Even if advances in technology mean that automated translations are perfect 90% of the time, enterprises can’t afford mistakes that can have serious consequences, so they are still sending every single AI-generated segment to a professional human translator for review at a cost about $1 per sentence, a price point that for big corporates can add up to millions or tens of millions of euros a year. ModelFront’s machine learning technology aims to predict which machine translations are correct and can be used as is, skipping unnecessary, costly human post-editing by enabling a hybrid translation workflow that combines human quality and machine speed.
“Our technology knows when to stop and let a human take over,” says Adam Bittlingmayer, CEO of the Silicon Valley-based company and a speaker at DLD AI in Munich September 7.
ModelFront’s technology enables a significant portion of content to be generated by AI (such as Google Translate or Microsoft Translator) and then “safely approved” by ModelFront’s AI. Bittlingmayer calls it “sensible automation.”
Customers include Fortune 1000 companies and startups.
“Our tech saves enterprises real time and money and every major translation management system (the unsexy software where humans do this work) has integrated ModelFront or is working on it,” he says.
ModelFront’s API supports more than 100 languages. That is key, says Bittlingmayer, because more than six billion people can’t understand English. They make up 80% of people on the Internet, and also most of the growth in users and spend.
“At one dollar a sentence companies don’t translate into local languages for many of the markets they serve,” he says. “By making this more efficient companies are going to be able to translate more content.”
The founders of ModelFront bring full-stack engineering experience building highly-scaled data-centric APIs, marketplaces, and SaaS platforms at Google Translate, Android and the growth-stage startups.The venture-backed company’s competitors include Google’s Translation Hub, which uses the search engine giant’s Neural Machine Translation, Phrase, which has this feature in its platform and Unbabel, an artificial intelligence-powered human translation platform.
Bittlingmayer says competitors add this service to enhance their offering. “We do not provide anything else,” he says. “We are fully focused on solving this problem.”
To read more of The Innovator’s Startup Of The Week stories click here.