The GermaParl Corpus of Parliamentary Protocols

The GermaParl Corpus of Parliamentary Protocols Andreas Bl¨atte, Andre Blessing University of Duisburg-Essen, University of Stuttgart andreas.blaette@uni-due.de, andre.blessing@ims.uni-stuttgart.de Abstract This paper introduces the GermaParl Corpus. We outline available data, the data preparation process for preparing corpora of parliamentary debates and the tools we used to obtain hand-coded annotations that serve as training data for classifying debates. Beyond introducing a resource that is valuable for research, we share experiences and best practices for preparing corpora of plenary protocols. Keywords: corpus creation, annotation, R, TEI, Digital Humanities 1. Introduction Parliamentary debates convey the arguments, interpreta- tions and disputes that shape political decision-making. They are recorded and transcribed by parliamentary admin- istrations with diligence and are published as plenary pro- tocols. These documents are available for long periods of time – for several decades and more – and they cover the full range of issues relevant to a political system. Plenary protocols are an indispensable resource for anybody inter- ested in the politics and policies of a democratic polity. If citizens’ access to plenary protocols is impeded, the norm of democratic transparency is violated. A corpus of ple- nary protocols is not just a language resource, it is a crucial building block of the public digital archive of democracy. The value of plenary protocols for research goes beyond disciplines genuinely interested in the substance of parlia- mentary activity. These documents are a great resource for studying the variation of language across political domains, and language change in time. Given the amount of data, a corpus of plenary protocols is a good basis for testing all kinds of algorithms. Social scientists, computational and corpus linguists, as well as data scientists may use plenary protocols productively in their research. There is a legal advantage to plenary protocols that deserves to be mentioned: There are no substantial legal barriers for using, processing and re-using of these documents. Restric- tive licensing conditions that inhibit working with various kinds of media (including social media) do not arise with plenary protocols. Corpora of plenary protocols that have been prepared can be shared, which is an essential basis for attaining the ideal of reproducible research. And as plenary protocols are open data, these documents are an outstand- ing resource for teaching purposes. The digital availability of plenary protocols is excellent and poor at the same time. Documents can be downloaded with- out technical or legal restrictions as txt, html or pdf docu- ments. A minimally annotated XML may be available “off the shelf”. But these data formats do not yet correspond to the requirements for digital-era data processing. To sub- stantively exploit the analytical potential of the data, origi- nal documents need to be converted into a semi-structured data format (XML) with a sufficiently fine-grained markup. Most importantly, speakers and their affiliation to parlia- mentary groups and parties need to be annotated to attain a useful resource. The GermaParl corpus as it has been pre- pared in the PolMine Project 1 implements this notion and is based on an XMLification …

Content truncated to 500 words