Website translation specialist SDL has launched a new system that learns in real-time and will create algorithms for social media formats and informal languages.
The XMT engine, referred to as ‘the next generation in machine translation’, is a complete re-write of the MT system. It was created in SDL’s own research labs and its main function is to improve the quality of language pairings that are difficult to translate, such as English to Chinese.
Speaking to Travolution at the Phocuswright conference in Dublin, ahead of the launch of the new system, SDL account director Kevin Ashbridge said that social networks had also changed translation needs.
“This is a global system that offers 100 language pairings – it even works on social media to change expressions that are difficult to translate, like acronyms or dialectical terminology,” said Ashbridge.
“The system remembers a correction made by human translators in the bedbank and learns in real-time. It is currently being modulised to translate to Asian languages from English.”
Ashbridge explained that word order may have to be changed to allow some translations to make sense but because XMT allows different algorithms to be used for different languages, you can make a change to just one language pair without affecting any others.
XMT will be made available via SDL’s suite of machine translation solutions, including SDL Language Cloud and SDL BeGlobal. The engines can be trained on individual organisations’ specific vocabulary and terminology to offer customized translation.
SDL counts Kuoni, Accor and Thomas Cook among its travel industry customers.