Name		Name	Last commit message	Last commit date
Latest commit History 10 Commits
README.md		README.md
README_incl_w_LRP		README_incl_w_LRP
align.py		align.py
bootstrapped_run.py		bootstrapped_run.py
chunk.py		chunk.py
features.py		features.py
gazetteer2lafs.py		gazetteer2lafs.py
generate_train_test_filelists.py		generate_train_test_filelists.py
io_.py		io_.py
logger.py		logger.py
makeLTFbaselines.py		makeLTFbaselines.py
rmABGbracketsFromLTF.py		rmABGbracketsFromLTF.py
run.sh		run.sh
score.py		score.py
seeABGproportions.py		seeABGproportions.py
seeABGtriples.py		seeABGtriples.py
seeLTFtokenABGdistr.py		seeLTFtokenABGdistr.py
tagger.py		tagger.py
train.py		train.py
util.py		util.py

Repository files navigation

uhhmm_NER

To run an experiment, use the following command:

./run.sh [language]

But this should be done after the directories have been configured inside run.sh.

Also, the features to be used should be properly configured inside features.py, starting on line 70.

All paths in run.sh should be changed to the directories or files containing the necessary data.

The commented experiments inside run.sh can be uncommented to be repeated. However, when running any experiment from

run.sh, it is important to ensure that features.py (l.70 and below) has been configured to include those features

and only those features which the experiment expects to be a part of its model.

Experiments will not produce errors if this is not the case; however, the results will not be accurate

since the features reported to have been used were not.

Any TRAIN_SCP and TEST_SCP text files will have to be created.

When training on all data, bash commands which concatenate the paths to the LAF and LTF files respectively are useful.

When training on all data, the following command could be used to create the TRAIN_SCP:

for fn in $(ls [LAF-DIR]); do echo "[LAF-DIR]${fn}" >> [TRAIN_SCP]; done

When testing on all data, the following command could be used to create the TEST_SCP:

for fn in $(ls [LTF-DIR]); do echo "[LTF-DIR]${fn}" >> [TEST_SCP]; done

When splitting into training and testing sets, generate_train_test_filelists.py will be helpful.

About

No description, website, or topics provided.

Report repository

Releases

No releases published

Packages

No packages published

Languages