Sentence Representations based on Dependency Graphs and Graph Convolutional Networks (DG-GCN)

Side project that I worked during the 2021 Xmas holidays to improve my pytorch-geometric skills.

The basic idea is to use GCN to build a vectorial representation of the sentence, starting from its dependency graph.

This representation can be used to carry out other tasks

Work is still in progress, the model was used for the PreTens22 challenge