Skip to main content

Table 14 Mean rank table for datasets of comments

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 3.33 (3.60) 1 SentiStrength 1.17 (1.50) 28.29 (24.02)
2 AFINN 4.33 2 Semantria 2.83 61.02
3 Opinion Lexicon 4.33 3 Sentiment140 4.17 36.49
4 Semantria 4.50 4 Opinion Lexicon 6.50 71.59
5 SO-CAL 5.17 5 LIWC15 6.67 65.80
6 LIWC15 6.17 6 AFINN 7.00 74.21
7 Umigon 9.50 7 SO-CAL 7.50 74.59
8 Emolex 10.33 8 VADER 9.50 (9.60) 81.98 (85.34)
9 Sentiment140_L 11.33 9 Umigon 10.50 57.87
10 Stanford DM 11.67 10 Emoticons 11.83 4.99
11 NRC Hashtag 12.00 11 Opinion Finder 13.00 55.66
12 Pattern.en 12.67 12 SenticNet 13.00 95.28
13 SenticNet 13.00 13 USent 14.00 45.66
14 Opinion Finder 13.17 14 NRC Hashtag 14.67 93.43
15 SentiWordNet 13.17 15 Emolex 15.00 69.69
16 SASA 14.67 16 PANAS-t 15.50 5.10
17 SentiStrength 15.17 (19.00) 17 Stanford DM 15.67 84.43
18 Sentiment140 15.50 18 Pattern.en 15.83 59.00
19 USent 15.83 19 Sentiment140_L 15.83 92.30
20 LIWC 17.67 20 SentiWordNet 17.00 63.32
21 ANEW_SUB 17.83 21 SASA 17.50 61.91
22 Emoticons DS 22.67 22 LIWC 19.67 62.24
23 PANAS-t 22.83 23 ANEW_SUB 22.00 94.31
24 Emoticons 23.17 24 Emoticons DS 23.67 99.31