Data Augmentation (DA) can be defined as any method for increasing the diversity of training examples without explicitly collecting new data. This repository contains the code for a research project focusing on coreference resolution in the field of natural language processing (NLP). The goal of the project was to investigate the impact of coreference resolution on classification of fake news. The research aimed to determine whether coreferencing the input data before classification is more effective than classifying them without coreference.
We used two availabe datasets:
- Getting real about fake news - training dataset.
- WELFake dataset - used for classification tasks.
The repository contains the following files:
- 📁 notebook: This directory contains the source code for the implementation of the proposed procedures.
- 📁 results: This directory stores the evaluation results and performance metrics of the implemented classifiers as .png and in .csv.
List of all authors who contributed to the project:
- Jozef Kapusta (supervisor)
- Dávid Držík (PhD. student)
- Kirsten Šteflovič (PhD. student)
- Kitti Szabó Nagy
You can read the article here.