Spatial Hashtags in Tweets

Twitter is an on-line social networking service which enables users to communicate by sending and reading up to 140 characters short text called ”tweets”. Users can attach any hashtag, starting with an ”#”, to tweets to indicate additional or summary information. tweets with explicit or implicit spatial intent can be classified as spatial tweets. Spatial tweets are commonly seen in practice, and it is important if we can discover such tweets and further understand the spatial information behind them. However, recognizing spatial tweets is very challenging since there is usually little spatial information in the short tweet text. In this paper, we design a Bootstrapping framework to automatically mine spatial hashtags from tweets corpus. This framework consists of three key components: PU Learning for classifier training, spatial tweets discovery and heuristic spatial hashtag extraction. Then we will show how to mine spatial hashtags from tweets with this framework and how to recommend spatial hashtags to tweets.