Adapting sentiment analysis system from English to Slovak

Cross-lingual adaptation is very important and challenging for minor languages. This paper proposes an adaptation system for sentiment analysis based on dictionaries. It is a difficult and time-consuming task to adapt a dictionary from major languages such as English to minor languages such as Slovak. This system uses evolutionary algorithms to replace a human annotator during the labeling process. We compared a human-labeled dictionary with the dictionaries labeled by Particle Swarm Optimization and Bare Bones Particle Swarm Optimization. These PSO labeled dictionaries outperformed the human labeled dictionary and assigned better polarity values to the words in the dictionaries.