Merge branch 'main' of https://github.com/Merterm/Modeling-Intensification-for-SLG

Author: sabithsn

code/README.md

@@ -0,0 +1,3 @@
+To run BERT-gloss-tagger.py make sure to install sklearn and ktrain: <br>
+pip install ktrain==0.28.3 <br>
+pip install sklearn==0.23.2 <br>

data/README.MD

@@ -0,0 +1,12 @@
+#### gloss-annotation.tsv:
+This file contains manual annotation of the gloss from PHOENIX 14T dataset.
+
+#### full.train, full.dev, full.test:
+These tab separated files contain manual annotation along with tags predicted by our classifier.
+The first column denotes whether the label is manual annotation or predicted by the classifier.
+The second column contains gloss data of PHOENIX 14T dataset infused with the intensification labels. Gloss tokens that are predicted/annotated to have lower intensity, are surrounded by <LOW-INT> and </LOW-INT>. Tokens with higher intensity are surrounded similarly with <HIGH-INT> and </HIGH-INT>. The tokens that do not have any associated intensity, are not surrounded by anything,
+Example:
+1. $manual-annotation$	<HIGH-INT> WOLKE </HIGH-INT> LOCH SPEZIELL NORDWEST
+- This means the intensity labels for this instance are manually annotated. The token WOLKE has high-intensity, other tokens have no intensity.
+2. $tagged$	SAMSTAG AUCH <LOW-INT> FREUNDLICH </LOW-INT>
+- This means the intensity labels for this instance are tagged by our classifier. The token FREUNDLICH has low intensity, other tokens have no intensity.