500,000 Tweets scanned
2,000+ assigned manually

Prof. Dr. Achim Rettinger’s research team is monitoring social media activity throughout the study using data from the microblogging platform Twitter, which is compared with media coverage data provided by the Event Registry media monitoring service. Over the entire study period, the team is collecting data on six aspects that correspond to the focal points of the representative survey. This is done using predefined search queries via the Twitter API, with a focus on statements related to the economic situation, migration, support for Ukraine, trust in the political system, media coverage, and the energy transition.

The aim of social media monitoring is to measure changes in attitudes (social sentiment) towards these aspects and compare them with the panel survey data.

To achieve this goal, a sub-corpus of the collected data is annotated manually according to the six aspects studied and the sentiment polarity of the respective tweet (positive / neutral / negative). This annotation is carried out collaboratively by multiple annotators using a specially developed tool in compliance with the Inter-Annotator Agreement. The current database, as of early March 2023, consists of 500,000 tweets, of which 2,000 have already been manually assigned to the aspects under investigation.

Using this data, the team can train classification models using pre-trained, transformer-based language models (bert-base-german-cased), which will be continuously improved through manual annotation and eventually support the annotation of the entire corpus. Based on the partly manually and partly mechanically annotated corpus, the collected sentiment polarities of the individual aspects will be compared with “events” of the media monitoring (events of the day) and the results of the panel survey.


We use cookies on our website to give you the most relevant experience by remembering your preferences and repeat visits. By clicking “Accept”, you consent to the use of ALL the cookies.

Decline Accept