CRIPCO corpus

The Cross-document Italian People Coreference (CRIPCO) corpus created at FBK is a subset of the news stories published by the local newspaper "L'Adige" from 1999 to 2006 annotated with information about person cross-document coreference. CRIPCO is composed of 43,328 documents and it was used in the NEws People Search (NePS) task (http://www.evalita.it/2011/tasks/NePS) at EVALITA 2011 (http://www.evalita.it/2011).To this purpose, the corpus was split into a development set and a test set, balanced according to the ambiguity of the Group Names and the fame of the related persons. The CRIPCO release maintains the subdivision into NePS development and test set, in order to allow comparisons with the results obtained in the evaluation campaign.<p>