This paper introduces Czech subjectivity lexicon – the new lexical resource for sentiment analysis in Czech. The lexicon is a dictionary of 4947 evaluative items annotated with part of speech and tagged with positive or negative polarity.
We describe the method for building the basic vocabulary and the criteria for its manual refinement. Also, we suggest possible enrichment of the fundamental lexicon.
We evaluate the current version of the dictionary by implementing it to the classifiers for automatic polarity detection and compare the results of both plain and supplemented system.