NLP_HZ at SemEval-2018 Task 9: a Nearest Neighbor Approach

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 909–913 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics NLP HZ at SemEval-2018 Task 9: a Nearest Neighbor Approach Wei Qiu Alibaba Group, China qiuwei.cw@alibaba-inc.com Mosha Chen Alibaba Group,China chenmosha.cms@alibaba-inc.com Linlin Li Alibaba Group, China linyan.lll@alibaba-inc.com Luo Si Alibaba Group, China luo.si@alibaba-inc.com Abstract Hypernym discovery aims to discover the hy- pernym word sets given a hyponym word and proper corpus. This paper proposes a simple but effective method for the discovery of hy- pernym sets based on word embedding, which can be used to measure the contextual simi- larities between words. Given a test hyponym word, we get its hypernym lists by comput- ing the similarities between the hyponym word and words in the training data, and fill the test word’s hypernym lists with the hypernym list in the training set of the nearest similarity dis- tance to the test word. In SemEval 2018 task9, our results, achieve 1st on Spanish, 2nd on Italian, 6th on English in the metric of MAP. 1 Introduction Hypernymy relationship plays a critical role in language understanding because it enables gener- alization, which lies at the core of human cogni- tion (Yu et al. (2015)). It has been widely used in various NLP applications (Espinosa Anke et al. (2016)), from word sense disambiguation (Agirre et al. (2014)) to information retrieval (Varelas et al. (2005)) , question answering (Prager (2006)) and textual entailment (Glickman et al. (2005)). To date, the hypernymy relation also plays an impor- tant role in Knowledge Base Construction task. In the past SemEval contest (SemEval-2015 task 171, SemEval-2016 task 132), the “Hypernym Detection” task was treated as a classfication task, i.e., given a (hyponym, hypernym) pair, deciding whether the pair is a true hypernymic relation or not. This has led to criticisms regarding its over- simplification (Levy et al., 2015). In the SemEval 2018 Task 9 (Camacho-Collados et al., 2018), the task has shifted to “Hypernym Discovery” , i.e., 1http://alt.qcri.org/semeval2015/task17/ 2http://alt.qcri.org/semeval2016/task13/ given the search space of a domain’s vocabulary and an input hyponym, discover its best (set of) candidate hypernyms. In this paper, the content is organized as fol- lows: Section 2 gives an introduction to the re- lated work; Section 3 describes our methods for this task, including word embedding projection learning as the baseline and the nearest-neighbour- based method as the submission result; The exper- imental results are presented in Section 4. We con- clude the paper with Section 5. 2 Related Work The work of identifying hypernymy relationship can be categorized from different aspects accord- ing to the learning methods and the task formuliza- tion. The earlier work (Hearst (1992)) formalized the task as an unsupervised hypernym discovery task, i.e., none hyponym-hypernyms pairs (x, y) are given as the training data. Hearst (1992) hand- crafted a set of lexico-syntactic paths that connect the joint occurrences of x and y which indicate hypernymy in a large corpus. Snow et al. (2004) trained a logistic regression …

Content truncated to 500 words