Since the launch of social media platforms politicians and parties are provided with an inexpensive tool for direct communication with voters. The large user base of the platforms produce an immense amount of unstructured data which has come to interest researchers as well as businesses. Machine learning algorithms have enabled an effective way of extracting meaningful information from such data, e.g. textdata, called text mining. Researchers have studied how, for example, Twitter can play a role in elections and election campaigns. However, it is still a rather unexplored area and a limited number of studies have been conducted in Sweden. In this thesis twitter data from the eight parties in parliament is examined using the unsupervised learning method hierarchical clustering. The aim is to explore what political questions the different parties are publishing about on twitter by looking at what parties are clustered together and the most important and frequent words for each cluster. First, seven clusters are decided on and discussed, followed by an expansion to eighteen clusters. The results show that for both seven and eighteen clusters one of the clusters is substantially larger than the others and there are no patterns of what parties are clustered together, yet it was possible to demonstrate what cluster most of a party's tweets belonged to.