Skip to content

Latest commit

 

History

History
45 lines (34 loc) · 4.15 KB

README.md

File metadata and controls

45 lines (34 loc) · 4.15 KB
Raw

From the fringes to the core

An analysis of right-wing populists’ linking practices in seven EU parliaments and Switzerland

This repository provides some data and scripts related to the paper:

For bug reports, comments and questions please use the issue tracker.

Related Software

  • tosca is used for managing and manipulating the text data to a structure requested by ldaPrototype.
  • ldaPrototype is used to determine a prototype from a number of runs of Latent Dirichlet Allocation.
  • ineq is used to calcaulate Gini coefficients.
  • longurl is used for expanding short urls and urltools is useful for extracting url cores from urls.
  • RCurl and RJSONIO are used for scraping.
  • batchtools is used for calculating (prototypes of) LDAs on the High Performace Compute Cluster LiDO3.
  • data.table is used for managing and storing tabular data, e.g. texts and meta information.
  • tm, lubridate and utf8 are useful packages for preprocessing and managing text data.
  • ggplot2, ggpubr, ggrepel and cividis are used to create the plots.
  • beanplot is used to visualize advanced boxplots.

Usage

Please note: For legal reasons the repository cannot provide all data. Please let us know if you feel that there is anything missing that we could add.

In the code folder you can view and trace the chronology of the R code used.

The folder countries contains the following structure for all examined countries:

  • the subfolder docs contains the used preprocessed texts (bag of words with indices of vocabulary - see vocab.txt/.rds),
  • the subfolder proto contains for all considered values K=20,25, ..., 75 the LDAPrototype models,
  • the subfolder tables contains some descriptive statistics in tabular form,
  • docs.rds contains the preprocessed texts in the form that can be used with the ldaPrototype package,
  • onepercent.txt specifies the parties of the country that would pass an artificial 1% hurdle,
  • parties.csv and parties_col.csv contain information about party abbreviations and used colors for the party,
  • topwords30.csv contains the 50 topwords for all K=30 topics of the corresponding LDAPrototype model,
  • vocab.txt contains the corresponding vocabulary, vocab.rds as RDS file (see docs.rds).

The folder countries/incl_UK also contains the (differing) results including the results of United Kingdom, which was omitted from the paper for interpretational reasons.

The misc folder contains various summary tables, e.g. in party_names.csv all party abbreviations are given - sorted by country - and in party_names2.csv sorted by abbreviation.

At last the pdf folder contains three PDFs:

  • statistics_general.pdf gives an insight into descriptive statistics concerning the parliament and the raw data set (see also code/3statistics_general.Rmd),
  • create_textmeta.pdf gives the process and statistics of the preprocessing (see also code/4create_textmeta.Rmd),
  • plots.pdf contains some descriptive plots and further analytical plots based on the LDA results.