People use Twitter for different reasons, e.g. business, personal, sharing information or emotion, etc, and broadcast tweets of different nature. You will analyse data on Twitter to develop a taxonomy of the basic types of tweets. You will then collect tweets that belong to the classes in your proposed taxonomy and investigate the language usage across different types (e.g. personal messages probably start with pronouns such as 'I' or 'my'). After collecting an initial set of lexical (words) and syntactic (phrases) clues, you will implement a classifier that will automatically map tweets to the most appropriate class in the taxonomy and evaluate it in terms of precision, recall and F-measure.