ADAPT Researcher Develops Political Twitterbot using Machine Learning

21 January 2021
ADAPT Researcher Develops Political Twitterbot using Machine Learning

Posted: 10/03/16

In antiticipation of the 2016 US Presidential elections, a new twitterbot has been created using artificial intelligence to generate tweets that mimic those some politicians might normally post. Created by Dr Killian Levacher, Research Scientist at the ADAPT Centre for Digital Content Technology at Trinity College Dublin, the ‘Trump Deep BS Quote Generator’ uses quotes harvested from select politicans’ official Twitter accounts to generate mimiced tweets.

Speaking about the project Dr Levacher said: “This project uses artificial intelligence (AI) Deep Learning techniques to learn the twitter user’s language patterns. The robot has been trained with official quotes spoken by political users that are fed into the system from the Twitter account. The training model finds natural language patterns and the algorithm then generates different combinations of the data that reflect what it has been taught in order to produce realistic mimiced quotes.”

The robot has so far created quotes such as “Jeb Bush is a sad sack”, “Republicans are cutting up their Macy’s credit card”, “Sanders is a dopey-clown”, “People know NOTHING”, and “Bill Clinton did always thing for debate like a rock”.  The system gradually learns character per character how to generate words and punctuation, eventually using the politician’s twitter vocabulary in order to formulate and express near-indentical sentences. Although not always clear, the language bears a significant resemblance to the politicians’ twitter public outpourings.

The political twitterbot will continue to generate quotes, some of which may be inappropriate and insensitive due to politician’s tweetable discourse, throughout the campaign on Twitter @TrumpBSQuoteGen.  As it progresses the project will be documented on Dr Levacher’s blog.

Share this article: