n this talk we present an annotation schema for domain-independent
tutorial dialogue. An annotation schema is a set of labels which can be
assigned to utterances and which are an abstract description of the
function of those utterances in the context that they appear. An
annotation schema both categorises the utterances in a corpus and
provides a reasoning level for the dialogue model.

General purpose dialogue annotation schemas have been proposed, for
instance DAMSL (Allen and Core, 1997), which are intended to be adapted
to different kinds of data. Previous work on tutorial dialogue in
intelligent tutoring systems does not account for the general dialogue
level functions of utterances, only their pedagogical goals and effects.
For these reasons we have decided to extend the existing DAMSL taxonomy
with tutoring-relevant content in order to link the dialogue level and
pedagogical level functions of utterances. Our schema is prepared based
on our analysis of data from two Wizard-of-Oz experiments, from which we
grouped utterances into abstracted functional categories.

In the talk we present the categorisation of utterances and how we
organised these into the annotation schema. We show to what extent it is
compatible with the DAMSL schema, and comment on a particular utterance
type which is difficult to categorise, namely clarification requests.