Text Classification Demonstration

I am an expert in classifying abstracts, e-mail, titles or phrases as being within the scope of

TM - Assessment and Evaluation
EA - Educational Management
PS - Early Childhood

Try me. Enter some text (title, question, abstract, phrase, word) in the box. I work best with abstracts. Please paste one in. Or, paste in a reference question, preferably within one of these scopes. You can get some abstracts at http://eric.ed.gov (in a new window).

This example was developed as to test my essay scoring software. The system is an application of the multinomial naive bayes model for text classification [see Mitchell, T. (1997). Machine Learning. New York: McGraw Hill. Chapter 6]. Words and two word phrases were coded, infrequent terms purged. The system was initially calibrated on 900 recent abstracts per category (clearinghouse). The classifiers were then applied to the trainning data, classification errors for the trainning set were noted and trimmed and the classifiers were recalibated. Calibration and trimming were iterated until the system had no errors with the trainning set.

Possible applications: routing reference questions, routing electronic submissions, suggesting where to transfer documents, flagging possible scope issues in the Acquisitions Data Report.   -   Larry