Humanist Discussion Group, Vol. 14, No. 262.
Centre for Computing in the Humanities, King's College London
<http://www.princeton.edu/~mccarty/humanist/>
<http://www.kcl.ac.uk/humanities/cch/humanist/>
[1] From: Wilhelm Ott <zrlot01@zdv.uni-tuebingen.de> (71)
Subject: Re: 14.0258 methodological primitives?
[2] From: lachance@chass.utoronto.ca (Francois Lachance) (68)
Subject: Re: 14.0258 methodological primitives?
[3] From: "Osher Doctorow" <osher@ix.netcom.com> (31)
Subject: Re: 14.0258 methodological primitives?
--[1]------------------------------------------------------------------
Date: Fri, 22 Sep 2000 09:41:49 +0100
From: Wilhelm Ott <zrlot01@zdv.uni-tuebingen.de>
Subject: Re: 14.0258 methodological primitives?
Dear Willard (and Humanist participants),
of course, the idea of "methodological primitives" is useful.
You may be aware that, with TUSTEP, we have tried (for more than
30 years now) to take this idea as a basis for a software which
is not only flexible but also guarantees that the user can really
take over the responsibility for the results - since he is able
(and has to do it) to define in every detail the "basic" operations
to be carried out by the computer. For a short description of the
concept, you are invited to consult
http://www.uni-tuebingen.de/zdv/tustep/tustep_ox92.html
(a paper given 1992 at the ALLC-ACH conference in Oxford).
I disagree however regarding two of the three examples for the
low-level end of "methodological primitives":
1. "lemmatizing the word forms of an inflected language":
Let me argue from my own experience:
It took us (Father Bonifatius Fischer and his team and myself)
more than two years to lemmatize the latin inflected word forms
occuring in the text of the Vulgate, though we had at our disposal
the "Lexicon Electronicum Latinum" which Father Busa had compiled
for his Index Thomisticus and though had written a software which
automatically added, with the help of this Lexicon, the lemma to
each inflected form - a procedure which produced an average of
2,5 proposals for each inflected form! We sorted the material back
into the text sequence, in order to have the context available for
the "manual" decision which of the proposed lemmatizations was the
correct one.
Expecting that a piece of software at the level of "methodological
primitives" could decide if "facies" is, at a given location, a
verb form or a substantive is far from reality in near future - though
30 years have gone since we lemmatized the Vulgate; also today,
even "higher level" software like SYSTRAN (which is e.g. offered for
translating web pages found by Altavista) fails to take the context
into account (and therefore translates the term "content provider"
into the German "der zufriedene Anbieter").
2. "Alphanumeric sorting":
Yes, if you have in mind a routine which rearranges a file according
to the sequence of bits at certain locations of every record,
irrespective if they are to be interpreted as numbers or as digits
and letters, I agree. But this is far from what I would call
"alphanumeric ordering". In TUSTEP, we did not even try to fully
automize the steps necessary to arrive at a required ordering.
Not only are there different rules for almost every language
(the EC has only recently established a norm for multilingual
ordering in EU documents, see http://www.stri.is/TC304/EOR), but
sometimes (as for German) even concurrent rules, depending on the
type of index you generate (e.g lexicon on the one hand, telephone
directory or library catalogue at the other). And for scholarly
purposes, also historical (or even unregulated) orthographies must
be covered. So, what is needed is more elementary than "alphanumeric
ordering": it is a routine which allows to construct arbitrarily
complex and multiple "ordering keys" from a given text or part of
text, which then allows to obtain the required order with the help
of a "mechanical" ordering procedure which has to do nothing than to
rearrange the records according to the ascending value of the bit
string contained in the ordering key. The scholar must however have
the possibility to define in every detail the rules for constructing
the ordering keys.
Therefore, also "alphanumeric sorting" is, for these purposes,
too complex a procedure to be put on the lower level of
methodological primitives.
A short summary of what we regard as "basic operations for text data
processing" which should be at the disposal of a textual scholar can be
found at http://www.uni-tuebingen.de/zdv/tustep/tdv_eng.html#b
Yours,
Wilhelm Ott
---------------------------------------------------------------------
Prof. Dr. Wilhelm Ott phone: +49-7071-2970210
Universitaet Tuebingen fax: +49-7071-295912
Zentrum fuer Datenverarbeitung e-mail: ott@zdv.uni-tuebingen.de
Waechterstrasse 76 http://www.uni-tuebingen.de/zdv/
D-72074 Tuebingen
--[2]------------------------------------------------------------------
Date: Fri, 22 Sep 2000 09:43:10 +0100
From: lachance@chass.utoronto.ca (Francois Lachance)
Subject: Re: 14.0258 methodological primitives?
Willard,
I just had to pour myself a glass of stout -- one of the most primitive of
scholarly processes -- to begin to reread and unravel that very intriguing
quasi-grant speak paragraph...
> I'll call these mechanical operations "methodological primitives" and
why not call them "mechanical operations"?
what is gained/lost in calling them "methodological primitives"?
As I sip on my stout, I wonder if this not close to reifying a process.
