Extending Twitter Real Time Search

I and my friends at Institute are Twitter fans. We are inspired by novel opportunities that amazing new medium - Twittersphere - naturally provides.

Today a lot of people talk about real-time search upon Twitter. Most of them understand real time search as a new possibility of getting breaking news quickly. Bright examples are January plain crash at Hudson and Mumbai terrorists attack. News about these events outrun CNN, and was received straight from the tin. Of course, that is a wonderful thing about Twitter, but we can get even more if consider real time search widely.

If something is happening in the world it makes buzz on Twitter. The buzz especially intensified due to Twitter-native retweet feature. So, events that happen in the world coincide with activity bursts on Twitter (I mean something more global than @mariagrineva :Just bought new mega cool gloves), and that not only true for events as big as terrorists attack, but also, for example, Damon Albarn presenting his new single on British radio. This feature of Twitter media should be used not only to capture breaking news, but to sort out important events related to the search query that took place any time ago.

We have built a TweetSieve demo system that improves search over Twitter: instead of presenting ALL tweets sorted by date to the user, it identifies important events related to the user query and present tweets that best describe these events.

Here is a snapshot of a TweetSieve that we are going to present at SIGIR conference this year. The user types in her query. The system then shows the period of events occurring for the  query subject and outputs tweets that best describe each of the events. Snapshot shows a case for “Semantic search” as a sample query.

 

Basically, TweetSieve does the following : (1) it identifies activity bursts in Twittersphere related the user query and (2) selects tweets that best describe news events using algorithm similar to this one. More details are in paper (Sifting Micro-blogging Stream for Events of User Interest), I will make it public after the conference.