2019
JHU LoResMT Shared Task System Description
Abstract
We describe the JHU submission to the LoResMT 2019 shared task, which involved translating between Bhojpuri, Latvian, Magahi, and Sindhi, to and from English. JHU submitted runs for all eight language pairs. Baseline runs using phrase-based statistical machine translation (SMT) and neural machine translation (NMT) were produced. We also submitted neural runs that made use of backtranslation and ensembling. Preliminary results suggest that system performance is reasonable given the limited amount of training data.