Date: Thu, 3 May 90 18:23 EDT
From:: Reid Rubsamen
Subject: GSL - GSL - GSL - GSL
To: *mac@MC.lcs.mit.edu

	      7.888: SEMINAR ON GASTRIC FOOD PROCESSING

          Sponsor: Ellen Hildreth and Shimon Ullman

4 May 1990
Noon
NE43 - 8th Floor Playroom


			     Julia Child

		      School of Culinary Science
		       Carved Melon University


	  Graduate Ingestion Without Taste Feature Detection


			       ABSTRACT

   One of the key failings  of current systems  for graduate ingestion
is the inability to extract reliable and useful  features from the raw
foodstuffs.  The problem is that the available techniques -- cold hors
d'oevres and  basic  food  grouping  --  depend  on the  assumption of
``gustatory coherence,'' i.e. the assumption  that the taste intensity
or  color is uniform  across a food category and  differs only at food
group  boundaries.  Real  flavors do  not  behave this way,  so  these
methods fail.

   In this talk, I will present work based  on a new approach  that we
call  ``menu coherence,''  in which  taste  features are based on  the
conformance of seasonings   and spices with  specific hypotheses about
the geographic locale that gave rise to a specific main course.  There
are two aspects of the local scene to be analyzed: ethnic authenticity
and piquantness properties.  I will begin with a review of our work in
ethnic authenticity analysis, particularly   in the analysis of  color
and  food combinations.  Next,   I  will  show our brand-new  work  in
analyzing piquantness properties, in   which we introduce the  n-alarm
spectrogram as a tool for spice potency and  "hotness"  analysis.  The
n-alarm spectrogram is a  space/frequency plot well known in szechuan,
mexican and creole cuisine; we  have  now shown  that it also captures
many important  relationships  for palate   analysis and  for  active,
purposeful menu choice by computer graduate students.

   The key thrust of this research effort is the direct application of
ethnic  and spice  properties to raw   foodstuffs, without traditional
``cooking problems'' getting in the  way.  Where most people begin  by
finding spices and then doing the n-alarm analysis, we propose instead
to begin with  n-alarm  analysis and use  the resulting  hypotheses to
indicate the boundaries of  ``allowed seasonings''.  I believe that it
will  one day lead  to low-level catering systems  that can apply many
physical laws in combination to  yield highly competent  and ``tasty''
low-level student feedings.

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

Hosts:
Humor: Tom Russ
Honcho: Reid Rubsamen