Euskarazko gertaeren etiketatze automatikoa
- Haritz Salaberri
- Olatz Arregi
- Beñat Zapirain
- Iñaki Alegria (ed. lit.)
- Ainhoa Latatu (ed. lit.)
- Miren Josu Omaetxebarria (ed. lit.)
- Patxi Salaberri (ed. lit.)
Publisher: Udako Euskal Unibertsitatea, UEU = Universidad Vasca de Verano
ISBN: 978-84-8438-627-8, 978-84-8438-628-5
Year of publication: 2017
Volume Title: Giza Zientziak eta Artea
Tome: 1
Volume: 1
Pages: 22-29
Congress: Ikergazte. Nazioarteko Ikerketa Euskaraz (2. 2017. Iruñea)
Type: Conference paper
Abstract
In this paper we present the results obtained by bEVENT, the first system developed for automaticallyannotating events in Basque that follows the ISO-TimeML standard. A two-step procedure is performedin order to conduct annotation: firstly, event-extents are identified, and secondly, values are given tothe linguistic attributes specified for these extents. The Euskal-TimeBank corpus has been used to trainand evaluate bEVENT and the system results have been calculated using train-test evaluation. bEVENTimplements a machine learning based approach and the precision, recall and F1 measures that correspondto the extraction of events are the following: 83.92, 72.76 and 77.94.