Named Entity Recognition in XLNet Cyberspace Security Domain Based on Dictionary Embedding

With the increase of network security incidents, network security analysts need to analyze massive log information. The introduction of knowledge graph into the field of network security can facilitate analysis by security analysts. NER (Named Entity Recognition) is the upstream task of knowledge graph construction, and the quality of the NER model determines the quality of the knowledge graph to a certain extent. However, the general domain named entity recognition model cannot extract the entities in the network security domain very well. For this phenomenon, this paper proposes an XLNet-Feature-Att model, which uses XLNet in the embedding layer to embed words into the vector space, the encoding layer uses the improved BILSTM FB structure and the Attention layer, and the decoding layer uses the CRF model to achieve sequence labeling and binding. Finally, the experimental comparison is carried out on the data set in the field of network security, and the F1-score reaches the highest 92.28%.