Aligning Bilingual Texts with Machine Learning

[kundalini]

I just came across a Python library that uses machine learning to generate an aligned bilingual text in HTML. I think it looks very promising! More details are in this article:
https://medium.com/@averoo/how-to-make-a-parallel-book-for-language-learning-part-1-python-and-colab-version-cff09e379d8c

The Google Colab notebook is here: https://colab.research.google.com/drive/1_ics0YzWg5qIZIPhA1X_Wbfg0XZzRO-p

If you aren't familiar with using Colab, and are interested in giving it a whirl, you just have to make a copy of the notebook into your own Google Drive, replace the Harper Lee text in the Colab notebook with bilingual texts of your own, and make sure that the language codes ('en' and 'ru' in the example) are correct. Then click on the play button in each cell in the notebook to run the code.

I gave it a go with the first chapter of The Count of Monte Cristo, using texts from https://www.gutenberg.org/cache/epub/17989/pg17989-images.html#I and https://archive.org/stream/thecountofmontec01184gut/crsto12.txt, and the result is shown below:

dumas.png

[Cainntear]

Oooh... I read too quickly and didn't realise that this wasn't a book translator thing, but actually a program that aligns human-translated texts with the original human text. That's a genuinely intriguing prospect that basically removes a massive labour-intensive step from the process of making these things. That's very interesting.

[kundalini]

Oooh... I read too quickly and didn't realise that this wasn't a book translator thing, but actually a program that aligns human-translated texts with the original human text. That's a genuinely intriguing prospect that basically removes a massive labour-intensive step from the process of making these things. That's very interesting.

Yes, I changed the thread title to clarify this point. It seems like lingtrain_aligner, the Python aligner program, uses machine learning to align sentences based on their likelihood of being semantically similar.