Overtone: Misinformation based on Article DNA
The technology provides insight by looking at the qualities of the text itself, rather than tertiary metrics such as clicks and shares. It uses the advances in Natural language processing (NLP) to make the type of data that teams of analysts could generate available to media companies.
The next step is sorting through articles based on what is in them on the paragraph level. Overtone can accurately label whether a paragraph is factual, opinionated, journalism, or potentially toxic. The work will begin with this model in the misinformation space, with the collaboration of citizens who want better control over what they see online.
The Overtone project will build ways to identify different types of misinformation based on the qualities of their text, using the new model to provide analysis paragraph by paragraph, the ‘DNA’ of an article. In a conversation with citizens, the project will create a taxonomy of potential misinformation that can be identified algorithmically.
Overtone uses advances in Artificia Intelligente (AI) to create data that powers a web based on qualitative signals. The solution is valuable not just in the context of misinformation, but in wide swaths of the media ecosystem that demand a new way of looking at content online.
The NLP products make qualitative data, which is intrinsic to the text, independent and immediate, available at speed and scale. This data is used for content strategy and distribution by businesses such as publishers who want to better distribute their content, or ad and PR professionals who want to better understand what’s out there.