- All Implemented Interfaces:
- CoreAnnotation<java.lang.Integer>, TypesafeMap.Key<java.lang.Integer>
- Enclosing class:
- CoreAnnotations
public static class CoreAnnotations.EmptyIndexAnnotation
extends java.lang.Object
implements CoreAnnotation<java.lang.Integer>
Some datasets - for example, the UD Estonian EWT dataset - use
"empty" nodes to represent words that were unspoken / unwritten
but can be inferred from the structure of the sentence. For
example, in English, one could say "Gimme" instead of "Give me
it", and "it" could be treated as an empty word. A more common
example is when it is used in a similar manner to the copy nodes,
but displaced in time. So, for example, a sentence which uses
them in the UD English EWT dataset (no relation) is:
"Over 300 Iraqis are reported dead and 500 wounded in Fallujah alone."
Here, one could build a dependency graph using "reported" as a
copy node, but instead the en_ewt dataset creates an "empty" node
and builds the enhanced dependencies using that node.
"Over 300 Iraqis are reported dead and 500 *reported* wounded in Fallujah alone."
Rather than the second "reported" being a copy of word 5, it is treated
as a separate word 8.1
As with IndexAnnotation, we count from 1.