Year: 2011
Type: conference-paper
Abstract: We use target-side monolingual data to extend\nthe vocabulary of the translation model\nin statistical machine translation. This method\ncalled “reverse self-training” improves the decoder’s\nability to produce grammatically correct\ntranslations into languages with morphology\nricher than the source language esp. in\nsmall-data setting. We empirically evaluate\nthe gains for several pairs of European\nlanguages and discuss some approaches of\nthe underlying back-off techniques needed to\ntranslate unseen forms of known words. We\nalso provide a description of the systems we\nsubmitted to WMT11 Shared Task.
Authors: Ondřej Bojar, Aleš Tamchyna
Institution: Charles University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W211509693