In our ever changing world, and with the excitement of new technology, it is clear that there have been great advances in the area of Machine Translation. Today, we read about Google Pixel’s ability to instantly translate speech for people having a conversation, and we use Microsoft’s Bing Translator and Google translate to speak with our friends from foreign countries and to learn new languages.
At the forefront of this technology is a new concept that has dramatically increased the abilities of Machine Translation: Neural Networks or Neural Machine Translation. This advancement allows Artificial Intelligence to learn from past translations and create new ones, always revising and improving. Just like Facebook and Instagram have learned to identify faces in pictures using AI, Machine Translation is now able to identify parts of the context in a sentence.
Although all of this seems very promising and on the surface, it sounds like the idea of a Human Translator is on the verge of extinction, the opposite is actually the case: Human Translators are more vital than ever.
Here’s a list of reasons why:
- Translators are needed to help format documents like Adobe Illustrator
- Translators understand the emotions behind the text
- Translators are able to understand the context: industry speciality
It doesn’t mean that the human translators will be perfect in every case, but the ability of a human to handle these sorts of complex tasks is much more robust and capable than a machine.
Firstly, we need to understand how the systems work. At the first stage, there needs to be data to work with. This data needs to be “good data”. If a poor translation is inserted into the machine, the output will either be consistently bad, or it will produce hit or miss results.
On a beginning and fundamental level, we need human translators and linguists to work with the developers to help refine and develop the actual code. Otherwise, the developers would be working in the dark and hoping that whatever the software spits out on the other side, that being linguistic data in the target language, is accurate. Translators and linguists are needed to feed quality linguistic data into the system. See Google’s research blog to see how this is done.