The maps on the table were created with some of the individual
Entity Linking annotators (see the
Linker column)
whose results are displayed on the
interface.
For details how the maps were created, see Ruiz Fabo
(2017)
1,
part III.
The point of the interface was to help a user choose annotations based on measures
indicating the likely quality of those annotations.
Two corpus-maps showing how
the interface can help in this respect are given below, from Ruiz Fabo
(2017)
1.
-
The top map (below) shows organizations identified by Entity Linking
tools in the corpus (without the weighted-vote filtering that applies in the
Auto-Selection tab).
Some organizations whose scores on the UI are low (suggesting a likely error)
are highligthed in the map in red or in blue (see Ruiz Fabo,
20171,
p. 157ff.
for details).
These annotations were verified to be errors.
-
The bottom map (below) is the result of correcting those erroneous organizations.