Table 2.

The annotation results for the training and evaluation datasets: date range, size (no. of comments annotated twice), and the annotator agreement in terms of interval Krippendorff’s $\mathit{Alpha}$  and accuracy ($\mathit{Acc}$).

Dataset	Dates	Size	$\mathit{Alpha}$	$\mathit{Acc}$
Training	Jan.–Apr. 2020	59,870	0.59	0.77
Evaluation	May 2020	10,543	0.56	0.82