And next up at the Social Media Access Days at the German National Library are Marco Wähner and Jan Dennis Gumz, exploring the further use of Wikipedia data on the early German federal election in 2025. Because of the unusual circumstances of the election, following the failure of the governing coalition, there was an increased need for information about the election amongst voters, and Wikipedia (as the only public-interest Very Large Online Platform classified by the EU) played an important role here.
But as a collaboratively edited online platform, Wikipedia represents a particularly special information ecosystem; editing activity here also points to key events during the election campaign. Data from the platform are provided under a Creative Commons licence and are therefore available for reuse, including by researchers; they are available via Wikipedia’s APIs. Such data contain information about access, editing, external and internal linking practices, page content, and other data points.
The Wikipedia API for the German-language Wikipedia edition was queried for such relevant information for the period from the end of the coalition to the day before the election, and data were made available via the Wiki4trends dashboard; this was also compared to a period of equivalent length before the end of the coalition. The project gathered data for the federal election page itself, and for pages one and two internal links further on from that page.
This process might overlook relevant content which was not linked to these initial articles, as well as overgather content which was linked but not relevant to the election context. Additional, various gathering errors might occur during the process, and lead to data losses.
A first analysis of the dataset shows a substantial increase in accesses to some such pages, both for mainstream content and obscure parties and individuals; some 219 new articles were created, and this includes especially also pages about key politicians. Some 15,000 Wikipedians edited these pages, and these include especially pages on terror attacks and other incidents, the monthly Nekrolog pages, and various others.
The ‘signal to noise’ ratio for this collection of data (where ‘signal’ means relevance to the election, and ‘noise’ relevance to everything else Wikipedia users might be interested in) is also interesting here – although of course ‘noise’ is very much in the eye of the beholder in this case.
