# Context [Emergent.info](http://www.emergent.info/) was a major rumor tracker, created by veteran journalist [Craig Silverman](https://twitter.com/CraigSilverman). It has been defunct for a while, but its well-structured format and well-documented content provides an opportunity for analyzing rumors on the web. [Snopes.com](http://www.snopes.com/) is one of the oldest rumors trackers on the web. Originally launched by Barbara and David Mikkelson, it is now run by a team of editors who investigate urban legends, myths, viral rumors and fake news. The investigators try to provide a detailed explanation for why they have chosen to confirm or debunk a rumor, often citing several web pages and other external sources. [Politifact.com](http://www.politifact.com/) is a fact-checker that is focused on statements made by politicians and claims circulated by political campaigns, blogs and similar websites. Politifact's labels range from "true," to "pants on fire!" --- # Content This dataset consists of three files. One file is a collection of all webpages cited in Emergent.info, and the second is a collection of webpages cited in Snopes.com, and the third is a similar collection from Politifact.com. The webpages were often cited because they had started a rumor, shared a rumor, or debunked a rumor. Emergent.info Emergent.info often provides a clean timeline of the rumor's propagation on the web, and identifies which page was for the rumor, which page was against it, and which page was simply observing it. Please refer to the image below to learn more about the fields in this dataset. ![The image displays a sample post from Emergent.info and highlights the corresponding fields in emergent.csv.][1] Snopes.com The structure of posts on **Snopes.com** is not as well-defined. Please refer to the image below to learn more about the fields in the Snopes dataset. ![This image displays a sample post from Snopes.com and highlights the corresponding fields in snopes.csv.][2] Politifact.com Similar to Emergent.info, Politifact.com follows a well-structured format in reporting and documenting rumors. There is a sidebar on the right side of each page that lists all of the sources cited within the page. The top link is the likeliest to be the original source of the rumor. For this link, page_is_first_citation is set to true. ![This image displays a sample post from Politifact.com and highlights the corresponding fields in politifact.csv.][3] --- # Inspiration I created this dataset in order to study domains that frequently start, propagate, or debunk rumors. By studying these domains and people who follow them, I hope to gain some insight into the dynamics of rumor propagation on the web, as well as social media. --- # Notes/Disclaimer When using the Snopes dataset, please keep the following in mind: * In addition to debunking rumors, Snopes.com occasionally reports news and other types of content. This collection only includes data from "[Fact Check](http://www.snopes.com/category/facts)" posts on Snopes. * Snopes.com was launched years ago. Some of the older posts on the website do not follow the current format of the site, therefore some of the fields might be missing. * Snopes.com used to use a service named "[DoNotLink.com](https://twitter.com/donotlink?lang=en)" for citation purposes. That service is no longer active and as a result some of the links are missing from older posts on Snopes. * In addition, some of the shortened links would time-out prior to resolution, in which case they would not be added to the dataset. * Occasionally, a website that has been cited has not maliciously started a rumor. For instance, Andy Borowitz is a humorist who writes for *The New Yorker*. His satirical column is sometimes mistaken for real news; as a result, *The New Yorker* may be cited as a source of fake news on [Snopes.com](http://www.snopes.com/trump-blasts-media-for-reporting-things-he-says/). This does not mean that *The New Yorker* is a fake news website. When using the Politifact dataset, please keep the following in mind: * The data included in this dataset are collected from the "[truth-o-meter](http://www.politifact.com/punditfact/statements/)" page of Politifact.com. * Politifact often fact-checks statements made by politicians. Since this dataset is focused on websites, I have ignored all the posts in which the rumor was attributed to a person, a political party, a campaign, or an organization. Instead, I have only included rumors attributed explicitly to websites or blogs. --- # Useful Tips for Using the Snopes collection As opposed to the Emergent collection where each page is flagged with whether it was for or against a rumor, no such information is available for the Snopes dataset. To avoid manually labeling the data, you may use the following heuristics to identify which page started a rumor: * Webpages that are cited in the "Examples" section of a post are often "observing" the rumor, i.e. they have not started it, but they are repeating it. In the snopes.csv file, these webpages have been flagged as "page_is_example." * Webpages that are cited in the "Featured Image" section of a post are often not related to the rumor. The editors on Snopes have simply extracted an image from those pages to embed in their posts. In the snopes.csv file, these webpages have been flagged as "page_is_image_credit." * Webpages that are cited through a secondary service (such as [archive.is](http://archive.is/)) are likelier to be rumor-propagators. Editors do not link to them directly so that a record of their page is available, even if it is later deleted. * If neither of these hints help, very often (but not always) the first link cited on the page (for which "page_is_example" and "page_is_image_credit" are false) is the link to a page that started the rumor. This link is identified by the "page_is_first_citation" field. Pages for which both "page_is_first_citation" and "page_is_archived" are true are very likely to be rumor propagators. * To identify satirical websites that are mistaken for real news, it's useful to inspect the way they are cited on Snopes. To demonstrate that a website contains satire or humor, Snopes writers often cite the "about us" page of the site. Therefore it's useful to see which domains often contain a URI to their "about" page (e.g. "http://politicops.com/about-us/"). [1]: http://imgur.com/JZPExar.png [2]: http://i.imgur.com/jFT6Vdb.png [3]: http://i.imgur.com/Z83JP7c.png