April 12, 2019
A Deeper Look at Neural Machine Translation, History and Related Technologies
Neural machine translation is showing a lot of promise in being the next big thing in translation. As such, it’s worth taking a deeper look at how this technology works and where it came from.
A History of Machine Translation
Neural machine translation (NMT) is one of the latest developments in automated machine translation. Like many technological advancements, NMT has come around due to a series of ideas and advancements that came before it. So, it’s worth taking a look at the history of automated machine translation.
The Georgetown Experiment
The idea of a machine translating language was a thought for a long time before it became a reality. One of the earliest and most well-known successful research came about in the 1950s with the Georgetown Experiment.
This experiment was on January 7, 1954, as a project involving Georgetown University and the information technology company IBM. The system was a simplistic one but impressive as it translated more than 60 sentences from Russian to English automatically. It did exactly what it set out to do - garner interest in the research of machine translation.
The Goal of Early Machine Translation
Nowadays, we tend to think of machine translation publicly, and on a much more daily basis. In other words, machine translation has started to become a integral part of our lives - a way to communicate with one another beyond a language barrier. This wasn’t the entire goal of the original machine translation developments. Much like the first computers and the internet, it was more of a government interest. It was spearheaded by Russia and the United States, mostly for the goal of translating items such as scientific journals.
Statistical Machine Translation (SMT)
The most common type of machine translation today is statistical machine translation or SMT. This is a machine translation that is dependent on a rule-based approach. In other words, it is set up with specific language rules and it uses those to translate one language into another. While useful, it does have its flaws. For instance, these rules may lead to an inaccurate or clunky translation. It may also struggle dealing with intricacies such as words and phrases that don’t have an equivalent in another language.
How This Has Led to Neural Machine Translation (NMT)
NMT is the product of wanting translations to be executed quickly, accurately, and seamlessly. Instead of the input style of the past, it is an AI-based system designed after the way human biology works. Rather than being limited to inputted rules, it is meant to “learn” the same way a human would. While this technology isn’t commonplace yet, it shows promise in being the next step of machine translation.
Read more articles about the history on the following sites: