Avo Muromägi
University of Tartu, Tartu, Estonia
Kairit Sirts
University of Tartu, Tartu, Estonia
Sven Laur
University of Tartu, Tartu, Estonia
Download article
Published in: Proceedings of the 21st Nordic Conference on Computational Linguistics, NoDaLiDa, 22-24 May 2017, Gothenburg, Sweden
Linköping Electronic Conference Proceedings 68:12, p. 96-104
NEALT Proceedings Series 29:12, p. 96-104
Published: 2017-05-08
ISBN: 978-91-7685-601-7
ISSN: 1650-3686 (print), 1650-3740 (online)
This paper explores linear methods for combining several word embedding models into an ensemble. We construct the combined models using an iterative method based on either ordinary least squares regression or the solution to the orthogonal Procrustes problem. We evaluate the proposed approaches on Estonian—a morphologically complex language, for which the available corpora for training word embeddings are relatively small. We compare both combined models with each other and with the input word embedding models using synonym and analogy tests. The results show that while using the ordinary least squares regression performs poorly in our experiments, using orthogonal Procrustes to combine several word embedding models into an ensemble model leads to 7-10% relative improvements over the mean result of the initial models in synonym tests and 19-47% in analogy tests.
