Xml format guidelines for the tuna corpus

Abstract

This document forms part of the 2008 distribution of the TUNA Corpus, Version 1.0. This is the first public release of the complete TUNA Corpus of Referring Expressions. A subset of the corpus was used in the first Shared Task and Evaluation Challenge for NLG, the Attribute Selection for the Generation of Referring Expressions Challenge (ASGRE), co-located with the Workshop on Using Corpora in NLG. A subset is also being used for the second edition of the Challenge (the REG Challenge 2008), to be held in Ohio in June 2008, co-located with the International Conference on NLG. Both of these previous releases consist exclusively of the singular referring expressions in the TUNA corpus; moreover, the annotation for both ASGRE 2007 and REG 2008 has a different format which was specifically designed for the tasks involved. This release contains the final version of the TUNA annotation, and includes the full corpus, that is, both singular and plural descriptions.