Texts have always been an important empirical resource for social scientists studying politics. In recent years, it has become an increasingly useful resource as large datasets of textual data have become available to the scholarly community, and as new technologies emerge that allow researchers to study this data. This project will use recent data collection efforts, and new methodologies developed in the field of Natural Language Processing (NLP) and Machine Learning, to study Swedish politics. Specifically, the project will use large language models, word embeddings and machine learning technologies to study polarization and representation in Sweden. The project will analyze a large number of texts, including parliamentary speeches and public social media posts. Some of the specific questions that will be analyzed is how geographical representation in the Swedish parliament has developed, and factors contributing to the polarization of the political debate.