... directly by the fissionmodule, but we are currently experimenting with using the OpenCCG realiser’s language models tochoose them, using example-driven techniques.4 Surface RealisationSurface realisation ... four main phases in the demonstra-tor. First, the user specifies the shape of theirown bathroom, using a combination of speech in-put, pen-gesture recognition and handwriting recog-nition. Next, ... adaptive output, COMIC uses information fromthe dialogue history and the user model throughoutthe generation process, as in FLIGHTS (Moore etal., 2004); both systems build upon earlier work onadaptive...