Chemical Entity Recognition and Resolution to ChEBI

Table 3

Evaluation of entity identification, subset of the gold standard composed by 9,696 chemical entities that contain a mapping to ChEBI. Results of entity identification (named entity recognition and resolution) for each alignment and method are shown in this table. The dictionary method recognized and mapped a total of 18,683 entities while the machine-learning method recognized and mapped 10,681 entities. True positives (TP) is the amount of entity recognitions that agree with the gold standard and for which the mapping also agrees with the gold standard. Values of precision, recall, and F-measure are presented.

Assessment	Method	TP	Precision	Recall	F-measure
Exact matching	Dictionary	4,530	24.25	46.72	31.93
	Machine learning	4,783	44.79	49.33	46.95
Left matching	Dictionary	4,559	24.40	47.02	35.13
	Machine learning	4,972	46.56	51.28	48.81
Right matching	Dictionary	4,592	24.58	47.36	32.36
	Machine learning	4,885	45.75	50.38	47.95
Left/right matching	Dictionary	4,621	24.73	47.67	32.57
	Machine learning	5,074	47.52	52.33	49.81
Partial matching	Dictionary	5,185	27.75	53.48	36.54
	Machine learning	5,202	48.72	53.65	51.07