Not a bad thing in itself when looks at a systems analysis of a complex
set of exchanges and communications across multiple nodes and
processing centres. Flowcharts of where something is produced, housed,
transfered, transformed and disposed, work for me. Stout is brewed, bottle
and shippped. Yet I wouldn't call any of these either "mechancial
operations" or "methodological primitives". Now the products and tools of
your Humanities Computing may be an entirely different type of entity than
a good bottle of stout.
> define the type as "an algorithmically specifiable transformation of
data
> that forms a recognisable component of multiple scholarly processes".
are not all transformation algorithmically specifiable?
(includinging random event generators such a pausing to sip on one's
breverage)?
recognisable to whom? to all the participants across all the disciplines
or to the "translators" at cross-disciplinary thresholds? Note, I have
indeed moved from your "scholarly processes" to "disciplines" thus
mimicking your initiall move from mechanical to methodological and from
process to primitive ... (I'm sure some ale drinkers could contribute some
fine words on the comparative etymology of "method" and "mecanic" and
there is scarely an imbiber of single malts who would not want to venture
and opinion as to the word roots that tangle "primitive" and "process" in
a heady concotion given several rounds of fine peaty Islay scotch.)
As an
> idea I would suppose the type to be a useful analytic tool with which to
> resolve what humanists do into a loosely bounded set of interoperable
> software components which could be assembled in whatever order by a scholar
> in order to aid his or her research.
I want to do a very primitive reader process and reparse this last bit so
I just might switch to grappa.
I would heartily raise a glass of fire-water to the aid to research any
thought about the use value of ideas (as long as it came from cows raised
on organic feed and free to range).
Actually I like the paragraph, I just would from a general systems
perspective change one word. The "into" should be changed to "with".
And I would add, draining my glass, that one of those components include
communications software to place the "with" as the hinge between a scholar
and the data but also between scholars. As I place my glass on the table
before me, I insist, comtemplating the empty glass, that behind much of
the lingo in that artfully crafted paragraph is a primitive all right, one
called the "means of production".
[this is where they haul me off before I rant and sputter and start
quoting long passages from a book with a long history and attempt
to reconcile that book's section on "the two fundatmental forms of
manufacture -- heterogenous and organic" with the elegant and freighted
word "interoperable"]
> My question is, how strong is the idea of methodological primitives? How
> useful? Is this a direction in which we should go?
If we were to tour Scotland, I could do the islands and you could do the
highlands, and we could meet in Thailand to compare notes. The year after
I could follow your route and yours mine and we could meet in Helsinki to
swap stories. OR we could swap stories while on the road. If we all went
in the same direction ... there would be no strength to any idea.
Promise to send me a postcard from where ever you plant the hops, harvest
the grapes or milk the cows.
f.
--
Francois Lachance, Scholar-at-large
http://www.chass.utoronto.ca/~lachance
Member of the Evelyn Letters Project
http://www.chass.utoronto.ca/~dchamber/evelyn/evtoc.htm
--[3]------------------------------------------------------------------
Date: Fri, 22 Sep 2000 09:44:05 +0100
From: "Osher Doctorow" <osher@ix.netcom.com>
Subject: Re: 14.0258 methodological primitives?
From: Osher Doctorow, osher@ix.netcom.com, Thurs. Sept. 21, 2000, 9:24PM
Professor McCarty, as usual your questions are ingenious. I do not know the
answer. I have been wondering something similar, although somewhat across
disciplines. Readers might be interested in one or two of my last few
contributions to evolutionary-computing@mailbase.ac.uk . Whatever the
answer, and I certainly will look into it, we need to remember (I think) to
keep our goals high. In fact, I am beginning to think that people's goals
should always be higher than they can even imagine - or maybe as high,
depending on the way one looks at it. I know that produces some strange
side effects, as one might call them, including the Man From La Mancha, but
I am beginning to change my view even of the Man From La Mancha. I think
that Cervantes could have had an intuition about the universe, whose
mysteries, whose questions and answers, go far beyond what is commonly
believed and held to be true, from prehistory and pyramids in Ancient Egypt
to the Magna Carta and Limelight and modern times on earth, not to mention
elsewhere. Occasionally some angry blamers get hold of goals beyond their
reasoning, like the Nazis who made the Holocaust and the bombing of Britain.
But, as I indicated on evolutionary-computing, there have to be questions
about viruses' viruses (viruses which destroy other viruses), things between
life and viruses, things between death and viruses, things between the
inorganic and viruses, and so on, before we can probably conquer cancer and
AIDS and even polio and tuberculosis whose symptoms we can eliminate but
whose causes we know nothing about. Let us have a type and a
methodological primitive or many for humanist computing and
interdisciplinary computing, by all means, but let it be types for all
seasons and all imaginings or beyond. Let us try not to merely computerize
machines to recreate human beings but to begin with what we know about
humanist computing and viruses and life and science and humanities and move
onward from there. A Type For All Seasons. A Type for the Viruses' Virus,
the cancer's cancer. Maybe even a type for the Old Man/Woman/Being, as
Einstein thought.
Osher Doctorow
This archive was generated by hypermail 2b30 : 09/22/00 EDT