After a few years of hard work, the Recognos ETI platform was released in the US. The platform got the name of ETI that aside of the acronym for Extract – Transform and Integrate is the name of the person who came out with the idea and first implemented it. His name was Iulian Margarintescu and his nickname was Eti. Iulian passed away last August but he is always with us in our thoughts and great memories.
This a human in the loop machine learning platform that extracts information from unstructured documents.
This is a story that is a good example how the ideas are born. Years ago Recognos started to work on a platform to extract data from unstructured legal documents using semantic technology, i.e. semantic NLP scripts. A system was implemented but Iulian, the technical lead had an idea to create a system and use humans to teach the system how to extract data. These documents are with known content, for example contracts, agreements, etc. The users know what fields are in them but don’t know where there are. So the business users define all the fields of interest in these documents and after that a user starts manually to extract them, i.e. highlight the fields and move them in the right data slot. The machine stores what the user did, looks what is before the extracted data and what is after it in the text (this is the context) . After that a machine algorithm starts applying the stored information to extract the data from the next document. Of course errors occur, but every error corrected by the user is also stored and creates an improvement in the extraction algorithm. This proprietary method was created by Iulian. Unfortunately Iulian passed away in an accident last year. It was a shock for all of us, he was a great mentor, amazing technical lead and great person. Due to the fact that he was a great mentor, the team members were able to continue his work. We didn’t have a name for the product. And all of the sudden, same time, two of us had a revelation, that Iulian’s nickname ETI can become the name for it. ETI is also an acronym for Extract Transform and Integrate, it is exactly what the product does with the unstructured data. And so ETI will be with us forever.