tnhaider / semantic-change-narrator Goto Github PK

This repository contains data and code to support the paper:

"The Ongoing Birth of the Narrator: Empirical Evidence for the Emergence of the Author-Narrator Distinction in Literary Criticism"

• figures
• data
  • all (all labeled instances and gold)
  • vanilla (folds 1-10 for vanilla model)
  • down (folds 1-10 with downsampling of B)
  • fic (folds 1-10 for coarse-grained model fictive-non-fictive)
  • predicted (predicted data test/large scale)
  • zeit (zeit text snippets for large scale)
  • dvjs (dvjs text snippets for large scale)
• scripts
  • text classification (doc_classification*)
  • inference (erzaehler_inference*)
  • extract_erzaehler (to extract text snippets from corpora)
  • get_erzaehler_embeddings (to plot token embeddings cf. Figure 4)
  • make_confusion_test (plot confusion matrix on predicted test data)

Training models is based on deepset FARM: https://github.com/deepset-ai/FARM

Clone their repository and make sure that the correct dependencies are used (sentence-transformers has since gotten a dedicated version number):

git clone https://github.com/deepset-ai/FARM.git
cd FARM
pip install -r requirements.txt
pip install --editable .

Make sure to set the correct fold number in the script!