Abstract
Twitter - well-established social network - becomes extremely popular worldwide and in the Kingdom of Saudi Arabia. The spread of smart phones and gadgets helps simplify the accessibility to Twitter and thus, increases the popularity overtime. In Twitter, users follow other users whom they are interested in their social contributions "tweets". At many occasions, users resort to go over other user's tweets to decide if the user is interesting enough to follow. Clearly, doing this manually does not scale when there are many suggested users to be followed. Thus, it would be very useful if there is a tool that would automatically profile users and extract different characteristics of them. In this paper, we target one sort of Twitter Profiling problem by attempting to extract the interests/topics of the user behind the tweets. Building an automatic interest/topic categorization tool would be helpful for users and companies alike. For users, it makes the decision to follow a user way easier for them as it gives them an easy way to know the topics of the accounts which could be an important factor when following. For companies, it gives them further information about the Twitter users, which could lead to sending relevant targeted ads or marketing related items. We devise a machine-learning based topics-classification technique of Twitter users from merely their tweets. We target tweets that are in the Arabic language with Saudi dialect. Our technique achieves an accuracy of 90%.