Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

NELA-GT-2022: A Large Multi-Labelled News Dataset for The Study of Misinformation in News Articles

Maurício Gruppi, Benjamin D. Horne, Sibel Adalı

TL;DR

NELA-GT-2022 addresses the need for a large-scale, time-consistent, multi-outlet news dataset with veracity labels to support misinformation research. The authors implement a stable data collection pipeline that scrapes RSS feeds twice daily, extracts embedded tweets via Goose3, and applies copyright-protective text transformations, releasing data in SQLite and JSON formats with MBFC ground-truth labels and event-based subsets. Key contributions include 1) a 2022 corpus of 1,778,361 articles from 361 outlets, 337 MBFC-labeled sources, and 346,283 embedded tweets, 2) accessible data formats and tooling, and 3) example subsets for the Russo-Ukrainian War and Roe v. Wade to facilitate analysis of coverage and messaging. The dataset supports longitudinal analyses, robust evaluation of veracity models, and studies of media manipulation across time and events, with implications for researchers, journalists, and policy makers.

Abstract

In this paper, we present the fifth installment of the NELA-GT datasets, NELA-GT-2022. The dataset contains 1,778,361 articles from 361 outlets between January 1st, 2022 and December 31st, 2022. Just as in past releases of the dataset, NELA-GT-2022 includes outlet-level veracity labels from Media Bias/Fact Check and tweets embedded in collected news articles. The NELA-GT-2022 dataset can be found at: https://doi.org/10.7910/DVN/AMCV2H

NELA-GT-2022: A Large Multi-Labelled News Dataset for The Study of Misinformation in News Articles

TL;DR

NELA-GT-2022 addresses the need for a large-scale, time-consistent, multi-outlet news dataset with veracity labels to support misinformation research. The authors implement a stable data collection pipeline that scrapes RSS feeds twice daily, extracts embedded tweets via Goose3, and applies copyright-protective text transformations, releasing data in SQLite and JSON formats with MBFC ground-truth labels and event-based subsets. Key contributions include 1) a 2022 corpus of 1,778,361 articles from 361 outlets, 337 MBFC-labeled sources, and 346,283 embedded tweets, 2) accessible data formats and tooling, and 3) example subsets for the Russo-Ukrainian War and Roe v. Wade to facilitate analysis of coverage and messaging. The dataset supports longitudinal analyses, robust evaluation of veracity models, and studies of media manipulation across time and events, with implications for researchers, journalists, and policy makers.

Abstract

In this paper, we present the fifth installment of the NELA-GT datasets, NELA-GT-2022. The dataset contains 1,778,361 articles from 361 outlets between January 1st, 2022 and December 31st, 2022. Just as in past releases of the dataset, NELA-GT-2022 includes outlet-level veracity labels from Media Bias/Fact Check and tweets embedded in collected news articles. The NELA-GT-2022 dataset can be found at: https://doi.org/10.7910/DVN/AMCV2H
Paper Structure (16 sections, 3 figures, 3 tables)

This paper contains 16 sections, 3 figures, 3 tables.

Table of Contents

  1. Introduction
  2. What's New in NELA-GT-2022?
  3. Data Collection
  4. News data and metadata
  5. Embedded tweet data
  6. Limitations
  7. Format of Data
  8. JSON Format
  9. Ground Truth Data Format
  10. Use Cases
  11. Analysis of news coverage during events
  12. Russo-Ukrainian War
  13. Overturning of Roe v. Wade
  14. Embedded Tweets
  15. Long-Term Use Cases
  16. ...and 1 more sections

Figures (3)

  • Figure 1: Number of articles (a, c) and embedded tweets (b, d) collected during each week of 2021.
  • Figure 2: Distribution of sources per reliability class (a) and factuality (b) score.
  • Figure 3: Number of articles related to (a) Russo-Ukrainian War, and number of articles related to (b) the overturning of the Roe v. Wade case as a fraction of the total number of articles in each week of 2022. Articles are found using a set of keywords shown in \ref{['tab:keywords']}.