The fnTBL toolkit


    Home
    News
    Documentation
    FAQ
    License
    Applications
    Download



  NLP logo




       


What is fnTBL

fnTBL is a customizable, portable and free source machine-learning toolkit primarily oriented towards
Natural Language-releated tasks (POS tagging, base NP chunking, text chunking, EOS detection, word
sense disambiguation). It can handle reasonably sized, discrete classification task (i.e. the samples can be
characterized as vectors with discrete components).

fnTBL is using a machine learning technique called transformation based learning, first introduced by
Eric Brill in 1992, mainly based on the idea of succesively transforming the data to correct the error that
gives the biggest win in terms of error rate. The transformation rules obtained are usually few and
meaningful.

fnTBL was created at the Johns Hopkins University NLP group by Radu Florian   and Grace Ngai .

Current News

09/12/2001 fnTBL version 1.0 is now available

fnTBL can be downloaded by accessing the download link on the side toolbar .

Comments

Please email us any comments or suggestions you have at rflorian@cs.jhu.edu .

Questions or comments at rflorian@cs.jhu.edu or the mailing list fntbl@nlp.cs.jhu.edu