In order to improve the performances of search engines for legal document managements systems such as Eunomos, the structural items of legal documents (articles, paragraphs, etc.) need to be classified, i.e., associated with tags or labels, taken from a restricted set of relevant ones. Of course, the task of automatically classifying (parts of) legal documents is needed in order to avoid intensive manual work, which is generally slow and costly.

Tags/labels may refer to the topic of the structural item, i.e., they inform what is the content of the structural item about, or to their function, e.g., if it is a modificatory provision (cancellation, update, insertion, etc.) of a previous norm.

University of Luxembourg, together with University of Turin developed a Support Vector Machine classifier able to classify structural items of legal text with respect to the EuroVoc multilingual thesaurus, as well as a rule-based classifier for modificatory provisions. A demo of the first is available here. Both tools have been used within the FP7 project EUCases (see (Boella et al, 2015)).

Main contributor(s): Livio Robaldo, Luigi Di Caro, Guido Boella