de-nncom-sem - Dataset of German noun-noun compounds annotated with semantic relations (properties) and prepositions

Contains 8005 compounds, annotated with a semantic relation (a property) and a preposition. Each line of the file contains the following information:                 
                 col1 - the data split (if train, test, dev or test-iaa); these data splits were used to produce the results reported in chapter 7 of Dima (2019).
                 col2 - the compound, e.g. Dreiecktuch
                 col3 - the modifier - the first constituent of the word, e.g. Dreieck (note that, like in the case of Dreieck, the modifier can be a compund itself)
                 col4 - the head - the second cosntituent of the word, e.g. Tuch
                 col5 - the collapsed property (German name) - e.g. Aussehen/*; the collapsed property does not take into account the direction of the semantic relation
                 col6 - the individual property (German name) - e.g. Aussehen; the individual property does take into account the direction of the semantic relation
                 col7 - the direction - can be 1 (read as modifier relation head - e.g. Träger Teil Rock for Trägerrock) or 2 (read as head relation modifier, e.g. Mitte Teil Brücke or Brücke Teil* Mitte for for Brückenmitte).
                 col 8 - the English name of the collapsed property, e.g. appearance/*
                 col 9 - the English name of the individual property, e.g. appearance
                 col 10 - the German preposition associated with the compound
The annotations were created in the A3 project of the SFB 833 at the University of Tübingen. See Telljohann et al. (2017) for a description of the annotation guidelines.