Modelling the Depths of a Literary Encoding, with an Example from Ovid

This paper explores some consequences of encoding a complex, intricate and lengthy poem for a densely
occurring set of literary phenomena. In it I dwell on one of these phenomena, personification, in order to
define and illustrate the idea of deep encoding and to reflect on its implications. I am not concerned with
details of the metalanguage as such, except in one particular point. Nor am I much concerned with the tagged
text or the scholarly product generated from it. Rather I focus on the epistemology of marking up a text: I ask,
how do we make new knowledge rather than simply record what we already, somehow know, or think we
know? In Wendell Piez's terms, in what sense may the markup of an ancient text be proleptic (Piez 2001)?
To establish the background of my research to date I briefly review material presented in prior
ACH/ALLC conferences (Paris 1994; Kingston 1997; Virginia 1999).
The poem in question is the Metamorphoses, written in classical Latin hexameter by the Roman poet
Publius Ovidius Naso at approximately the beginning of the Christian era. The Met is a problematically
organized collection of mythological stories that became the primary means by which much of the
Greco-Roman heritage was handed down through the Middle Ages to the present day. The literary
cohesiveness of the Met—how it works as a poem—is therefore an important problem, but it has proven a
considerable challenge to critics. My approach to this problem, in the Analytical Onomasticon project
(McCarty 2002), is to encode all devices of language that refer to persons (hence the Onomasticon, or “book
of names”), and from the resulting tags to generate a set of indexes for constructing interrelations among
stories. The Onomasticon is thus intended to aid further literary-critical study, especially along the lines of
Wheeler 2000 and to a somewhat lesser extent Schmidt 1991.
In McCarty 1998 and 1999 I made first attempts to consider the epistemological consequences of
markup by isolating treatment of personification (i.e. the creation of persons by rhetorical means, Paxson
1994). In this paper I take issue with these attempts, in which I applied the term “grammar” loosely to a set of
criteria for encoding without giving any regard to implementation. I argue here, rather, that these criteria
comprise only a first step. I adapt the Chomskian sense of a grammar as a computable set of rules by which
the linguistic phenomena in question may be generated (Thorne 1972: 184–6). Hence, in these terms, for there
to be a grammar of personification, individual instances must be computable from the primitive elements
which can be said to be responsible for them. These primitive elements are described in some detail in
McCarty 1998—for example, apostrophe, familial relationship and mental activity. How, then, might they be
implemented such that instances of personification could be generated from a true grammar?
In the paper I give a few examples from the Metamorphoses to indicate the complexity that a
grammatical perspective on Ovidian personification involves. When viewed in such a way, personification
comprises not only the linguistic and rhetorical factors local to a candidate, but also five larger contexts that,
according to my proto-grammar, affect the success of these factors: onomastic, narrative, poetic-mythological,
ontological, personal. Furthermore, moving from a declarative phenomenology of personification to an actual
grammar means that the rules by which success is computed themselves become part of the problem.
In the first instance the subject of this paper is not how to write these rules. Rather it is how best to
conceptualize the kind of problem for encoding that personification exemplifies. How we compute a complex
textual phenomenon, whether from the metadata or directly from the data, is an important aspect of this
question, but I approach it indirectly with the help of two ideas: depth and modelling.
At some level markup must be simply declarative: thus, “X exists here”. Let us define a shallow
encoding as one in which only such declarative statements are possible or are admitted. One might say in such
a case that the phenomena of interest are the primitives of the system— in a builder's terms, for example, the
bricks, mortar, timbers, tiles, plaster and paint. Contrast the encoding of personification as described. It is an
example of a deep (or at least deeper) kind because the object of study is not simply declared as a primitive
but may be computed from declarable elements. Thus, if our builder were to take a particular interest in paint,
it would cease to be a primitive in his system, rather would be something to be mixed from primary pigments
and a carrier base.
Note that an encoded work may not be uniformly deep: for example, some objects of study in the
Onomasticon, such as proper names, are simply declared as such, and within the work as conceived, it makes
no sense whatever to compute them from more primitive elements. Hence an encoding may have depths but
not be simply deep. Note also that a deep encoding is not necessarily a thick one, to borrow Wendell's term:
the density of tagging is a separate matter.
