File /Humanist.vol22.txt, message 286


Date:         Sun, 26 Oct 2008 10:00:52 +0000
From: Humanist Discussion Group <willard.mccarty-AT-MCCARTY.ORG.UK>
Subject: 22.296 text-analysis in the news
To: humanist-AT-Princeton.EDU


               Humanist Discussion Group, Vol. 22, No. 296.
       Centre for Computing in the Humanities, King's College London
                        www.princeton.edu/humanist/
                     Submit to: humanist-AT-princeton.edu



         Date: Sun, 26 Oct 2008 09:40:49 +0000
         From: Humanist Discussion Group <willard.mccarty-AT-mccarty.org.uk>
         Subject: Cohmetrix/LIWC

Dear Willard,

I saw both demonstrated at the July IGEL2008 Conference and had a hands
on workshop in Coh-Metrix. I haven't tested both on substantials corpora
though.

I was somewhat impressed by the sheer functional possibilities of
Coh-Metrix. Given the right preparation, Coh-Metrix seems to be able to
put your texts through about any corpus linguistic algorithm out there.
And it seems preparation wasn't very cumbersum. But I wasn't
particularly impressed by scalability and robustness. At least I was
able to crash the system by a few mean but simple hacker tricks. These
border to simple vandalism, but if a system isn't resistant to such
simple 'fool-proof' testing, what does this say for the internal
stability of the algorithms? But I'd give it the benefit of the doubt.

LIWC I only saw demonstrated. It seemed to be able to sort of reliable
model the word usage behavior indicative of affective speech (assuming I
understood correctly). Cindy Chung sure made for a splendid
presentation, with lots of interesting and hilarious conclusions - like:
you're bound to be neurotic if you're an over user of the ellipsis
(three dots)...

Kind regard,
Joris van Zundert


   

Humanist Main Page

 

Display software: ArchTracker © Malgosia Askanas, 2000-2005