Concept recognition as a machine translation problem

BMC Bioinformatics

Table 18 Exact match results for the unseen and seen text mentions (relative to the training data) for the core + extensions evaluation annotation set of the 30 held-out documents

Ontology	Total/unique # unseen Mentions	% Unseen OpenNMT class ID (%)	% Seen OpenNMT class ID (%)	% Unseen OpenNMT character (%)	% Seen OpenNMT character (%)
ChEBI_EXT	476/188	32	92	67	85
CL_EXT	775/209	36	99	77	97
GO_BP_EXT	861/431	26	89	57	78
GO_CC_EXT	339/113	39	99	57	98
GO_MF_EXT	515/146	31	83	45	78
MOP_EXT	21/10	67	98	85	> 99
NCBITaxon_EXT	123/79	1	86	68	94
PR_EXT	3114/429	0	25	66	75
SO_EXT	318/183	51	94	69	92
UBERON_EXT	1609/532	23	96	79	95

Reporting the total number of mentions and the number of unique mentions along with the percent exact match on the class ID level and character level for both unseen and seen text mentions

ISSN: 1471-2105