README.md

Data Mining Project

Context Aware Category Prediction Using Semantic Article Graphs

How to Run Demo

1. Download the dataset from huggingface
2. Place the downloaded dataset in the demo/input directory.
3. Download the models from:
   • GCN model: GCN
   • GCN without edges: GCN_no_edges
   • Place the downloaded models in the demo/models/ directory.
   • Rename the folders as undirected_gnn and no_edge_gnn respectively.
4. Set PYTHONPATH to the parent directory:

   export PYTHONPATH=demo/
   

5. Navigate to the demo directory in your terminal.
6. Install the required dependencies using:

   pip install -r requirements.txt
   

7. Run the Streamlit app using:

   streamlit run src/demo.py
   

Alternatively you can visit the hosted demo at: Demo

How to Train the GNN model or Run Heuristic Methods

1. Download the dataset from huggingface
2. Place the downloaded dataset in the modeling/ directory.
3. Install the required dependencies using:

   pip install -r demo/requirements.txt
   

4. Navigate to the modeling directory in your terminal.
5. Run gnn.ipynb or heuristic_methods.ipynb using Jupyter Notebook or Jupyter Lab.

How to view Preprocessing Steps

1. Navigate to the preprocessing directory in your terminal.
   • preprocess.ipynb: Contains all the preprocessing steps to create the dataset from raw wikipedia articles collected using Wikipedia API.
   • generate_embeddings.ipynb: Contains steps to generate document embeddings for the articles using a pre-trained embedding model.

---

• Metin Usta
• 504251504