Dataset of conversation graphs from 2019 Anti CAA protest in India

To facilitate future research in online social movements,  we are publishing this dataset of conversation graphs in the from the 2019 anti-CAA protest tweets.  The conversation graph dataset has two types of conversations:

• Mention Graphs: In this, vertices are users and (directed) edges represent any interaction between the users. Within the context of our dataset, these interactions are limited to mentions alone when a user refers to another user directly by username.

• Reply Graphs: In this case, vertices are users, and a (directed) edge exists if one user replies to another.

The reply graph is built as follows:

• Since replies often form chains, find the root 'destination' user nodes that have been replied to at least once.
•  Find all the immediate next-level users who directly replied to the root nodes (these are 'source' nodes)
• Iteratively map the immediate next level of users who reply to the previous level of corresponding users, stopping when all of the users in the next level have not been replied to (leaf users, so to say).

The mention graph is built as follows:

• Since replies often form chains, find the root 'destination' user nodes that have been replied to at least once.
• Find all the immediate next-level users who directly replied to the root nodes (these are 'source' nodes)
• Iteratively map the immediate next level of users who reply to the previous level of corresponding users, stopping when all of the users in the next level have not been replied to (leaf users, so to say).

Additionally, the dataset has the users and conversation graphs labeled for emotion and toxicity.  Motif count has been provided for each of the graph. 

To maintain confidentiality as per Twitter guidelines, user ids are anonymized  and Tweet IDs and Tweet texts are not part of the dataset.