Masterarbeit Max Hampel: Evaluating the Importance of Profile Descriptions for the Sentiment Analysis of Tweets in Protest Movements

Abstract: Plenty of research has been done to analyze and classify the sentiment (i.e., the positive, neutral or negative stance) of social media posts during crises, conflicts, and protests using machine learning techniques. However, fewer works have examined the impact of the user profile description on the outcome of sentiment analysis for different post types, including tweets, quotes and replies. Using the case of the occupation and evacuation of Lützerath, this thesis presents the collection and annotation of a respective Twitter/X data set, the model creation for sentiment analysis based on post content and user profile descriptions, as well as its evaluation using traditional machine learning methods (e.g., Random Forest, SVM) and more recent language models (e.g., DistillBERT, RoBERTa). Since the thesis is still a work in progress, evaluation findings and implications will be disclosed in the presentation announced in this mail.

Betreuer: Dr. Marc-André Kaufhold

Prüfer: Prof. Dr. Dr. Christian Reuter