Currently, topic modelling is an effective analytical tool for the automated investigation of text data. However, identifying the underlying topics is still a challenging task that is dependent on the selection of the proper technique. Moreover, due to the considerable number of topic modelling techniques reported
in the literature, uncertainty about the application of the techniques arises for both researchers and practitioners. Therefore, we conducted a comparison of three different topic modelling techniques (LDA, PAM, DMR) to give recommendations for three use cases identified in the literature: content extraction, trend analysis and content structuring. For each of them, we identified several requirements and by conducting the method ‘Goal Question Metric’, we derived several comparison metrics. We applied these metrics to a real-world Facebook data set (4,155,992 posts) to highlight the differences between the three topic modelling techniques and to give recommendations for our defined use cases.