Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!neat.cs.toronto.edu!marina
From: marina@ai.toronto.edu (Marina Haloulos)
Newsgroups: ont.events
Subject: Dr. Sergei Nirenburg, Tuesday 24 October 1989: COLLOQUIUM
Message-ID: <89Oct17.132107edt.3623@neat.cs.toronto.edu>
Date: 17 Oct 89 17:21:28 GMT
Lines: 24


           Department of Computer Science, University of Toronto
         (SF = Sandford Fleming Building, 10 King's College Road)

       -------------------------------------------------------------

                                COLLOQUIUM
              SF1105, at 11:00 a.m., Tuesday 24 October 1989

                           Dr. Sergei Nirenburg
                        Carnegie Mellon University

              "Issues in Knowledge-Based Machine Translation"

Knowledge-based MT stresses the use artificial intelligence techniques.
KBMT systems are traditionally designed around a cascaded analyzer
containing a syntactic parser and a semantic/pragmatic interpreter for the
input language.   Until recently relatively less attention has been paid to
other modules of a KBMT-system, notably the generation component.  In my
talk I will describe the KBMT-89 system developed at CMU CMT.  After
describing the architecture of the system, I will concentrate on a) the
requirements of knowledge acquisition for KBMT; b) the ways of invoking
human assistance in cases of analyzer failure; and c) the peculiarities of
the generation module in a KBMT system.