Combined Annotations of Misinformation, Propaganda, and Fallacies Identified Robustly and Explainably (CAMPFIRE)

Author(s):

Austin Blodgett, Claire Bonial, Taylor Hudson, and Clare Voss

Abstract:

This report details annotation guidelines for marking misinformation indicators, propaganda, and logical fallacies in text, culminating in the novel Combined Annotations of Misinformation, Propaganda, and Fallacies Identified Robustly and Explainably (CAMPFIRE) corpus. We give the motivation for developing this annotation schema, as well its relation to other parallel annotations. We describe the collection of a corpus of COVID-19 related texts to which the schema has been applied and supply sample annotations. We close with remarks on how this corpus is being leveraged to develop interactive question-answering and information extraction systems that are able to automatically detect and explain potential misinformation in queried documents.