... to much priorwork in NLG, e.g. (Walker et al., 2004; Dembergand Moore, 2006; Polifroni and Walker, 2008).In this problem, there there are many decisionsavailable for exploration. For instance, ... realizer. Both corpora contain data from“overhearer” experiments targeted to InformationPresentation in dialogues in the restaurant domain.While we are ultimately interested in how hearersengaged ... COM-PARE. The strategies realise varying numbers ofattributes according to different “conciseness” val-ues: concise (1 or 2 attributes), average (3 or 4), and verbose (4,5, or 6). Overhearersrate...