TwitterGraphDataset

Repository contents

This release provides a single archive, dataset.zip, containing four files.
Each file is a node-level table with community assignments computed using the Leiden community detection algorithm (optimizing CPM) on a specific underlying graph.

dataset.zip files

• rt_nodes_with_communities
  Community assignments computed on the retweet graph (user–user interaction network).

• reply_nodes_with_communities
  Community assignments computed on the reply graph (user–user interaction network).

• rt_and_hashtags_nodes_with_communities
  Community assignments computed on the retweet + hashtag graph, a mixed network that includes:

  • the retweet interaction network, and
  • user–hashtag links, where hashtags are those used in retweeted content.

• reply_and_hashtags_nodes_with_communities
  Community assignments computed on the reply + hashtag graph, a mixed network that includes:

  • the reply interaction network, and
  • user–hashtag links, where hashtags are those used in replied-to content.

In the interaction-only graphs (rt_nodes_with_communities, reply_nodes_with_communities), nodes are users only.
In the mixed graphs (rt_and_hashtags_nodes_with_communities, reply_and_hashtags_nodes_with_communities), nodes include users and hashtags.

---

File format

All four files share the same schema. Each row represents a node (user or hashtag, depending on the file).

Each file contains:

• id: a sequential identifier for each node
• name: the unique identifier of the node
• type: node type
  • u = user node
  • h = hashtag node
• 0.1, 0.2, …, 1.0: community assignments (community IDs) produced by Leiden (CPM) for different values of the resolution parameter

Community assignment columns (Leiden + CPM)

For each resolution column:

• the value is an integer community ID assigned to that node at that resolution
• community IDs are only meaningful within the same resolution (e.g., community 3 at 0.2 is not necessarily related to community 3 at 0.8)