Continuous Similarity Learning with Shared Neural Semantic Representation for Joint Event Detection and Evolution

Table 4

Comparison of our model with variants.

Task	Event detection					Event evolution
Model	P (%)	R (%)	F1 (%)	(%)	ACC (%)	P (%)	R (%)	F1 (%)	ACC (%)
JEDE w/o shared	53.73	59.42	56.43	68.02	55.23	56.09	64.74	60.11	53.98
JEDE w/o stack	57.57	63.83	60.54	67.35	56.39	63.52	64.68	64.09	68.06
JEDE w/o time	68.36	70.19	69.26	51.11	64.94	69.41	70.75	70.07	70.04
JEDE	75.14	71.75	73.41	50.06	77.21	73.83	68.63	71.14	72.18

The bold values highlight the performance of the proposed JEDE model on several metrics. It can be seen from this ablation study that shared semantic representation, neural stacking, and time information help the proposed JEDE model achieve the optimal performance. We have described these results in Section 5.5.