It’s not a secret why machine translation (MT) keeps on attracting the attention of the localization & translation service buyers: Significantly increasing productivity and reducing costs, machine translation seems like a magical solution to the impediments of translation, but is it?
In terms of quality, there has always been a big question mark around how reliable MT is in producing good final quality.
It is well known that the way machine translation works are based on three main approaches, including rule-based, statistical, and neutral, but in this blog, we will go a little beyond the technical aspects and begin from where it all started.
How Machine Translation (MT) Works?
The process is done through what’s known as a computer algorithm. The mechanism of machine translation (MT) varies; though it could be based on word-by-word translation through a basic automated search process (using a database to search for identical equivalents for each separate word from one language to another), MT has recently evolved into a more complex search process aiming to match and connect phrases together in order to come out with the most appropriate and natural translation into the target language. In the later model, the database is more likely to be collected and archived from previously translated materials that have been done by human translators.
Machine Translation: Expectations vs Reality
The recent significant improvement in technology might have made us think there’s an easier way to achieve a high-quality outcome that’s even better than human translation, not to mention that some people out there might also think that MT can smartly tweak translation and consider complicated cultural diversity aspects the same way do while actually MT merely operates based on algorithms that form and consolidate data from a certain database that has been collected and built from different sources. So, it is untrue to assume that MT fails to translate; the real problem lies in our expectations, as “users,” of machine translation.
Oh wait, so what does “algorithm” mean then?
Here comes the anachronism: etymologically, the
word algorithm is named after Al-Khwarizmi, a 9th-Century Persian mathematician who lived in Baghdad, Iraq, during the Abbasid Caliphate.
In mathematics and computer science, an algorithm is a self-contained sequence of actions to be performed. Algorithms to perform the calculation, data processing, and/or automated reasoning tasks.
The concept was adopted and later on turned into a ground-breaking innovation in the field of machine translation when an enthusiastic group of researchers and scholars was trying to find a way to quickly translate information back in the 19th Century. It wasn’t until some American universities began for MT that computational linguistics field of study came into existence.
What are the core issues with (MT) and why is the impact huge with RTL languages?
Basically, there are technical and non-technical reasons behind the core inefficiencies with (MT) in both Right-to-Left (RTL) and Left-to-Right (LTR) languages some of which are language-specific, others are attributed to the lack of proper usage for some RTL languages in addition to the inaccurate inputs and data, but why is the impact huge in RTL languages? Probably because of the different sentence structure and language direction.
Machine Translation: To Use It or Not to Use It? That’s the Question!
Concluding this blog, those who are fully counting on machine translation–especially with large text either for RTL or LTR languages–need to be aware that the outcome will be missing essential quality elements–which are the main factors for maintaining best quality translation–such as having the text gone through the main translation phases resulting in an unfinished and unreviewed outcome. Machine Translation (MT) can be good for limited small text volume. Moreover, as a technology, it is significantly developing, and even now it is playing a vital aiding role for translators during the translation process, helping them to come up with a variety of equivalents for different terminology.
In the meantime, it is not recommended to use MT with content that is going to be published online or going public considering, the high potential of mistakes that might occur when using MT, even with the most developed software. There must always be machine translation post-editing phase (MTPE) performed by a human translator to ensure the quality outcome is error-free.
The good news is that there’s no real/valid alternative for human translators. I can tell they will remain for quite some time and the machine will not take their jobs.
Get more details about how you can leverage the optimum blend of the latest, most advanced MT technology and the indispensable creativity of human translators to get the high-quality translations
