Unsupervised Domain Adaptation for Same Task/ Same Domain

[jk7599]

Domain Adaptation

PDF: https://arxiv.org/pdf/2004.10964.pdf
Implementation
https://github.com/allenai/dont-stop-pretraining

Things to Consider

• Are we going to use Roberta like the paper or BERT or some other model?
• How are we gonna store the trained model? Is there free storage that we can use?

Dataset

• Same Task

1. TREC-8 QA Dataset
   https://trec.nist.gov/data/qa/t8_qadata.html
2. Yahoo Answer Dataset
   https://www.kaggle.com/datasets/jarupula/yahoo-answers-dataset

• Same Domain

1. Covid 19 Dataset
   https://github.com/davidcampos/covid19-corpus

Plan

-[] Implement to train the model with domain adaptation
-[] Store the model for used by active learning later