18.629 David Hoover at CHUG, Brown

               Humanist Discussion Group, Vol. 18, No. 629.
       Centre for Computing in the Humanities, King's College London
                   www.kcl.ac.uk/humanities/cch/humanist/
                        www.princeton.edu/humanist/
                     Submit to: humanist_at_princeton.edu

         Date: Fri, 11 Mar 2005 07:13:20 +0000
         From: "Mylonas, Elli" <Elli_Mylonas_at_BROWN.EDU>
         Subject: David Hoover Wed, Mar 16 4pm

The Computers and the Humanities Users Group [Brown University] presents

What are Electronic Texts Good For?

David L. Hoover, Department of English, NYU

4pm, Wednesday, March 16
STG Conference Room
Ground Floor, Grad Center, Tower E

The Brown Women Writers Project and many other archives, text
collections, and web sites clearly demonstrate one of the most
valuable uses of electronic texts, but I want to concentrate on some
techniques and approaches that make more active use of the
manipulability of electronic texts. Text analysis, corpus analysis,
authorship attribution, and statistical approaches to style are not
new, but they have not yet had a very great impact on mainstream
humanities disciplines, largely because of recent trends in literary
theory and linguistics. The continued rapid increase in the
availability of electronic literary texts, and the recent
availability of huge natural language corpora provide an opportunity
for reinvigorating and rehabilitating these and other techniques that
take full advantage of the characteristics of electronic texts. It
seems time to return to the text, specifically the electronic,
computable text, to see what corpora, text-analysis, statistical
stylistics, and authorship attribution can reveal about meaning and
style.

Over the past twenty years, I've been using electronic texts in an
expanding variety of ways, from computing Anglo-Saxon metrical
patterns to applying authorship attribution tools to the analysis of
style variation within a literary text. My recent work combines many
of these techniques. For example, doing text-analysis on large
electronic corpora provides norms against which to evaluate the style
of a single text or author. Using highly automated text-analysis
tools and multivariate authorhsip attribution methods on specially
created corpora and has lead to innovations in authorship attribution
and to new questions about its theoretical bases. And applying these
same tecnhiques to variations within the style of a single author or
a single text have suggested modifications in authorship attribution
tools that help to clarify how these tools can both correctly
attribute all of an author's texts and at the same time clearly
differentiate the styles of single texts, sections of texts,
characters, or narrators within that author's texts. Finally, the
easy and infinite revisability of electronic texts can be put to a
constructive, if very simple use: altering a literary text is often
the most effective way of understanding it.
Received on Fri Mar 11 2005 - 02:22:36 EST