Wednesday, June 1

Twitter Releases Search+ Relevance based search

Today marks a significant milestone for real-time search. The results are now ranked based on relevance instead of purely based on recency. The announcement was made by CEO Jack Dorsey at the All Things D conference earlier today. A key important feature is that the new search incorporates rich media results, as mentioned on their blog announcement,
Not only will it deliver more relevant Tweets when you search for something or click on a trending topic, but it will also show you related photos and videos, right there on the results page. It's never been easier to get a sense of what's happening right now, wherever your curiosity takes you.
Danny Sullivan has an article covering the release and what "relevance" means in the context of real-time search:
Relevance for us today is using a combination of signals, your follower graph, who you follow, who’s following you. Another aspect is just looking at the content itself and the resonance of the content,” Mike Abbott, Twitter’s vice president of engineering.
The Twitter Engineering blog has on the update has more detail on the evolution of Twitter search since the original Summize days. Here is a small excerpt on what is needed to provide personalized relevance and filtering:
  • Static signals, added at indexing time
  • Resonance signals, dynamically updated over time
  • Information about the searcher, provided at search time
The post has more details worth reading on the infrastructure that goes into the search
For more news on twitter search be sure to follow @twittersearch.


  1. Remember our conversation at UMass back in 2008, in which you were lamenting the state of traditional IR? You talked about how traditional IR focused on topical relevance, when what was really important nowadays was user relevance. Right?

    What is your view of Twitter search with respect to user relevance? Is it meeting that goal? I'm not asking whether or not it is using ranking signals other than topical signals. Of course it is. It's using who you follow, who follows you, recent activity/popularity, etc. My question is whether those signals, and the algorithms that tie them together, are correctly oriented toward the appropriate Twitter searcher user model. In other words, what is a Twitter searcher really looking for, when they search? Are they looking for recent events? Are they looking for popular events? And is their information need precision-oriented (looking for that one, known-item, "best NDCG@3" tweet)? Or is it recall-oriented (looking for the best four dozens tweets, that all together tell a story of a conference, a natural disaster, etc.)?

    In short, before you begin to design an algorithm to satisfy user relevance (rather than topical relevance), you have to understand what it is the user is actually trying to do. And I don't think I've ever seen a study on what Twitter searchers are actually trying to do, what sort of information they're trying to find, and why.

    Do you know of any such studies?

  2. Jeff, I tried Twitter search after the update and have to say I find the filtering a bit unpredictable. One of my main use cases for Twitter is a vanity query, and I find that mentions return different results than a search for my user name -- and a search for a boolean expression that ORs my user name loses some of those results. Perhaps this is an undocumented feature, but at best it's confusing.

    Would love to hear what you can share -- but of course I understand that you have limits there.

  3. Jeremy - thank you for the comments. There was a recent paper by MSR on some analysis of twitters searches. Many are focused on people (what ashton kutcher is doing now...) and events like WWDC.

    There's still a lot that needs to be done!

    Daniel - I'll look into it. In the meantime, you can switch between search modes using the dropdown menu - instead of "top" select "all" in the UI. Let me know how that works.

  4. Many are focused on people (what ashton kutcher is doing now...) and events like WWDC.

    And so what is the open research question? Because it seems like if the focus is on @people and #events, then the @ and # solves 99% of the problem, does it not? e.g. a simple boolean match for a single word query @akutcher or #wwdc, with a little bit of stemming (#wwdc2011, #wwdc11) is all that is needed in order to satisfy what it is that people are actually doing with the system, correct? So if that is what the primary user need is, that actually means that not a lot needs to be done...

    Or are you thinking more along the lines of user needs starting to appear, once capabilities appear? That adding search options will actually drive new and interesting user needs, instead of the other way around?

    p.s.: I don't get broadcast television, but would still really like to see you on Fox. Any idea how I can do that?