Code-switching input for machine translation: a case study of Vietnamese–English data

Multilingualism presents both a challenge and an opportunity for Natural Language Processing, with code-switching representing particularly interesting problem computational models trained on monolingual datasets. In this paper, we explore how code-switched data affects the task of Machine Translation, which only recently has started to tackle multilingual data. We test three Translation systems from Canberra Vietnamese–English Codeswitching Speech Corpus (CanVEC) evaluate translation output using automatic human metrics. find that, perhaps counter-intuitively, performs better input than input. particular, comparison evaluation suggests that codeswitching may boost semantic faithfulness output, effect term lexico-semantic enrichment. also report two cases where is most least clear in Vietnamese–English, namely gender-neutral 3SG pronouns interrogative constructions respectively. Overall, suggest Processing more generally, ought view multilingualism as rather obstacle.Abbreviations: 1: First person; 2: Second 3: Third CLF: Classifier; COP: Copula; DET: Determiner; PL: Plural; POSS: Possessive marker; PRT: Particle; PST: Past tense; Q: Question SG: Singular