9.383 weights and measures

Humanist (mccarty@phoenix.Princeton.EDU)
Fri, 15 Dec 1995 08:56:03 -0500 (EST)

Humanist Discussion Group, Vol. 9, No. 383.
Center for Electronic Texts in the Humanities (Princeton/Rutgers)

[1] From: "C. M. Sperberg-McQueen" (41)
Subject: Re: 9.358 encoding, accents, interpretation

With regard to your query: it seems to me that not using weights is
effectively the equivalent of using weights of 0 and 1 (or 0 and 100,
however one delimits one's scale). Using numeric weights is a way to
get a bit more nuance into the representation. But one ought to
recognize even without weights that while all paragraphs may be equally
paragraphs, and all tables of contents equally tables of contents,
nevertheless it is unlikely that everything marked as, say, an echo of
another writer is equally an echo. Some are stronger and clearer than
others. If they are marked without weights (and thus implicitly with
the same weight), they are nevertheless not claimed to be identical.
Instead, they are simply not claimed to be distinguishable.

Similarly, if following the editors of the Dead Sea Scrolls I
distinguish four levels of certainty in a reading, or in a tagging, and
call them 95, 75, 50, 25, this should almost surely be read only as a
claim that I think the ones marked 25 are all less certain than the
others, not as a claim that they are all equally certain among

Even if I use a mechanical process which generates all the numbers
between 0 and 100 as values, I may not wish to claim that a value of 56
is substantially distinct from a value of, say, 55. (When my brother
took the SAT, his scores had, apparently, three significant figures:
756, 798, what-have-you. But 798 and 799 were not significantly
distinct from each other. By the time I came through, they were only
using two significant figures: all scores ended in zero. But even so,
I'm not sure that the difference between 790 and 780 is statistically
significant. I would expect this type of value to arise only out of a
mechanical process; parsers do weight alternative interpretations of a
text, and it might conceivably be of interetst to capture those weights,
without claiming more for them than they can deliver.