Messages from alternative Spanish Telegram channels, 2019-2024

This dataset contains processed data extracted from Telegram channels using pytopicgram from 2019-12-01 to 2024-08-31. It includes anonymized channel information, sampled messages, and topics identified using BERTopic. The data has been anonymized and structured for ease of analysis. The dataset comprises two main CSV files:

1. Topics (topics.csv)

This file contains topics extracted from the full dataset using BERTopic. Each topic is described by a concise text generated by OpenAI o1.

Column Name
Description

Topic
Numeric identifier for each topic. -1 is the generic topic for non-assignable messages.

Name
Human-readable name summarizing the topic.

Representation
List of representative keywords for the topic.

Description
Concise description of the topic generated by OpenAI.

2. Messages (messages.csv)

This file contains a 25% stratified sample of messages (on topic column) from Telegram channels.

Column Name
Description

channel_id
Anonymized identifier for the Telegram channel.

week_year
Week and year when the message was posted (format: week_year).

media_type
Type of media included in the message (txt, img, video, audio, doc, web).

reach
Number of users reached by the message.

virality
Virality score of the message.

is_viral
Boolean indicating whether the message is considered viral.

topics
Topic identifier associated with the message.

probs
Probability scores for topic assignment.