Combining the External Medical Knowledge Graph Embedding to Improve the Performance of Syndrome Differentiation Model

Table 3

Performance of different embedding models on KDSD.

Methods	P@1	P@3	P@5
BERT	0.89	0.876	0.845
+TransE	0.9	0.879	0.852
+DistMult	0.878	0.856	0.829
+ComplEx	0.868	0.852	0.825
+ConvKB	0.886	0.862	0.833