Thesis work:

Information Enriched Constituents in Dialogue

Information enriched constituents is a term I have chosen for a phenomenon in human-human and human-machine dialogue, a phenomenon which has much in common with what has been studied within approaches to ellipsis, but which also, and importantly, embrace differences. A crucial contribution to the differences is that I take my point of origin in spontaneously spoken dialogue (information enriched constituents in bold face):

A: at what time does it get to Toronto? *
B: ah three thirty five p.m.

U: Is this avoiding the banana tree? **
W: Avoiding... No, avoiding gorillas.

whereas studies on ellipsis often concern constructed one-sentence examples:

Edith ordered asparagus from the waiter, and bread too

Approaches to fragments, short answers and non-sentential utterances have all looked at spoken dialogue, but through the term information enrichment I want to emphasise that these examples are not reduced in any sense. Rather, as you find when studying dialogue, in the context in which these utterances are produced they contain just the right amount of information. Exploring the full complexity of spontaneous spoken dialogue, I also try to investigate the impact that the presence of a potentially large dynamic context and several speakers with utterances distributed across turns (which gives overlap, interruptions, etc) have on information enrichment.

My analysis of dialogue corpora uses theories of Information Structure, and my thesis work also includes formalisation and implementation using the Information State approach to dialogue.

* Dialogue excerpt from the Amex travel corpus
** Dialogue excerpt from the HCRC map task corpus