Thesis Work

Open maboberlin opened this issue 9 years ago • 0 comments

ParsCit-Trainingssystem-Extension

Additional Tools for training and testing CRF++ models for ParsCit. Using BibTeX-Data and LaTeX (customized .bst files) for automatical building and testing models.

Usage (in /bin folder):

train.pl (gets a bibtex-file and optional additial crfpp-formated data as input and builds a crfpp model)
test.pl (gets testdata in form of a bibtex file or a tagged references file plus the name of the model file and outputs data about the accuracy of the model)

Additional Requirements:

LaTeX installation (to run latex programmatically set: /usr/share/texlive/texmf/web2c/texmf.cnf -> change openout_any = p to = r )

Extension structure:

/lib/Trainer: Modules for converting BibTeX data to tagged references and for converting tagged references to plain references + configuration module
/resources/tex: original and modificated bibtex style files for converting BibTeX data to tagged and untagged references
/resources/traindata: example train data
/resources/traindata: example test data

Feb 10 '17 19:02 maboberlin