Machine Translation – Future Reality or a dream well sold

Recently I attended the Localization World, Bangkok and found that the theme of the whole event revolved around Machine Translation and how it has helped many organizations in leveraging content for multiple languages in a very short time span.

From distance, the whole approach of Machine translation and its success looks a reality but on close inspection it feels that there is more to Machine Translation then what has been sold to an average customer. For those who are unaware of what Machine Translation is, here is a layman’s explanation of it. Machine Translation is a standard name for computer based systems which are responsible for producing translations from one language into another with or without the help from humans.

Machine translation has been commercially used in the Localization industry for over a decade and has seen fair share of success in recent past. It has definitely matured over the past few years with the incorporation of new tools to leverage the content available in the market, but we have to be realistic of the extent where we can utilize Machine Translation in our daily Localization Project cycles. It cannot become our one stop solution to all localization content requirements. If you are focused on producing vast content of marketing collateral and trans-creational content, then Machine Translation is definitely not the answer. It would be unwise for us to expect a machine to give us content in a creative dimension that can be expected to be accepted by the end users. MT definitely has the capability to produce fairly acceptable (~85% success rate) literal translations, but it will be unfair for us to expect any MT system to deliver content which displays artistic and creative flair. For organizations who deliver reiterated content with high repeat value, configuring machine translation into their localization workflow would certainly be a fruitful enterprise, if managed with great precaution. Machine Translation certainly has the potential to define the Technology aspect of Localization Industry, but what about Supply chains involving Humans and processes? There are few glaring issues which are yet to be tackled for Machine Translation.

What happens to Machine Translation efforts for Asian Languages where success rate is less than 30%? Majority of the published content is in its native form, and converting this content for corpus generation is the biggest challenge. Apart from corpus creation, the contextual flexibility of the languages adds to the challenge. Formulating process flows for these languages deviates from the widely used workflows in the Machine Translation world.
Defining Quality Assurance process for MT content which will be acceptable for the end consumers is a bigger challenge. Who will define what level of quality for MT will be acceptable to the end customers?
At what level MT can be incorporated in the Localization workflow involving TEP and PE of MT. How will this impact the whole localization project cycle will be worth noting. It is important to be realistic of what can be achieved with Machine Translation and how. Optimizing your Machine Translation processes and activities are vital to determine productivity metrics to manage in house resources in collaboration with all “live” project work to contend with MT as well.

The writer is associated with Braahmam as Marketing Manager – Communications and the views expresses here are hers and hers only. The company is not responsible for the accuracy, completeness, suitability, or validity of any information on this blog.