[ont.events] AI SEMINAR

gh@utai.UUCP (Graeme Hirst) (09/06/84)

	  Conjunctions in natural language understanding

			    Vicky Snarr
		  Department of Computer Science
		      University of Toronto


	 Tuesday, 11 September, 3pm, Sandford Fleming Bldg 1105

Although the conjunction "and" appears to have a simple function in the
English language, it has proved to be a stumbling block for both theoretical
and computational linguists.

One of the theoretical problems of conjunction is to determine what governs the
acceptability of a structure in which two elements are connected by "and".
The corresponding computational problem is, given this knowledge, to
incorporate it into an efficient parser for English.

I propose a solution to the theoretical problem which is in the form of two
general constraints -- a syntactic constraint and a semantic one; and then
show how these constraints were incorporated into a ``strictly deterministic''
parser for English.
-- 
\\\\   Graeme Hirst    University of Toronto	Computer Science Department
////   utcsrgv!utai!gh	/  gh.toronto@csnet-relay  /  416-978-8747

eas@utcsrgv.UUCP (Ann Struthers) (01/07/85)

	        	ARTIFICIAL INTELLIGENCE SEMINAR

		           Tuesday, January 8, 1985

	            3:00 P.M. Sandford Fleming Building 1105





			      Dr. Jack Edwards

                     Longwoods Research Group Ltd., Toronto

		 "Evaluating the Intelligence in Dialogue Systems"


The talk presents a set of criteria for evaluating intelligence in AI
computer systems which engage in dialogue with human users. It is divided
in five sections as follows:

(1) A definition of dialogue, characterization of a dialogue system, and
    brief discussion of the concept of intelligence in dialogue systems.

(2) An elaboration of eight aspects of dialogue and dialogue systems.

(3) A discussion of four evaluative dimensions and an illustration of how
    they may be combined with the aspects above to generate criteria for
    designing and evaluating dialogue systems.

(4) Applications of some of these criteria to three example dialogue 
    systems: ELIZA, GUS, and GUIDON.

(5) A discussion of the current state of evaluation and our evaluative
    methodology, and directions for future work.