Table 6.

Test results for both subtasks using the final hybrid solution: rule-based engine combined with deep learning module for observation extraction.

Subtask and annotation type		Precision	Recall	F1 score
Subtask 1
	Family members	0.7887	0.8307	0.8092
	Observations	0.7523	0.8332	0.7907
	Overall	0.7662	0.8322	0.7979
Subtask 2
	Living status	0.5964	0.6462	0.6248
	Observations	0.4635	0.4371	0.4499
	Overall	0.5100	0.5063	0.5082