Assessing Authenticity and Anonymity of Synthetic User-generated Content in the Medical Domain

Tomohiro Nishiyama (1) , Lisa Raithel (2, 3) , Roland Roller (2) , Pierre Zweigenbaum (3) , Eiji Aramaki (1)

1 NAIST - Nara Institute of Science and Technology - Graduate School of Information Science
2 DFKI Lab Berlin - Deutsches Forschungszentrum für Künstliche Intelligenz GmbH = German Research Center for Artificial Intelligence
3 STL - Sciences et Technologies des Langues - LISN

Since medical text cannot be shared easily due to privacy concerns, synthetic data bears much potential for natural language processing applications. In the context of social media and user-generated messages about drug intake and adverse drug effects, this work presents different methods to examine the authenticity of synthetic text. We conclude that the generated tweets are untraceable and show enough authenticity from the medical point of view to be used as a replacement for a real Twitter corpus. However, original data might still be the preferred choice as they contain much more diversity.

Mots clés: Natural Language Processing; Natural Language Generation; Biomedical domain; Adverse Drug Effects

Domaines: Informatique et langage [cs.CL]; Intelligence artificielle [cs.AI]; Traitement du texte et du document

Tomohiro Nishiyama, Lisa Raithel, Roland Roller, Pierre Zweigenbaum, Eiji Aramaki. Assessing Authenticity and Anonymity of Synthetic User-generated Content in the Medical Domain. Workshop on Computational Approaches to Language Data Pseudonymization (CALD-pseudo), Mar 2024, St. Julian’s, Malta. pp.8-17. ⟨hal-04528240⟩