A Novel Threat Intelligence Information Extraction System Combining Multiple Models

Table 1

The performance of the entity extraction task.

Model	Precision	Recall	F1	Accuracy
EEMAP-WE	69.23	67.71	68.46	61.12
BiLSTM [11]	69.80	66.62	68.17	62.11
BiLSTM-CRF [13]	70.10	68.40	69.24	62.77
EEMAP-BERT (our model)	79.02	77.22	78.11	74.13
No POS	78.43	76.46	77.44	73.39
No attention	78.56	76.84	77.69	73.57
No POS + no attention	75.97	74.74	75.35	70.68