“Beware of Safety” What?!
No doubt, everyone reading this has used Google Translate. No doubt, we have all received a very strange translation from time-to-time, and wondered “Why is Google Translate so inaccurate?” Well, this blog will tell you why, from top to bottom.
In short, firstly, Google Translate cannot recognise idioms, or cases where the same word may take a different meaning, depending on the context (mononyms). Secondly, there is no checking of the translation. There may be an error in the translation, in a language that you do not speak, and it would go unnoticed. Also, confidentiality! Whatever is put into that box, is then stored & used by Google, however they so wish.
So, why exactly is Google Translate so inaccurate..?
- No Context (or Idiom Recognition)
- No Human Revision
- It Lacks Any Subtlety
- So, How does Google Translate Work?
- When NOT to use it?
- When Should I Use Google Translate?
1. No Context (or Idiom Recognition)
Google Translate can see text, but it cannot understand it.
It can see that X equals Y, but it cannot understand why X equals Y. This is especially problematic when X can equal both Y & Z, but the choice between the two, is context-dependent.
One of the above, would be the case of mononyms (words that look & sound the same, but have differing meanings). For example, with “bat”, how is one to know the difference between the animal, or the piece of baseball apparatus? Without context, humans do not know, so neither can Google.
Other examples include: “desert”, “fair”, “lie”, “lead”, “row”, “wind”, and “project”. We could go on and on.
If we were to translate “I have many bats” into French, say, there is no telling whether Google will pick “une chauve souris” (the animal), or “une batte” (the sporting object). Without knowledge of French, one may be claiming that they play baseball with a small blind creature, rather than a wooden club!
Friends will laugh, and help you make a correction. But what something more important? To avoid costly misinterpretations, translations need to be perfect.
Another context issue, is the problem of idioms.
These are metaphorical sentences that make very little sense literally, but carry a deeper separate interpretation. Let’s take one commonplace French one:
Le pauvre, tout le monde casse du sucre sur son dos.
How did Google Translate fare…
Poor guy, everyone is breaking sugar on his back.
To a native English speaker, this makes no sense whatsoever. Other than a very literal meaning; a very bizarre and improbable image! So, what’s the actual translation..
Poor guy, everyone is talking about him behind his back.
Now, that makes sense. Clearly, the difference from GT’s effort, and the actual translation, is a gulf apart. There is simply no understanding the true idiomatic meaning with the translation on offer.
2. No Human Revision
Another big one as to why google translate is inaccurate, is the lack of revision. This means that any problem with the translation, will go unnoticed, because you cannot understand it!
In professional translation, one of the highest standards is the ISO 17100. With this, the translation is looked at by two native speakers & experts on the content (mechanics, for an instruction manual for a car, for example). After this, it is again checked by another person. This eliminates any possible issues, which Google Translate cannot.
Having said this, Google have updated their translator to include an editing tool. This has helped improve the quality of translation, but it still suffers from many of the same pitfalls. At the moment, in the world of translation, nothing beats an expert human mind.
3. Confidentiality Problems
There is also a more sinister side to this equation; as the oft quoted saying goes: “if it’s free, you’re the product”. The question changes from “why is Google Translate inaccurate?”, to “why does Google Translate want my data?”
One major drawback to GT is the complete lack of confidentiality. Anything that enters the GT box is stored for later use by Google in whatsoever fashion it pleases. If one is requiring the translation of confidential documentation, is it crucial that this does not get parsed by GT.
This can have severe ramifications, legally. Non-disclosure Agreements, for example, can be broken with a simple copy & paste; an action that is so ingrained, that it is instinctual.
4. It Lacks Any Subtlety
Proper translations are not word-for-word recitations. They are contextually relevant, subtly engineered images, specifically tailored to a given audience. Even if Google Translate gets past the issues of idioms & mononyms as outlined in section 1, it can still fall foul to the problem of emphasis.
Take this example:
I’ll see you at 9pm.
In monochronic cultures, such as the US, UK, or Germany, “9pm” means exactly 9pm. In polychronic cultures, such as Arabian or South American countries, “9pm” means “Around 9pm. Maybe I’ll be there at that time, maybe not”. They simply view things differently.
This is one micro-example that, on the surface, might seem a bit trivial. But it’s not. Understanding culture is at the essence of translation & localisation; something that Google Translate simply cannot do.
5. So, How Does Google Translate Work?
Google Translate is an example of a Neural Machine Translation (NMT) service and is the most popular free at-the-point-of-use translator on the market. But how does it work?
In 2016, Google Translate switched from being a Statistical Machine Translation (SMT) service, to being an NMT service. The former works as such: using the probability distribution that a given sentence in English, has the same meaning as a given sentence in French. The most probable, according to the algorithm, is the selected translation.
The latter is “example-based”. This is, essentially, a translation by analogy: a method which breaks phrases down into their constituent parts, translates them, and reuses them. For example: “My dog is very big” = “Mon chien est très grand”, and “My cat is very small = Mon chat est tres petit”.
From these two example phrases, Google can easily make more: “My cat is very big = Mon chat est très grand”, and “My dog is very small = Mon chien est très petit”.
The more examples that Google Translate is fed, the more that it “learns”, and the more ‘chunks’ it has at its possession to deploy.
6. When NOT to use it?
As I mentioned earlier, there are some instances where one should not use Google Translate. Essentially, the rule of thumb is ‘do not use it if the stakes of miscommunication are high’. It might just get you into deep trouble. Some examples are:
- Legal Contracts
- Medical Documents
- Health & Safety Documents
- Where your reputation is concerned:
7. When SHOULD I use Google Translate?
Having said all of the above, Google Translate is still absolutely brilliant in some situations.
In terms of when to use it, the rule of thumb is the inverse of the last. If the stakes of miscommunication are low, absolutely use it!
- If you just need a sense of what the foreign language is saying – known as gisting (soft G) – go right ahead. Furthermore, if you are trying to communicate in a basic way with another person, Google Translate can make that happen! Essentially, GT is very useful when the stakes are low: when it does not matter if there is a slight error in the translation.
- Secondly, Google Translate is – for the most part – absolutely fantastic for translating single words. I leave a slight qualification there, as we have already discussed mononyms! If used like a dictionary, GT is extremely helpful.
- Professional translation, despite it being needed when the stakes are high, is also expensive! There are several service options, but none beat GT on price: it’s free!
So, there we are, a complete guide as to “Why Google Translate is Inaccurate?” Thank you for reading. If you would like to read more from our blogs, click here!
Should you be interested in our services, please find them here.