BiLSTM-CRF with Compensation Method for Spatial Entity Recognition

As a basic task, named entity recognition (NER) plays a very important role in the field of natural language processing. In recent years, the neural network method has achieved excellent results in NER. However, the NER method is not very effective in fine-grained entity recognition tasks in the subdivision field. In order to solve this problem, we proposed the BiLSTM-CRF model with compensation method (BiLSTM-CC) by increasing the vector representing the semantic information of the word and compensating the model output. In the task of spatial entity recognition, the improved algorithm shows excellent performance.