Service: NER-based extraction of contract data

Responsible organisation: RSZ/ONSS (Private sector)

Smals is an ICT provider for government with new technologies and seeing where they can be applied for in government. One of them is the RSZ/ONSS with many siloed data. They wanted to centralize the database. The question is to extract some kind of information out of incoming documents with 2 proofs of concepts. The first one is to see if emails can be analysed to extract persons, functions, phones. Especially structured information are possible. Functions are difficult as it depends on the context. Another Proof of Concepts is to extract information from forms. First, these letters have to be scanned before they can be analysed. However, the pilots showed that AI for everything is not always possible and 100% accuracy is not possible which requires validation by people. AI cannot understand text because they only correlate patterns. It may therefore be more useful for mailroom automation software.

Additional information

Source Open Innovation Regione Lombardia
Web site  https://www.smals.be/nl/content/named-entity-recognition-une-application-pratique-du-nlp
Start/end date 2019 -
Still active?

Related cases