During the last decade Twitter has undergone a massive growth and several algorithms have been proposed for analysing its content. A common challenge is identifying user interests. In this work, we focus on personalizing the consumption of tweets by taking into consideration the tweets profile of each user (average tweet). In particular for each user an average vector is estimated that maps her/his interests. Then the distance between tweets that are posted on a user’s wall is estimated. In this process the average tweet is also considered. Finally a graph of tweets is constructed and the minimum distance path is estimated. Experimental results on real life data exhibit the advantages and limitations of the proposed scheme.