Automatic Classification of Communicative Functions of Definiteness

Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers, pages 1059–1070, Dublin, Ireland, August 23-29 2014. Automatic Classification of Communicative Functions of Definiteness Archna Bhatia∗,‡ Chu-Cheng Lin∗ Nathan Schneider∗ Yulia Tsvetkov∗ Fatima Talib Al-Raisi∗ Laleh Roostapour∗ Jordan Bender† Abhimanu Kumar∗ Lori Levin∗ Mandy Simons∗ Chris Dyer∗ ∗Carnegie Mellon University †University of Pittsburgh Pittsburgh, PA 15213 Pittsburgh, PA 15260 ‡archnab@cs.cmu.edu Abstract Definiteness expresses a constellation of semantic, pragmatic, and discourse properties—the communicative functions—of an NP. We present a supervised classifier for English NPs that uses lexical, morphological, and syntactic features to predict an NP’s communicative function in terms of a language-universal classification scheme. Our classifiers establish strong baselines for future work in this neglected area of computational semantic analysis. In addition, analysis of the features and learned parameters in the model provides insight into the grammaticalization of definiteness in English, not all of which is obvious a priori. 1 Introduction Definiteness is a morphosyntactic property of noun phrases (NPs) associated with semantic and pragmatic characteristics of entities and their discourse status. Lyons (1999), for example, argues that definite markers prototypically reflect identifiability (whether a referent for the NP can be identified by the discourse participants or not); other aspects identified in the literature include uniqueness of the entity in the world and whether the hearer is already familiar with the entity given the context and preceding discourse (Roberts, 2003; Abbott, 2006). While some morphosyntactic forms of definiteness are employed by all languages—namely, demonstratives, personal pronouns, and possessives—languages display a vast range of variation with respect to the form and meaning of definiteness. For example, while languages like English make use of definite and indefinite articles to distinguish between the discourse status of various entities (the car vs. a car vs. cars), many other languages—including Czech, Indonesian, and Russian—do not have articles (although they do have demonstrative determiners). Sometimes definiteness is marked with affixes or clitics, as in Arabic. Sometimes it is expressed with other constructions, as in Chinese (a language without articles), where the existential construction can be used to express indefinite subjects and the ba- construction can be used to express definite direct objects (Chen, 2004). Aside from this variation in the form of (in)definite NPs within and across languages, there is also vari- ability in the mapping between semantic, pragmatic, and discourse functions of NPs and the (in)definites expressing these functions. We refer to these as communicative functions of definiteness, following Bhatia et al. (2014). Croft (2003, pp. 6–7) shows that even when two languages have access to the same morphosyntactic forms of definiteness, the conditions under which an NP is marked as definite or indefinite (or not at all) are language-specific. He illustrates this by contrasting English and French translations (both languages use definite as well as indefinite articles) such as: (1) He showed extreme care. (unmarked) Il montra un soin extrême. (indef.) (2) I love artichokes and asparagus. (unmarked) J’aime les artichauts et les asperges. (def.) (3) His brother became a soldier. (indef.) …

Content truncated to 500 words