The Twitter Bot Vanellus

The Vanellus (zool. peewit) project was started in April 2018. This work was inspired by news about Cambridge Analytica. My reaction to these news was "this can't be so difficult, lets try it yourself" (I still think so).
The bot downloads all German tweets according the search pattern "lang:de". These are about 1 million tweets/day. Technically one can download tweets arbitrarily. German tweets are selected because this is the upper limit of tweets one can handle on a personal account with a plain PC. Austrian-German is my native language and I have some knowledge about the cultural and political context.
Vanellus is written in Java and uses the twitter4J library. The tweets are stored in real-time in a very efficient custom database. On top of this database is a powerful query-language and a host of analysis tools.
So far the Kiebitz enjoys his silence and keeps - on Twitter - his knowledge by itself. The papers below describe the functionality of Vanellus and present some of the analytic results.
If you are interested in the progress of the project, send me an Email and I put you on the distribution list for forthcoming papers. The papers are so far in German, English versions are coming soon. Alternatively keep an eye on this page.

  1. Die automatische Identifizierung von AfD Anhänger auf Twitter: Der Vanellus-Follower-Classification-Index VaFCI. (In German).
    This paper describes the fully automatic classification of AfD Followers with the Vanellus-Follower-Classification-Index VaFCI. The VaFCI misses sometimes an AfD-supporter. But a user classified as supporter is almost always indeed a fan of the AfD.
    The full set of results can be downloaded from:
    Table of all active Followers
    Table of all active AfD-Supporters

  2. In German: Wie misst man den Einflusses von MdBs auf Twitter? Der Vanellus-Twitter-Index VaTI Version 1.1.
    English version: Measuring the influence of German Members of Parliament on Twitter: The Vanellus-Twitter-Index VaTI.
    This paper describes the Vanellus-Twitter-Index VaTI. The VaTI is applied to all German Members of Parliament (MdB) and to the heads of the Austrian political parties. VaTI considers the actual footprint of users on Twitter. The VaTI is considerable more informative than the plain "number of followers".
    The German Version 1.1 and the English Version are extended with the July 2018 results. The full data-set of this investigation - including a list of the Twitter-Accounts of all MdBs - can be downloaded from VaTI Data-Archive
    VaTI MdBs, July 2018
    VaTI Austrian Party Leaders, July 2018
    The actual VaTI can be found on the VaTI page.

  3. Highland Analytica oder der Twitter-Bot Vanellus(in German).
    This paper was first published in the May-2018 issue of the popular-science publication Chrillys Goldpreis Report. An interesting new feature is the automatic classification of users in supporters and opponents of the new right-wing party AfD. The classification can be used for other classifications like fans of FC Bayern München. But it's principal working was demonstrated for this political party.

  4. Der Twitterbot Vanellus zur vollständigen Analyse deutschsprachiger Tweets (in German).
    This paper describes the state of the art of version V0.29. V0.29 was a considerable progress in relation to V0.18. Since this version all German tweets are downloaded in real-time and an efficient custom database was build. A query language for retrieving tweets from this data collection was implemented and some mechanism for finding bots and bot-gangs were implemented.

  5. Vanellus V0.18 (in German).
    This paper documents the first - very rudimentary - version of Vanellus. It describes the general philosophy of the project, the programming tools and some first results. Vanellus V0.18 did not read the full set of German tweets.