Evaluation - task1-intertass-PE
1428 tweets in qrel, 1424 tweets in run.

Accuracy: 0.355

Confusion matrix
Real	Pred.	#
N	NEU	266
N	P	39
N	NONE	64
N	N	103
NEU	NEU	197
NEU	N	45
NEU	NONE	73
NEU	P	51
NEU	?	1
NONE	NEU	79
NONE	NONE	32
NONE	P	26
NONE	N	22
P	N	36
P	NEU	165
P	?	3
P	NONE	51
P	P	175

Category evaluation
Cat	P	R	F1
N	0.500	0.218	0.304
NEU	0.279	0.537	0.367
NONE	0.145	0.201	0.169
P	0.601	0.407	0.485

Macro-Precision: 0.381
Macro-Recall: 0.341
Macro-F1: 0.360