A spin-off of the Hieronymus agency, specialized in legal and financial translation, Neur.on is breaking new ground with the creation of Corrext, a machine translation platform based on Natural Language Processing (NLP). A challenge in a world where data security is key and vocabulary extremely complex. Here's a look back at a technological and human success story.
Hieronymus, an agency specializing in legal and financial translation, embarked on the artificial intelligence adventure following an initial Innosuisse project supported by Alliance, which introduced them to the potential of data and Natural Language Processing (NLP). Paula Reichenberg, director of Hieronymus, then decided to create Neur.on, a spin-off specializing in neural machine translation for legal and financial professionals. This LegalTech start-up has surrounded itself with the right people to take on major challenges, and has launched the Corrext platform, which optimizes translation processes for complex documents, while ensuring quality, expertise and security.
Transforming a traditional profession
Following the arrival and widespread adoption of neural machine translation solutions, Hieronymus was inspired to take a further essential step towards automation. "We discovered that it was also possible to automate the preparatory phase of translations, from identifying lexical fields to checking formats, while ensuring maximum data security," explains Paula Reichenberg. But the challenge is considerable: to open the door to a complex technical world, while ensuring perfect integration with the very specific needs of legal and financial translation. Artificial intelligence must support legal specialists in an extremely controlled environment, where each document is analyzed and adapted according to precise criteria, with human intervention to validate technical and sensitive passages.
From the Innosuisse project to the birth of a start-up
While Paula Reichenberg was convinced that she needed to open up her SME to the world of data, she also needed someone to bridge the gap between the legal world and the scientific and technical world. It was during a course at EPFL that she heard about Alliance and Robert van Kommer, innovation advisor for the IT and AI fields. "From the very first contact, Robert was able to understand our needs and direct us to the right people," says Paula Reichenberg. He put her in touch with Jean Hennebert, from the iCoSys Institute at the Fribourg School of Engineering and Architecture (HEIA-FR), and supported the first Innosuisse project, which was accepted. Paula Reichenberg then decided to launch a second Innosuisse project, again with the support of Robert van Kommer, to create the startup Neur.on. This was followed by a third SIP project, for which Robert van Kommer acted as consultant. "These Innosuisse projects require a lot of hard work, rigor and collaboration. Robert's help has enabled us to make leaps forward, and the story doesn't end there," confides Paula Reichenberg.
Putting together a good Innosuisse dossier requires not only a great deal of rigor, but above all close collaboration and genuine trust between everyone on the team. You can't do it alone.
NLP at the service of human expertise
With Corrext, Neur.on has designed a disruptive platform for the legal translation market. The innovation of this platform lies not only in the use of NLP to automate translations, but above all in the assistance it offers users. "Contrary to what you might think, machine translation has not made the translator's job disappear - quite the contrary," emphasizes Paula Reichenberg. The platform developed by Neur.on identifies the most complex passages, pinpoints sector-specific vocabulary and generates a reliability score for each part of the text. This assistant role guides users towards elements requiring manual verification, and enables them to go much further than standard tools, while ensuring strict control of data security.
Collaboration, the key to success
Today, Neur.on continues to innovate with a reinforced team. The start-up is in the product-market fit phase and is making rapid progress. The dossiers put together with the HEIA-FR team and the support of Alliance pour les projets Innosuisse have enabled them to establish clear objectives, which they still use today as a roadmap. The research is progressing, the product is evolving and several students who worked on the projects have been recruited, ensuring a mixed team of linguists, lawyers and engineers, all of whom are very involved.
Neur.on in brief
Founded: 2022
Head office: Fribourg
Team: 11 people
Website: neur-on.ai