Publication at Faculty of Mathematics and Physics |
2013
Abstract
This research focuses on determining semantic compositionality of word expressions using word space models (WSMs). We discuss previous works employing WSMs and present differences in the proposed approaches which include types of WSMs, corpora, preprocessing techniques, methods for determining compositionality, and evaluation testbeds.
We also present results of our own approach for determining the semantic compositionality based on comparing distributional vectors of expressions and their components. The vectors were obtained by Latent Semantic Analysis (LSA) applied to the ukWaC corpus.
Our results outperform those of all the participants in the Distributional Semantics and Compositionality (DISCO) 2011 shared task.