With the burgeoning development of Web-space, it usually takes a few seconds for the news to be widespread. Especially, for some sensitive or widely popular events, they probably draw attention of a large amount of people who create a storm of public opinion online. Mean- while, for journalists, organizations and relevant parties, it’s necessary and obliged to flexibly analyze the user’s activities and comprehend the general public opinion over specific topics. YouTube as the most famous and representative social platform, which provides us such possibility. In this paper, authors mainly propose a general solution to user discussion analysis on YouTube. Combined analytic techniques and different approaches including the state-of-the-art models are being applied. In our experiment, the case of American police misconducts in dealing with a black man is taken as our example for analysis. Temporarily our model only supports English language text.