Pr.A.T.I.D: a coding scheme for pragmatic annotation of dialogues.

Our purpose is to propose and discuss the latest version of an integrated method for dialogue analysis, annotation and evaluation., using a set of different pragmatic parameters. The annotation scheme Pr.A.Ti.D was built up on task-oriented dialogues. Dialogues are part of the CLIPS corpus of spoken Italian, which consists of spoken material stratified as regard as the diatopic variation. A description of the multilevel annotation scheme is provided, discussing some problems of its design and formalisation in a DTD for Xml mark-up. A further goal was to extend the use of Pr.A.Ti.D to other typologies of task-oriented texts and to verify the necessity and the amount of possible changes to the scheme, in order to make it more general and less oriented to specific purposes: a test on map task dialogues and consequent modifications of the scheme are presented. The application of the scheme allowed us to extract pragmatic indexes, typical of each kind of text types, and to perform both a qualitative and quantitative analysis of texts. Finally, in a linguistic perspective, a comparative analyses of conversational and communicative styles in dialogues performed by speakers belonging to different linguistic cultures and areas is proposed