Named Entity Recognition in the Tourism Domain Based on Attention Mechanisms

Aiming at the problems of long length, complex structure, serious nesting and many types of entities in the field of tourism, a method of tourism entity recognition based on BERT-BiLSTM-TextCNN-Attention-CRF model was proposed. The BiLSTM model and TextCNN combined model were used to further extract the medium and long term dependency of text and enhance the ability to capture local features. Introduction of Attention attention mechanism to further improve the attention to key information; Finally, the tags are optimized by conditional random field (CRF), and the optimal tag sequence is obtained. Through the experiment on the established tourism data set, it is found that the BERT-BiLSTM-TextCNN-Attention-CRF model has achieved a good effect in practical application, which verifies the effectiveness of this method for named entity recognition.