Detecting Asymmetric Semantic Relations in Context: A Case-Study on Hypernymy Detection

Proceedings of the 6th Joint Conference on Lexical and Computational Semantics (*SEM 2017), pages 33–43, Vancouver, Canada, August 3-4, 2017. c⃝2017 Association for Computational Linguistics Detecting Asymmetric Semantic Relations in Context: A Case-Study on Hypernymy Detection Yogarshi Vyas and Marine Carpuat Department of Computer Science University of Maryland yogarshi@cs.umd.edu and marine@cs.umd.edu Abstract We introduce WHIC1, a challenging testbed for detecting hypernymy, an asym- metric relation between words. While pre- vious work has focused on detecting hy- pernymy between word types, we ground the meaning of words in specific contexts drawn from WordNet examples, and re- quire predictions to be sensitive to changes in contexts. WHIC lets us analyze com- plementary properties of two approaches of inducing vector representations of word meaning in context. We show that such contextualized word representations also improve detection of a wider range of se- mantic relations in context. 1 Introduction Language understanding applications like ques- tion answering (Harabagiu and Hickl, 2006) and textual entailment (Dagan et al., 2013) bene- fit from identifying semantic relations between words beyond synonymy and paraphrasing. For instance, given “Anand plays chess.”, and the question “Which game does Anand play?”, suc- cessfully answering the question requires know- ing that chess is a kind of game, i.e. chess entails game. Such lexical entailment relations are asym- metric (chess =⇒ game, but game ̸=⇒ chess), and detecting their direction accurately is a chal- lenge. While prior work has defined lexical entailment as a relation between word types, we argue that it is better defined between word meanings illus- trated by examples of usage in context. Ignoring context is problematic since entailment might hold between some senses of the words, but not others. Consider the word game in the following contexts: 1https://github.com/yogarshi/whic 1. The championship game was played in NYC. 2. The hunters were interested in the big game. Given the sentence, Anand is the world chess champion, chess =⇒ game in the first context, while chess ̸=⇒ game in the second context. Lexical entailment encompasses several seman- tic relations, with one important relation being hypernymy (Roller et al., 2014; Shwartz et al., 2016). In this work, we focus on hypernymy de- tection in context, and show that existing resources can be leveraged to automatically create test beds for evaluation. We introduce “Wordnet Hyper- nyms in Context” (WHIC, pronounced which), a large dataset, automatically extracted from Word- Net (Fellbaum, 1998) using examples provided with synsets. Crucially, WHIC includes challeng- ing negative examples that assess the ability of models to detect the direction of hypernymy. We use WHIC to determine the effectiveness of existing supervised models for hypernymy detec- tion (Roller and Erk, 2016) applied to represen- tations, not only of word types, but of words in context. Such contextualized representations are induced in two ways: the first is based on Con- text2Vec, a BiLSTM model that embeds contexts and words in the same space (Melamud et al., 2016); the second aims to capture geometric prop- erties of the context in a standard word embedding …

Content truncated to 500 words