Preface to Papers for GenChal

Generation Challenges 2012 (GenChal’12) was the sixth of the annual Generation Challenges umbrella events for shared-task evaluation activities in natural language generation. It followed five previous events: the Pilot Attribute Selection for Generating Referring Expressions (ASGRE) Challenge hosted by UCNLG+MT in Copenhagen, Denmark in 2007; Referring Expression Generation (REG) Challenges at INLG’08 in Ohio, US; Generation Challenges 2009 at ENLG’09 in Athens, Greece; Generation Challenges 2010 at INLG’10 in Trim, Ireland, and, most recently, Generation Challenges 2011 at ENLG’11 in Nancy, France. More information about all these events can be found via the links on the Generation Challenges homepage (http://www.nltg.brighton.ac.uk/research/genchal12). Like its predecessor events, GenChal’12 had a Future Task Proposals Track where researchers were invited to submit papers describing ideas for shared tasks to be run in the future. The responses to this call are the last three papers in the GenChal’12 part of the present volume. White’s paper on a Syntactic Paraphrase Reranking task is explicitly proposed as a further development on the outcomes of the Surface Realisation task, with the emphasis on methods to rank realisation alternatives. Banik et al.’s KBGen task proposal focuses on generation of text from structured knowledge sources, while Bouayad-Agha et al.’s proposal focuses on methods for content determination in NLG from semantic web data. GenChal’12 also included a progress report on the Surface Realisation shared task, currently organised by Anja Belz, Bernd Bohnet, Simon Mille, Leo Wanner and Michael White. The report, which can be found at the beginning of the GenChal’12 part of this volume, describes current work on the SR Task’s commonground input representations and plans for the next edition of the task, SR’13. The GenChal’12 session at INLG’12 included two further presentations, one an update report on the Question Generation Challenge organised by Vasile Rus, Arthur Graesser and Paul Piwek, which did not run this year. The other was a summary report on the second edition of the Helping Our Own task (HOO’12) organised by Robert Dale, whose full results session was this year hosted at BEA’12. HOO’12 was based on a much more focused task definition than its 2011 predecessor. While HOO’11 addressed the automatic correction of a very broad range of errors in scientific articles in English by non-native authors, this year’s edition was centred on errors related to the use of determiners and prepositions. This focus was partly motivated by the results of HOO’11. HOO’12 attracted 14 participating teams. Preparations are already underway for a seventh NLG shared-task evaluation event next year, Generation Challenges 2013, which may be organised as part of ENLG’13. Hopefully, this will include new tasks from among those proposed in