Machine learning – what it means to us in the language industry
One thing we’ve noticed from visiting a number of conferences outside the language industry recently is just how quickly technology is advancing. Take robotics, for example – what was once something you’d only find in a SciFi movie has now got people worried about the future of their livelihoods. While the language industry is innovating at an admittedly somewhat slower pace, it’s still evolving, especially with the help of machine learning.
AI and machine learning have helped us speed up the translation process, and improve quality at the same time. This has allowed companies to pump money into new markets, reinvesting the funds they once would have spent cracking just one market. Even though we’ve only just started harnessing the power of machine learning, here’s how it’s already making an impact.
Although neural machine translation is no longer “new” news, it’s still playing a massive role in driving the language industry forward. Deep learning is helping to give context to source sentences, thus improving the quality of output. It’s even been proven to do a better job than a human in identifying complex patterns in datasets.
QA is a vital part of the translation quality process, but how do you provide QA on a budget? The answer is automated QA. Automated QA engines use pattern recognition and other technologies to identify potential issues such as missing content and inconsistencies. If used early enough in the process, it can help translators go back and fix their mistakes before the delivery deadline. In case you’re worried about leaving QA to technology, it’s actually been proven to be able to detect more errors than a human QA specialist.
Machine learning has also been impacting on the way vendor managers work, too, by helping them find the best linguist for a given job. Using big data, machine learning can determine who has the most relevant experience for a certain project, and even who does a better job of translating certain things.
Machine learning can also help with translation workflows as a whole. Starting off with vendor management, they can help teams decide which terminology to use, for example, before going on to help with machine translation. ML can also alert translators to potential issues before they’ve even got to that particular point so they are fully aware of what they’re dealing with.