Skip to content
Inductive Learning of Concept Representations from Library-Scale Corpora with Graph Convolution
Python Makefile
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
models
.gitignore
Makefile
README.md
classify.py
cluster.py
data.py
preprocessing.py
requirements-stable.txt
requirements.txt
train.py
utils.py

README.md

Concept Representation Learning

We consider a graph of papers and concepts. Paper nodes have a title and are connected when they have at least one common author. Concept nodes are featureless. The task is to learn a representation for the featureless concept nodes.

Quick start

Run make to setup a virtual environment, download sample data, and run the experiments. If there are any issues, or the methods should be applied to other data, follow the detailed steps below.

Setup

Set up a virtual environment (i.e. virtualenv -p /usr/bin/python3 venv && source venv/bin/activate) and then run:

pip install -r requirements.txt

In case, the required packages are not backwards compatible, run pip install -r requirements-stable.txt instead.

Preparation

A graph directory graph_dir with three csv files:

  • paper.csv with columns: paper_id, year, title
  • annotation.csv with columns: paper_id, subject
  • authorship.csv with columns: paper_id, author

Usage

python3 train.py gcn_cv_sc graph_dir -o model_dir

For more information on hyperparameters, consult python3 main.py -h. Then you can evaluate the resulting embedding model_dir/embedding.csv with the scripts cluster.py and classify.py.

Reproduction

Run make experiment to reproduce the experiments from the paper. This will download a sample of 100k research items to construct the graph.

Contributing

If you are interested in contributing, please notify us.

You can’t perform that action at this time.