Proposed work
The research community has formalized the above problem as: retrieving relevant fact checks given an input claim.
The key challenge is identifying a resurfaced claim as it is often formulated differently than the one already checked;
novel formulations may use paraphrasing, introduce sarcasm, eliminate or replace context, and so on.
Our goal is to design and develop a complete approach for streaming content analysis
which, on one hand, given an input claim, identifies past fact-checks (FCs, in short), and on the other hand,
tracks the evolution and origin of claims, across media content of many forms (social media messages, short
like tweets or longer such as Facebook posts, blogs, news-propagating websites, etc.)
While we will rely as necessary on existing English benchmarks, our focus will
be primarily on French, for which we plan to build and share corpora and models. This work is part of our participation
to ELIAS Work Package 2, related to tools for online democracy.
