A Corpus of Literal and Idiomatic Uses of German Infinitive-Verb Compounds

A Corpus of Literal and Idiomatic Uses of German Infinitive-Verb Compounds Andrea Horbach◦, Andrea Hensler⋆, Sabine Krome⋆, Jakob Prange◦, Werner Scholze-Stubenrecht×, Diana Steffen◦, Stefan Thater◦, Christian Wellner◦, Manfred Pinkal◦ ◦Department of Computational Linguistics, Saarland University, Saarbr¨ucken, Germany, {andrea, jprange, dsteffen, stth, chwell, pinkal}@coli.uni-saarland.de ⋆Redaktion Wahrig bei Brockhaus, G¨utersloh, Germany, {BMBF-Wahrig, Sabine.Krome}@bertelsmann.de ×Bibliographisches Institut GmbH Dudenverlag, Berlin, Germany, Werner.Scholze-Stubenrecht@duden.de Abstract We present an annotation study on a representative dataset of literal and idiomatic uses of infinitive-verb compounds in German newspaper and journal texts. Infinitive-verb compounds form a challenge for writers of German, because spelling regulations are different for literal and idiomatic uses. Through the participation of expert lexicographers we were able to obtain a high-quality corpus resource which is offered as a testbed for automatic idiomaticity detection and coarse-grained word-sense disambiguation. We trained a classifier on the corpus which was able to distinguish literal and idiomatic uses with an accuracy of 85%. Keywords: Corpus Annotation, Semantics, Idiom Detection 1. Introduction This paper presents a study on German infinitive-verb com- pounds with an inflected head verb and an infinitive modifier. Specifically, we consider infinitive-verb compounds with the head verbs bleiben and lassen which allow for an id- iomatic, i.e., conventionalized figurative interpretation in addition to a literal interpretation. Sentence (1) illustrates the lexicalized idiomatic sense “to have to repeat a year in school” of the compound sitzenbleiben, while (2) illustrates the semantically transparent use of the compound (“remain seated”). (1) Die Atlas-Werte bemessen sich unter anderem danach, wie gut Grundsch¨uler lesen k¨onnen, wie viele Sch¨uler sitzenbleiben, wie viele Hochschul- absolventen eine Region z¨ahlt. (2) Einmal schildert Grosz das Beispiel einer Frau na- mens Marissa Panigrosso, die in ihrem B¨uro im New Yorker World Trade Center nach dem Ein- schlag des ersten Flugzeugs am 11. September 2001 vom Schreibtisch aufsprang und durchs Treppen- haus auf die Straße fl¨uchtete, w¨ahrend ihre liebsten Kolleginnen einfach im B¨uro sitzen blieben. The distinction between literal and idiomatic uses is a prere- quisite for correct spelling. According to the current German spelling rules1, literal uses of infinitive-verb combinations must be written as two separate words (as in (2)), while idiomatic uses of compounds with bleiben and lassen as inflected head verbs can also be written in one word (as in (1)). Our motivation for addressing this task is two-fold: the problem of idiom recognition is an interesting challenge from the point of view of computational semantics; due to the influence of the idiomaticity status on spelling, there is a societal and political motivation as well. Spelling became a political issue in Germany through the German orthography reform, which was decreed in 1996 1http://www.rechtschreibrat.com and caused a fierce controversy carried out in public as well as uncertainty even among professional writers about the correct spelling. As a consequence of this tumult caused by the reform, the Council for German Orthography (Rat f¨ur deutsche Rechtschreibung) was established in 2004. Its central responsibility is to observe and analyze whether the writing community follows the official orthographic regu- lations and to make …

Content truncated to 500 words