Towards Unsupervised Few-Shot Text Transfer: A Method with Pretrained Language Model and Style Extractor

UCSD LIGN 167 WI22 Final Project

Personal implementation for TextSETTR: Few-Shot Text Style Extraction and Tunable Targeted Restyling

t5_extractor.ipynb contains most of the code used for the final model.
Other notebooks are either for data processing or for preliminary experiments
writeup/ contains the final submission and the poster for the UCSD LIGN 167 course project
data/ includes part of the data used in the training. However, the amazon review dataset is not included since it is too large.

Name		Name	Last commit message	Last commit date
Latest commit History 4 Commits
.vscode		.vscode
data		data
writeup		writeup
.gitignore		.gitignore
bart_discriminator.ipynb		bart_discriminator.ipynb
flickr8k_data.ipynb		flickr8k_data.ipynb
readme.md		readme.md
t5_extractor.ipynb		t5_extractor.ipynb
t5_vanilla.ipynb		t5_vanilla.ipynb
yelp_preprocess.ipynb		yelp_preprocess.ipynb

Provide feedback