Research on textual geographic entity relationship extraction techniques based on DeBERTa

Geographic entity relationship extraction is a key step in constructing a geographic knowledge graph, and text data often implies a large amount of geographic entity relationship information. Aiming at the problem of complex semantic relationships of geographic entities in text data, this paper proposes a geographic entity relationship extraction model based on DeBERTa, which takes the geographic domain text on the web as the data source, introduces a BiLSTM layer to capture the bidirectional context semantics, and optimizes the sequence labelling by using a CRF layer to complete the geographic entity relationship extraction task. Experiments demonstrate that the model has excellent performance in terms of precision, recall, and F1 value compared with the baseline model, thus verifying its effectiveness in reducing the cost of manual annotation and improving the extraction accuracy.