Sentiment Analysis of a German Twitter-Corpus

The amount of easily accessible texts in different languages on the internet grows daily and so the effort and need to organize these texts grows as well. An automated process is needed to extract useful information, like a sentiment, from this amount of published text. This paper deals with the extraction of a sentiment, divided into three classes, from tweets written in German language. Different machine learning algorithms and a variety of preprocessing steps are compared to find the optimal combination. While most work in this field aims at Englishlanguage tweets, this paper adapts and transfers these ideas to Germanlanguage tweets. The results are compared to the findings of other projects designed for English-language tweets. Further it is shown that the impact of the chosen feature encoding on the results is most significant.