
This dataset consits of 901 sentences sampled from Polish Dependency Treebank "Skladnica" (http://zil.ipipan.waw.pl/Składnica).

Selected sentences were identified as containing at least one sentiment word from sentiment dictionary from http://zil.ipipan.waw.pl/SlownikWydzwieku

The sentences were verified for presence of sentiments and their targets. Many of the sentences do not contain in fact any sentiment word according to human annotator.

In a sentence, there is only up to one sentiment-target pair. In consequence, there are 975 annotated sentences as some of them appear twice (eg. there are two sentence-target tuples). 

Annotation of sentiments and targets is contained in the last CONLL column. Sentiment words are marked with an "S", target words with an "A". 

The "isAtrRelToSent" means as follows:
   1    - attribute/opinion target is related to / modifies sentiment word
  -1    - no attribute word selected for sentiment word matched with an "S" 
   0    - no sentiment word is present
