CaMEL: Case Marker Extraction without Labels

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 5506 - 5516 May 22-27, 2022 c⃝2022 Association for Computational Linguistics CaMEL: Case Marker Extraction without Labels Leonie Weissweiler*, Valentin Hofmann†*, Masoud Jalili Sabet*, Hinrich Schütze* *Center for Information and Language Processing, LMU Munich †Faculty of Linguistics, University of Oxford {weissweiler,masoud}@cis.lmu.de valentin.hofmann@ling-phil.ox.ac.uk Abstract We introduce CaMEL (Case Marker Extraction without Labels), a novel and chal- lenging task in computational morphology that is especially relevant for low-resource languages. We propose a first model for CaMEL that uses a massively multilingual corpus to extract case markers in 83 languages based only on a noun phrase chunker and an alignment system. To evaluate CaMEL, we automatically construct a silver standard from UniMorph. The case markers extracted by our model can be used to detect and visualise similarities and differences between the case systems of different languages as well as to annotate fine-grained deep cases in languages in which they are not overtly marked. 1 Introduction What is a case? Linguistic scholarship has shown that there is an intimate relationship between mor- phological case marking on the one hand and se- mantic content on the other (see Blake (1994) and Grimm (2011) for overviews). For example, the Latin case marker -ibus1 (Ablative or Dative Plural) can express the semantic category of location. It has been observed that there is a small number of such semantic categories frequently found cross- linguistically (Fillmore, 1968; Jakobson, 1984), which are variously called case roles or deep cases. Semiotically, the described situation is complicated by the fact that the relationship between case mark- ers and expressed semantic categories is seldom isomorphic, i.e., there is both case polysemy (one case, several meanings) and case homonymy or case syncretism (several cases, one marker) (Baer- man, 2009). As illustrated in Figure 1, the Latin Ablative marker -ibus can express the semantic 1In this paper, we use italic when talking about case mark- ers as morphemes in a linguistic context and monospace (accompanied by $ to mark word boundaries) when talking about case markers in the context of our model. Translitera- tions of Cyrillic examples are given after slashes. -ibus ABL DAT I L R -ами INST I -ах LOC L -ам DAT R Figure 1: Morpho-semiotic foundation of this study. The Latin case marker -ibus is used for both the Abla- tive (ABL) and the Dative (DAT), which in turn express the three semantic categories of instrument (I), location (L), and recipient (R). This is an example of both case polysemy (one case: ABL, several meanings: I and L) and case syncretism (several cases: ABL and DAT, one marker: -ibus). Russian, on the other hand, has an isomorphic relationship between Instrumental (INST), Locative (LOC), and Dative (DAT), the case markers corresponding to them (-ами/-ami, -ах/-ax, -ам/-am), and the expressed semantic categories (I, L, R). category of instrument besides location (case poly- semy), and it is also the marker of the Dative Plural expressing a recipient …

Content truncated to 500 words