Arg0:creator
Arg1:thing created
Arg2:source
Arg3:benefactive
Arg4:attribute of arg1
transitive (-) | |
---|---|
| |
Arg0: | John |
REL: | redeveloped |
Arg1: | a strong argument |
transitive-1 (-) | |
---|---|
| |
Arg0: | UCLA |
REL: | redeveloped |
Arg1: | a new promoter gene |
but (-) | |
---|---|
| |
Arg0: | John |
REL: | redeveloped |
Arg1: | a new promoter gene |
ArgM-LOC: | at UCLA |
creation as DO (-) | |
---|---|
| |
Arg0: | *trace* -> Mitsubishi |
REL: | redevelop |
Arg2: | Marunouchi |
Arg1-into: | a business center of high-tech buildings |
with source (-) | |
---|---|
| |
Arg0: | John |
REL: | redeveloped |
Arg1: | a strong argument |
Arg2-from: | from the footnotes of the paper |
with end-result (-) | |
---|---|
| |
Arg0: | John |
REL: | redeveloped |
Arg2: | the footnotes |
Arg1-into: | into a strong argument |
with benefactive (-) | |
---|---|
| |
Arg0: | Bill Gates |
REL: | redeveloped |
Arg1: | DOS |
Arg3-for: | for Microsoft |
these are also -PRPs, but the tag -PRD can be changed to -PRD-PRP during postprocessing
with secondary predication (-) | |
---|---|
| |
Arg0: | Alber M. Kligman |
REL: | redeveloped |
Arg1: | Retin-A |
Arg4: | to combat acne |
with secondary predication (-) | |
---|---|
| |
Arg0: | Alber M. Kligman |
REL: | redeveloped |
Arg1: | Retin-A |
Arg4: | for combating acne |
The above structure should be used in all cases, even when it feels like the Arg1 and the purpose clause (secondary predicate) are very closely linked and feel like they should be a single argument. Treebank is inconsistent in it treatment of these; sometimes it's split into two nodes, sometimes there's a single node holding both. If we are consistent in dividing these arguments, it's relatively easy for someone else to merge them. The reverse is not true.
Manner (-) | |
---|---|
| |
Arg0: | John |
REL: | redeveloped |
Arg1: | the beach |
ArgM-MNR: | through a trust |
Plus the usual assortment of TMPs, LOCs (easily confusable with MNRs, nota bene), ADVs, etc.