Collocation Classification with Unsupervised Relation Vectors

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 5765–5772 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 5765 Collocation Classification with Unsupervised Relation Vectors Luis Espinosa-Anke1, Leo Wanner2,3, and Steven Schockaert1 1School of Computer Science, Cardiff University, United Kingdom 2ICREA and 3NLP Group, Universitat Pompeu Fabra, Barcelona, Spain {espinosa-ankel,schockaerts1}@cardiff.ac.uk leo.wanner@upf.edu Abstract Lexical relation classification is the task of predicting whether a certain relation holds be- tween a given pair of words. In this pa- per, we explore to which extent the current distributional landscape based on word em- beddings provides a suitable basis for classi- fication of collocations, i.e., pairs of words between which idiosyncratic lexical relations hold. First, we introduce a novel dataset with collocations categorized according to lexical functions. Second, we conduct experiments on a subset of this benchmark, comparing it in particular to the well known DiffVec dataset. In these experiments, in addition to simple word vector arithmetic operations, we also in- vestigate the role of unsupervised relation vec- tors as a complementary input. While these relation vectors indeed help, we also show that lexical function classification poses a greater challenge than the syntactic and semantic rela- tions that are typically used for benchmarks in the literature. 1 Introduction Relation classification is the task of predicting whether between a given pair of words or phrases, a certain lexical, semantic or morphosyntactic re- lation holds. This task has direct impact in down- stream NLP tasks such as machine translation, paraphrase identification (Etzioni et al., 2005), named entity recognition (Socher et al., 2012), or knowledge base completion (Socher et al., 2013). The currently standard approach to relation classi- fication is to combine the embeddings correspond- ing to the arguments of a given relation into a meaningful representation, which is then passed to a classifier. As for which relations have been targeted so far, the landscape is considerably more varied, although we may safely group them into morphosyntactic and semantic relations. Morphosyntactic relations have been the focus of work on unsupervised relational similarity, as it has been shown that verb conjugation or nomi- nalization patterns are relatively well preserved in vector spaces (Mikolov et al., 2013; Pennington et al., 2014a). Semantic relations pose a greater challenge (Vylomova et al., 2016), however. In fact, as of today, it is unclear which operation per- forms best (and why) for the recognition of indi- vidual lexico-semantic relations (e.g., hyperonymy or meronymy, as opposed to cause, location or ac- tion). Still, a number of works address this chal- lenge. For instance, hypernymy has been modeled using vector concatenation (Baroni et al., 2012), vector difference and component-wise squared difference (Roller et al., 2014) as input to linear regression models (Fu et al., 2014; Espinosa-Anke et al., 2016); cf. also a sizable number of neu- ral approaches (Shwartz et al., 2016; Anh et al., 2016). Furthermore, several high quality semantic relation datasets are available, ranging from well- known resources such as WordNet (Miller, 1995), Yago (Suchanek …

Content truncated to 500 words