Table 3.

Performance of the algorithms employed

Tested algorithm	Performance of the algorithm	Accuracy	Precision	Recall	F1-score
rule-based 1	in identifying smoking related sentences	94.3	99.0	93.3	96.1
ULMFiT-based language model	in classifying smoking statuses current smoker	85.9	66.2	96.2	78.5
	ex-smoker	89.9	97.6	68.3	80.4
	never smoker	93.4	94.9	89.2	91.9
rule-based 2	in identifying sentences related to smoking cessation interventions	87.0	94.0	82.5	87.9