“SocioFillmore” (named after the American linguist Charles Fillmore) formed the core of my PhD project. Together with my collaborators and supervisors (Sara Gemelli, Malvina Nissim, Tommaso Caselli, Chiara Zanchi & others) we investigated how computational applications of Frame Semantics can help us understand, from a critical perspective, how the same event can be perspectivized (for example in the media) in different ways and with different socio-political meanings.
Much of the project centered on femicide: what does it mean when a newspaper writes that “a family drama occurred” instead of “a man murdered his wife”? How are these constructions used, how can we detect them, and how do readers perceive these constructions?
project deliverables
Some of the main outcomes of the project come in the form of software for organizing & analyzing newspaper corpora. I’m currently in the process of improving the documentation and accessibility of this software, but the main demo is on this huggingface page.
media appearances
SocioFillmore was featured in the media several times, including in the Netherlands:
- NRC: ‘Krantenkoppen schetsen hoe er over een moord gedacht wordt’ [original] [archived]
[“‘Newspaper headlines draw an image of how a murder is talked about‘”] - WomenInc Podcast: #32 – Schrijven over femicide [spotify] [apple]
[“Writing about femicide“]
… and in Italy:
- Corriere della Sera: Femminicidio: «morta» o «uccisa»? Un algoritmo calcola (l’enorme) differenza [original] [archived]
[“Femicide: ‘dead’ or ‘murdered’? An algorithm calculates the (enormous) difference”] - Treccani: L’algoritmo, il femminicidio e il giornalismo [original] [archived]
[“Algorithms, femicide and journalism”] - Ansa.it: L’IA impara a prevedere l’interpretazione delle notizie [original] [archived]
[“AI learns to predict the interpretation of news articles“]
future of the project
SocioFillmore is currently (at least from my side) semi-on-pause, but the project will continue in one way or the other! In particular, I would be very interested in working on continuously monitoring news about femicide: creating information centers that are updated with information about new cases (similar to e.g. the NonUnaDiMeno observatory) coupled with news articles about these. This can help make femicide less invisible, and also provide insights into how discourse on femicide is developing in different places.
A tiny proof-of-concept of this is www.sociofillmore.nl, which currently hosts a PeARS instances that indexes recent Italian news articles about femicide along with a rudimentary frame annotation.
My dream for the hopefully not-too-distant future would be to work on expanding this to Dutch news, automatically index new articles, group together related cases and link to structured data where available, and improve the interface and frame visualizations.