The Discrepancies of Online Translation-Machine Performances: A Mini-Test on Object Language and Metalanguage

This research has explored the discrepancies in online translation-machine performances through a mini-test on object language and metalanguage translation. Within this framework, object language is the verbal language with its literal meaning or denotation, as it factually is, while metalanguage is the language with its figurative meaning or connotation, which results from human creative imagination. The words sitting duck as a duck which is sitting (bebek duduk) is an object language while sitting duck as an easy target (sasaran empuk) is a metalanguage. This research has gone through six methods: online observation, online-machine translation, auto-expert judgment, verification, classification, and interpretation. The discrepancies within this quasi-qualitative research are obtained from verification on four set-up aspects of 10 corpus data, i.e. object language alone, object language within context, metalanguage alone, and metalanguage within context. Upon the analyses of English-Indonesian translation performances by Google Translate, Bing Microsoft Translator, Yandex Translate, and Systran Translate, the high-percentage discrepancies of translation machine performances mostly happen in the translations of metalanguage, while the translations of object language are successful enough with low-percentage discrepancies. Upon the findings of this research, online translation should be developed and improved within the aspects of metalanguage in the target language. Theoretically, the findings propose the inclusion of object language and metalanguage in the online translation machines, while empirically they challenge more metalanguage words to exercise the appropriacy of online translation machines for their better service ahead. This study can also be a research model for other languages in the world in the context of online translationmachine performances.