Adapting polarised disambiguation to surface realisation

The aim of surface realisation is to produce a string from an input semantics. The task may be viewed as the inverse of parsing, and like parsing, it is made more difficult by the problem of lexical ambiguity in natural language. (Bonfante, Guillaume, and Perrier 2004) propose a polarisation technique for parsing that greatly reduces the effects of ambiguity. We show how polarisation can be adapted for surface realisation and how to address two complications which arise: multi and zero-literal semantic input.