Prepositional phrase attachment candidates (AmbiPP)

This dataset includes prepositions with their head candidates, with 3.15 candidates per preposition on average. The data set was extracted from TüBa-D/Z, using topological field distributions of prepositions and their heads. Lines have the following columns: TüBa-D/Z sentence number Preposition Preposition tag Preposition topological field Preposition object Preposition object tag Preposition object topological field This is then followed by the candidate heads. For each candidate head, there are the following columns: Head token Head tag Head topological field Absolute distance from the preposition (negative means before the preposition) Rank distance from the preposition Head/non-head (1/0)