Studying the issue communication of political parties through the analysis of texts has a longstanding tradition in political science. Most of the research focuses on a single type of text (e.g. manifestos). Parties, however, communicate through various different communication channels simultaneously. This aspect is not reflected in research as comparative studies of different types of text are rare.

Computer-based text analysis offers several opportunities to drive such a comparative study of political texts forward. Nowadays the computational toolbox contains a wealth of methods and approaches, with each of them offering benefits but also limitations. Furthermore, the different types of political texts provide specific challenges for the application of computer-based tools.

This study provides an examination of computer-based tools for the comparative study of party texts. We apply unsupervised (LDA), semisupervised (Newsmap) and supervised (BERT) approaches to different types of party texts. These approaches are used to perform an issue classification task. The issue categories are based on an adapted codebook from the Manifesto Project (MARPOR). We analyse manifestos, press releases, parliamentary speeches and tweets from political parties in Austria, Germany and Switzerland.

The performance of the applied approaches is evaluated against a "gold standard" in the form of manual coding. The results of this analysis will show the applicability of text analysis approaches to study different types of party communication. This should equip researchers with knowledge about available tools for the comparative study of party communication channels as well as about the opportunities and limitations coming with each of them.