the data is generated in analysis-btw17-candidates-topics
data source is abgeordnetenwatch.de (and google "Kandidaten in der Google Suche" for the top candidates)
the candidate cloud is based on 8 abstract topics, that were found using NMF (nonzero matrix factorization). The top words for each candidate are their tf-idf (term frequency - inverse document frequency) scores.
The topic filters (for the five topics "arbeit", "natur", ...) are based on manually selected wordlists. the sources for the wordlists are in analysis-btw17-candidates-topics.
Installation:
npm install
Start Development Server:
npm run dev
Build Production Version:
npm run build