Politics, Morals and Tweets: Identifying Political Affiliation Utilizing Moral Foundations Theory and Contextual Embeddings

This project seeks to extract moral preferences from US politicians' tweets and predict their political ideology based on these moral preferences. The underlying theoretical framework is the Moral Foundations Theory.

---

Notes

• Dependencies can be found in environment.yml.
• Starting BERT server for inferring seed phrases for moral values (i. e. we want one embedding vector for the whole phrase): bert-serving-start -model_dir ~/Development/data/BERT/base/ -num_worker=1 -cpu
• Starting BERT server for inferring words in tweets (i. e. we want one embedding vector per token): bert-serving-start -model_dir ~/Development/data/BERT/base/ -num_worker=1 -cpu -max_seq_len=40 -pooling_strategy NONE
• Created for the course "Social Media Data: Quantitative Text Analysis of Big Data", University of Vienna, Winter Semester 2018/2019.