This obvious dependency feature is not yet included in the frame identification. (the other depLabs already in place are: those between head-children and those between head's parent and grandparent)
In my experiments using the standard datasets, this gives me +0.15 in frameid accuracy using the basic IdFeatureExtractor.
This obvious dependency feature is not yet included in the frame identification. (the other depLabs already in place are: those between head-children and those between head's parent and grandparent)
In my experiments using the standard datasets, this gives me +0.15 in frameid accuracy using the basic IdFeatureExtractor.