telegram_forwards — Pushshift Telegram channels and users (2015-2019)

Description

Networks of Telegram channels and groups from the Pushshift Telegram dataset of Baumgartner, Zannettou, Squire and Blackburn, who collected all messages (317 million, from 2015 to 2019) of 27,801 public channels and groups (supergroups), which they found by a snowball crawl starting from about 250 channels on politics, cryptocurrencies and other topics, through the channels referenced in their messages (mostly right-wing and cryptocurrency channels). The data are as returned by the Telegram API (Telethon). The entry contains two networks:

forwards: a directed edge from channel u to channel v means that messages of u were forwarded to v (v republished messages first posted in u), weighted by the number of forwarded messages. Besides the crawled channels, the nodes include all channels whose messages were forwarded to them. Forwards of messages of individual users (and of hidden ones, given only by name) are left out.
posts: a bipartite network of users and channels, where an edge connects a user to a channel (or group) in which they posted messages, weighted by the number of messages. Messages posted by the channels themselves (in broadcast channels, without user) are left out, and so are service messages (e.g. users joining). All crawled channels are included, also those without messages of users.

In posts, the users have the information of their profiles (names, usernames, phone numbers and flags) from the file of the 2.2 million accounts of the dataset, retrieved in 2019; Telegram users have no biographies in these data, and most phone numbers are hidden (about 13,000 accounts show one). Left out are the profile photos and the online status of the users, and the contents of the messages, replies and mentions.

The networks have the following properties:

(Node) channel_id (channels): the id of the channel on Telegram.
(Node) crawled (channels): whether the channel is one of the crawled ones (otherwise only a source of forwarded messages).
(Node) title, username, created (channels): its title, its username (its address is https://t.me/<username>) and its time of creation (Unix time), known for the crawled channels and for the channels linked to them (empty or -1 otherwise).
(Node) broadcast (channels): whether it is a broadcast channel (1), a group (0) or unknown (-1).
(Node) participants_count (channels): the number of members when it was crawled (-1 if unknown).
(Node) about (channels): its description (crawled channels only).
(Node) retrieved (channels): the time when its information was retrieved (Unix time; -1 for channels not crawled).
(Node) is_user (posts): whether the node is a user (otherwise a channel).
(Node) user_id (posts): the id of the user on Telegram (-1 for channels).
(Node) has_profile (posts): whether the user is in the file of accounts.
(Node) first_name, last_name (posts): the display name of the user.
(Node) phone (posts): the phone number of the user, if visible (international format without "+").
(Node) bot, deleted, verified, scam, restricted (posts): the flags of the account (false for channels).
(Node) username, retrieved (users, in posts): the username of the account (its address is https://t.me/<username>) and the time when the profile was retrieved (Unix time). The other channel properties are empty or -1 for users.
(Edge) count: the number of forwarded messages (forwards) or of posted messages (posts).
(Edge) first_time and last_time: the times of the first and of the last of these messages (Unix time; in forwards, the times when they were forwarded).

Tags

Social Online Communication Political Bipartite Weighted Timestamps Metadata

Added

2026-10-31

Citation

J. Baumgartner, S. Zannettou, M. Squire, J. Blackburn, "The Pushshift Telegram Dataset." Proceedings of the International AAAI Conference on Web and Social Media 14, 840-847 (2020) 10.1609/icwsm.v14i1.7348 Sci-Hub

Upstream URL OK

https://doi.org/10.5281/zenodo.3607497

Upstream license

Creative Commons Attribution 4.0 International (CC BY 4.0), https://creativecommons.org/licenses/by/4.0/

Networks

Tip: click on a column header to sort the list, or hover over it for a description.

Column legend

$\left<k\right>$: Mean degree
$\sigma_k$: Standard deviation of the degrees
$\lambda_h$: Largest eigenvalue of the Hashimoto (non-backtracking) matrix
$\tau$: Random walk mixing time
$r$: Degree assortativity
$c$: Global clustering coefficient
$\oslash$: (Pseudo-) diameter
$S$: Relative size of the largest component
NPs: Node properties
EPs: Edge properties

Name	Nodes	Edges	$\left<k\right>$	$\sigma_k$	$\lambda_h$	$\tau$	$r$	$c$	$\oslash$	$S$	Kind	Mode	NPs	EPs	gt	GraphML	GML	csv
forwards	348,789	1,338,172	3.84	47.98	223.50	457.86	-0.13	0.04	11	1.00	Directed	Unipartite	channel_id crawled title username created broadcast participants_count about retrieved	count first_time last_time	18.5 MiB	28.5 MiB	28.2 MiB	27.7 MiB
posts	1,197,990	1,325,088	2.21	75.91	46.72	979.99	-0.28	0.00	12	0.98	Undirected	Bipartite	is_user user_id channel_id crawled title username created broadcast participants_count about retrieved has_profile first_name last_name phone bot deleted verified scam restricted	count first_time last_time	32.1 MiB	50.0 MiB	48.3 MiB	51.5 MiB