Wikipedia hoax and non-hoax articles from Kumar, West & Leskovec, 'Disinformation on the Web' (WWW 2016), distributed by SNAP. The README describes 64 hoaxes, each paired with a non-hoax article created the same day with similar appearance; the parsed folders hold those 64 + 64 articles, while the cleaned-markup folder holds 76 hoaxes and 64 non-hoaxes (140 articles). The archive is SNAP's own and also contains 65 macOS '._' resource-fork files, which are not articles. Released under the CRAPL licence (CRAPL-LICENSE.txt is included).
A Wikipedia (online encyclopedia) dataset of 11,377 users and 196,614 directed signed votes (143,879 support, 40,667 oppose, 12,068 neutral) with 189,541 vote rationale texts, collected by West and Leskovec from Wikipedia's public adminship election records spanning 2003–2013, distributed by SNAP. Each record contains directed signed edges from voter to candidate: +1 support, -1 oppose, 0 neutral; election outcome per candidate (0/1); natural language vote rationale text per vote. The dataset is notable for its unique combination of signed network structure with natural language rationale text per edge, enabling joint graph and text analysis; one of the few real-world signed networks with explicit textual justifications for positive and negative links. Intended for research into signed network analysis, trust and reputation modeling, text-attributed graph research, opinion mining, link sign prediction with text features, and community detection in collaborative platforms.
A Wikispeedia (Wikipedia-based online navigation game) dataset of 4,586 Wikipedia article nodes, 119,767 hyperlinks, and 76,193 human navigation paths, collected by West & Leskovec via the Wikispeedia online game (2008–2012). Each record contains article hyperlink graph (source → target), navigation path sequences, path duration in seconds, path length in clicks, back-click count, and optional difficulty rating (1=easy to 5=brutal). Unique dataset combining a Wikipedia hyperlink graph with 76,193 human navigation paths from an online game where players navigate from a random start article to a target article using only hyperlinks — 67.4% of paths were successfully completed.. Suitable for semantic distance estimation, human wayfinding behavior, hyperlink graph analysis, information foraging, Wikipedia structure research.
Wikipedia