The aim of this project is to imrpove the performance and accuracy in NER and LM adaptation tasks but artificial data augmentation techniques.
The data folder has train, test and full data csv files. Perform all training/generation/augmentation on the train set and for evaluation, use the test set.
create_templates.py can be run to generate thrax templates.
Install the following libraries to run the Pegasus code:
!pip install sentence-splitter
!pip install transformers
!pip install SentencePiece
!pip3 install num2words
!pip3 install pandas