- Posted by Stephen Whiteley
- On 17/11/2011
- machine translation, Memorias de Traducción, MT, TM, Traducción Automática, translation memory
What is the difference between Machine Translation and Translation Memory?
There seems to be some confusion about what, precisely, distinguishes Machine Translation from Translation Memory. This may be the result of the similarity of their acronyms in English – MT and TM, respectively – or the fact that they are both ways of using technology in the translation process.
But the question ‘what is the difference?‘ really doesn’t make any sense in this context: Machine Translation and Translation Memory are two totally different things.
What is Translation Memory?
In order to appreciate this difference, it is important to be sure where translation memory fits in the modern translation cycle.
Computer Aided Translation (CAT) software is a sort of generic name for the various bits of software that in combination make the translation process more streamlined and effective.
In (very) brief, what CAT software does is split a source text into manageable units known as “segments”, and builds databases of equivalent segments in different languages. A segment is the basic semantic unit of a text: ‘the red house’, for example, or ‘eighty-three’.
The databases of these matching segments form what are called Translation Memories (TM). TMs store all of the segments pertaining to a particular client.
CAT makes the translation process faster and more cost-effective, as it means that our team of translators don’t need to spend time or energy on something which has been translated before – savings which are, of course, passed on to our customers.
How Translation Memory works
Now, we have looked at TM at length on this very blog. Here is a selection of articles on the nuts and bolts:
Translation Memory: the basics. What it is, how it works, and what it does.
Using TM to optimise the translation cycle. How TM makes the translation process quicker, cheaper and better.
Potential pitfalls involved in using TM.
Data protection issues in relation to TM
So…what’s machine translation?
Machine translation (MT) is exactly what it sounds like: translation done by a machine.
To be more specific: it is the translation you get when you feed a text through an online translation engine. In the past, a variety of these engines vied for dominance, but in recent months Google Translate ,with tedious inevitability, has come to dominate the market.
Again, we have looked at MT at length on this blog.
“The inner workings of Google Translate are shrouded in a characteristic mix of Willy Wonka bonhomie and mildly Orwellian obfuscation…” Here is examination of how Google Translate works (as far as we can tell), and the various data-protection issues that surround it.
Here is an article about Google Translate’s version of the opening sentence of Don Quixote – quite good, apart from the fact that, by omitting one tiny word, it inverts the sense of the piece: the achilles heel of MT.
Why you shouldn’t use MT to translate marketing collateral.
What’s wrong with machine translation?
There is absolutely nothing wrong with MT, if you want to get the gist of some document quickly and for free. Feeding an email, say, into Google Translate or Babelfish will enable you to understand more or less what it says.
However. If what you want is a translation of anything like an acceptable standard, MT just isn’t an option; it always makes mistakes, sometimes terrible ones. For every sentence that it renders more or less accurately and appropriately, there will be a sentence of gibberish, or one in which the meaning has been totally lost or changed.
And heaven forfend that anyone try to use MT for documentation which a customer will see – marketing experts are unanimous – you might as well cover it with a big red stickers that say ‘DON’T BUY THIS’.