Salomé Do
Welcome to my page!
Home • Teaching
Current work
I am Salomé Do, a PhD student working at LATTICE (Ecole Normale Supérieure) and médialab (Sciences Po). I’m working at the crossroads between Natural Language Processing, Mass Media and Political Communication.
The aim of my dissertation is to explore and assess computational methods for news framing analysis. More specifically, I investigate how supervised machine learning can help automate some parts of content analysis methods, and the requirements needed for SML to work. For instance, SML can automate the coding of simple lexical indicators, but up to which level of linguistic abstractness can we push the algorithms? How much manual coding do we really need for SML algorithms to be able to learn to automate the task? What should be the quality of this manual coding? Is it better to have crowdsourcers, or a few expert coders? Ultimately, my aim is also to question the utility of using such methods: when is SML really needed, and when can we just use manual content analysis methods?
Through this methodological lens, news framing analysis is a particularly interesting object to study. The definition of frames and framing is still debated, and very difficult to operationalize. It draws on concepts in linguistic, rhetorics, stylistics, discourse analysis, etc., and involves multiple facets of language. Meanwhile, its effects on how we perceive the world are fascinating.
A joint project with Etienne Ollion and Rubing Shen uses SML to try to understand the transformation of strategic news framing in french written press since 1950.
UPDATE : Article accepted in Sociological Methods and Research:
Some of my past projects also include:
- A project on Bruno Latour’s questionnaire on world-to-come after Covid-19
- A project on denial, delation schemes and controversies during Covid-19 crisis in France (Published in Le monde d’aujourd’hui. Les sciences sociales au temps de la Covid., Presses de Sciences Po)
- A contribution to the project ” Mapping AI issues in media through NLP methods” (accepted at CHR2021)
Teaching
Apart from that, I :
- Worked as a teaching assistant for Xavier Dupré’s Python for data scientist course at ENSAE for 2 years. I wrote some introductory material for my students, that you can find on my teaching page
- Taught a course named “Learning text representation for Social Sciences” at Université Gustave Eiffel. A short syllabus can be found here
- Worked as a teaching assistant for the ML for NLP course at ENSAE.
- Taught an introductory course Learning Text Representations for Social Sciences for Etienne Ollion at Ecole Polytechnique
Previous work
On previous projects during my master’s degree at ENSAE IP Paris, I created biomedical named entity recognition systems for a startup called Posos.