With the increasing popularity of online machine translations, one question that is increasingly being asked is this: Will there be a time when human translators become obsolete?

Having worked as a translator for a few years, I have had some experience using online machine translations. Let me answer this question with a confident NO! I will explain why below.

Firstly, the key reason why machine translations work badly is because they do not understand ‘context’. While humans intuitively understand what this means, ‘context’ is a difficult word to explain properly. Teaching it to a machine, or writing it into a programme, is consequently fraught with difficulties. For those with some background in IT and programming, a common way to do so, is to search the words before and words after in order to figure out the correct translation (think optimization or probability), but this method is still not too accurate and more often a case of ‘hit-and-miss’. As anyone who works long enough with languages know, there are all sorts of exceptions with regards to the rules of any language, let alone translating into another language. How to optimize and increase the accuracy of such a procedure is currently a matter of intense research.

Hence, we find that translations work fairly well for singular words used generally, where context is not so critical. On the other hand, translations of whole sentences leave much to be desired.

The source language and target language also greatly affect the accuracy of translation. Since English is the most common source language (and also target language) of translation, let me use this language to illustrate my point. From my experience, I have found that if the target language is an “anglicized” language (like German, Malay, French, Spanish, etc), the translations tend to work better.

However, if we translate into “non-anglicized” languages (like Chinese, Japanese, Korean, etc), we often find that the translations do not make much sense. This has to do with the fact that the structures of such languages vary quite differently from English and most translation programmes were first written primarily to translate between European languages, not to mention that most European languages are rooted in Latin.

In conclusion and to reiterate the question posed:  Will there be a time when human translators become obsolete? My answer is a resolute NO!

PS: All opinions expressed above reflect the views of the writer only and do not in any way represent the views of Whizwordz International. While all due care and attention have been made with regards to the content, the writer cannot be held responsible for any inaccuracy or factual error.

Have you attended wedding dinners before? Ever wondered what the usual programs or activities are during the dinner? It is usually the emcees doing the opening, then introducing a video clip of photos of the bride and groom from young till now, before the couple into the ballroom. This is a typical kind of opening for a wedding dinner.

Last night, I went to my cousin’s wedding. Her husband and her are quite “arty farty” kind of people, so instead of doing the photo collage, they came up with a comic strip -- hand drawn by themselves. It was based on the lyrics of a song. As the song went on, the story develops. It was a different feel although I reckon that our elders understood what the comic strip was talking about.

It got me thinking, does a picture really paint a thousand words, like how the song goes? Somehow if you put many pictures together and have a song with the lyrics to go them, it does paint -- well, maybe not a thousand words -- but at least a story that everyone is able to relate to. Just like translation.

Machine Translation (MT) is generally understood as the process in which a “machine” (specificaly a computer program or software) is substituting words and phrases in one langauge to another.

But if you do understand a little bit of the language in which you have translated your document into, you will quickly notice the translation problems.  And the end result is definitely not what you have desired and clearly not of publishable quality.

MT Softwares consist of dictionaries and algorithms to compose a sentence according to grammar rules of a language. The inherent difficulty in machine translation is that it does not discern the context and intended meaning the way humans can.  It doesn’t take into the meanings of words, word combinations, sentences, paragraphs, or the whole text structure. It knows nothing about rhetoric and style. It only diligently substitutes words and expressions taken from a dictionary into another language.

Of course, nobody wants their document content to be turned into a meaningless word mesh.  And there are specific styles of translations that machine tools cannot handle properly.  The basic purpose is to translate the message and meaning of the document and not just words and phrases. For this, it is always better to engage the services of professional translation agencies that employ native speaking human translators.

A human translator is still needed to correctly convey the exact meaning of the source text. No matter how advanced translation tools have become, they cannot be regarded as a substitute for a competent human translators.

It is true that MT can offer better alternatives in term of speed and cost. But if you intend to have a good translated document or marketing collaterals, accuracy is still of utmost important.  As of now, MT has not achieved that kind of high accuracy yet.   In a way, you would not want to submit a document full of syntax and grammar errors to your clients.  The “translated” copy will not make any sense to your clients.  Eventually, it will cost more if the documents need to re-translate again by a human translator.

Developers of Machine Translation software recognize these limitations and advise to use these translations as drafts only.  This is to enable the users to check the meaning of the foreign language document quickly and inexpensively.