Visualising discourse coherence in non-linear document generation

In order to produce coherent linear documents, Natural Language Generation (NLG) systems have traditionally exploited the structuring role played by textual discourse markers such as relational and referential phrases. However, these coherence markers of the traditional notion of text do not work as efficiently in non-linear informative documents, where ambiguity and indeterminacy are usually undesirable. A new set of devices, not only textual but graphic, is therefore needed together with formation rules to govern their usage, supported by sound theoretical frameworks. While, in linear documents, graphic devices such as layout and formatting play a complementary role to textual devices in the expression of discourse coherence, we contend that in non-linear documents their role is much more important and needs to be adequately studied. In this paper, we present our work in progress, which explores new possibilities for achieving coherence in the automatic generation of hypertext documents.