Analysis of a Citizen Consultation "Covid-19 Crisis: How to invent together the world of tomorrow?"

The unprecedented crisis we are facing could be a new signal of the urgency to change our societal models. It is from this observation, and because structuring decisions for our common future are already under discussion, that the French Red Cross and WWF France are partnering with Make.org and the SOS Group, in collaboration with Unis-Cité and the UP Movement, to invite you to answer this crucial question: “Covid-19 Crisis: How can we all together invent the world of tomorrow?”.

This project aims to collect proposals and concerns from French citizens on the Make.org platform and apply Natural Language Processing (NLP) methods to automatically group them into themes (Topic modeling). The themes are varied and include questions about health, employment, environment, etc. This work was previously done manually by administrators, so training an NLP model to predict the themes related to the proposals constitutes a considerable time-saving for the platform. We also used vectorization methods (Word to Vec and Doc to Vec) as well as machine learning techniques (the K Nearest Neighbors) to create groups of proposals that are as close as possible and compared this to the results of topic modeling. Overall, the results obtained were consistent and coherent. This project was carried out using Python as programming language and the NLTK along with scikit-learn packages.

This project was conducted as part of the statistical project at ENSAI and was co-authored by Véronique Carelle and Batourou Sano.