Re: Corpora: generalisation in text

Kenneth W. Church (kwc@research.att.com)
Thu, 10 Sep 1998 15:09:18 -0400

No, no. Tom is a good guy. Tom was at AT&T Bell Labs about the time of
Writer's Workbench. George Heidorn and Karen Jensen were doing similar things
at IBM at about the same time. We'll probably find similar stuff showing up in
Microsoft products now that Karen and George have moved to Microsoft, but that
is just a guess. There was a tradition in the 1950s to develop metrics for
everything including metrics for books to determine which books were of
appropriate difficulty for which grades. You could look at vocabulary size,
sentence length, etc., etc. Writer's workbench combined some of this tradition
along with spelling checkers/grammar checkers and even checks for sexism. My
guess is that Tom is more interested in the first set of issues and that the
folks at Microsoft are more interested in the second. Both seem like fun
issues to study, with lots of good upside potential value for society. I would
be more concerned about the question of how well these things work than whether
they are bad for society. Of course, I would hope that teachers would
understand that there is much more to creative writing than merely the types of
things that these programs are likely to be good at, e.g., correct
spelling/grammar, vocabulary size, sentence structure/complexity. But in any
case, it wouldn't be a bad thing for both students and teachers to be routinely
using spelling checkers and similar such programs (assuming that they work
reasonably well).

Ken Church

john bouchlis wrote:

> Shades of 1984...<shiver>...
>
> ----------
> > From: John Milton <lcjohn@uxmail.ust.hk>
> > To: CORPORA@HD.UIB.NO
> > Subject: Re: Corpora: generalisation in text
> > Date: Thursday, September 10, 1998 1:05 AM
> >
> > Is this rlevant?
> >
> > THE INTELLIGENT ESSAY ASSESSOR
> > A psychology professor at the University of Colorado at Boulder is
> > spearheading the creation of an Intelligent Essay Assessor, a
> computerized
> > tool to assist professors in grading students' written essays. Thomas
> > Landauer says that to use the program, a professor must first teach it to
> > recognize both good and bad essay writing by feeding it examples of both,
> > which have been manually graded. The program can also be trained using
> > what he calls a "gold standard" -- passages from textbooks or other
> > materials written by experts on the same subject as the essay to be
> > graded. While earlier digital essay graders work by analyzing essays
> > mechanically -- looking at sentence structures and counting commas,
> > periods and word lengths -- Landauer says his program can actually
> > "understand" the student's writing using sophisticated artificial
> > intelligence technology called "latent semantic analysis." It does so by
> > comparing the patterns of word usage in student essays with the usage
> > patterns it has learned from the initial samples, enabling the computer
> > "to a good approximation, to understand the meanings of words and
> passages
> > of text." If an essay appears to convey the same knowledge as those used
> > in the examples, the computer gives it a high score. The Intelligent
> > Essay Assessor is not meant to be used to grade essays in
> > English-composition or creative-writing assignments, where a student is
> > being graded more on writing skill than subject knowledge.
> > (Chronicle of Higher Education 4 Sep 98)
> >
> >
> > .............................................
> > John Milton
> > Hong Kong University of Science & Technology
> > lcjohn@usthk.ust.hk
> >
> >
> >
> >