The current work collected a dataset on the interaction between people to be used in
future research on sentiment analysis. Based on messages sent from an individual to
others, a crawler is build to able to identify individual with high likelihood of response.
Based on a random forest model that analyzes features in message and frequent term
count analysising the text body, the crawler was able to detect replyied individuals
with 75% of acurracy. This allowed us to build a dense and strong connected social
network and thus can works for more detailed analysis social researches.
