Research Article

Few-Shot Relation Extraction via the Entity Feature Enhancement and Attention-Based Prototypical Network

Table 3

Experimental results on the TinyRel-CM dataset.

ModelsGroupD-DD-SD-FD-UF-US-F
NN = 5N = 2N = 6N = 6N = 2N = 6

GNNK = 526.0266.2237.4844.4755.0237.13
K = 1027.9658.7537.8845.1555.5338.39
K = 1528.3670.1043.2551.4056.4540.37

SNAILK = 524.6458.5528.5729.0553.3027.88
K = 1027.3857.1227.8533.5055.6030.52
K = 1523.8457.4030.6335.7556.3530.43

ProtoK = 531.5069.5634.3644.7657.4340.36
K = 1035.4572.5437.1148.2858.7143.55
K = 1538.2872.7937.9050.0360.0944.97

HATTK = 538.2175.0046.6754.1773.7544.17
K = 1044.4571.2449.3751.6775.0049.17
K = 1549.6180.4144.3356.6772.5849.17

MLMANK = 538.1764.9845.3251.3064.2448.34
K = 1044.8970.0250.8356.6070.7953.69
K = 1549.2372.4154.2559.6774.5856.67

BERT-PAIRK = 526.8759.5252.5443.2863.3349.95
K = 1028.1962.3255.8646.1366.7052.92
K = 1529.2563.9857.6047.8168.8354.16

OursK = 552.4766.9869.2571.0371.2165.20
K = 1060.6372.1771.4175.7678.3669.15
K = 1565.4073.7371.7678.2080.3871.22

Acc. %. The bold values represent the state-of-the-art results under the current settings.