Re: Corpora: generalisation in text

Ken Litkowski (ken@clres.com)
Thu, 10 Sep 1998 16:05:22 -0400

I would certainly second Ken Church's comments and add a little of my
own.

In my software, I have implemented a content analysis program known as
Minnesota Contextual Content Analysis, which has been used extensively
over 25 years by many students of a Professor of Sociology at UMinn, Don
McTavish. It does pretty well at characterizing "emphasis" and
"context" in a text (from sentence to book) as a result of a decent
category dictionary. Recently, some licensees of mine have been
exploring its use in distinguishing how well student essays show depth
of conceptualization, with reasonably decent correlations to human
grading.

I have been working with this category dictionary and trying to
understand why it works well and have made several observations about
issues of generalization and conceptual depth. (See papers at my site.)

My explorations are very much a work in progress, but they are
suggestive. And they should be taken very much in conjunction with
current research in information retrieval and text summarization. That
Karen and George are working in this direction as well can be seen in
their beginnings in Word97 with its identification of significant
sentences. Certainly, much is on the way.

Ken

-- 
Ken Litkowski                     TEL.: 301-482-0237
CL Research                       EMAIL: ken@clres.com
9208 Gue Road
Damascus, MD 20872-1025 USA       Home Page: http://www.clres.com