Filtering of attributes in the Text Classification API

8 views
Skip to first unread message

julien nioche

unread,
Jun 14, 2007, 3:45:35 AM6/14/07
to DigitalPebble
Our TextClassification API (http://www.digitalpebble.com/
TextClassificationAPI.pdf) has a new functionality!
It is now possible to filter out some of the attributes based on log
likelihood ratio. This substantially improves the quality of the
results and at the same type speeds up the training and
classification.

The TextClassification API is available as a GATE plugin. A wrapper
for UIMA (http://incubator.apache.org/uima/) will be available during
the summer.

Reply all
Reply to author
Forward
0 new messages