Cropping outperforms dropout as an augmentation strategy for training self-supervised text embeddings

In this repository you can find the code associated to the paper "Cropping outperforms dropout as an augmentation strategy for training self-supervised text embeddings"

How to use this repository

The notebooks in scripts/ contain the code to reproduce all the experiments and analyses performed in the paper. In particular, the notebooks 02-03 contain all the steps for generating the 2D embedding from raw text. The notebook 21-rgm-figures.ipynb contains the code to generate the final figures included in the paper. All figures generated with the notebooks will be stored in the results/figures/updated_dataset/final_figures/last_version folder.

Installation

This project depends on Python ($\geq$ 3.7). The project script can be installed via pip install . in the project root, i.e.:

git clone https://github.com/ritagonmar/text-embeddings
cd text-embeddings
pip install -e .