text_gcn

Pytorch implementation to paper "Graph Convolutional Networks for Text Classification".

Running

You can run the whole process very easily. Take the R8 corpus for example,

Step 1: Clean the corpus.

./scripts/R8.sh clean

Step 2: Build dataset.

./scripts/R8.sh build_dataset

Step 3: Build graph.

./scripts/R8.sh build_graph

Step 4: Train and inference.

./scripts/R8.sh train

Performance Comparision

-	Dataset	Accuracy
Original	R8	97.07%
This Implementation	R8	96.21%

-	Dataset	Accuracy
Original	R52	93.56%
This Implementation	R52	92.52%