Online Image Analysis to Classify Activity of Politicians in Twitter Data (Working title) (Bachelor Thesis, Ongoing)
Social media channels such as Twitter have gained huge importance in political processes such as votes and elections over the past few years. Consequently, Twitter has become a treasure trove of data and information that can be processed and analysed to try to answer research questions in political sciences and similar areas.
This project approaches one such question in a proof-of-concept manner. The high-level idea is to gather information on how politicians and candidates for public office present themselves in pictures on social media during elections. For this purpose, we will work with a dataset of Tweets collected during the 2019 Swiss Federal Elections (“National- und Ständeratswahlen”) between July and October 2019. Naturally, we are mainly interested in the visual content contained in those Tweets.
The goal of the project is twofold. In a first step, we will try to detect the presence of known politicians or candidates in such images using state-of-the art facial recognition algorithms. In a second step, and once such an image has been identified, we’ll try to classify the images in terms of objects or concepts so as to gather information on its more general context. It will be up to the candidate to identify potential classification and/or summarization techniques.
Start / End Dates
2020/10/26 - 2021/01/25