Skip to main content
. Author manuscript; available in PMC: 2022 Mar 8.
Published in final edited form as: Proc AAAI ACM Conf AI Ethics Soc. 2020 Feb 7;2020:337–342. doi: 10.1145/3375627.3375841

Table 2:

SVM performance trained on hand-labeled vs distantly-labeled data. The difference between F1 scores is statistically significant with p=0.001.

Loss Other Aggression
Labels p r f p r f p r f Macro F1
Gold 77.08 56.92 65.49 88.04 95.76 91.74 50 27.59 35.56 64.26
Distant 50.00 48.46 49.22 85.63 84.50 85.06 19.72 24.14 21.71 52.00