Abstract
This paper surveys the current state of the art in Natural LanguageGeneration (NLG), defined as the task of generating text or speech fromnon-linguistic input. A survey of NLG is timely in view of the changes that thefield has undergone over the past decade or so, especially in relation to new(usually data-driven) methods, as well as new applications of NLG technology.This survey therefore aims to (a) give an up-to-date synthesis of research onthe core tasks in NLG and the architectures adopted in which such tasks areorganised; (b) highlight a number of relatively recent research topics thathave arisen partly as a result of growing synergies between NLG and other areasof artificial intelligence; (c) draw attention to the challenges in NLGevaluation, relating them to similar challenges faced in other areas of NaturalLanguage Processing, with an emphasis on different evaluation methods and therelationships between them.