GPBP: Pipeline Extraction of Entities and Relations for Construction of Urban Utility Tunnel Knowledge Graph

In order to ensure the stable operation of the urban utility tunnel, the knowledge graphs can be applied to its inspection and maintenance process. Knowledge extraction is a key algorithm in the construction of knowledge graphs, including named entity recognition (NER) and relation extraction (RE) from textual data. For the construction of knowledge graphs in professional fields, it is necessary to design knowledge extraction algorithm based on text features. We collate extensive corpus from maintenance manuals and official documents related to urban utility tunnel, and create the dataset Standards for Operation and Maintenance of Urban Utility Tunnel (SOMUUT). In addition, we proposed a Global Pointer Based Pipeline approach (GPBP) for entity and relation extraction. we adopt global pointer to relation extraction and use two independent neural network structures to finish NER task and RE task, which achieves a better performance comparing with mainstream algorithms on SOMUUT.