Ever bosses to do related projects, the use of VADER to tweet feature extraction or textblob pack? I saw some paper illustrates the need for pretreatment, some feel do not need again, can you give some advice