A German Corpus on Topic Classification and Success of Social Media Posts from Facebook

Abstract

We provide a corpus consisting of 6,000 posts of German food delivery services from five brand pages on the online social network Facebook. The brand pages include Call a Pizza, Deliveroo, Domino’s, Lieferando, Mundfein and Smiley’s. A group of social media marketing experts annotated each post with one or more topic labels from eleven marketing related categories describing its content. Additionally an assessment on the success of the social media post is provided as a binary label. The inter- rater reliability over all annotators according to Fleiss’ Kappa is 0.4835 for the topics and 0.6674 for success. Furthermore, baseline measurements with machine learning based text classification with an F1-score up to 0.7173 are presented as a first experiment on this new corpus. The data set of the corpus on German topic classification and success (GTCS6k) is publicly available here: https://ccwi.github.io/corpus-gtcs6k

Keywords

• topic classification
• machine learning
• natural language processing
• facebook
• online social network
• social media
• inter-rater reliability
• corpus
• data set