Commemorating the Great War
on Twitter

Frédéric Clavert, Asst Professor, C2DH

Institute for Historical Research, 7 January 2020

Introduction
the #ww1 project

The WW1 Centenary and social media

The Centenary as the first large-scale commemoration in the era of social media

The #ww1 project: harvesting data

  • Twitter data only
    • Started 1st April 2014
    • Ended 1st December 2019

⇒ One 11th November outside the Centenary

The ww1 project: the state of the database

  • 9,1 millions tweets
    • 2/3 of RTs - 1/3 of original tweets
  • 1 million users

Top 1000 hashtags
Top 1000 hashtags

  1. Methodology
  2. Global overview
  3. Unoriginal content, new practices?

Methodology

Why Twitter?

Because we can.

  • The Twitter Streaming API is relatively open
  • Pieces of Software exist (analysis / harvesting),
    relatively easy to use

The technical dispositif

  • PHP streaming API scripts: 140dev (2014-september 2017) / DMI-TCAT (September 2017-December 2019)
  • «in-between» tools: basic text editors / spreadsheet software / OpenRefine / Dataiku DSS
  • Analysis tools: IRaMuTeQ / Gephi

The technical dispositif: IRaMuTeQ

Hierachical clustering

  • unsupervised
  • good old statistical method (1983 / French school of data anlysis)
  • clustering of segment of text and not words (in case of tweets: texts = segments of text = tweets)
Reinert Max, « Les “mondes lexicaux” et leur “logique” à travers l’analyse statistique d’un corpus de récits de cauchemars », Langage et société 66 (1), 1993, pp. 5‑39. En ligne: https://doi.org/10.3406/lsoc.1993.2632; Ratinaud Pierre et Dejean S., « IRaMuTeQ : implémentation de la méthode ALCESTE d’analyse de texte dans un logiciel libre. », in: Modélisation Appliquée aux Sciences Humaines et Sociales, Toulouse, 2009. En ligne.

Questionning the corpus

  • Where are Jean Jaurès, Georges Clémenceau and the battles of the Marne?
  • the #lestweforget issue

Global overview

General temporality

Number of tweets per day
Number of tweets per day

Linguistic temporality

Number of tweets per day per language without RTs
Number of tweets per day per language without RTs

French corpus clustering…

Hierachical clustering (French corpus)
Hierachical clustering (French corpus)

…and its temporality

Hierachical clustering (French corpus) through time
Hierachical clustering (French corpus) through time

English corpus clustering…

Hierachical clustering (English corpus)
Hierachical clustering (English corpus)

…and its temporality

Hierachical clustering (English corpus) through time
Hierachical clustering (English corpus) through time

Unoriginal contents, new practices?

Amteur historians: the #1j1p case

Controversies: the commemoration of the battle of Verdun

French historians on Twitter

Conclusion

General conclusions

Digital bricolage

The allure of the archive in the digital era

Bibliography

Pictures

  • Monument aux Morts, Place de la République Strasbourg