Recognizing fine-grained expressway location references

This project aims to recognize fine-grained expressway-related location references from unstructured text. The project shares a corpus of 5322 data and 31,954 labeled entities, and a deep learning method that can simultaneously recognize flat, nested, and discontinuous location references. The labeled entities contain 12 types of entities such as expressway name, road section, direction, tunnel, and flyover. The dataset can be used for several tasks such as Geoparsing, construction of knowledge graphs, named entity recognition, and social sensing.