Learning Word Embeddings for African Languages

[ToluClassics]

The goal here is to create a Word2vec (CBOW and SkipGram) Colab tutorial to learn word representations for African languages. We would start with English and then migrate to other languages like Yoruba, Igbo, Hausa, Swahili e.t.c. All that needs to be changed would be the corpora anyway.

• Video to understand word2vec:: https://www.youtube.com/watch?v=ERibwqs9p38&t=582s
• Word2vec paper:: https://arxiv.org/pdf/1301.3781.pdf
• sample pytorch word2vec tutorial:: https://rguigoures.github.io/word2vec_pytorch/

We should only use datasets from huggingface:: here's one https://huggingface.co/datasets/mc4

LMK if you need any help along the way

[Seun-Ajayi]

Alright boss, I would let you know when I need help.

[Seun-Ajayi]

@ToluClassics
Boss, I've gone through the materials you put up there, I would be going through them again for better assimilation. Just thought to update you. Many thanks.

And can I get to know what's next?

[ToluClassics]

I added a skeleton notebook for the word2vec tutorial; that is what we need to flesh out::

the idea is to learn the word embeddings and visualize them like they did here or here

[Seun-Ajayi]

Am I to go ahead and attempt something rough with the dataset from huggingface after mastering the concept? I mean, that's a good way to evaluate my understanding of the concept.

[ToluClassics]

Oh you don't have to master it; just do something we'll review and take it from there