We present an approach for tackling the tweet quantification problem in SemEval 2016. The approach is based on the creation of a co-occurrence graph per sentiment from the training dataset and a graph per topic from the ...