GhoSt-NN: A Representative Gold Standard of German Noun-Noun Compounds

GhoSt-NN: A Representative Gold Standard of German Noun-Noun Compounds Sabine Schulte im Walde, Anna H¨atty, Stefan Bott, Nana Khvtisavrishvili Institute for Natural Language Processing University of Stuttgart, Germany {schulte, anna.haetty, stefan.bott, nana.khvtisavrishvili}@ims.uni-stuttgart.de Abstract This paper presents a novel gold standard of German noun-noun compounds Ghost-NN including 868 compounds annotated with corpus frequencies of the compounds and their constituents, productivity and ambiguity of the constituents, semantic relations between the constituents, and compositionality ratings of compound–constituent pairs. Moreover, a subset of the compounds containing 180 compounds is balanced for the productivity of the modifiers (distinguishing low/mid/high productivity) and the ambiguity of the heads (distinguishing between heads with 1, 2 and >2 senses). Keywords: noun–noun compounds, compositionality ratings, semantic relations 1. Introduction Compounds are combinations of two or more simplex words. They have been a recurrent focus of attention within theoretical, cognitive, and in the last decade also within computational linguistics. Our focus of interest is on Ger- man noun-noun compounds,1 such as Ahornblatt ‘maple leaf’, Feuerwerk ‘fireworks’, and Obstkuchen ‘fruit cake’, where both the grammatical head (in German, this is the rightmost constituent) and the modifier are nouns. More specifically, we are interested in the degrees of composi- tionality of German noun-noun compounds, i.e., the seman- tic relatedness between the meaning of a compound (e.g., Feuerwerk) and the meanings of its constituents (e.g., Feuer ‘fire’ and Werk ‘opus’). The compositionality of noun-noun compounds has at- tracted much research across languages over the years. From a psycholinguistic point of view, researchers are in- terested in finding out how compound words are cognitively processed and represented in the mental lexicon. There is an ongoing debate about whether morphologically complex words are stored and processed as single units (full listing approach (Butterworth, 1983)), whether they are decom- posed into their morphemes (Taft and Forster, 1975; Taft, 2004), or whether they can be accessed both ways: as whole forms and componentially (dual route models, cf. Cara- mazza et al. (1988), Baayen and Schreuder (1999)). From a computational point of view, addressing the compo- sitionality of noun compounds (and multi-word expressions in more general) is a crucial ingredient for lexicography and NLP applications, to know whether the expression should be treated as a whole, or through its constituents, and what the expression means. For example, studies such as Cholakov and Kordoni (2014); Weller et al. (2014); Cap et al. (2015); and Salehi et al. (2015) have integrated the pre- diction of compositionality into Statistical Machine Trans- lation. Accordingly, research across languages has aimed for predicting the compositionality of noun compounds au- tomatically. For example, Reddy et al. (2011) predicted 1See Fleischer and Barz (2012) for a detailed overview and Klos (2011) for a recent detailed exploration. the compositionality of 90 English noun–noun compounds via distributional information. Similarly, Schulte im Walde et al. (2013) assessed various types of distributional fea- tures to predict the compositionality of 244 German noun– noun compounds. Salehi and Cook (2013) and Salehi et al. (2014) explored multi-lingual dictionaries and distribu- tional evidence to …

Content truncated to 500 words