[ont.events] UW AI Semi., Prof. Cohen on "An Incremental Model for Discourse Analysis".

ylfink@water.UUCP (ylfink) (09/30/86)

DEPARTMENT OF COMPUTER SCIENCE
UNIVERSITY OF WATERLOO
SEMINAR ACTIVITIES

ARTIFICIAL INTELLIGENCE SEMINAR

                    - Thursday, October 2, 1986.

Professor  Robin Cohen of this department will speak on
``An Incremental Model for Discourse Analysis''.

TIME:                1:30 PM

ROOM:              CHEM2 RM 079

ABSTRACT

We   propose   a   processing   model   for   analyzing
discourse with the following   properties:  (i)  deter-
mining    the   plan   of   the   speaker  (intentional
structure)   drives   the  process  (ii)  the  analysis
is  incremental   -   each new utterance is interpreted
and  a  representation  successively  built  (iii)  the
process   is   fine-grained   -  the  main decision  is
discerning  the  "purpose" of each utterance, its func-
tion  with   respect   to   the   overall   plan of the
speaker.   We then provide definitions  for  the  "pur-
pose"   of an utterance, leading to proposed operations
for the model.

The   model   is   intended   to   provide  a basis for
interpreting  goal- oriented   discourse  where  possi-
ble  underlying  plans  may  not  be available  before-
hand.    The  case where the top level goal is "get the
hearer  to  believe  p"  is  used for illustration, but
goals  of  action  and informing are also discussed. In
particular,  the model can be shown to provide  a  pos-
sible  basis  of  processing for the model of Grosz and
Sidner.