Resolving Shell Nouns

34694

Authors

nan

Abstract

Shell nouns, such as fact and problem, oc- cur frequently in all kinds of texts. These nouns themselves are unspecific, and can only be interpreted together with the shell content. We propose a general approach to automatically identify shell content of shell nouns. Our approach exploits lexico- syntactic knowledge derived from the lin- guistics literature. We evaluate the ap- proach on a variety of shell nouns with a variety of syntactic expectations, achiev- ing accuracies in the range of 62% (base- line = 33%) to 83% (baseline = 74%) on crowd-annotated data. 1

Keywords

nan

Full Content

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 499–510, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics Resolving Shell Nouns Varada Kolhatkar Department of Computer Science University of Toronto Toronto, ON, M5S 3G4, Canada varada@cs.toronto.edu Graeme Hirst Department of Computer Science University of Toronto Toronto, ON, M5S 3G4, Canada gh@cs.toronto.edu Abstract Shell nouns, such as fact and problem, oc- cur frequently in all kinds of texts. These nouns themselves are unspecific, and can only be interpreted together with the shell content. We propose a general approach to automatically identify shell content of shell nouns. Our approach exploits lexico- syntactic knowledge derived from the lin- guistics literature. We evaluate the ap- proach on a variety of shell nouns with a variety of syntactic expectations, achiev- ing accuracies in the range of 62% (base- line = 33%) to 83% (baseline = 74%) on crowd-annotated data. 1 Introduction Shell nouns are abstract nouns, such as fact, issue, idea, and problem, which facilitate efficiency by avoiding repetition of long stretches of text. The shell metaphor comes from Schmid (2000), and it captures the various functions of these nouns in a discourse: containment, signalling, pointing, and encapsulating. Shell nouns themselves are unspe- cific, and can only be interpreted together with their shell content, i.e., the propositional content they encapsulate in the given context. The process of identifying this content in the given context is referred to as shell noun resolution or interpreta- tion. Examples (1), (2), and (3) show usages of the shell nouns fact and issue. The shell noun phrases are resolved to the postnominal that clause, the complement wh clause, and the immediately pre- ceding clause, respectively.1,2 (1) The fact that a major label hadn’t been at liberty to exploit and repackage the material on CD meant that prices on the vintage LP market were soaring. (2) The issue that this country and Congress must address is how to provide optimal care for all without limiting access for the many. (3) Living expenses are much lower in rural India than in New York, but this fact is not fully captured if prices are converted with currency exchange rates. Observe that the relation between shell noun phrases and their shell content is similar to the relation of abstract anaphora (or cataphora) (Asher, 1993) with backward- or forward-looking abstract-object antecedents. For anaphoric shell noun examples, the shell content precedes the shell noun phrase, and for cataphoric shell noun examples the shell content follows the shell noun phrase.3 Shell nouns as a group occur frequently in argu- mentative texts (Schmid, 2000; Flowerdew, 2003; Botley, 2006). They play an important role in or- ganizing a discourse and maintaining its coher- ence (Schmid, 2000; Flowerdew, 2003), and re- solving them is an important component of var- ious computational linguistics tasks that rely on 1Note that the postnominal that-clause in (1) is not a rela- tive clause: the fact in question is not an argument of exploit and repackage. 2All examples …

Content truncated to 500 words

Labeled Text

Shell nouns, such as fact and problem, occur frequently in all kinds of texts. These nouns themselves are unspecific, and can only be interpreted together with the shell content. We propose a general approach to automatically identify shell content of shell nouns. Our approach exploits lexicosyntactic knowledge derived from the linguistics literature. We evaluate the approach on a variety of shell nouns with a variety of syntactic expectations, achieving accuracies in the range of 62% (baseline = 33%) to 83% (baseline = 74%) on crowd-annotated data. 1

Publication Details

Paper ID:
34694
Year:
2014
Publication:
TIPSTER
Publication Type:
conf
DOI:
nan