A Named Entity Recognition Method Combining Keyword and Self attention Mechanism

As the basis of natural language processing tasks such as information extraction and machine question answering, named entity recognition has important research significance. However, traditional models have less consideration for utilizing the feature information of the text. To address this issue, this paper proposes a named entity recognition model that integrates keyword and self attention mechanism based on the HBT model.The model has made two improvements from the perspective of enhancing the utilization of text's own feature information: one is to introduce keyword feature vectors, and the other is to use multi head self attention mechanism for encoding layer vectors. The experimental results show that the improved model achieves an F1 value of 70.74% on the DuIE2.0 dataset, which is significantly improved compared to the benchmark model.