Hyperlink Prediction on Hypergraphs of Text

This repository contains datasets and code for the paper Hyperlink Prediction on Hypergraphs of Text. Alessia Antelmi, Tiziano Citro, Dario De Maio, Daniele De Vinco, Valerio Di Pasquale, Mirko Polato, Carmine Spagnuolo.

About The Project

Hypergraphs have recently emerged as powerful tools for modeling high-order relationships in complex real-world data. Nevertheless, existing hyperlink prediction methods primarily emphasize the structural connectivity encoded by these structures, often overlooking the rich semantic information associated with nodes and relations, which can often be expressed in the form of text.

In this work, we propose a novel framework for hyperlink prediction on Hypergraphs of Text (HoTs), where both nodes and hyperedges are enriched with textual attributes. Our model jointly leverages these semantic and structural signals by combining hypergraph convolutional operators with cross-attention mechanisms that iteratively refine node and hyperedge representations.

Experimental results demonstrate that integrating semantic information from nodes and hyperedges with structural properties consistently improves performance over baselines relying solely on topology, hence highlighting the effectiveness of contextual representations for hyperlink prediction and opening new directions for semantic-aware hypergraph learning.

Built With

Getting Started

Follow these steps to set up and run HPHoT (Hyperlink Prediction on Hypergraphs of Text) locally.

🧩 Prerequisites

Before you begin, ensure you have the following installed:

• Python 3.10+
• Git
• (Optional) CUDA toolkit (if using GPU)

You can check your Python version with:

python --version

If you don’t have pip or venv, install them using:

python -m ensurepip --upgrade

⚙️ Installation Clone the repository

git clone https://github.com/hypernetwork-research-group/hphot.git
cd hphot
(Recommended) Create and activate a virtual environment

python -m venv venv
source venv/bin/activate       # macOS / Linux
venv\Scripts\activate          # Windows

Install dependencies

pip install -r requirements.txt

Usage

You can run all the experiments, including baselines and ablation studies, using the provided script:

./run.sh

This will:

• Train and evaluate all baseline models
• Perform ablation studies on different hypergraph configurations
• Save the results and logs in the runs/ directory

License

Distributed under the project_license. See LICENSE.txt for more information.

(back to top)

Contact

Project Link: https://github.com/hypernetwork-research-group/hphot

(back to top)