ABSTRACT
Algorithms inspired by the search processes of natural evolution have generated several robust search methods. These so-called evolutionary algorithms have been applied to a wide range of problems. This chapter discusses their application to a problem in natural language dialogue processing. It examines the possibility of using evolutionary algorithms to automatically carry out the tuning of large scale, object based, linguistic interactor, translator and analyser (LOLITA) dialogue module to particular situations. The parameters that control the plan boxes contained within the dialogue module of LOLITA have been fine tuned by hand to give a particular behaviour. A method by which the usefulness of a tuning algorithm could be determined would involve first of all tuning the parameters to give a particular series of utterances. Then starting with a random set of parameters, use the tuning algorithm to try and optimise the parameters so that the same series of utterances is generated.
