i3 seminar “Analysis of online participation” will welcome
Mathieu Brugidou and Philippe Suignard
What can word embedding algorithms do
for the sociological analysis of texts?
Mathieu Brugidou is a senior researcher at EDF R&D, Energy Technology and Society Research Group, and a research associate at the Pacte laboratory (UMR 5194), CNRS-Université Grenoble Alpes- Sciences Po Grenoble.
Philippe Suignard is an Engineer and Research Expert at the ICAME Department (Business Innovation, Market and Environmental Analysis), EDF R&D.
Papers to read in preparation for the session:
Brugidou Mathieu, Suignard Philippe, Escoffier Caroline, et Charaudeau Lou (2020). Un discours et un public “Gilets Jaunes” au cœur du Grand Débat National ? Combinaison des approches IA et textométriques pour l’analyse de discours des plateformes “Grand Débat National” et “Vrai débat”. JADT 2020. https://halshs.archives-ouvertes.fr/halshs-03132818/document
Suignard Philippe, Escoffier Caroline, Charaudeau Lou et Brugidou Mathieu. Que peuvent les algorithmes de plongement de mot pour l’analyse sociologique des textes ? Analyser les discours et caractériser les locuteurs des plateformes “Grand Débat National” et “Vrai débat”. Paper to appear in Statistique et Société.
Abstract: In this presentation, we propose to further the evaluation of the contribution of so-called “word embedding” algorithms to the sociological analysis of texts: on the one hand, by confronting the results of the semantic analyses of these algorithms to the now well-known approaches of textual data analysis or textometry; on the other hand, by focusing on what is one of the main obstacles to the sociological analysis of the web: the difficulty to characterize sociologically the authors of statements from the web. To this end, we analyze statements from “civic tech” platforms – i.e. the “Grand Débat National” [Great National Debate], a governmental platform, and its political and algorithmic riposte provided by a collective of Gilets jaunes, the “Vrai Débat” [The Real Debate]. A third corpus from the “Entendre la France” [Hearing France] platform, with the same design as the “Grand Débat National” and documented in terms of socio-political properties, allows us to characterize the speakers according to their discourses and to attempt to predict, through machine learning approaches, “pseudo-properties” assigned to the contributors of the “Grand Débat National”.
i3 seminar “Analysis of online participation”
The “Analysis of Online Participation” (APeL) seminar aims to develop reflection on approaches to the use of data about the online usage and participation. The seminar is a place for researchers from various backgrounds to meet and exchange knowledge, practices and know-how. At each session, the seminar guests present and discuss the “making of” of one of their researches: construction of the survey, specificities of the data engineering implemented, tools for capturing / building corpora, analysis strategies and tools used, etc.
Information and registration
The seminar is open to all. Please register here to participate in this session.