Computer Philology: 'Wissenschaft' or 'Hilfswissenschaft'?

Computer Philology: 'Wissenschaft' or
'Hilfswissenschaft'?

Jan
Christoph
Meister

University of Hamburg
jan-c-meister@uni-hamburg.de

2002

University of Tübingen

Tübingen

ALLC/ACH 2002

editor

Harald
Fuchs

encoder

Sara
A.
Schmidt

This paper discusses the problem of the theoretical-methodological status of
Humanities Computing (HC). More particularly it focuses on Computer Philology
(CP), i.e. the computational concepts and techniques dedicated to aspects of a
philological analysis of literary texts. The question guiding my deliberations
is whether CP is a proper ‘Wissenschaft’ in the sense of an emerging
fully-fledged discipline, or rather an ancillary discipline, a
‘Hilfswissenschaft’ similar to Statistics or Library Science?
One can safely say that the practice of HC has begun to establish itself as an
integral part of research and teaching in various disciplines in the Arts and
Humanities. But one must also concede that a consensus regarding the
methodological-theoretical calibre of our undertakings seems to have evaded us
thus far. As a result our own debate seems to retreat more and more onto the
safe ground of pragmatism and technology, or in other words: who cares about the
methodological and philosophical appraisal of HC - as long as it works?
I believe that this is a dangerousely narrow focus - if nothing else, it is at
least strategically detrimental to the aim of furthering the
institutionalization of Humanities Computing (HC) at universities. The fact that
Computer Philology (CP) concerns a more narrowly defined and homogenous array of
disciplines than HC makes matters even worse. Any attempt to introduce CP vis à
vis the well defined philological disciplines only accentuates the problem of
the as yet vague methodological status of this ‘newcomer’.
For example, at Hamburg University we recently (October 2001) managed to formally
institute a joint ‘Arbeitsgruppe Computerphilologie’ in the Faculties of
Computer Science and the Faculty of Languages. We are clear as to our brief:
introduce modularized course components in CP into the language and literature
curricula. Pragmatic considerations, such as, the aim to equip students with new
skills, as well as the current ‘sexyness’ of computer technology work in our
favor. But we are nevertheless divided as to the more profound philosophical and
methodological arguments on which to ‘sell’ this initiative to colleagues and
students. And or differences do indeed boil down to the very question: Is CP a
proper ‘Wissenschaft’ - or is it rather a ‘Hilfswissenschaft’?
Proponents of the latter position hold that CP cannot be a true discipline as it
is neither defined by an exclusive subject matter, or a particular perspective
onto such matter, nor ─ and this would seem the more difficult verdict to
counter - has it as yet produced a new theory of literature, not to mention a
meta-theory reflecting its own axioms and practices. This, for example, is the
opinion of my colleagueWalther von Hahn who approaches the question from the
perspective of the Linguist and Computer Scientist. Von Hahn illustrates his
argument with the following process model:
Figure 1:

