The ML PR doesn't delete the features from the old annotations, it simply creates new annotations in parallel with the existing ones.
I'm planning to change that so it will add the new feature to an existing annotation (if it can find one) but I can't promise when this will happen.
| -----Original Message-----
| From: owner-gate@... [mailto:owner-gate@...] On
| Behalf Of Partha Lal
| Sent: Monday, July 14, 2003 11:48 AM
| To: gate-crashers@...
| Subject: Feature Request for MachingLearningPR
| I'm using the MachineLearningPR, and whats really infuriating is that way
| it deletes all features from Annotations it classifies. (so after
| classification the only feature in the Annotation is the class). I've got
| other useful information stored as features, and I'm having to recompute
| So, could MachineLearningPR just edit the value of the class feature only
| and leave the others? (for the case when the class is a feature rather
| than the presence/absence of an Annotation)
| Partha Lal
| MPhil Student (Computer Speech, Text & Internet Technology)
| University of Cambridge