Visualization

[GiovanniPioDelvecchio]

It is necessary to create some functions to plot sailent elements of the project such as:

• Dependecy graph of a sentence with the right labels (corresponding to the tokens/words or, alternatively, the node indexes);
• How values for each node change after each GCN layer, meaning that it is needed a function to plot the embeddings (just like colorful tiles) and a function to plot the hidden states of the GCN.

Some plotting functions are already present in the notebook, however it is needed to extend them in order to get a more meaningful graph, where after aggregating the node features (the feature vector of a single node should become a color) it is possible to comprehend better what happens after the convolutions and thus what the network is learning.

[GiovanniPioDelvecchio]

These have been added in the issue#2 branch