The positioning of the CP-block in the bottom left rectangle, and in particular
the orientation of the unilinear arrows in this diagramm clearly demonstrate an
hierarchical organisation: CP is primarily conceived of by von Hahn as a
research practice that is governed by the preceeding formulation of philological
desiderata. These desiderata are identified in the course of the inspection of
textual data (1) which leads to a formulation of specific research questions and
interests (2). Only then can the selection of particular CP-tools (= practices)
take place (3). These are now applied to the data (4) and generate a new
representation of the textual data which in turn are passed back to the
philologist (5) for their eventual interpretation. This interpretation then
feeds back into philological theory and methodology.
I would now like to juxtapose the underlying hypothesis - e.g., CP as a
‘Hilfswissenschaft’ - with conclusions drawn from my own research into computer
based analysis of narrated action, a 7-year project concluded in early 2001
(components of this project were presented at the ALLC conferences Paris 1994 /
Virginia 1999.)
I initially set out to analyse the structural features of narrated action, with a
view to formulating a model of ‘minimal action structures’, that is, the least
complex yet logically coherent and context-idenpendent sequence of narrated
events. I had found that there was no hard and fast theory on what actually
constitutes a coherent piece of narrated action ─ the problem was either being
discussed in terms of the conditions for achieving narrative ‘closure’, or in an
altogether impressionistic and intuitive way. The only models and partial
definitions useful to me were those developed in Formalism and Structuralism,
schools of thought to whom the graphic respresentation of narrated action in the
form of tree-and-nodes diagrams is fairly common (and owed to generative models
imported from structural linguistics). Then I happened to stumble upon an
article by Alain Colmerauer which gave a brief description of the AI-language
PROLOG and illustrated it by way of a tree diagram. The visual analogy between
an action- and a PROLOG-tree pointed me to the conceptual analogy, and the
possibility to use a computer to model a minimal coherent action in a narrative,
i.e., an EPISODE in the logical sense.
Up to this point my project had thus proceeded more or less along steps 1 to 3 of
van Hahn’s process model. The real problem ─ and challenge! ─ however was that
the two blocks in the bottom left box turned out to be - empty containers: there
was no developed theory or model dealing with my particular research problem
available in HC/CP; there were also no established practices to apply. The
bottom line was that I had to design and program a mark-up tool before I could
even begin to model action structures by running a combinatory PROLOG-algorithm
on the meta-data.
In more abstract terms, I had found myself confronted with what I now regard as
the most important methodological principle to be elaborated upon in the
methodological-theoretical appraisal of CP. The type of questions considered
relevant in the philologies normally require a semantic mark-up of the ‘raw’
textual data; the mere digital representation of non-numeric data (unless it is
confined to a purely statistical distribution analysis) does not provide access
to semantic phenomena. However, the process of data mark-up, subsequent analysis
and modeling of meta-data, and eventual evaluation of the model in itself also
presupposes that the philologist has identified not just a philological, but
also a ‘computational’ frame of reference - in other words, the computer
philologist is not just picking tools from a box and applying these to old
questions. He or she must rather reconceptualize the research problem in a new
light before the tool’s aspect even comes into play.
As for my own project, this is how I would therefore schematize its
methodological architecture:
Figure 2:

In comparing von Hahn’s to my own diagram I come to the following
conclusions:
The interdependency between the tools and practices of CP on the one
hand, and the textual data on the other involves a hidden ‘third party’
- a conceptual frame of reference novel to the philologies.
The two-step representational transformation from textual data to
meta-data, and then from meta-data to philological interpretation,
therefore cannot be thought of as governed exclusively by conceptual
models drawn from philological theory and methodology. The import of the
‘foreign’ conceptual model informs the entire research architecture
throughout; it is as essential to the design of the transformational
processes as it is to the eventual interpretation of the transforms
generated.
The key intellectual prospect of the development and integration of CP
into the methodological ambit of the philologies proper is not the fact
that, because of deploying computer technology, one can in certain cases
analyze and model textual data faster, more coherently, or on the basis
of more explicit (and thus transparent) interpretive and/or cognitive
algorithms. What seems far more important is that we are offered new
frames of reference as to how to conceptualize and model text-based
phenomena.
Computer Philology undisputedly ‘has’ tools and practices - but it
‘is’ more than these: neither a proper ‘Wissenschaft’, nor an
a-theroretical ‘Hilfswissenschaft’, its methodological status is that of
a new philological heuristics.

Part of the problem of how to adequately define the methodological status of HC
and CP might, in fact, stem from the tendency to confound two meanings of the
term ‘computer’. When we focus on tools and practices in CP we refer to the
machine ‘computer’ in a very literal sense - hence the pragmatists’ attempt to
define CP per se in this vein. But ‘computer’ is also a metaphor, like ‘book’ or
‘image’, that stands for a particular mode and method of symbolic
representation. In the philologist’s case it is a metaphor for a very specific
reconceptualization of the phenomenon of ‘text based meaning’: one that aims at
bridging the gap between the qualifiable, and the quantifiable. Whether that
‘metaphor’ is run on a Cray or on paper and pencil is not really the key
question.