Automatic Idiom Identification in Wiktionary

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 1417–1421, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Automatic Idiom Identification in Wiktionary Grace Muzny and Luke Zettlemoyer Computer Science & Engineering University of Washington Seattle, WA 98195 {muznyg,lsz}@cs.washington.edu Abstract Online resources, such as Wiktionary, provide an accurate but incomplete source of idiomatic phrases. In this paper, we study the problem of automatically identifying idiomatic dictio- nary entries with such resources. We train an idiom classifier on a newly gathered cor- pus of over 60,000 Wiktionary multi-word definitions, incorporating features that model whether phrase meanings are constructed compositionally. Experiments demonstrate that the learned classifier can provide high quality idiom labels, more than doubling the number of idiomatic entries from 7,764 to 18,155 at precision levels of over 65%. These gains also translate to idiom detection in sen- tences, by simply using known word sense disambiguation algorithms to match phrases to their definitions. In a set of Wiktionary def- inition example sentences, the more complete set of idioms boosts detection recall by over 28 percentage points. 1 Introduction Idiomatic language is common and provides unique challenges for language understanding systems. For example, a diamond in the rough can be the literal unpolished object or a crude but lovable person. Un- derstanding such distinctions is important for many applications, including parsing (Sag et al., 2002) and machine translation (Shutova et al., 2012). We use Wiktionary as a large, but incomplete, ref- erence for idiomatic entries; individual entries can be marked as idiomatic but, in practice, most are not. Using these incomplete annotations as super- vision, we train a binary Perceptron classifier for identifying idiomatic dictionary entries. We intro- duce new lexical and graph-based features that use WordNet and Wiktionary to compute semantic re- latedness. This allows us to learn, for example, that the words in the phrase diamond in the rough are more closely related to the words in its literal defi- nition than the idiomatic one. Experiments demon- strate that the classifier achieves precision of over 65% at recall over 52% and that, when used to fill in missing Wiktionary idiom labels, it more than dou- bles the number of idioms from 7,764 to 18,155. These gains also translate to idiom detection in sentences, by simply using the Lesk word sense disambiguation (WSD) algorithm (1986) to match phrases to their definitions. This approach allows for scalable detection with no restrictions on the syn- tactic structure or context of the target phrase. In a set of Wiktionary definition example sentences, the more complete set of idioms boosts detection recall by over 28 percentage points. 2 Related Work To the best of our knowledge, this work represents the first attempt to identify dictionary entries as id- iomatic and the first to reduce idiom detection to identification via a dictionary. Previous idiom detection systems fall in one of two paradigms: phrase classification, where a phrase p is always idiomatic or literal, e.g. (Gedigian et al., 2006; Shutova …

Content truncated to 500 words