Artificial intelligence translator Lengoo has recently ended the Series B round with a new $20 million funding, reported Tech Crunch. The company seeks to automate the translation process to cater to the growing need for such cases.
The company will use the fund to expand into North America and other European markets. The round was participated by investors Redalpine, Creathor Ventures, and Techstars, as well as angels Matthias Hilpert and Michael Schmitt.
This service follows several other language-focused technologies such as DeepL and Lilt in addition to services already offered by companies like Google Translate. Tech Crunch noted that there are cases in which translation services can only be “good enough.”
Lengoo CEO and founder Christopher Kranzler said, “The next step to take obviously was automating the translation itself.”
However, it is worth noting that humans will remain relevant when it comes to translating from one language to another. This was also the case for DeepL and Lilt, especially as there are niche cases that can only be accurately translated with the help of humans.
Kranzler shares the same opinion, saying, “We’ll still need humans in the loop for a long time – the goal is to get the models to the level where’s they’re actually usable and the human has fewer translations to make.”
What Lengoo seeks to achieve is to provide speedy and specific translations by creating a model that combines jargon, style, and formats given by a client in the translation process.
The company creates a custom machine model and trains it using the requirements and specifications of its clients. It is also fed with documents and websites by customers, as well as feedback on its performance.
“We have an automated training pipeline for the models. The more people contribute to the correction process the faster the process gets. Eventually, we get to be about three times faster than Google or DeepL,” said Kranzler.
To start, customers will need a custom model based on the existing documents from the past years. Machine learning comes into play every time users make corrections, which the system will remember and integrates for the rest of its training and operation.
The outputs are bound to have some errors, but the AI has a built-in quality check mechanism, resulting in fewer corrections. However, this does not mean that the results are guaranteed to be high-quality, but it emphasizes the speed of the process.