dl-translate: a python library for text translation between 50 languages using Deep Learning

I'm happy to share the release of my new library called dl-translate, which has been my personal software project for the past few weeks. Before I get started, here are some quick links:

• ?? GitHub Repository
• ?? Documentation / Readthedocs
• ?? PyPi project
• ?? Colab Demo / Kaggle Demo

The dl-translate library can be installed with pip and it lets you translate text into one of 50 different languages. This code snippet (derived from the repository's README) gives a concise overview:

The goal of the project is to make it easier for any Python user to start translating text, without necessarily having been exposed to Deep Learning and Natural Language Processing. For this reason, I also provided utility functions and methods to help you quickly find the language you want to use. For example, if you use an editor like VSCode with Python plugins, then you will be able to leverage the auto-complete feature after you type dlt.lang:

Although the library is designed to be used in an abstracted way, I wanted to make it useful for advanced Python users and those that are familiar with Deep Learning and NLP (more specifically, transformers). Thus, I included options to customize the underlying deep learning model and tokenizer, as well as access them through getter methods. This is all explained in-depth in the advanced section of the readme and documented in the API reference. Here's a preview:

If you already know Machine Learning, then I recommend checking out Hugging Face's transformers library and the multilingual BART model (trained on 50 languages), along with the paper behind it.

If you are not familiar with any of this, then I hope that by using this library, you will be encouraged to try out the transformers library and learn more about mBART, both of which helped simplify neural machine translation for the greater ML/Python community.