This repository contains codes for my paper:
Requirements:
python: 3.x
pytorch: 1.5+
gensim: 3.x
The data pre-processing part (processor.py and data.py) were adapted based on Aditya and Suraj's Transformers-For-Negation-and-Speculation, but added my personal label augmentation options and changed the dataloader structure.
You will need to store the downloaded data (please collect from the corresponding dataset owners) into /Data/ folder, in the forms listed in params.py (data_path), in order to run the experiments.
Various settings are listed in params.py, you may try to adjust different hyperparameters.
Sorry for the codes contains a lot of line irrelevant to the paper (which were used for previous exploration of solutions)