BERT-BiLSTM-CRF-DRL-A Named Entity Recognition Method for Knowledge Graph

This paper conducts research around the key technologies in the construction of the cybersecurity knowledge graph, with a focus on the naming entity recognition technology. It analyzes the problems that the traditional naming entity recognition methods have low accuracy in entity recognition and cannot accurately extract entities and their corresponding versions in sample data, thereby resulting in the “wrong answers” problem in the constructed knowledge graph. To address this issue, this paper proposes to improve the model using technologies such as pre-trained models and deep reinforcement learning. And a comparative experiment based on sample data, the results of the traditional model and the improved model is designed. Through the comparison of the experimental results and the verification of the effects, it is proved that the improved model solves the problem of “wrong answers” existing in the traditional model.