Yoshinobu KANO wrote:
We have developed a comparison tool which seems related to what you are
looking for.
Our tool compares annotations inside a CAS, and visualizes results.
However, generally speaking, it is always the problem which annotations
could/should be compared.
Could you explain your purpose in a bit more detailed?

Well, in a simple scenario, say NER, we just want to compare the entity mentions found by different components (either different analysis components, or e.g. a gold standard). Features of the annotation types should also be considered during comparison. I thought that such a comparator would make new annotations to the CAS for TP/FP/... so that from such annotations overall scores (R/P/F) can be calculated.

Katrin

Reply via email to