/sa_election

Analysis of social media posts related to the May 29, 2024 South Africa elections.

Primary LanguageHTML

Social media analysis of May 29 South African elections using LLM-driven frame extraction

Analysis of social media posts related to the May 29, 2024 South Africa elections.

Time series of frames expressed in social media posts

Below is a list of dates for each of several topics. For each topic, every day we collect the tweets that reference that topic. Then we feed each tweet to an AI pipeline that identifies any and all claims of broad social significance expressed in the tweet – which we call “frames”. E.g., “Ready to vote to keep Ramaphosa at the helm where he belongs” expresses the frame “Ramaphosa should win reelection”, whereas “Whoa, just served Ramaphosa’s son in the Starbucks I work in” does not express any frames. Once we have the frames expressed by each tweet, we group them together into clusters that all express variants of roughly the same frame. Then we track the activity of these clusters over time, to see what frames are being discussed online. The link for each date goes to a page with plots showing topic activity for that date, as well as for the week preceding that date. For each topic, a separate page provides a list of all clusters of frames found for that topic.

AfricanNationalCongress (Cluster labels)

DemocraticAlliance (Cluster labels)

EconomicFreedomFighters (Cluster labels)

IndependentElectoralCommission (Cluster labels)