Unit:
Κατεύθυνση / ειδίκευση Διαχείριση Πληροφορίας και Δεδομένων (ΔΕΔ)Πληροφορική
Supervisors info:
Μανόλης Κουμπαράκης, Καθηγητής, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ
Original Title:
Named Entity Recognition and Linking in Greek Legislation
Translated title:
Named Entity Recognition and Linking in Greek Legislation
Summary:
We show how entity recognition in Greek legislation texts can be achieved by utilizing a
named entity recognizer (NER). Our work is the first of its kind for the Greek language in
such an extended form and one of the few that examines legal text. We apply grid search
on multiple neural network architectures and combination of hyper-parameters to maxi-
mize the efficiency of our approach. We show that, utilizing a big legal corpus we built
word/token-shape embeddings using Word2Vec, and finally achieve 86% accuracy on av-
erage in recognition of organizations, legal references, geographical landmarks, persons,
geo-political entities (GPEs) and public documents. The evaluation of our methodology is
based on the metrics of precision, recall, f 1 -score per entity type for each neural network.
Finally, we measure the ratio of correctly guessed links for the interlinking of RDF datasets
produced by our approach with well-known public datasets and how new knowledge can
be inferred indirectly by our approach from DBpedia, ELI (Europeal Legislation Identifier)
and GAG (Greek administrative geography) of Kallikratis.
Main subject category:
Technology - Computer science
Keywords:
Named Entity Recognition and Linking, Legislative Knowledge Representation, Entity Reference Representation, Linked Open Data, Deep Learning, Entity Generation
