MIT Press

Skip to Main Content

Table 5

Performance of the CRF classifier on the development set on event mention detection and on event mention detection + classification with different context windows.

CONTEXT	MENTION DETECTION ONLY			DETECTION + CLASSIFICATION
CONTEXT	P	R	F1	P	R	F1
0	81.41%	70.06%	74.10%	38.29%	23.99%	27.73%
±1	85.42%	79.04%	81.95%	37.72%	25.99%	29.27%
±2	86.74%	80.60%	83.41%	41.00%	25.46%	29.13%
±3	87.89%	79.72%	83.34%	35.79%	24.21%	27.65%
±4	87.06%	79.49%	82.87%	32.37%	23.28%	25.78%

CONTEXT	MENTION DETECTION ONLY			DETECTION + CLASSIFICATION
CONTEXT	P	R	F1	P	R	F1
0	81.41%	70.06%	74.10%	38.29%	23.99%	27.73%
±1	85.42%	79.04%	81.95%	37.72%	25.99%	29.27%
±2	86.74%	80.60%	83.41%	41.00%	25.46%	29.13%
±3	87.89%	79.72%	83.34%	35.79%	24.21%	27.65%
±4	87.06%	79.49%	82.87%	32.37%	23.28%	25.78%