More on Writer's Workbench Responses to survey
Gary Perlman
perlman at wanginst.UUCP
Wed Oct 9 10:47:01 AEST 1985
I have some comments about the Writer's Workbench software.
Few, if any of these ideas are my own, but have been pointed
out to me by others, so I'll take no credit, but will accept
responsibility. I should point out that I am working with a
knowledge of the System V software of 1984, and many of my
criticisms may have been addressed. If so, someone "in the
know" should follow up.
The Writer's Workbench, WWB, is a collection of programs to
aid writers. Although I use the programs, I think it is
important to know their limitations. Despite having
limitations, they can still be useful.
My main reservations about WWB are:
(1) The programs do not deal with what I think are primary
variables affecting writing quality, which is what WWB is
supposed to help me with. WWB tools, notably the "style"
program and its readability scores, ignore structure
(sections, paragraphs, lists, tables) and look at sentences.
Certainly, sentence structure is important, and long
sentences and long words make a text harder to read, but the
outline structure of a document is primary for me. No, I
don't have any data on this. I am always struck by the
paradox of how style uses lists--you know, those structural
entities showing a greater depth of knowledge communicated
in a clear format--style cuts them out because they make the
text look like long sentences. The style program is good
for Reader's Digest kind of text, but for technical text, I
think it is useless.
(2) The programs can give people a false sense of security.
Some people write text and run the programs and get that
warm feeling when they have removed all the spelling errors
found by spell (spellwwb) and get a good readability score.
WWB is dumb enough (or general enough if you prefer that
term) to work on C programs, implying that you can do well
with WWB but still have complete garbage. This is not a
problem with WWB, it is a problem that some users of WWB do
not understand its limitations and think it is a solution.
WWB is an aid.
(3) WWB is written for the troff text formatter, mainly in
that it uses deroff to remove the formatting commands. It
also assumes that you are using the -mm or -ms macro
packages (used in part to remove those nasty lists). To use
the WWB tools with something like Scribe or TeX, you need to
write your own pre-processor.
(4) The programs tend to run slowly. They are hacks in the
worst sense in that they contain the cruddiest code and have
the most ugly and undecipherable output formats of any UNIX
programs.
WWB is still a useful sidekick for writers. I find problems
after I have edited a document to my satisfaction: doubled
words, spelling errors, punctuation problems, and wordy
phrases. I never use sexist language. Besides being
useful, WWB is cheap and private. When I give a draft of a
paper to others for comments, they do not spend their
valuable time to point out low level blunders I should have
caught. And hopefully, they do not need to know that I
can't spell, or that I tend to use a very unique number of
meaningless phrases in drafts. :-)
Still, I find myself in a bind when I am told by WWB
programs that something is wrong and I disagree. I know I
am right, but I wish the programs would stop telling me that
"man" is sexist when I am writing about the UNIX "man"
command. I know that "sexist" is just pattern matching, as
it does in "diction", but I wish I could get it to shut up.
The programs are good at pointing out problems, but not so
good about suggesting solutions. This is especially true
for the style program. You are told that your document has
a readability grade level of 18, but you are only helped out
by being presented with the long sentences. So you madly
chop up your sentences to make them more readable and style
is made happy. Writing to "make the grade" sometimes works,
but better organization might be a better strategy.
The programs in WWB I like the most are ones that help me
view my documents in ways that are difficult for me. The
diction program highlights (in a pathetic way, by
bracketing) candidate meaningless words and phrases. The
punct program checks my punctuation faster than I can. The
analyses from WWB that I am most uncomfortable with, as
though you could not tell yet, are ones that try to make
quality judgements. The output from the wwb program
(saturation bombing with most of the WWB programs) seems
authoritative as it tells me that in my C program, I have
appropriately limited the number of passives, and that my
readability grade level is good for "this type of document."
What a charade.
What I would like to see come out of the WWB group is some
software for novel views of documents, and less cosmetic
adjustments like in the wwb program (I would not mind some
cosmetic adjustments to their output formats). This was one
of the motivations behind me and Tom Erickson writing some
tools for structural displays (not analysis) of documents.
(Oh, so here is the self-plug!) My "headings" program prints
out a section and paragraph headings outline for troff
documents using any macro package--in about 100 times the
speed of the WWB org program. Our "punc" program prints out
punctuation graphs of sentences so that you can view their
length and structure in novel ways. Here are the punc
graphs for this paragraph:
____________________,_________(___________).
_________________(__)__.
(_,_____-_!)
_"_"________________--___________.
_"_"___________________.
You can read more about our views of graphical abstractions
of technical documents and how they differ from some of the
WWB measures in:
Perlman, G., & Erickson, T. D. (1983) Graphical Abstractions
of Technical Documents. Visible Language, 17, 380-389.
Perlman, G., & Erickson, T. D. (1984) Abstraction Program
Aids To Documentation. Asterisk, 10:2, 13-16.
You can get reprints by sending me your postal address. If
there is demand, I may post the programs to net.sources.
--
Gary Perlman Wang Institute Tyngsboro, MA 01879 (617) 649-9731
UUCP: decvax!wanginst!perlman CSNET: perlman at wanginst
More information about the Comp.unix
mailing list