Skip to main content

Table 19 Percentage of predicted non-existent class IDs out of the total number of predicted mismatch class IDs for the core set for the training, validation and evaluation sets

From: Concept recognition as a machine translation problem

Ontology

% Non-existent class IDs in training (%)

% Non-existent class IDs in validation (%)

% Non-existent class IDs in evaluation (%)

ChEBI

3

4

11

CL

0

0

0

GO_BP

2

2

2

GO_CC

2

4

1

GO_MF

1

1

50

MOP

0

2

0

NCBITaxon

7

7

11

PR

10

10

2

SO

0

1

0

UBERON

0

1

2