text_gcn
Graph Convolutional Networks for Text Classification
Require
Python 2.7 or 3.6
Tensorflow >= 1.4.0
Reproduing Results
-
Run
python remove_words.py
-
Run
python build_graph.py
-
Run
python train.py
Example input data
-
/data/20ng.txt
indicates document names, training/test split, document labels. Each line is for a document. -
/data/corpus/20ng.txt
contains raw text of each document, each line is for the corresponding line in/data/20ng.txt
-
Change
dataset = '20ng'
inremove_words.py
,build_graph.py
andtrain.py
when producing results for other datasets.