk or memory of the computer device 20. The memory 22 may also be an external storage device of the computer device 20, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device 20. Further, the memory 22 may also include both an internal storage unit of the computer device 20 and an external storage device. The memory 22 is used to store computer programs and other programs and data required by the computer device. The memory 22 may also be used to temporarily store data that has been output or is to be output.

[0177] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0178] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0179] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0180] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0181] The above-mentioned integrated unit implemented in the form of a software functional unit can be stored in a computer-readable storage medium. The above-mentioned software functional unit is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor (Processor) to perform some steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, ROM), random access memory (Random Access Memory, RAM), disk or optical disk and other media that can store program codes.

[0182] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for constructing a knowledge graph, characterized in that: include: Build a database of data related to the set field; Constructing an ontology model of a knowledge graph based on the data database; Extracting knowledge from the data database based on a deep neural network model to generate extraction information; Performing knowledge fusion on the extracted information and the ontology model to generate fusion information; The fusion information is visualized to construct the knowledge graph.

2. The method according to claim 1, characterized in that: The ontology model of constructing the knowledge graph according to the data database includes: Acquire a plurality of first attribute information corresponding to the component body based on the data database; Determine a first attribute relationship for each of the first attribute information; The ontology model is constructed according to the first attribute information and the first attribute relationship.

3. The method according to claim 1, characterized in that The step of extracting knowledge from the data database based on a deep neural network model to generate extraction information includes: Based on the BIESO method, the fault entity boundary is marked and the marking information is generated; Based on the annotation information, knowledge extraction is performed according to the acquired entity extraction model to generate entity information and second attribute information corresponding to the component failure; Associating the entity information with each other to generate entity relationships; Associating the entity information with the second attribute information to generate a second attribute relationship; The extraction information is generated according to the entity relationship and the second attribute relationship.

4. The method according to claim 3, characterized in that: Before extracting knowledge based on the label information according to the acquired entity extraction model to generate entity information and second attribute information corresponding to the component failure, the method includes: Based on the deep neural network model, an entity extraction model is constructed according to the fault text data in the data database. The entity extraction model includes a BERT layer, a BiLSTM layer, an Attention layer and a Softmax layer. The fault text data is encoded by the BERT layer, the BiLSTM layer generates hidden values ​​in forward and reverse directions, the Attention layer learns to obtain the distribution probability value of attention, and the entity extraction model is generated based on the Softmax layer.

5. The method according to claim 1, characterized in that The step of fusing the extracted information with the ontology model to generate fused information includes: The extracted information and the ontology model are fused with knowledge, and based on the linking technology of entity popularity, the extracted entity information is processed to generate fused information.

6. The method according to claim 5, characterized in that The entity information includes component failure entities, and the linking technology based on entity popularity performs data processing on the extracted entity information to generate fusion information, including: The entity information and the ontology library corresponding to the ontology model are fused, processed, disambiguated and updated to generate fused information.

7. The method according to claim 1, characterized in that The data database includes: component drawing data, component maintenance logs recorded by maintenance personnel, component maintenance operation instructions, component professional books and / or component failure maintenance cases published on the web page.

8. A device for constructing a knowledge graph, characterized in that: include: The first construction module is used to construct a data database related to the set field; A second construction module is used to construct an ontology model of the knowledge graph according to the data database; A first generating module is used to extract knowledge from the data database based on a deep neural network model to generate extraction information; A second generating module is used to perform knowledge fusion on the extracted information and the ontology model to generate fusion information; The third construction module is used to visualize the fusion information to construct the knowledge graph.

9. A storage medium, characterized in that: include: The storage medium includes a stored program, wherein when the program is running, the device where the storage medium is located is controlled to execute the method for constructing a knowledge graph as described in any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, wherein the memory is used to store information including program instructions, and the processor is used to control the execution of the program instructions, characterized in that: When the program instructions are loaded and executed by the processor, the steps of the method for constructing a knowledge graph as described in any one of claims 1 to 7 are implemented.

Images