The Construction Method Of Automobile Fault Event-logic Graph Based On Internet Knowledge Extraction

In the background of the information age,the continuous enrichment of functions and the increasing complexity of systems of intelligent integrated automotive equipment make fault diagnosis more difficult,resulting in low efficiency of automotive fault repair.This paper introduces the concept of event-logic graph in the field of automotive fault,based on the sample characteristics of automotive repair cases,and uses natural language processing related technology to build an automotive fault event-logic graph oriented to Internet knowledge extraction,which provides data for automotive fault analysis and decision support for fault diagnosis and fault repair.The main contents of this paper are as follows:（1）To propose a set of Internet knowledge extraction-based automotive fault event-logic framework suitable for the vertical domain.A set of methodological framework suitable for this study is proposed for the reasoning event-logic concept and construction method framework of existing studies are significantly different from this study and cannot be completely migrated to this paper.In view of the fact that previous ontologies in the field of automotive fault are mainly oriented to the construction of static knowledge,lacking the consideration of combining the aspect of event-logic and the value of event dynamic information of unstructured text of Internet repair cases,this paper uses the top-down ontology modeling idea to construct the ontology model of automotive fault event-logic.To address the problem of lack of relevant corpus in the field of automotive fault to build eventlogic,this paper designs the corresponding crawler program according to the characteristics of automotive repair-related websites,obtains the document knowledge required to build the knowledge base from the Internet,and uses professional dictionaries combined with expert experience to realize the extraction of meta-knowledge of automotive fault theory.（2）Automotive fault event extraction.Aiming at the problem that the automobile fault domain lacks a perfect lexicon and corpus and the automobile fault event sentences have distinctive domain characteristics,this paper adopts the regularization-based trigger word matching method to identify the extraction of fault phenomenon events and the regularizationbased trigger word combined with trigger word matching method to identify the extraction of fault cause events,and the effectiveness of the algorithm is proved through experiments.（3）Automotive failure event value classification.This paper converts relationship extraction to event value classification.To address the problem that automotive fault domain knowledge has distinctive domain characteristics and lacks corpus that can be used for model training,this paper adopts a word vector-based network model to mine the fine-grained meaning of automotive fault words and realize event value classification in the automotive fault domain.By analyzing five mainstream deep learning classification models,labeling the relevant corpus,and conducting classification performance tests respectively,the results show that the FastText model has obvious superiority in feature extraction in the automotive fault domain,with 82% classification accuracy.（4）Event generalization and visualization.In order to facilitate the management of valuable fault event knowledge and the reasoning of automotive fault event-logic,this paper adopts the simple matching method to divide the fault phenomenon events into six categories and the fault cause events into four categories,and combines the semantic features of the sentences of automotive fault cause events,mines the fault parts of each fault cause event,takes the fault parts as the fault event family,and generalizes the events with the same fault parts The events with the same faulty parts are generalized into the same class,Final,visual storage using Neo4j graph database.