New generative AI startup Eleven Lab Speech translation is the hottest category for translation services. But it pre-empted another market that startups have been targeting for some time now: content translation. This market remains large, as international businesses need to translate content around the world, as evidenced by the $106 million raised to date by companies like Portugal’s Unbabel (which recently raised $60 million).
Easy TranslationContent translation specialist has been around since 2010 by using machine learning models to identify the best freelance translators to translate specific types of content. But now the company is taking a new direction with a new generative AI-driven platform it’s calling “HumanAI.”
“We’ve shifted our entire business model from a human service-based business model to being an AI technology provider, reducing costs and speeding up processes,” company founder Frederik R. Pedersen told TechCrunch.
Most translation services provide machine-translated content, some of which is edited by humans. However, translators often need to evaluate the entire machine-generated translation to understand the context and meaning of the content. EasyTranslate’s HumanAI platform reverses this, absorbing content and blending it with large language models (LLMs), and using the short-term memory of LLMs to translate content more accurately. Additionally, humans are involved only when necessary, reducing translation time and costs.
To achieve this, HumanAI uses a combination of multiple LLMs, including those provided by OpenAI, and a proprietary recommendation system. The platform uses proprietary algorithms and customer data to provide customized content translation.
The secret to this transformation, according to Pedersen, is using LLM to generate short-term memory, allowing the platform to read a general English translation and convert it into specific English. By “vectorizing” content into a database, it can do a semantic search, find similarities between pieces of content, and use that to generate the LLM (which is Search Extension Generation).
This means that the platform can use any number of LLMs to translate, for example, the English used in marketing copy and the English employed in financial reports, all the while preserving the meaning of the text.
“More traditional neural machine translation engines can be combined with customer-specific data to form the foundation of the localization and translation process – moving from a general language to a customer-specific language, for example,” he said.
Why is that important? Pedersen explains: “Machine translation might give you a grammatically perfect translation, but it might still not sound right. So we identify which parts of your content have a low confidence score and a human fixes it. This combination makes us much more productive.”
Pederson claims HumanAI can cut translation costs by 90%, pricing its services as low as 0.01 euros per translated word. The company’s clients include global companies such as Wix and Monday.com.
Pricing is a particularly important issue in this sector, as companies have large amounts of content that needs to be translated.
“If you look at Adobe, they have a team dedicated to researching how terminology aligns across markets, and if you look at global brands, there’s a lot of effort put into making sure it’s recognized correctly locally,” Pedersen says.
But the question is, what will help EasyTranslate compete with pure AI-based solutions that are likely to improve over time?
“Our goal is not to become a pure AI. [service]”We believe our goal is to create added value by combining humans and AI and provide it to our customers. To further improve the AI, we still need human feedback,” he said.
“It’s one thing to say you want to implement all the content creation, all the translation, but it’s another thing to make sure you can actually control the model. You need humans to control the model because humans are not machines and language is always changing.”
EasyTranslate has raised a total of €3M to date and is backed by private equity, debt financing, Copenhagen angel investors and the Danish Innovation Fund.