Phrasal Substitution of Idiomatic Expressions

Proceedings of NAACL-HLT 2016, pages 363–373, San Diego, California, June 12-17, 2016. c⃝2016 Association for Computational Linguistics Phrasal Substitution of Idiomatic Expressions Changsheng Liu and Rebecca Hwa Computer Science Department University of Pittsburgh Pittsburgh, PA 15260, USA {changsheng,hwa}@cs.pitt.edu Abstract Idioms pose a great challenge to natural lan- guage understanding. A system that can auto- matically paraphrase idioms in context has ap- plications in many NLP tasks. This paper pro- poses a phrasal substitution method to replace idioms with their figurative meanings in literal English. Our approach identifies relevant re- placement phrases from an idiom’s dictionary definition and performs appropriate grammat- ical and referential transformations to ensure that the idiom substitution fits seamlessly into the original context. The proposed method has been evaluated both by automatic metrics and human judgments. Results suggest that high quality paraphrases of idiomatic expressions can be achieved. 1 Introduction An idiom is a combination of words that has a fig- urative meaning which differs from its literal mean- ing. Idioms pose a great challenge to many NLP tasks, such as machine translation, word sense dis- ambiguation, and sentiment analysis (Volk, 1998; Korkontzelos et al., 2013; Zanzotto et al., 2010; Williams et al., 2015). Previous work (Salton et al., 2014) has shown that a typical statistical ma- chine translation system might achieve only half of the BLEU score (Papineni et al., 2002) on sentences that contain idiomatic expressions than on those that do not. Idioms are also problematic for second lan- guage learners. In a pilot study we have surveyed seven non-native speakers on 100 Tweets containing idioms; we have found that, on average, the partici- pants had trouble understanding 70% of them due to the inclusion of idioms. This work explores the possibility of automati- cally replacing idiomatic expressions in sentences. The full pipeline of a successful system has to solve many problems. First, it has to determine that an expression is, in fact, being used as an idiom in a sentence (Fazly et al., 2009; Korkontzelos et al., 2013; Sporleder and Li, 2009). Moreover, the sys- tem has to sense disambiguate the idiom – it has to pick the correct interpretation when more than one is possible. Second, it has to generate an appropriate phrasal replacement for the idiom using literal En- glish. Third, it has to ensure that the replacement phrase will fit seamlessly back into the original sen- tence. This paper focuses on the second and third problem, which have not been studied as extensively in previous works. We propose to extract the phrasal replacement for an idiom from its definition, assuming the existence of an up-to-date dictionary of broad coverage and high quality.1 Because a typical definition is quite long, it cannot directly serve as a replacement for the idiom. A major challenge of our work is in identi- fying the right nugget to extract from the definition. Another major challenge is the smooth integration of the substitution phrase into the sentence. We con- sider both grammatical fluency as well as references …

Content truncated to 500 words