erwanm/clg-authorship-analytics
The package contains a set of scripts and libraries to perform author-identification related tasks.
HTMLNOASSERTION
Issues
- 0
- 0
keep only filename as id in DocCollection
#17 opened by erwanm - 0
parameters minFreqIndiv and nbStopWords
#16 opened by erwanm - 0
- 0
about on-demand computing and parallel processing
#14 opened by erwanm - 0
improve method to return best configs
#13 opened by erwanm - 0
remove disk read/write access options everywhere?
#12 opened by erwanm - 0
multiple datasets/languages in run-std-training
#11 opened by erwanm - 0
wait for files in distributed process
#10 opened by erwanm - 0
- 0
pickNSloppy (and possibly others variants) should not be called with an empty list (divide by zero error)
#8 opened by erwanm - 1
Training process final output
#7 opened by erwanm - 1
Assuming independent cases in current version
#6 opened by erwanm - 0
- 0
why not use DocCollection instead of array of DocProvider objects in verif-author etc.?
#4 opened by erwanm - 0
Impostors and minDocFreq: check behaviour consistent with minDocFreq for probe docs in verif-author.pl
#3 opened by erwanm - 0
bug in train-multi-stages.sh: call to train-test for final retraining might not follow options like prefered data location
#2 opened by erwanm - 1
Bug in train-multi-stages.sh (individual strategy training), in the last part: re-training selected models using all cases and re-cross-validating in order to use predictions for meta-training stage
#1 opened by erwanm