Initial foray into Machine Translation began in the 1950s and 60s but it wasn’t until recent years that it came to be widely considered as commercially ‘feasible’. Machine Translation, now has the distinction of being both excellent, yet a cost effective advancement our industry is fast adopting. No wonder why everyone has their attention fixated on Machine Translation (MT). But in spite of having come ahead a long way from its early days, not all is perfect with piece of technology.
In favour of MT:
Companies all over have been adopting Machine Translation in hopes of an improved bottom line. MT has tremendous potential applications in visual media, news publications, online chatting and numerous other areas.
Google, currently pioneering in the R&D of machine translation, recently came out with their NMT backed real time translation earpiece, Pixel Buds. This earpiece more than hints at the revolution heading our way in the near future. Not only did the earpiece blow us away, their NMT also excelled at real world checks performed in 10 different languages. You can see the results here for yourself.
AI will continue to keep industry on its toes. It will bring about the largest reduction in the involvement of humans in translation. However, given that most machine translated work is considered ‘awkward’, outsourcing translation work is your only alternative, for now.
In favour of Human Translators:
MT has come a long way forward. But it still has poorer understanding of contexts. It makes mistakes, a lot of them. Humans make mistakes too, but another human in the loop can rectify such a mistake. Human translators also make edits for localisation of your content. For instance, it might be necessary to edit certain words and phrases to alter the style— to make sure the tone being conveyed to the audience, is the one originally intended to. While ensuring the meaning hasn’t changed. Translators take into consideration the cultural subtleties, helping the client improve their connect with their readers.
Modern MT systems use large quantities of previously translated texts to build a translation model. If these texts don’t exist, the system can’t either. So, paradoxically, machine assisted translation is dependent on human translators, for now. There are many languages we do not have such texts for, and there are many for which such texts have never existed in the first place. You’d have no option but to hire someone to do a translation of Old Japanese, if you are ever in need of one that is.
Is it perfect yet?
MT lacks the human touch. We are quite some time away from creating feasible, commercial AI translations of text books and poetry. When compared to alternative AI solutions, MT still produces inferior translations. As of now, MT does the bulk of the translation work and humans later loop in to make necessary edits, to rectify the mistakes and localise the content. This symbiotic relationship will likely continue for some time. Standalone works of machine translation will take some time to be widely accepted and recognised.
Fact is, AI is not yet able to truly and effective translate and localize content. AI’s only shortcoming is that it’s not human. While MT can more than take on the basic parts of translation work, it’ll be years (may be decades) before it can fully replace skilled human translators.
We at Virtual Lookup would love to hear your opinions and views on the subject. Write to us at firstname.lastname@example.org or ring us on 14156709780.