Abstract
Hash tag recommendation is the problem of finding interesting hash tags for a user, which are not easily found via Twitter search. Searching a hash tag simply shows a list of tweets, each contains the query hash tag string. To find even more relevant hash tags, we propose to use a graph-based approach to find similar hash tags by using the social network graph around hash tags. We start by using a heterogeneous social graph that contains users, tweets, and hash tags, then we summarize the graph to a hash tag graph that shows the similarity between different hash tags. Finally, we rank the vertices in respect to a query hash tag using a random walk with restart and a content similarity measure. The experimental work demonstrates the effectiveness of our approach compared to baselines.