You would possibly think latest generative AI startups like Eleven Labs are the most well liked game on the town for translation services. But voice translation was way back preceded by one other market, targeted a while ago by startups: content translation. Any company with a world presence must have their content translated all over the world, so this stays a giant market. This was evidenced by the $106 million raised so far by the likes of Unbabel in Portugal (which last raised $60 million).
EasyTranslate, which makes a speciality of content translation, has been around since 2010, using machine learning models to discover which freelance translators were best suited to translate specific sorts of content. But now it’s headed in a brand new direction with a brand new, generative AI-driven platform that it calls ‘HumanAI’.
“We have now pivoted the entire business model from a human service-based business model towards being an AI technology provider, driving down the price and speeding up the method,” the corporate’s founder Frederik R. Pedersen told TechCrunch.
Most translation services offer machine-translated content, with a small portion edited by humans. Nonetheless, translators often must assess the complete machine-generated translation to know the context and make sense of the content. EasyTranslate’s HumanAI platform flips this on its head, absorbing content, mixing it with large language models (LLMs) and employing short-term memory within the LLM to translate content more accurately. What’s more, it’ll only involve humans where it must, thus reducing translation times and costs.
To do that, HumanAI uses a mixture of LLMs, including the one offered by OpenAI, in addition to its own advice systems. The platform runs off its own algorithms and customer data to offer customized content translation.
The key to the pivot, Pedersen said, is using LLMs to generate short-term memory so the platform can read a translation in generic English and switch it into specific English. It “vectors” content right into a database, enabling it to do a semantic search and find similarities between content, which is then used to create a short-term memory with an LLM (this can be known as retrieval augmented generation).
This implies the platform can use any variety of LLMs to translate between, for instance, the English utilized in marketing copy or English employed in finance reports, and preserve the meaning within the text all of the while.
“We are able to mix the more traditional, neural machine translation engines with customer-specific data to create a foundation for the localization and translation process. So, moving from generic language towards customer-specific language, as an example,” he said.
Why is that vital? Pedersen explained: “You would possibly get a grammatically perfect machine-based translation, nevertheless it still may not sound correct. So we discover which a part of the content has a low confidence rating after which use humans to correct it. The mixture massively increases our productivity.”
Pederson claimed HumanAI can drive-down translation costs by 90%, and finally ends up pricing its services at €0.01 per translated word. Its customers include global businesses similar to Wix and Monday.com.
And pricing is an especially crucial puzzle to unravel on this space because firms have an amazing deal of content that needs translating.
“Should you have a look at Adobe, they’ve a full team just taking a look at how the terminologies align across markets. And if we have a look at global brands, there’s a major amount of effort put into ensuring that you just are perceived in the correct way locally,” Pedersen said.
The query is, though, what is going to help EasyTranslate compete against pure-play AI-based solutions, that are prone to recuperate with time?
“Our goal shouldn’t be to develop into a pure AI [service]. I believe our goal is to create the added value of getting humans combined with AI, and supply this service to customers. AI still needs human feedback to be improved,” he said.
“It’s one thing to say you prefer to to implement all content creation, all translation, and one other to be sure you could actually control the model. You may have to have some humans to regulate the models, because humans aren’t machines and language changes continuously.”
EasyTranslate has raised a complete of €3 million so far, and is backed by private equity, debt financing, some angel investors in Copenhagen and the Danish Innovation Fund.
