Textual Genre Analysis and Identification - Ambient Intelligence for Scientific Discovery - page 133

cruit the textual surface, despite its indeterminacy, into reliable micro-indicators
about how texts perform work. These micro-indicators, in turn, fit into cultural
macro-understandings of texts at work and, especially, understandings of the
cultural work that texts perform within genre functions.
In the next section, we make a small illustrative argument on behalf of the
claim that readers have learned how to recruit language in the small to recognize
genre functions in the large. In the final section, focusing on the culturally-rich
genre of the technology review, we make a more extended case.
3 The Surface Text as a Probabilistic Constraint
on Genre: An Illustrative Case
How might surface features of texts play a role in how we understand genre?
Consider four conventional textual genres: (1) corporate letters that make com-
plaints (see above); (2) corporate letters that field complaints; (3) corporate
letters that offer apologies; (4) corporate letters that seek to collect money from
a delinquent account. We collected a database of 300 texts containing each of
the four types. We found that we had no trouble, as readers, telling the genres
apart intuitively. Even without knowing in advance what kind of letter we were
reading, it was easy to classify it accurately.
Yet it was a different matter when we asked ourselves, with the naked eye, to
back up our classifications with a close reading of the surface text. The surface
text available to us as everyday readers - making conventional segmentations of
the surface text into clauses and sentences - often provides random hints but
not smoking-gun evidence that one letter had to be in one category or another.
In essence, we could be very confident in our holistic classification of letters,
but not very confident in our ability to articulate the elements in the surface
language that prompted these holistic judgments.
Nonetheless, by systematically coding smaller runs of language (strings) into
separate categories and performing statistical comparisons across genres, it be-
comes possible to describe each letter type as a distinctive combination of string
families. In Table 1, we show a profile of each letter type based on string families
depicted in the first column. For our limited illustrative purposes here, we don't
define these categories further and assume many of them are self-explanatory,
or self-explanatory enough for the reader to see the point.
A positive number in a cell indicates that the string family is more active in
that genre relative to the other three genres. A negative number indicates that
the string family is less active relative to the other genres. The absolute size of
the number indicates the strength of this relative activity or relative inactivity. A
“1” indicates that the genre is significantly more active or inactive with respect
to one other genre. A “2” indicates significance relative to two other genres. A
“3” indicates significant difference relative to all three remaining genres.
An inspection of Table 1 clarified our own hidden decision rules about how we
were able to classify the genres. Compared to other letter genres, letters of apol-
ogy contain high amounts of first person, autobiographical, self-disclosing, apol-