Objective 1: Analysis of annotations

The first objective is to see how well the accountability annotations correspond to features in the text, to determine if there could be any issues with the given annotations, such as human errors or bias. This will be investigated with:

  • Basic statistics about the text annotations, e.g. word distributions
  • Semantic analysis with framenet: investigate using the causation frame, consider proposing a new frame

Objective 2: Train a classifier

The objective is to see if a classifier could be used to automate this type of annotation for future research. To do this, I will train a classifier directly on the given datasets, this classifier could be applied sentence by sentence to articles in the future to identify accountability in larger datasets.