The sentiment data consists of wordnet synsets 
 annotated for subjectivity. 
The format of the dataset is the following: Each line contains a wordnet
synset with its subjectivity annotation of either subjective (S),
objective (O) or both (B). The line is structured as follows: 
First,the WordNet synset number of Wordnet 2.0 is given,
then the part of speech, then
the subjectivity annotation and then the synset with its gloss.  
All information refers to WordNet version 2.0. 


The use of this data
is free for research purposes. Annotation scheme and  agreement
are described in Su and Markert: From words to senses: a case
study in subjectivity recognition. Coling 2008. All studies that do use
our data have to cite this paper.