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)