FrEX: Extracting Property Expropriation Frame Entities from Real Cases

We describe FrEX, a Frame Entity eXtractor for real estate property expropriation cases rooted in Frame Semantics and heavily exploiting Natural Language Processing approaches. FrEX has been tested on 24 real, non anonymized cases shared with us by the Tribunal of Milan, described by almost 1000 PDF documents. FrEX results were compared with the relevant entities associated with those cases, namely debtors, creditors, lawyers, judges, experts, cadastral data of the property to be expropriated, manually inserted by domain experts. Although FrEX’s development is still under way, the results are very encouraging and suggest that it can effectively relieve lawyers and judges from the highly repetitive task of looking for entities relevant for expropriation cases, when retrieving, filtering, and classifying legal documents.