🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

#ParlSpeech update now available: Jan Schwalbach and I can provide more than 6.3 million parliamentary speeches from nine representative democracies, covering periods between 21 and 32 years. #TextAnalysis #TextAsData (1/3)

Video of my first hour with a Kaggle dataset: youtube.com/watch?v=S6GVXk… #python #programming

Something for data gourmets: You can now download the government declaration dataset, a classic that was originally published with MPP I (on a CD!). If you are interested in policy preferences of governments you should definitely take a look: manifesto-project.wzb.eu/information/do…

This is mind blowing. With GPT-3, I built a layout generator where you just describe any layout you want, and it generates the JSX code for you. W H A T

New update of the Manifesto Dataset (2020a)! First Highlight: We added 9 elections to the Main Dataset: BIH 2018 CHE 2019 ESP 2019(April) FIN 2019 GBR 2019 MEX 2018 POL 2015 SVN 2018 UKR 2012 Also included are the already coded Flemish manifestos of the Belgium 2019 election.

New blogpost on #Trumps agenda for #USElection2020. Tobias Burst, Pola Lehmann, Sven Regel, Bernhard Weßels & Lisa Zehnter (all Manifesto Project) analyze the presidents current agenda compared to the Republican manifesto from 2016. --> bit.ly/3oLyhUz WZB

Text generation from decoder models like GPT is a wildly inefficient way to do zero-shot classification. It's not cost effective at scale, and these models have a whole host of pathologies that affect results in ways we don't understand. 1/n