When Hearst Is not Enough: Improving Hypernymy Detection from Corpus with Distributional Models

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 6208–6217, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 6208 When Hearst Is not Enough: Improving Hypernymy Detection from Corpus with Distributional Models Changlong Yu1∗ Jialong Han2 Peifeng Wang3 Yangqiu Song1 Hongming Zhang1 Wilfred Ng1 Shuming Shi4 1HKUST 2Amazon 3University of Southern California {cyuaq, yqsong, hzhangal, wilfred}@cse.ust.hk4Tencent AI Lab jialonghan@gmail.com, peifengw@usc.edu, shumingshi@tencent.com Abstract We address hypernymy detection, i.e., whether an is-a relationship exists between words (x, y), with the help of large textual corpora. Most conventional approaches to this task have been categorized to be either pattern-based or distributional. Recent studies suggest that pattern-based ones are superior, if large-scale Hearst pairs are extracted and fed, with the sparsity of unseen (x, y) pairs relieved. How- ever, they become invalid in some specific sparsity cases, where x or y is not involved in any pattern. For the first time, this paper quan- tifies the non-negligible existence of those spe- cific cases. We also demonstrate that distribu- tional methods are ideal to make up for pattern- based ones in such cases. We devise a com- plementary framework, under which a pattern- based and a distributional model collaborate seamlessly in cases which they each prefer. On several benchmark datasets, our frame- work achieves competitive improvements and the case study shows its better interpretability. 1 Introduction A taxonomy is a semantic hierarchy of words or concepts organized w.r.t. their hypernymy (a.k.a. is-a) relationships. Being a well-structured re- source of lexical knowledge, taxonomies are vital to various tasks such as question answering (Gupta et al., 2018), textual entailment (Dagan et al., 2013; Bowman et al., 2015; Yu et al., 2020b), and text generation (Biran and McKeown, 2013). When automatically building taxonomies from scratch or populating manually crafted ones, the hyper- nymy detection task plays a central role. For a pair of queried words (xq, yq), hypernymy detec- tion requires inferring the existence of a hyponym- hypernym relationship between xq and yq. Due to ∗ Work done when C. Yu, J. Han and P. Wang were with Tencent AI Lab. Figure 1: The overall framework of complementary methods for hypernymy detection from corpus. Dif- ferent sparsity types of queried pairs are handled with pattern-based and distributional models respectively. the good coverage and availability, free-text cor- pora are widely used to facilitate hypernymy detec- tion, resulting in two lines of approaches: pattern- based and distributional. Pattern-based approaches employ pattern pairs (x, y) extracted via Hearst-like patterns (Hearst, 1992), e.g., “y such as x” and “x and other y”. An example of extracted pattern pairs from corpus are shown in Figure 1. Despite their high precision, the extracted pairs suffer from sparsity which comes in two folds i.e., Type-I: xq and yq separately appear in some extracted pairs, but the pair (xq, yq) is ab- sent e.g., (dog, animal); or Type-II: either xq or yq is not involved in any extracted pair e.g., (crocodile, animal). Although matrix factorization (Roller et al., 2018) or embedding techniques …

Content truncated to 500 words