First things first, have "abstract.txt", "abstract_paper.txt" and "coauthorship.edgelist" in the directory, then run:
python inverted_to_normal.py
python remove_stopwords.py
To obtain a txt file "sentences_line.txt" which is "abstract_paper.txt" converted to paragraphs and a "filtered_sentences.txt" which are paragraphs without stopwords.
Then, run
python createWeightedGraph.py
python createD2Vmodel.py
To create D2V model, you need to install fastnode2vec, and if you have the latest gensim version, you need to edit the node2vec.py file in your python site-packages/fastnode2vec and change size to vector_size for it to work.
python createSimMatrix.py
python createNode2Vec.py
to create the graphs and embeddings.
With gensim=3.8.3 and deepwalk=1.0.3, you can run a commandline to generate deepwalk embeddings, make sure to make the DeepWalkEmbeddings folder:
deepwalk --format edgelist --input coauthorship.edgelist --number-walks 20 --representation-size 64 --walk-length 20 --window-size 10 --output DeepWalkEmbeddings/coauthor.embeddings
Now with all the data ready, you can load the notebook createData.ipynb to create the X_train, X_test and Y_train csv files. Inside the file you will also find our current best model.
You can also use the notebook learning.ipynb to train the different models.