Depth, then, is not a given but is created by the encoding scholar's focus in combination with
available theory, the analytical tools at hand and the nature of the phenomena. It is expressed grammatically
by a synthesis involving primitive elements and rules for their combination. It is a relative, not an absolute
term. It implies no value-judgment, although a completely shallow encoding, however useful to scholarship,
is hardly itself a scholarly work. And a thin encoding, however deep, would not give scholarship much scope.
The idea of depth has value in that it expresses a qualified imperative to extend what can be seen to a
finer level of detail. The idea of a grammar has value because a grammar is independent of the text to which it
is applied; it is exportable, with the result that either it is universalized or the phenomenon in question is
resolved into two or more related kinds. Thus, for example, if depth is created in encoding personification
within the Metamorphoses, the resulting grammar can be applied to other Ovidian works, others in classical
Latin and in various languages across different historical periods. What is personification, exactly? Perhaps
plumbing its depths through encoding will yield us much more revealing questions than that one.
Depth, I suggested, has a temporal dimension contingent upon developing ideas and interests in
literary texts. Its method of expression, however, can aid that development directly by the way in which
computational grammars are designed and published. If, that is, we conceive our grammars as devices for
modelling phenomena of interest—i.e., manipulating representations of these phenomena—and if we publish
these modelling tools, then our colleagues will be able to continue the work we begin, not simply agree or
disagree with it. Such a modelling tool, e.g. for personification, would not simply allow for differing views of
the trope to be expressed, rather more for them to be explored, particularly fruitful ones identified and so the
phenomenon itself better to be understood.
36
In conclusion I return to the epistemological question with which I began: how may markup be used
to get beyond the recording of old knowledge, with consequent reliance on sorting and formatting what we
already know? How is new knowledge made through markup? Deep encoding, I argue, provides a powerful
answer so long as we understand that depths are to be modeled and that successful modelling turns them into
shallows while revealing further depths beneath.
This paper, then, constitutes a criticism of previous work and a proposal for a deepening of it. The
existing proto-grammar is, as far as I can determine, the first of its kind for personification but is not in a form
that can be tested. In the paper I sketch a plan for this work and show an example of how the markup might be
done.
REFERENCES
Craik, F. and R. Lockhart. 1972. “Levels of processing: A framework for memory research“. Journal of
Verbal Learning & Verbal Behavior 11: 671–684.
McCarty, Willard. 1998. “A provisional grammar of personification for the Met”. In “What is humanities
computing? Toward a definition of the field”.
http://www.kcl.ac.uk/humanities/cch/wlm/essays/what/grammar.html (23/11/02).
---. 1999. “Thinking with markup: the case of personification”. ACH/ALLC annual conference, University of
Virginia, 9–13 June 1999.
--- (with John Bradley, Monica Matthews, Aara Suksi, Burton Wright). 2002. An Analytical Onomasticon to
the Metamorphoses of Ovid. http://www.kcl.ac.uk/humanities/cch/wlm/analyticalonomasticon/
(24/11/02).
Paxson, James J. 1994. The Poetics of Personification. Literature, Culture, Theory 6. Cambridge.
Piez, Wendell. 2001. “Beyond the ‘descriptive vs. procedural’ distinction”. Extreme Markup Languages
2001: 210ff. http://www.piez.org/wendell/papers/beyonddistinction.pdf (24/11/02).
Schmidt, Ernst. 1991. Ovid's Poetische Menschenwelt: Die Metamorphosen als Metapher und Symphonie.
Heidelberg.
“Semantic encoding.” Glossary. Non-Roman Script Initiative, SIL International.
http://www.sil.org/nrsi/glossary.htm#semenc (23/11/02).
Thorne, James Peter. 1972. “Models for Grammars”. In Teodor Shanin, ed., The Rules of the Game:
Cross-Disciplinary Essays on Models in Scholarly Thought. London: Tavistock: 179-205.
Wheeler, Stephen M. 2000. Narrative Dynamics in Ovid's Metamorphoses. Classica Monacensia, Bd. 20.
Tübingen: Gunter Narr Verlag. (See the helpful review by Alden Smith, Bryn Mawr Classical
Reviews 2001.11.23 (http://ccat.sas.upenn.edu/bmcr/2001/2001-11-23.html), and Wheeler's
response in BMCR 2001.12.18 (http://ccat.sas.upenn.edu/bmcr/2001/2001-12-18.html).