Relation induction in word embeddings revisited

Bouraoui, Zied, Jameel, Mohammad ORCID: https://orcid.org/0000-0002-3707-4367 and Schockaert, Steven ORCID: https://orcid.org/0000-0002-9256-2881 2018. Relation induction in word embeddings revisited. Presented at: 27th International Conference on Computational Linguistics (COLING 2018), Santa Fe, NM, USA, 20-26 August 2018. Published in: Bender, Emily M., Derczynski, Leon and Isabelle, Pierre eds. Proceedings of the 27th International Conference on Computational Linguistics. pp. 1627-1637.

Preview

PDF - Accepted Post-Print Version Available under License Creative Commons Attribution. Download (551kB) | Preview

Abstract

Given a set of instances of some relation, the relation induction task is to predict which other word pairs are likely to be related in the same way. While it is natural to use word embeddings for this task, standard approaches based on vector translations turn out to perform poorly. To address this issue, we propose two probabilistic relation induction models. The first model is based on translations, but uses Gaussians to explicitly model the variability of these translations and to encode soft constraints on the source and target words that may be chosen. In the second model, we use Bayesian linear regression to encode the assumption that there is a linear relationship between the vector representations of related words, which is considerably weaker than the assumption underlying translation based models.

Item Type:	Conference or Workshop Item - unpublished
Date Type:	Completion
Status:	Unpublished
Schools:	Schools > Computer Science & Informatics
Date of First Compliant Deposit:	18 July 2018
Last Modified:	27 May 2026 10:34
URI:	https://orca.cardiff.ac.uk/id/eprint/112688

Citation Data

Cited 22 times in Scopus. View in Scopus. Powered By Scopus® Data

Actions (repository staff only)

Edit Item