Information Technology Reference
In-Depth Information
ambiguous or incorrect. In cases where they found errors, students proposed
changes to the software's internal dictionaries as part of their log-work. If their
proposals were verified by the course instructors, the internal dictionaries were
changed to reflect them. It is beyond the scope of this paper to say more about
these methods. Further discussion about them is available elsewhere [4].
5 Coding Heuristics:
Accommodating Surface Language Indeterminacy
Thus far, we have made a case for manually coding surface language to uncover
genre functions. There are reasons, some we have hinted at already, for a healthy
skepticism toward this project. The surface stream of natural language is fraught
with ambiguity, irregularity, and contextual contingency. It is futile to hope that
communities of speakers will segment surface strings in the perfectly predictable
and convergent way that machines can algorithmically. So in what do we place
some confidence that our string segmentations of English provide reliable grist
for macro-level analysis of texts?
We place this confidence in the fact that we have coded strings from an
idealized perspective of the competent writer. Whatever the hazards of surface
language, competent writers fearlessly take on the hazard, and typically end up
with reasonable success in pinning down the experiences for readers they wish
to pin down. How do competent writers do it? We don't pretend to have direct
psychological answers to this question. We did, however, benefit, we believe, by
using this question, and possible answers to it, as a basis for establishing our
coding rules. More to the point, we idealized the role of a competent writer
as an underlying basis for our coding. Unlike the competent language-user in
the Chomskian sense, who brings to sentence generation an implicit standard
of grammaticality, the competent writer, in the sense we coded for, brings to
the generation process an expertise in pinning down experience for other human
beings in contiguous runs of language.
We further imagined an idealized competent writer as always involved in
an ongoing mental game we called rhetorical scrabble. In conventional scrabble,
players can be dealt any subset of 26 letters and must form legal words with them.
In rhetorical scrabble, we imagined, the writer is dealt in background any legal
word of English (anywhere from 15,000 to 100,000 depending upon the writer's
vocabulary) and is challenged to select a word from the list or create a word series
that completes a felt-experience that another human (a reader) can recover.
The shorter the series, the more the writer needs to worry about ambiguity,
communicating too many possible experiences without communicating one. The
longertheseries,themorethewriterneeds to worry about recurrence and re-
usability, the chances that the series in question, called upon once, will continue
tobecalleduponinothertextsandcontexts. The ideal player of rhetorical
scrabble, it is important to keep in mind, boasts not just the largest conceivable
store of English strings. He or she boasts the largest conceivable store of diverse
and re-usable strings.
Search WWH ::
Custom Search