电网技改与检修工程造价文本的信息抽取与知识图谱构建OA

To address the challenges of massive,various data formats,prominent unstructured characteristics,and difficulties in effective mining and utilization of power grid engineering cost documents,this paper designs a format conversion method to preprocess semi-structured tables and extract the cost information by manually defined rules.A BERT-Bi-LSTM-CRF model integrates bidirectional long short-term memory(Bi-LSTM),bidirectional encoder representations from transformers(BERT),and conditional random field(CRF).It is developed to perform named entity recognition(NER)on unstructured textual paragraphs and generate structured cost knowledge triplets.Then,these triplets are imported into the Neo4j graph database to construct a knowledge graph for power grid technical reform and maintenance engineering costs.Experimental results demonstrate that the proposed method achieves a precision rate of over 89.52%in recognizing diverse named entities from engineering cost paragraphs.The constructed knowledge graph effectively transforms tabular and textual data from various cost documents into nodes,entities,and attribute relationships,enabling intelligent retrieval of cost information.This study provides structured data support for engineering cost estimation,analysis,and decision-making in power grid.