BA-IKG: BiLSTM Embedded ALBERT for Industrial Knowledge Graph Generation and Reuse

As the industrial production mode is shifting towards digitalization and intelligence in the new era. Enterprises put forward higher requirements for efficient processing and utilization of accumulated unstructured data. At present, the knowledge and data contained in a large number of unstructured documents are scattered. The types of entities and relationships are diverse. And the constraints of production rules are complicated, which increases the difficulty of knowledge management and utilization. Therefore, this paper studies the semantic knowledge graph generation and reuse method for industrial documents, which can form standardized production resources, the knowledge related to the industry, and question and answer strategies for industrial processing. The challenge of the research is to explore a feasible process knowledge model and efficient industrial information extraction method to effectively provide structured knowledge of process documents. We build process knowledge representation models and information extraction models and algorithms based on process knowledge representation model and natural language processing. The entities and relations of the main production factors are extracted. The knowledge representation model associates the extracted entities and relations to form an industrial knowledge graph, which provides information support for processing knowledge retrieval and question answering methods. Finally, the approach is evaluated by employing the aerospace machining documents. And the proposed method can obtain valuable information in the document and improve utilization of industrial unstructured data.