SAFT: Structure-aware Transformers for Textual Interaction Classification

SIGIR 2025 paper, available at arXiv

This repository contains the source code and datasets for running the experiments.

Requirements

The code is written in Python 3.12.3. To run it, install the required packages using the following commands (a virtual environment is recommended):

pip install scikit-learn==1.5.0
pip install tqdm==4.66.2
pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu121
pip install pandas==2.2.2
pip install transformers==4.41.2

---

Datasets

Download processed data. To reproduce the results in our paper, you need to first download the processed datasets. Place the edge_data folder in the same directory as the SAFT folder (if placed elsewhere, update the dataset path in the scripts).

Raw data & data processing. Raw data can be downloaded from Amazon, Goodreads and Google directly.

Running

Train

bash script/$dataset/train_SAFT_{GNN/GAU}_$dataset.sh

Test

bash script/$dataset/test_SAFT_{GNN/GAU}_$dataset.sh