"text processing for sentiment analysis"

i would like to process my text file(which contains messgaes posted from twitter). this file has as features ID, text and  class( category).
this processing consists on eliminating    #topic, @username, then for each message, eliminate no english word, and tokinze it i don't like to consider file as one document, i would like to process this file line per line  because i need it to classify message into categories (classification)
thanks for your help