Re: DH Idiocy 1


Herman Rubin wrote:
In article <1149814188.258097.148690@xxxxxxxxxxxxxxxxxxxxxxxxxxxx>,
Shlomo Argamon <argamon@xxxxxxxxxxx> wrote:

Dan Kimmel wrote:
"Lisa" <lisa@xxxxxxxxxxxx> wrote in message
news:1149802804.202741.180820@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

...................

Indeed - and "deriving meaning from the text" using the tools of DH is
precisely what R'Dr Mordechai Breuer does. On the other hand, there is
a great difference between using a new critical tool for litcrit or
hermeneutics, and using it to make declarations about matters of
empirical fact. I think that few, if any, of the "anti-DH" folks in
this thread would object if the only claim that DH-ers were making was
that "DH methods lend some insight into the meaning of the text". But
the claim is that DH work proves (or demonstrates, or shows) that the
Torah was redacted from some set of previously existing documents
written by people, perhaps divinely inspired, perhaps based on old
legends (or some combination of the above). The fact is that it does
nothing of the kind.

As it happens, today I attended the first day of a 2-day workshop on
text-based authorship attribution, and I spent the day discussing the
state of the art in this field with top linguistics, literary scholars,
statisticians, and computer scientists working in this area. The clear
and entirely undisputed consensus view was that nothing even
approaching an empirically validated theory of text generation and
construction exists, which could be used to derive methods for
comparing the authorship properties of two texts. That is, there is NO
WAY that DH-type methods (or any other methods currently known) could
possible conclude, based on internal textual evidence, that the Torah
was edited together from multiple source documents. This is ENTIRELY
separate from the issue of Divine authorship, note.

-Shlomo-

What is being stated is that there are rarely TOOLS for doing
this. However, Mosteller and Wallace did find a way of deducing
the authorship of several Federalist Papers whose authors were
not known, and who had very similar styles of writing.

Mosteller and Wallace's work was seminal and immense, however, FAR from
unassailable. The interpretation you place on it here, while not
unreasonable, rests upon a number of entirely unproven assumptions
about the nature of authorship and language.

We still do not have reasonable tools for understanding speech
or handwriting.

Or word choice - I work in this field, it is NOT simple! And the Torah
is a very small corpus, and there are NO ground-truthed comparison
texts of even similar language and genre (let alone the same).

We do not have tools for recognizing the type
of diseases on has, or coming up with reasonable treatments in
many cases. But individuals can do these things.

And we trust them because they have been successful in many cases. Can
you name even ten cases of documents in ancient semitic languages
whose authorship has been determined by DH methods, such that the
determinations were found post facto to be correct, according to
unassailable evidentiary standards? And that would just be for a
start!

The major problem is not that of finding differences, but the
differences in pieces of text made by one author, and the
differences between that and other authors. Tell me, did any
of the statisticians discuss using the Kolmogorov-Smirnov test?
I have tried it on some samples, and it works too well, rejecting
the hypothesis that they had the same author even if they did.

All you are able to tell by just applying a statistical test is "these
look different" or "these do not look different" - what that difference
MEANS, in terms of authorship, is not a statistical question AT ALL.
It is a linguistic (& sociological, historical, psychological)
question.

I have looked at the problem, and I would like to know what was
tried. I suspect that it was too simplistic.

I agree with this, but likely not for the reasons that you do. More
sophisticated statistics are NOT what is needed.

Anyhow, people can distinguish authors even when machines cannot.

Indeed, in some cases they can, with sufficient expertise in the field
of a particular type of text, and sufficient extrinsic (to the text)
reasons to accept the textual assumptions that they work from.
However, it must first be established that they can do so based on
clear evidentiary standards. This standard is not (to my knowledge)
met for DH work.

-Shlomo-

.

Relevant Pages

  • The ethics of authorship and the systematic omission of statisticians
    ... Authors, Ghosts, Damned Lies, and Statisticians ... Elizabeth Wager is a publications consultant, Princes Risborough, ... issued guidance on who qualifies for authorship, ... A New Study on Ghost Authors ...
    (sci.stat.edu)
  • Re: DH Idiocy 1
    ... comparing the authorship properties of two texts. ... of the statisticians discuss using the Kolmogorov-Smirnov test? ... are those of the Statistics Department or of Purdue University. ... Herman Rubin, Department of Statistics, Purdue University ...
    (soc.culture.jewish.moderated)