An example of the progress that has been made comes via an EU project to translate languages using machine learning. It’s a task that has challenged the AI community for years. Finnish, for instance, has 15 cases, so translating it to other languages is incredibly difficult.
To overcome this obstacle, the team are taking a different tact and have developed their system to recognize patterns in a huge text repository that it can then learn from.
“This machine learning strategy has nothing to do with natural intelligence, but it does have similarities with the processes that occur in the human brain when we control the muscles in our bodies. Children have to learn to pick up their feet when walking in the woods so as not to trip over roots or stones. In adults, this sort of mental process runs automatically in the background, as the brain has learnt how their feet have to be placed,” the team say.
Learning to translate
The project is undertaken by QT21, which is a consortium of leading machine learning research institutions across Europe and Hong Kong.
“Our common goal is to exploit machine learning to significantly improve automatic translation, particularly of more complex languages such as Latvian or Czech,” the group say.
The project is currently pulling together a wide range of datasets from places such as European government ministries to help train the software and put it through its paces.
Suffice to say, whilst the team are bullish over the prospects of the software, they don’t think it will reduce the need for official translators as they don’t believe computers will be able to work to the accuracy required for a little while to come yet, thus requiring translators to come along and post-edit any work done in bulk by the software.
It’s certainly a fascinating project however, and a good indication of the progress that’s being made. Whether they’ll crack the nut before I master the rudiments of Czech however remains to be seen.