The folder contains 1 programe, 
one for preprocessing the data -> transliterate_and_preprocess back transliterates the given data to roman script.

The graph convoutional networks have been addressed for text classification tasks, known as Text-GCN. text-gcn works well for low resource datasetes. 
We have used the same Text-GCN code to test our use of GCN for cmsa. The code is available at https://github.com/yao8839836/text_gcn and modified the program by adding a simple attention layer. 

