torch_geometric.datasets.UPFD

class UPFD(root: str, name: str, feature: str, split: str = 'train', transform: Optional[Callable] = None, pre_transform: Optional[Callable] = None, pre_filter: Optional[Callable] = None, force_reload: bool = False)[source]

Bases: InMemoryDataset

The tree-structured fake news propagation graph classification dataset from the “User Preference-aware Fake News Detection” paper. It includes two sets of tree-structured fake & real news propagation graphs extracted from Twitter. For a single graph, the root node represents the source news, and leaf nodes represent Twitter users who retweeted the same root news. A user node has an edge to the news node if and only if the user retweeted the root news directly. Two user nodes have an edge if and only if one user retweeted the root news from the other user. Four different node features are encoded using different encoders. Please refer to GNN-FakeNews repo for more details.

Note

For an example of using UPFD, see examples/upfd.py.

Parameters:

  • root (str) – Root directory where the dataset should be saved.

  • name (str) – The name of the graph set ("politifact", "gossipcop").

  • feature (str) – The node feature type ("profile", "spacy", "bert", "content"). If set to "profile", the 10-dimensional node feature is composed of ten Twitter user profile attributes. If set to "spacy", the 300-dimensional node feature is composed of Twitter user historical tweets encoded by the spaCy word2vec encoder. If set to "bert", the 768-dimensional node feature is composed of Twitter user historical tweets encoded by the bert-as-service. If set to "content", the 310-dimensional node feature is composed of a 300-dimensional “spacy” vector plus a 10-dimensional “profile” vector.

  • split (str, optional) – If "train", loads the training dataset. If "val", loads the validation dataset. If "test", loads the test dataset. (default: "train")

  • transform (callable, optional) – A function/transform that takes in an torch_geometric.data.Data object and returns a transformed version. The data object will be transformed before every access. (default: None)

  • pre_transform (callable, optional) – A function/transform that takes in an torch_geometric.data.Data object and returns a transformed version. The data object will be transformed before being saved to disk. (default: None)

  • pre_filter (callable, optional) – A function that takes in an torch_geometric.data.Data object and returns a boolean value, indicating whether the data object should be included in the final dataset. (default: None)

  • force_reload (bool, optional) – Whether to re-process the dataset. (default: False)