Knowledge Guidance Based Work Ticket Intelligent Generation of Electric Power Equipment Inspection

Abstract

Electric power inspection work ticket is an important text document in the electric power inspection business. The current methods of generating work tickets have problems such as irregular manual filling, low efficiency, and poor knowledge library portability, etc. To solve the problems, this paper constructs the knowledge graph (KG) in the field of power inspection and proposes a knowledge-guided method for intelligently generating work tickets of electric power inspections. Firstly, the named entity of the component or part in the work order is recognized by applying a bidirectional long and short-term memory network (Bi-LSTM) and conditional random field (CRF). Secondly, the bi-encoder is applied for entity disambiguation to make the entity-mention with the description text of defect in the work order corresponds to the node of the component or part defect in the KG. Finally, the relevant paths of the target linked entity are inquired in the KG, and the semantic similarities based on the cosine distance between the text of work order and the texts of paths are calculated to select the optimal path, and the nodes under this path are filled in the slot to generate work tickets. This paper conducts simulation experiments and analysis results for named entity recognition, entity disambiguation, and semantic similarity comparison to verify the effectiveness of work ticket intelligent generation of electric power inspection.