Fig. 2

The annotation dataset distribution and brat windows. (a) The perovskite literature dataset annotations. The colored blocks depict the 40-snippet subgroups of different materials: yellow for CsPbI₃, blue for FAPI, green for CsPbBr₃, violet for MAPI and red for MAPB. Annotators are marked as A1 for annotator 1, A2 for annotator 2 etc. Every snippet was annotated by two annotators. (b) Labeled snippet (from article⁵⁹) in brat annotation tool. Name CsPbBr₃ is labeled and in this snippet the material name is mentioned four times. (c) The annotation dialog of brat. From the dialog the user can select bandgap type from four different choices. Under the entity type dialog there is a type of NOT CLEAR answer, which could be selected if the annotator was not sure what to annotate.