Text2Vec as node feature inputs

alecokas / swahili-text-gcn

Graph Convolutional Network for Swahili News Classification: https://arxiv.org/abs/2103.09325

https://arxiv.org/abs/2103.09325

MIT License

8 stars 4 forks source link

Text2Vec as node feature inputs #26

Closed alecokas closed 3 years ago

alecokas commented 3 years ago

Doc2vec and word2vec as inputs for the document and word nodes
Wrapped the directory checks in a util function check_df_and_stemming_paths
word_list index order checking function (ordered_word_list)
Checking function to make sure there are no Nones in the word list (_check_order)
Use the correct dimension to get the number of input features

Just need to do one last check on this before I request review

tyler-martin-12 commented 3 years ago

Awesome work adding the W2V and D2V feature inputs! I think the functions in word_features.py and doc_features.py could use a slight restructure but overall really like it

alecokas commented 3 years ago

@tyler-martin-12 I made the suggested changes - let me know if this makes more sense.