USAAR at SemEval-2016 Task 13: Hyponym Endocentricity

Proceedings of SemEval-2016, pages 1303–1309, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics USAAR at SemEval-2016 Task 13: Hyponym Endocentricity Liling Tan1, Francis Bond2 and Josef van Genabith1 Universit¨at des Saarlandes1 / Campus, Saarbr¨ucken, Germany Nanyang Technological University2 / 14 Nanyang Drive, Singapore liling.tan@uni-saarland.de, bond@ieee.org josef.van genabith@dfki.de Abstract This paper describes our submission to the SemEval-2016 Taxonomy Extraction Evalua- tion (TExEval-2) Task. We examine the en- docentric nature of hyponyms and propose a simple rule-based method to identify hyper- nyms at high precision. For the food domain, we extract lists of terms from the Wikipedia lists of lists by using the name of each list as the endocentric head and treating all terms in the extracted tables as the hyponym of the en- docentric head. Our submission achieved competitive results in taxonomy construction and ranked top in hypernym identification when evaluated against gold standard taxonomies and also in manual evaluation of novel relations not cov- ered by the gold standard taxonomies. 1 Introduction Semantic taxonomies provide structured world knowledge to Artificial Intelligence (AI) and Natu- ral Language Processing (NLP) systems. Traditional broad-coverage taxonomies such as CYC (Lenat, 1995), SUMO (Pease et al., 2002; Miller, 1995), YAGO (Suchanek et al., 2007) and Freebase (Bol- lacker et al., 2008) have been manually created or curated with much effort. With the rapid technological evolution, it is more feasible to construct a domain-specific taxonomy that caters to domain or company specific terminol- ogy (Lefever, 2015). This motivated the move to- wards unsupervised approaches to taxonomy extrac- tion (Berland and Charniak, 1999; Lin and Pantel, 2001; Snow et al., 2006) and specifically focused to- wards particular domains (Velardi et al., 2013; Bor- dea et al., 2015). The aim of the Taxonomy Extraction Evaluation (TExEval) task is to automatically find lexical rela- tions between pairs of terms within several specified domains. Previously, we have developed a hyper- nym extraction system using word embeddings by exploiting the frequent occurrence of the ‘X is a Y’ pattern in encyclopedic text (Tan et al., 2015). We have achieved competitive results SemEval-2015 and as a follow up to our study,in we would like to explore the endocentric nature of hyponyms that contributed substantially to the system performance in the previous TExEval task. Below, we will briefly (i) describe related work on different approaches to taxonomy induction, (ii) explain the linguistic phenomenon of endocentricity, (iii) present our endocentric hypo-hypernym identi- fication system and the results of our submission to the TExEval-2 task in SemEval-2016. 2 Related Work The hierarchical structure of domain concepts is made of hypo-hypernymy relations between terms. Different approaches have been proposed to in- duce these relations automatically ranging from pattern/rule-based approaches (Hearst, 1992; Girju, 2003; Kozareva et al., 2008; Ceesay and Hou, 2015) to clustering and frequency based approaches (Lin, 1998; Caraballo, 2001; Pantel and Ravichan- dran, 2004; Grefenstette, 2015), classification ap- proaches (Snow et al., 2004; Ritter et al., 2009; Espinosa Anke et al., 2015) and graph-based ap- 1303 proaches (Kozareva and Hovy, …

Content truncated to 500 words