Improving hypernymy extraction with distributional semantic classes

Panchenko, Alexander ; Ustalov, Dmitry ; Faralli, Stefano ; Ponzetto, Simone Paolo ; Biemann, Chris

URL:	http://www.lrec-conf.org/proceedings/lrec2018/summ...
Additional URL:	https://arxiv.org/abs/1711.02918
Document Type:	Conference or workshop publication
Year of publication:	2018
Book title:	LREC 2018, 11th International Conference on Language Resources and Evaluation : 7-12 May 2018, Miyazaki (Japan)
Page range:	1541-1551
Conference title:	International Conference on Language Resources and Evaluation 2018
Location of the conference venue:	Miyazaki, Japan
Date of the conference:	May 7-12, 2018
Publisher:	Calzolari, Nicoletta
Place of publication:	Paris
Publishing house:	European Language Resources Association, ELRA-ELDA
ISBN:	979-10-95546-00-9
Publication language:	English
Institution:	School of Business Informatics and Mathematics > Information Systems III: Enterprise Data Analysis (Ponzetto 2016-)
Subject:	004 Computer science, internet
Keywords (English):	semantic classes , distributional semantics , hypernyms , co-hyponyms , word sense induction
Abstract:	In this paper, we show for the first time how distributionally-induced semantic classes can be helpful for extraction of hypernyms. We present a method for (1) inducing sense-aware semantic classes using distributional semantics and (2) using these induced semantic classes for filtering noisy hypernymy relations. Denoising of hypernyms is performed by labeling each semantic class with its hypernyms. On one hand, this allows us to filter out wrong extractions using the global structure of the distributionally similar senses. On the other hand, we infer missing hypernyms via label propagation to cluster terms. We conduct a large-scale crowdsourcing study showing that processing of automatically extracted hypernyms using our approach improves the quality of the hypernymy extraction both in terms of precision and recall. Furthermore, we show the utility of our method in the domain taxonomy induction task, achieving the state-of-the-art results on a benchmarking dataset.