Within-Between Lexical Relation Classification

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 3521–3527, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 3521 Within-Between Lexical Relation Classification Oren Barkan∗,1 Avi Caciularu∗,2,3 Ido Dagan2 1Computer Science Department, Ariel University, Israel 2Computer Science Department, Bar-Ilan University, Ramat-Gan, Israel 3Allen Institute for Artificial Intelligence {barkanoren1,avi.c33}@gmail.com, dagan@cs.biu.ac.il Abstract We propose the novel Within-Between Rela- tion model for recognizing lexical-semantic re- lations between words. Our model integrates relational and distributional signals, forming an effective sub-space representation for each relation. We show that the proposed model is competitive and outperforms other baselines, across various benchmarks. 1 Introduction and Related Work Recognizing lexical-semantic relations between words is beneficial for a variety of NLP tasks such as machine translation (Thompson et al., 2019), relation extraction (Shen et al., 2018), natural lan- guage inference (Chen et al., 2018), and question answering (Yang et al., 2017). The lexical relation classification task assigns a word-pair (pair of words) to its corresponding rela- tion out of a finite set of relations. This set contains lexical relations, including the random relation (in- dicating that the words are unrelated). Two main lexical relation classification techniques are studied in the literature: Path-based methods (Hearst, 1992; Snow et al., 2005; Nakashole et al., 2012; Riedel et al., 2013) and distributional methods (Mikolov et al., 2013a; Pennington et al., 2014; Bojanowski et al., 2017; Glavaˇs and Vuli´c, 2018a), with some ef- fort for integrating the two (Shwartz et al., 2016a). In this work we follow the distributional ap- proach, which was shown to improve upon path- based methods. This approach considers static word embeddings such as word2vec (Mikolov et al., 2013b), GloVe (Pennington et al., 2014),and Fast- Text (Bojanowski et al., 2017), which produce out-of-context vector representation for each word. Note here that while contextualized embeddings (Devlin et al., 2019; Peters et al., 2018) have re- placed the use of non-contextualized embeddings ∗ Equal contribution, order determined randomly. in many settings, static word embeddings remain the standard choice for lexical relation classifi- cation, since in this task the input word-pair is given out-of-context. Taking the word embeddings as input, a classifier is trained while considering each word’s representation in the pair. The recent SphereRE method (Wang et al., 2019), a purely distributional method that learns hyperspherical relation representation, presented state-of-the-art lexical relation classification results. While presenting state-of-the-art performance, prior distributional methods suffer from the “lex- ical memorization” problem Levy et al. (2015). This problem arises when a test word-pair includes a rather frequent word in the training set, which is labeled by a dominant category in training. In such cases, the supervised model often ignores the sec- ond word in the input pair and resorts to the domi- nant training label according to the frequent word. Notably, lexical memorization is common for pro- totypical hypernyms — “category” words that are frequently labeled as hypernyms. For example, the vast majority of training examples that include the word fruit are labeled as hypernymy (fruit is the hypernym of apple, …

Content truncated to 500 words