Table 7. Comparison of fine-tuned BERT with baseline using Wilcoxon’s signed-rank test (damage vs non-damage).

Disasters	Binary classification
	Fine-tuned BERT	Second best	Diff	Rank
Iraq–Iran earthquake	95.12	78.48	16.64	7
Sri Lanka floods	85.13	86.89	−1.76	1
Mexico earthquake	88.80	79.26	9.54	6
California wildfires	80.03	73.24	6.79	5
Hurricane Harvey	78.89	76.28	2.61	3
Hurricane Maria	79.79	79.1	0.69	2
Hurricane Irma	82.47	77.1	5.37	4