Transcription of WORD TRANSLATION WITHOUT PARALLEL DATA - arXiv
{{id}} {{{paragraph}}}
Example: biology
WORD TRANSLATION WITHOUT PARALLEL DATA - arXiv
Published as a conference paper at ICLR 2018 WORD TRANSLATION WITHOUT PARALLEL DATAA lexis Conneau , Guillaume Lample ,Marc Aurelio Ranzato , Ludovic Denoyer , Herv e J egou methods for learning cross-lingual word embeddings have reliedon bilingual dictionaries or PARALLEL corpora. Recent studies showed that the needfor PARALLEL data supervision can be alleviated with character-level these methods showed encouraging results, they are not on par with theirsupervised counterparts and are limited to pairs of languages sharing a commonalphabet. In this work, we show that we can build a bilingual dictionary betweentwo languages WITHOUT using any PARALLEL corpora, by aligning monolingual wordembedding spaces in an unsupervised way.
Published as a conference paper at ICLR 2018 WORD TRANSLATION WITHOUT PARALLEL DATA Alexis Conneauyz, Guillaume Lampleyx, Marc’Aurelio Ranzatoy, Ludovic Denoyerx, Herv´e J ´egou y faconneau,glample,ranzato,rvjg@fb.com ludovic.denoyer@upmc.fr
Loading..
Tags:
Information
Domain:
Source:
Link to this page:
Please notify us if you found a problem with this document: