Trained model(Nov.19)
(BERT trained with BODY) https://duke.box.com/s/6clw9gx2vqpu26s4p7yh7z64gg3vi8yg
(BERT trained with HEADLINE) https://duke.box.com/s/8bax76my1wfbdu2715xtrp662zu6fl72
(BERT trained with SUMMARY) https://duke.box.com/s/zq5bu72d29k83tbuen5b24deqojzhiaj -> The accuracy is 62%. Using 10% (15K lines) of the full dataset (150K) Summarization was conduced with the T-5 'small' model. Would using the T-5 'Base' model improve the results......?
https://www.kaggle.com/code/mikiota/data-augmentation-csv-txt-using-back-translation
https://www.kaggle.com/code/nkitgupta/text-representations
https://huggingface.co/docs/transformers/en/tasks/summarization