Dank Memes is one of the tasks selected for the 7th edition of EVALITA, the contest of Natural Language Processing and Speech for Italian. The final workshop of the evaluation campaign will be held on December 3rd in Bologna, as a co-located event of the Italian Conference on Computational Linguistics – CLiC-it 2020 (Nov 30th – Dec 2rd 2020).
Internet memes established themselves as widespread visual framing devices, playing a key role in influencing users’ opinions and imaginary, as they are pervasively employed on social media to comment on current events. Their ubiquity increases the necessity to further a multimodal approach of analysis, considering textual and visual cues together and not as separate entities. This task will provide 2000 images, among which 1000 political memes. For every image will be provided text transcription, image embeddings, and manual annotation such as offensive language and formal characteristics (i.e. if the meme derives from a standardized macro), to stimulate the identification of meme instances and of potentially offensive content.
The task is organized by Guido Anselmi (University of Milan), Giulia Giorgi (NASP, University of Milan, University of Turin), Gianluca E. Lebani (Ca’ Foscari University of Venice), Martina Miliani (University of Pisa, University for Foreigners of Siena), Ilir Rama (NASP, University of Milan, University of Turin).