Bilbo help
——————————————
BILBO : automatic reference labeling
——————————————
Usage: python src/bilbo/Main.py [options] <input data folder> <output data folder>
e.g. training: python src/bilbo/Main.py -T -t bibl Data/train/ Result/train/
labeling: python src/bilbo/Main.py -L -t bibl -d Data/test/ Result/test/
Options
(mode)
-T : –Training
Bilbo training, default=’False’
-L : –Labeling
Data labeling, default=’False’
(training, labeling)
-t : –typeref <string>
Input reference type
bibl => bibliography with a heading, wrapped by <bibl></bibl> (default)
note => notes wrapped by <note></note>
impl => implicit citations
-i : –informat <string>
Input reference format
tei => xml following tei guidelines including manual annotation (defalut)
xml => simple xml without tree in manual annotation
plain => plain text (for labeling only)
-m : –model <string>
Bilbo model name, default=’revues’
create a folder under “model/” folder and save models when training / load models when labeling
-g : –gradetag <string>
Grade of tag detail when using tei (option -f or -o)
simple => simple title and biblscope (default)
detail => detailed title and biblscope
-k : –keeptmp <string>
Decide which temp files are kept, among those created for svm crf training or test
none => keep nothing (default)
primary => keep primary temp files
all => keep all files
-v : –validatexml <string>
Decide if we validate files,
none => do not validate (default)
input => validate input files with tei_all.dtd
output => validate output files with tei_openedition3.xsd
all => validate input and output files
-s : –svmfilt
Training a svm model or classifying notes to filter out non-bibliographical notes, default=’False’
-u : –undopuncsep
Undo punctuation separation, default=’False’
(labeling)
-o : –outformat <string>
Output data format
tei => xml following tei guidelines (default)
xml => simple xml
simple => only labeled references without article contents or original tags
-d : –doi
Digital object identifier (doi) extraction via crossref site, default=’False’
-e : –exterdata
Labeling data different from training set. Do not use svm filtering when using this option. default=’False’
Arguments
arg1 : <string>
input data folder where the data files are (training or labeling)
arg2 : <string>
output data folder where the result files are saved
OpenEdition vous propose de citer ce billet de la manière suivante :
Young-Min Kim (11 avril 2013). Bilbo help. OpenEdition Lab. Consulté le 22 mars 2025 à l’adresse https://doi.org/10.58079/qnqd