Skip to content

Latest commit

 

History

History
8 lines (7 loc) · 709 Bytes

README.md

File metadata and controls

8 lines (7 loc) · 709 Bytes

FER TAR course project

Detection of propaganda techniques in news articles

Repo for our team's solution (WallstreetBets)

Original shared task data and labels are located in data/articles and data/labels.

  • data_parsing.py is responsible for generating labeled words from original data, suitable for sequence labeling task
  • split_data.py splits processed data (located in data/labeled_articles) into train-validate-test split. Also, it merges all articles into one file for each split, which is needed for Flair datasets
  • training.py is example of Flair usage for sequence labeling tasks. Its output can be some simple baseline, and it is stored in folder resources (ignored by git)