Sem-K-BERT, which can enhance BERT by using knowledge graph and semantic role tag information , and we used location coding to ensure that the normal work of external information would not be interfered with each other. Experiments show that our model has achieved better performance on a wide range of Chinese nlp tasks. This shows that the simultaneous use of knowledge graph and semantic role labeling information can bring greater improvement to BERT.
Sem-K-BERT, which can enhance BERT by using knowledge graph and semantic role tag information , and we used location coding to ensure that the normal work of external information would not be interfered with each other. Experiments show that our model has achieved better performance on a wide range of Chinese nlp tasks. This shows that the simultaneous use of knowledge graph and semantic role labeling information can bring greater improvement to BERT.
Sem-K-BERT, which can enhance BERT by using knowledge graph and semantic role tag information , and we used location coding to ensure that the normal work of external information would not be interfered with each other. Experiments show that our model has achieved better performance on a wide range of Chinese nlp tasks. This shows that the simultaneous use of knowledge graph and semantic role labeling information can bring greater improvement to BERT.