Figure 2.

Precision–recall plot for the dictionary-based and the Transformer-based taggers on the seven S1000 journal categories. The task on which both methods are assessed is NER of “strain,” “species,” and “genera.” F-score contours are presented with grey dots in the plot.