Arguments and Adjuncts in Universal Dependencies

Proceedings of the 27th International Conference on Computational Linguistics, pages 3837–3852 Santa Fe, New Mexico, USA, August 20-26, 2018. 3837 Arguments and Adjuncts in Universal Dependencies Adam Przepiórkowski Institute of Philosophy University of Warsaw and Institute of Computer Science Polish Academy of Sciences ul. Jana Kazimierza 5 01-248 Warszawa, Poland adamp@ipipan.waw.pl Agnieszka Patejuk Institute of Computer Science Polish Academy of Sciences ul. Jana Kazimierza 5 01-248 Warszawa, Poland aep@ipipan.waw.pl Abstract The aim of this paper is to argue for a coherent Universal Dependencies approach to the core vs. non-core distinction. We demonstrate inconsistencies in the current version 2 of UD in this respect – mostly resulting from the preservation of the argument–adjunct dichotomy despite the declared avoidance of this distinction – and propose a relatively conservative modification of UD that is free from these problems. 1 Introduction Universal Dependencies (UD; Nivre et al. 2016) has recently become a de facto standard for the rep- resentation of dependency treebanks. An important feature of UD is that it attempts not to represent the argument–adjunct distinction. This is stated explicitly in the paper introducing UD (Nivre et al. 2016: 1661): “The scheme [. . . ] makes a distinction between core arguments (e.g., subject and object) and other dependents, but does not attempt to distinguish complements vs. adjuncts” (recall that com- plements are non-subject arguments), and it is confirmed on the current official UD webpage:1 “[The UD taxonomy] does not make a distinction between adjuncts (general modifiers) versus oblique argu- ments (arguments said to be selected by a head but not expressed as a core argument).” This decision is not justified in Nivre et al. 2016, but it is briefly motivated on the UD webpage (cf. fn. 1): “[T]he argument/adjunct distinction is subtle, unclear, and frequently argued over. For instance, syntacticians at certain times have argued for various obliques to be arguments, while at other times arguing that they are adjuncts, particularly for certain semantic roles such as oblique instruments or sources. We take the dis- tinction to be sufficiently subtle (and its existence as a categorical distinction sufficiently questionable) that the best practical solution is to eliminate it.” This view goes against the theoretical-linguistic received wisdom – linguistic theories invariably as- sume a fundamental argument–adjunct dichotomy (AAD). However, it has been repeatedly observed that AAD has been flawed ever since its conception: the three criteria for distinguishing arguments from ad- juncts given by Tesnière (1959) – the father of modern valency theory – are pairwise incompatible (Vater 1978a, 1978b, Przepiórkowski 1999). Multiple tests proposed since then have often been inconsistent with linguists’ intuitions about AAD (and with each other) and short-lived; as cautiously noted by Tu- tunjian and Boland (2008, 633), “[t]he sheer number of [purported argument/adjunct] tests underlines the fact that no single test is entirely satisfactory”. The existence of such a universal dichotomy is also called into doubt by typologists (see, e.g., Croft 2001: 272–280 and Haspelmath 2014). We have argued recently that even those theories which put much emphasis …

Content truncated to 500 words