text_gcn

Graph Convolutional Networks for Text Classification

Require

Python 2.7 or 3.6

Tensorflow >= 1.4.0

Reproduing Results

1. Run python remove_words.py

2. Run python build_graph.py

3. Run python train.py

Example input data

1. /data/20ng.txt indicates document names, training/test split, document labels. Each line is for a document.

2. /data/corpus/20ng.txt contains raw text of each document, each line is for the corresponding line in /data/20ng.txt

3. Change dataset = '20ng' in remove_words.py, build_graph.py and train.py when producing results for other datasets.