ChrEnTranslate: Cherokee-English Machine Translation Demo with Quality Estimation and Corrective Feedback

We introduce ChrEnTranslate, an online ma chine translation demonstration system for translation between English and an endangered language Cherokee. It supports both statistical and neural translation models as well as pro vides quality estimation to inform users of re liability, two user feedback interfaces for ex perts and common users respectively, exam ple inputs to collect human translations for monolingual data, word alignment visualiza tion, and relevant terms from the Cherokee English dictionary. The quantitative evalu ation demonstrates that our backbone trans lation models achieve stateoftheart transla tion performance and our quality estimation well correlates with both BLEU and human judgment. By analyzing 216 pieces of expert feedback, we find that NMT is preferable be cause it copies less than SMT, and, in gen eral, current models can translate fragments of the source sentence but make major mistakes. When we add these 216 expertcorrected paral lel texts into the training set and retrain mod els, equal or slightly better performance is ob served, which demonstrates indicates the po tential of humanintheloop learning. 1