TEG Datasets

Dataset Format

Each dataset is a PyG Data object and is stored in the processed subdir following a unified format with each attribute defined as follows:

• edge_index: Graph connectivity in COO format with shape [2, num_edges] and type torch.long.
• text_nodes: List contains textual information for each node in the graph.
• text_edges: List contains textual information for each edge in the graph.
• node_labels: List contains text labels for each node in the graph. We use -1 to represent nodes without labels

Embedding Data Format

The embedding data is thrived from text_nodes and text_edges through PLM including:

• GPT
• BERT-base
• BERT-large

We will provide more TEG datasets and PLM embedding in the futrue!