Table 3. Analysis of the proposed method with previous approaches.

	Method	Precision (%)	Recall (%)	F-score (%)
Figure Extraction	Approach used in Li, Jiang & Shatkay (2018)	90.24	88.10	85.64
	Approach used in Clark & Divvala (2015)	71.09	68.21	70.74
	The approach used in Choudhury et al. (2013)	88.12	61.45	74.62
	The approach used in Naiman, Williams & Goodman (2022)	91.08	90.24	88.84
	Proposed approach	96.73	94.21	95.86
Caption Extraction	Approach used in Li, Jiang & Shatkay (2018)	73.16	84.74	79.59
	Approach used in Clark & Divvala (2015)	33.10	40.34	37.27
	The approach used in Choudhury et al. (2013)	81.52	76.95	75.62
	The approach used in Naiman, Williams & Goodman (2022)	90.14	89.76	85.29
	Proposed approach	92.87	87.14	89.94
Figure-Caption Pair Extraction	Approach used in Li, Jiang & Shatkay (2018)	88.59	83.17	86.13
	Approach used in Clark & Divvala (2015)	51.14	60.98	54.11
	The approach used in Choudhury et al. (2013)	64.17	61.30	62.84
	The approach used in Naiman, Williams & Goodman (2022)	89.76	85.49	87.18
	Proposed approach	91.76	88.12	90.17