[Humanist] 22.685 text-analysis speculations

Humanist Discussion Group willard.mccarty at mccarty.org.uk
Sat Apr 11 09:24:12 CEST 2009

                 Humanist Discussion Group, Vol. 22, No. 685.
         Centre for Computing in the Humanities, King's College London
                Submit to: humanist at lists.digitalhumanities.org

        Date: Fri, 10 Apr 2009 10:51:04 -0500
        From: Devin Griffiths <devin.griffiths at rutgers.edu>
        Subject: RE: [Humanist]  22.678 text-analysis speculations


If you accept that the majority of logical operators in a language like
English are expressed by the core set of function words, then a logical
stylistics of natural language would have to limit it self to some modified
set of those words (perhaps with some additional operators thrown in).  In
addition, the signature which is developed to describe the stylistics of a
given author would have to be composed of some specific, unique, and
logically coherent relation between how those terms are used by the author.
I guess this would look like a kind of syntactical analysis.  I would argue
that the combination of various statistical measures which underpins some
stylistics is not formal or logical, but rather qualitative.  This to me is
the strength of statistical approaches; they allow us to dig empirically
into the rich "alogical aspects in language" (which, for me, constitute the
inspired elements of a writer like Poe or Swift).

Devin Griffiths

-----Original Message-----
From: humanist-bounces at lists.digitalhumanities.org
[mailto:humanist-bounces at lists.digitalhumanities.org] On Behalf Of Humanist
Discussion Group
Sent: Thursday, April 09, 2009 12:19 AM
To: humanist at lists.digitalhumanities.org

                 Humanist Discussion Group, Vol. 22, No. 678.
         Centre for Computing in the Humanities, King's College London
                Submit to: humanist at lists.digitalhumanities.org

  [1]   From:    Willard McCarty <willard.mccarty at mccarty.org.uk>
        Subject: Turing, McCulloch & Pitts, Weaver and text-analysis

  [2]   From:    "Ian.Lancashire" <ian.lancashire at utoronto.ca>
        Subject: Re: [Humanist]  22.676 text-analysis in the news

        Date: Wed, 08 Apr 2009 13:51:22 +0100
        From: Willard McCarty <willard.mccarty at mccarty.org.uk>
        Subject: Turing, McCulloch & Pitts, Weaver and text-analysis

In the context of the possibilities for text-analysis, consider the

1. In a letter to the editor in the Times Literary Supplement for 4 May
1962, Karen Spärk Jones and T. R. McKinnon Wood pointed out that the
problems and arguments concerning machine translation can be generalised to
"any field which is concerned with handling language". In other words, what
has happened in MT should, at least in principle, concern us.

2. In his memorandum of 15 July 1949, "Translation", Warren Weaver wrote the
> A more general basis for hoping that a computer could be designed 
> which would cope with a useful part of the problem of translation is 
> to be found in a theorem which was proved in 1943 by McCulloch and 
> Pitts. This theorem states that a robot (or a computer) constructed 
> with regenerative loops of a certain formal character is capable of
deducing any legitimate conclusion from a finite set of premises.
> Now there are surely alogical elements in language (intuitive sense of 
> style, emotional content, etc.) so that again one must be pessimistic 
> about the problem of literary translation. But, insofar as written 
> language is an expression of logical character, this theorem assures one
that the problem is at least formally solvable.
In other words, MT links us, "insofar as written language is an expression
of logical character", back to the work on the physiology of thought pursued
by McCulloch and Pitts by means of networks of idealised neurons.

3. This work on the physiology of thought had a number of sources, but
certainly one of them was the Turing Machine, which linked the behaviour of
humans and machines. As Tara Abraham puts it, in "(Physio)logical
circuits: The intellectual origins of the McCulloch-Pitts neural networks",
Journal of the History of the Behavioral Sciences 38.1 (2002),
> Simply put, Turing was able to define the complicated process of 
> computation in "mechanical" terms, with the notion of a simple 
> algorithm so exhaustive, rigorous, and unambiguous that the executor 
> would need no "mathematical knowledge" to carry out its task. Turing 
> had linked the behavior of humans and machines: in both cases, 
> "computing numbers" involved a finite number of "states of mind" or 
> "configurations." These "states of mind," according to Turing, were
> irreducible: the "operations" performed by a logic machine or a human 
> computer can be split up into "simple operations" so elementary they 
> cannot be further divided.

So, we have the human person in one of his or her many roles (here doing
calculations) rendered as a machine, which in turn serves in the design of
an idealised scheme to explain how humans think. That scheme then
contributes to another for rendering the strictly machine-like aspects of a
specific expression of thinking in one language into another. And it lives
on to teach us about the limits of what we can expect from text-analysis.

Or does it -- teach us about limits, that is? Propositional statements are
one thing, but when stylometric techniques applied to a literary text
demonstrate consistency in the style, say, of Swift, or Poe, or whomever,
then is it fair to say that what is found is an "expression of logical
character" in that literary language? Is, then, that "logical character"
extractable in some way from the statistical patterning of language?

I have the sense that I'm going around in circles. Can anyone see around the
next bend?

Willard McCarty, Professor of Humanities Computing, King's College London,
staff.cch.kcl.ac.uk/~wmccarty/; Editor, Humanist,
Interdisciplinary Science Reviews, www.isr-journal.org.

        Date: Wed, 08 Apr 2009 12:32:49 -0400
        From: "Ian.Lancashire" <ian.lancashire at utoronto.ca>
        Subject: Re: [Humanist]  22.676 text-analysis in the news
        In-Reply-To: <20090408082145.B161F2F662 at woodward.joyent.us>

Thanks to Willard for this note. The Agatha Christie paper and poster 
were quietly presented at the 19th annual Rotman Institute conference 
(on cognition and aging) last month. You can find both at


Somehow Anne Kingston found out and wrote a fine article on them in 
Maclean's Magazine last week at


which in turn attracted the attention of the Guardian and UPI.

The paper argues that my Christie findings confirm Peter Garrard's 
evidence (in analysing Iris Murdoch's novels) that a sudden decline in 
vocabulary richness is an early marker of AD. I used Rob Watt's 
Concordance for vocabulary and indefinite nouns, and TACT (still usable 
in a virtual OS on my PC) to collect repeating phrases. Graeme and his 
students are now developing software to analyze for degraded syntactic 
features. We are fortunate to collaborate with Dr Regina Jokel, a 
post-doctoral researcher at Baycrest in speech pathology and AD.

Anyone interested in this subject should read Peter Garrard's recent 
paper, "Cognitive Archaeology: Uses, Methods, and Results," in Journal 
of Neurolinguistics 22.3 (May 2009): 250-65 ... a study of Harold 
Wilson's contributions to questions period. Garrard also uses Rob Watt's 
Concordance and John Burrows' delta.

Ian Lancashire
University of Toronto

List posts to: humanist at lists.digitalhumanities.org
List info and archives at at: http://digitalhumanities.org/humanist
Listmember interface at:
Subscribe at: http://www.digitalhumanities.org/humanist/membership_form.php

More information about the Humanist mailing list