EPJ Data Science

Table 8 Mean rank table for all datasets

From: SentiBench - a benchmark comparison of state-of-the-practice sentiment analysis methods

3-classes			2-classes
Pos	Method	Mean Rank	Pos	Method	Mean Rank	Coverage (%)
1	VADER	4.00 (4.17)	1	SentiStrength	2.33 (3.00)	29.30 (28.91)
2	LIWC15	4.62	2	Sentiment140	3.44	39.29
3	AFINN	4.69	3	Semantria	4.61	62.34
4	Opinion Lexicon	5.00	4	Opinion Lexicon	6.72	69.50
5	Semantria	5.31	5	LIWC15	7.33	68.28
6	Umigon	5.77	6	SO-CAL	7.61	72.64
7	SO-CAL	7.23	7	AFINN	8.11	73.05
8	Pattern.en	9.92	8	VADER	9.17 (9.79)	82.20 (83.18)
9	Sentiment140	10.92	9	Umigon	9.39	64.11
10	Emolex	11.38	10	PANAS-t	10.17	5.10
11	Opinion Finder	13.08	11	Emoticons	10.39	10.69
12	SentiWordNet	13.38	12	Pattern.en	12.61	65.02
13	Sentiment140_L	13.54	13	SenticNet	13.61	84.00
14	SenticNet	13.62	14	Emolex	14.50	66.12
15	SentiStrength	13.69 (13.71)	15	Opinion Finder	14.72	46.63
16	SASA	14.77	16	USent	14.89	44.00
17	Stanford DM	15.85	17	Sentiment140_L	14.94	93.36
18	USent	15.92	18	NRC Hashtag	17.17	93.52
19	NRC Hashtag	16.31	19	Stanford DM	17.39	87.32
20	LIWC	16.46	20	SentiWordNet	17.50	61.77
21	ANEW_SUB	18.54	21	SASA	18.94	60.12
22	Emoticons	21.00	22	LIWC	19.67	61.82
23	PANAS-t	21.77	23	ANEW_SUB	21.17	94.20
24	Emoticons DS	23.23	24	Emoticons DS	23.61	99.36

Back to article page