Analyzing US Election troll tweets with Datasette
6th August 2018
FiveThirtyEight published nearly 3 million tweets from accounts associated with the Russian “Internet Research Agency”, based on research by Darren Linvill and Patrick Warren at at Clemson University.
FiveThirtyEight’s tweets were shared as CSV, so I’ve used my csvs-to-sqlite tool to convert them and used Datasette to publish them in a searchable, browsable interface: https://russian-troll-tweets.datasettes.com/
The data is most interesting if you apply faceting. Here’s the full set of tweets faceted by author, language, region, post type and account category:
The minimal source code for this Datasette instance is on GitHub.
More recent articles
- An Introduction to Google’s Approach to AI Agent Security - 15th June 2025
- Design Patterns for Securing LLM Agents against Prompt Injections - 13th June 2025
- Comma v0.1 1T and 2T - 7B LLMs trained on openly licensed text - 7th June 2025