Complementing RRL for dialogue summarisation
Norton Trevisan Roman, Ariadne Maria Brito Rizzoni Carvalho · 2010
Abstract. This paper describes the use of neca’s rrl – Rich Repre-sentation Language – in the aid of automatic dialogue summarisation. We start by defining a set of summary acts, and then present a mapping between these acts, taken from human generated summaries, and their corresponding speech acts from the source dialogues. This mapping can then be used by a summary planner to identify portions in the dialogue with higher probability to support some argument it is about to intro-duce in the summary. Depending on the application, it could also be used to gather, in the source dialogue, specific information to fulfill summary templates. In following this procedure, we hope researchers will be able to build more human-like summaries, by linking corpora summaries to their source dialogues. 1