YouTube modelling labelled data

This repository contains the YouTube data used in the project CLINT (https://clint-project.github.io/) during the task that developed prediction models for content classification, link generation, and analysis of polarization evolution.

Summary

This repository contains a labelled dataset derived from the Recfluence project (https://recfluence.net/), where YouTube channels have been annotated according to their political bias (Left / Center / Right) and content type tags, such as Mainstream News or Conspiracy.

We focused on two subnetworks with opposing tags:

1. Partisan Network: Videos from channels tagged as either partisan left or partisan right.

2. Content-Type Network: Videos from channels tagged as mainstream news or conspiracy.

Data for both networks was collected using the YouTube Data API, following the official documentation and adhering to quota limitations. We queried videos from the available channels related to climate change and labelled each video according to the channel’s evaluation.

Data description

The dataset (data/video_data.csv) contains the following columns:

• network: Indicates which subnetwork the video belongs to. There are two subnetworks in this dataset:

  1. Partisan Network – videos from channels tagged as either partisan left or partisan right.

  2. Content-Type Network – videos from channels tagged as mainstream news or conspiracy.

• video_id: The unique identifier for the YouTube video, as provided by the YouTube platform. This can be used to access the video directly on YouTube.

• node_label: The label assigned to the video based on the channel it belongs to. For the partisan network, this is either left or right. For the content-type network, this is either mainstream or conspiracy.

• publishedAt: The publication date and time of the video, as reported by YouTube. The format follows the standard ISO 8601 timestamp.

Statistics overview