Paper: Domain and Dialect Adaptation for Machine Translation into Egyptian Arabic

ACL ID W14-3627
Title Domain and Dialect Adaptation for Machine Translation into Egyptian Arabic
Venue Workshop on Arabic Natural Language Processing
Session
Year 2014
Authors

In this paper, we present a statistical ma- chine translation system for English to Di- alectal Arabic (DA), using Modern Stan- dard Arabic (MSA) as a pivot. We cre- ate a core system to translate from En- glish to MSA using a large bilingual par- allel corpus. Then, we design two separate pathways for translation from MSA into DA: a two-step domain and dialect adap- tation system and a one-step simultane- ous domain and dialect adaptation system. Both variants of the adaptation systems are trained on a 100k sentence tri-parallel cor- pus of English, MSA, and Egyptian Arabic generated by a rule-based transformation. We test our systems on a held-out Egyp- tian Arabic test set from the 100k sen- tence corpus and we achieve our best per- formance using the two-step domain and dialect adaptat...