SynSetExpan: An Iterative Framework for Joint Entity Set Expansion and Synonym Discovery

Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, pages 8292–8307, November 16–20, 2020. c⃝2020 Association for Computational Linguistics 8292 SynSetExpan: An Iterative Framework for Joint Entity Set Expansion and Synonym Discovery Jiaming Shen1⋆, Wenda Qiu1⋆, Jingbo Shang2, Michelle Vanni3, Xiang Ren4, Jiawei Han1 1University of Illinois Urbana-Champaign, IL, USA, 2University of California San Diego, CA, USA 3U.S. Army Research Laboratory, MD, USA, 4University of Southern California, CA, USA 1{js2, qiuwenda, hanj}@illinois.edu 2jshang@ucsd.edu 4michelle.t.vanni.civ@mail.mil 4xiangren@usc.edu Abstract Entity set expansion and synonym discovery are two critical NLP tasks. Previous studies accomplish them separately, without exploring their interdependences. In this work, we hy- pothesize that these two tasks are tightly cou- pled because two synonymous entities tend to have similar likelihoods of belonging to var- ious semantic classes. This motivates us to design SynSetExpan, a novel framework that enables two tasks to mutually enhance each other. SynSetExpan uses a synonym discovery model to include popular entities’ infrequent synonyms into the set, which boosts the set expansion recall. Meanwhile, the set expan- sion model, being able to determine whether an entity belongs to a semantic class, can gen- erate pseudo training data to fine-tune the syn- onym discovery model towards better accuracy. To facilitate the research on studying the in- terplays of these two tasks, we create the first large-scale Synonym-Enhanced Set Expansion (SE2) dataset via crowdsourcing. Extensive experiments on the SE2 dataset and previous benchmarks demonstrate the effectiveness of SynSetExpan for both entity set expansion and synonym discovery tasks. 1 Introduction Entity set expansion (ESE) aims to expand a small set of seed entities (e.g., {“United States”, “Canada”}) into a larger set of entities that belong to the same semantic class (i.e., Country). En- tity synonym discovery (ESD) intends to group all terms in a vocabulary that refer to the same real- world entity (e.g., “America” and “USA” refer to the same country) into a synonym set (hence called a synset). Those discovered entities and synsets in- clude rich knowledge and can benefit many down- stream applications such as semantic search (Xiong *Equal Contributions. Connecticut CT SynSetExpan Framework Illinois IL Land of Lincoln Texas TX Lone Star State California CA Golden State User Provided S Seed Synsets Text Corpus D Vocabulary V derives part of Wisconsin WI America’s Dairyland Washington WA Evergreen State …… …… …… Discovered Synsets SVC of Semantic Class C inputs inputs inputs outputs Set Expansion Synonym Discovery Figure 1: An illustrative example of joint entity set expansion and synonym discovery. et al., 2017), taxonomy construction (Shen et al., 2018a), and online education (Yu et al., 2019a). Previous studies regard ESE and ESD as two independent tasks. Many ESE methods (Mamou et al., 2018b; Yan et al., 2019; Huang et al., 2020; Zhang et al., 2020; Zhu et al., 2020) are developed to iteratively select and add the most confident en- tities into the set. A core challenge for ESE is to find those infrequent long-tail entities in the target semantic class (e.g., “Lone Star State” …

Content truncated to 500 words