H-TFIDF: What makes areas specific over time in the massive flow of tweets related to the covid pandemic?

Decoupes Rémy, Kafando Rodrique, Roche Mathieu, Teisseire Maguelonne. 2021. H-TFIDF: What makes areas specific over time in the massive flow of tweets related to the covid pandemic?. In : Proceedings of the 24th AGILE Conference on Geographic Information Science, 2021 - Volume 2. Partsinevelos Panagiotis (ed.) , Kyriakidis Phaedon (ed.), Kavouras Marinos (ed.). Technical University of Crete, AGILE. Göttingen : Copernicus Publications, 1-8. (AGILE: GIScience Series, 2) AGILE Conference on Geographic Information Science (AGILE 2021). 24, s.l., 8 Juin 2021/11 Juin 2021.

https://doi.org/10.5194/agile-giss-2-2-2021

Communication avec actes

Prévisualisation

Version publiée - Anglais
Sous licence

.
agile-giss-2-2-2021.pdf
Télécharger (4MB) | Prévisualisation

Url - jeu de données - Entrepôt autre : https://www.kaggle.com/smid80/coronavirus-covid19-tweets-early-april / Url - autres données associées : https://github.com/echen102/COVID-19-TweetIDs/

Note générale : Le congrès s'est tenu en ligne

Résumé : Data produced by social networks may contain weak signals of possible epidemic outbreaks. In this paper, we focus on Twitter data during the waiting period before the appearance of COVID-19 first cases outside China. Among the huge flow of tweets that reflects a global growing concern in all countries, we propose to analyze such data with an adaptation of the TF-IDF measure. It allows the users to extract the discriminant vocabularies used across time and space. The results are then discussed to show how the specific spatio-temporal anchoring of the extracted terms make it possible to follow the crisis dynamics on different scales of time and space.

Mots-clés libres : Text Mining, Spatio-temporal analysis, Social network analysis, Tweets, One Health, COVID-19

Auteurs et affiliations