In this work, we present a novel counter-fitting method which injects
antonymy and synonymy constraints into vector space representations in order to
improve the vectors' capability for judging semantic similarity. Applying this
method to publicly available pre-trained word vectors leads to a new state of
the art performance on the SimLex-999 dataset. We also show how the method can
be used to tailor the word vector space for the downstream task of dialogue
state tracking, resulting in robust improvements across different dialogue
domains.
%0 Conference Paper
%1 mrksic2016counterfitting
%A Mrkšić, Nikola
%A Séaghdha, Diarmuid Ó
%A Thomson, Blaise
%A Gašić, Milica
%A Rojas-Barahona, Lina
%A Su, Pei-Hao
%A Vandyke, David
%A Wen, Tsung-Hsien
%A Young, Steve
%B HLT-NAACL
%D 2016
%K learning metric semantic
%T Counter-fitting Word Vectors to Linguistic Constraints
%U http://arxiv.org/abs/1603.00892
%X In this work, we present a novel counter-fitting method which injects
antonymy and synonymy constraints into vector space representations in order to
improve the vectors' capability for judging semantic similarity. Applying this
method to publicly available pre-trained word vectors leads to a new state of
the art performance on the SimLex-999 dataset. We also show how the method can
be used to tailor the word vector space for the downstream task of dialogue
state tracking, resulting in robust improvements across different dialogue
domains.
@inproceedings{mrksic2016counterfitting,
abstract = {In this work, we present a novel counter-fitting method which injects
antonymy and synonymy constraints into vector space representations in order to
improve the vectors' capability for judging semantic similarity. Applying this
method to publicly available pre-trained word vectors leads to a new state of
the art performance on the SimLex-999 dataset. We also show how the method can
be used to tailor the word vector space for the downstream task of dialogue
state tracking, resulting in robust improvements across different dialogue
domains.},
added-at = {2017-04-13T17:50:18.000+0200},
author = {Mrkšić, Nikola and Séaghdha, Diarmuid Ó and Thomson, Blaise and Gašić, Milica and Rojas-Barahona, Lina and Su, Pei-Hao and Vandyke, David and Wen, Tsung-Hsien and Young, Steve},
biburl = {https://www.bibsonomy.org/bibtex/2e827d868563879fc5b08c8039cea9836/thoni},
booktitle = {HLT-NAACL},
interhash = {8bde3944c92a0192802aa164e5c53f67},
intrahash = {e827d868563879fc5b08c8039cea9836},
keywords = {learning metric semantic},
timestamp = {2017-05-16T09:45:53.000+0200},
title = {Counter-fitting Word Vectors to Linguistic Constraints},
url = {http://arxiv.org/abs/1603.00892},
year = 2016
}