Table 9.

F-score on the Sarcasm detection task

		Baselines		Our resources				Combinations
Class	Model	N	L${}_{\mathit{tr}}$	L${}_g$	L${}_{\mathit{tw}}$	L	E	NL	NE	LE	NLE
Non-sarcastic	MC	88.47	88.47	88.47	88.47	88.47	88.47	88.47	88.47	88.47	88.47
	LR	92.75	88.48	88.76	91.00	91.21	90.87	92.79	91.97	91.33	91.85
	RF	92.93	88.51	88.73	90.11	90.42	93.01	91.59	92.65	92.96	92.81
	SVM	92.34	88.49	88.59	87.20	87.22	92.64	92.30	93.46	92.28	93.40
Sarcastic	MC	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.00
	LR	70.94	0.77	22.43	57.70	59.05	64.52	71.37	67.93	66.21	67.92
	RF	71.61	12.11	33.43	50.72	52.10	68.50	59.72	65.53	68.84	67.04
	SVM	72.32	11.79	21.70	33.99	39.31	68.63	71.50	73.14	68.50	73.10
Macro-average	MC	44.23	44.23	44.23	44.23	44.23	44.23	44.23	44.23	44.23	44.23
	LR	81.85	44.63	55.59	74.35	75.13	77.69	82.08	79.95	78.77	79.88
	RF	82.27	50.31	61.08	70.41	71.26	80.76	75.65	79.09	80.90	79.93
	SVM	82.33	50.14	55.14	60.60	63.26	80.64	81.90	83.30	80.39	83.25

The best performing feature set per algorithm is highlighted in bold