Working on the conversion of the Swedish treebank, I have started to wonder about the treatment of lists and the use of the "list" relation. Part of the Swedish treebank consists of information brochures, which contain a lot of sentences like this one.
Special rules apply for
- people who live in Sweden
- people who live abroad
- people who live part of the year in Sweden
The original annotation in the Swedish treebank treats all three noun phrases as "nmod" of "apply", but without any overt indication that they are coordinated. Note also that the preposition "for" is shared by all of them, which suggests that it is an instance of coordination. So one possible analysis would be:
apply -nmod-> people_1
people_1 -case-> for
people_1 -conj-> people_2
people_1 -conj-> people_3
At the same time, this is a list, if I ever saw one, so another plausible analysis is:
apply -nmod-> people_1
people_1 -case-> for
people_1 -list-> people_2
people_1 -list-> people_3
I also think that many treebanks who do not have a special "list" relation annotate lists as coordination, so we probably need some guidelines here anyway.
Working on the conversion of the Swedish treebank, I have started to wonder about the treatment of lists and the use of the "list" relation. Part of the Swedish treebank consists of information brochures, which contain a lot of sentences like this one.
Special rules apply for
The original annotation in the Swedish treebank treats all three noun phrases as "nmod" of "apply", but without any overt indication that they are coordinated. Note also that the preposition "for" is shared by all of them, which suggests that it is an instance of coordination. So one possible analysis would be:
apply -nmod-> people_1
people_1 -case-> for
people_1 -conj-> people_2
people_1 -conj-> people_3
At the same time, this is a list, if I ever saw one, so another plausible analysis is:
apply -nmod-> people_1
people_1 -case-> for
people_1 -list-> people_2
people_1 -list-> people_3
I also think that many treebanks who do not have a special "list" relation annotate lists as coordination, so we probably need some guidelines here anyway.