TRACES is an applied research project, whose originality is in detecting traces of both human and deepfake disinformation (intentionally spread fake information), with human disinformation being detected by its deliberate intent. TRACES will develop a new combination of interdisciplinary methods, including approaches from Natural Language Processing (NLP), Machine Learning (ML), Linguistics, Psycholinguistics, and Journalism. The Use Case of TRACES is Bulgarian, and the project will create guidelines on how to develop such technologies and create resources for other (low-resourced) languages. Due to the sensitivity of the topic, the project has a strong legal basis.
Funding organisation: AI4Media (https://www.ai4media.eu)
Funding programme: Open Call 1 (https://www.ai4media.eu/open-call-1/)
Duration: 1 year (1 March 2022 – 28 February 2023)
Dr Irina Temnikova
• Linguistic and psycholinguistic resources for Bulgarian (lists of linguistic expressions, signalling lies, deception, and manipulation);
• New balanced social media datasets
• Discovery of whether textual deepfakes in Bulgarian exist in social media.
• New tool for highlighting potential human and deepfake disinformation in Bulgarian text.
• Raising Bulgarian society awareness of linguistic aspects of disinformation.
• Winter school in January 2023.