Identify aggressive comments on youtube using LSTM model.
We used Selenium and Beautiful Soup to crawl comments from youtube videos based on keywords. We manually graded each comment in to 3 scales: (1) normal (2) slightly aggressive (3) very aggressive.
In order to train model on the collected dataset, we use the following steps to tokenize natural language into numerical vectors.
