NERsuite version 1.2.1
NERsuite is a named entity recognition toolkit.
This toolkit includes and uses a modified version of the GENIA 3.0.1 tagger, and the GENIA tagger uses the information extracted from the WORDNET. The NERsuite also requires the installation of libLBFGS (http://www.chokkan.org/software/liblbfgs/) and CRFsuite (http://www.chokkan.org/software/crfsuite/). Dictionary compiler and tagger include CDBPP library (http://www.chokkan.org/software/cdbpp/). Therefore, the use of this toolkit need to follows the copyrights and the license terms of these softwares. For other parts of this toolkit, it follows the BSD license.
Naoaki Okazaki for his invaluable help in developing and distributing this toolkit Yoshimasa Tsuruoka for the GENIA tagger And other collegues who willingly tested NERsuite and gave comments and advices
If you do make use of NERsuite or its components please cite as follows:
@misc{,
author = {Cho, Han-Cheol and Okazaki, Naoaki and Miwa, Makoto
and Tsujii, Jun'ichi},
title = {NERsuite: a named entity recognition toolkit},
howpublished = {https://github.com/nlplab/nersuite},
}
Minor updates (2015.01.20)
Minor updates (2012.07.06)
Version 1.2.1 (2012.07.02)
Change: -multidoc [SEP] option for all programs now regards lines "beginning with" [SEP] as comment lines and print them without any modification, whereas it previously dealt with the lines "exactly same" to [SEP]. In addition, multiple comment lines can be used from this version.
Change: nersuite_tokenizer computes token offsets at document level, whereas it used sentence level token offsets. When it encounters a document separator, which is given by -multidoc [SEP] option, the tokenizer resets the offset to 0.
New function: correct_mp program which removes named entities having mismatched parenthesis is added. Documentation will be added to the homepage later.
Minor updates (2012.06.28)
Change: the standoff and brat output option now use unique IDs for entities regardless of their semantic types. The IDs are also counted in document level, whereas the conll option still counts it in sentence level.
Bug fix: a bug of the string normalization part of the dictionary compiler and tagger is fixed.
New function: brat format for output is added (by Sampo Pyysalo)
Version 1.2
Version 1.1
Version 1.0