Skip to main content

Table 1 Evaluation of the trained model on a test set of 200,000 molecules

From: Translating the InChI: adapting neural machine translation to predict IUPAC names from a chemical identifier

Subset

Accuracy (whole name)

Normalized edit distancea

All

0.91

0.02 ± 0.03

Organic

0.91

0.02 ± 0.03

Inorganic

0.14

0.32 ± 0.20

Organometallicb

0.20

0.37 ± 0.24

Other organic–inorganic mixture

0.50

0.15 ± 0.18

  1. aAverage over subset, with dispersion indicated by mean absolute deviation
  2. bDefined as having an explicit carbon–metal bond