Robust Cross-Lingual Hypernymy Detection Using Dependency Context

Proceedings of NAACL-HLT 2018, pages 607–618 New Orleans, Louisiana, June 1 - 6, 2018. c⃝2018 Association for Computational Linguistics Robust Cross-lingual Hypernymy Detection using Dependency Context Shyam Upadhyay1∗, Yogarshi Vyas2∗, Marine Carpuat2, Dan Roth1 1 Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA 2 Department of Computer Science, University of Maryland, College Park, MD shyamupa@seas.upenn.edu, yogarshi@cs.umd.edu, marine@cs.umd.edu, danroth@seas.upenn.edu Abstract Cross-lingual Hypernymy Detection involves determining if a word in one language (“fruit”) is a hypernym of a word in another language (“pomme” i.e. apple in French). The abil- ity to detect hypernymy cross-lingually can aid in solving cross-lingual versions of tasks such as textual entailment and event coreference. We propose BISPARSE-DEP, a family of un- supervised approaches for cross-lingual hyper- nymy detection, which learns sparse, bilingual word embeddings based on dependency con- texts. We show that BISPARSE-DEP can sig- nificantly improve performance on this task, compared to approaches based only on lexical context. Our approach is also robust, show- ing promise for low-resource settings: our dependency-based embeddings can be learned using a parser trained on related languages, with negligible loss in performance. We also crowd-source a challenging dataset for this task on four languages – Russian, French, Arabic, and Chinese. Our embeddings and datasets are publicly available.1 1 Introduction Translation helps identify correspondences in bilingual texts, but other asymmetric semantic re- lationships can improve language understanding when translations are not exactly equivalent. One such relationship is cross-lingual hypernymy – identifying that ´ecureuil (“squirrel” in French) is a kind of rodent, or ворона (“crow” in Russian) is a kind of bird. The ability to detect hypernyms across languages serves as a building block in a range of cross-lingual tasks, including Recogniz- ing Textual Entailment (RTE) (Negri et al., 2012, ∗ These authors contributed equally. 1https://github.com/yogarshi/ bisparse-dep/ 2013), constructing multilingual taxonomies (Fu et al., 2014), event coreference across multilingual news sources (Vossen et al., 2015), and evaluating Machine Translation output (Pad´o et al., 2009). Building models that can robustly identify hy- pernymy across the spectrum of human languages is a challenging problem, that is further com- pounded in low resource settings. At first glance, translating words to English and then identify- ing hypernyms in a monolingual setting may ap- pear to be a sufficient solution. However, this ap- proach cannot capture many phenomena. For in- stance, the English words cook, leader and super- visor can all be hypernyms of the French word chef, as the French word does not have a exact translation in English covering its possible usages. However, translating chef to cook and then deter- mining hypernymy monolingually precludes iden- tifying leader or supervisor as a hypernyms of chef. Similarly, language-specific usage patterns can also influence hypernymy decisions. For in- stance, the French word chroniqueur translates to chronicler in English, but is more frequently used in French to refer to journalists (making journalist its hypernym).2 This motivates approaches that directly detect hypernymy in the cross-lingual setting by extend- ing distributional methods for detecting monolin- gual hypernymy, as in …

Content truncated to 500 words