“Innovation has nothing to do with how many R&D dollars you have. When Apple came up with the Mac, IBM was spending at least 100 times more on R&D. It's not about money. It's about the people you have, how you're led, and how much you get it.” - Steve Jobs

Functioning Form – Looks at Clustering of search results

Luke in one of his latest pieces talks about how Google is adding subtle clustering models to their results. I think it is all well and good, but what about the already good examples of clustering that are already out there?

Now someone could say these don’t work that well .. and btw, i don’t use them, so that must say something about HOW they are doing the clustering, but it sounds like the big boys are just ignoring the good work of others.

But I would ask the question, is “clustering” really the next way to make search results more meaningful and lead me to the end result of finding what I’m looking for. It does help to narrow the results, but does clustering help in the way that make findability better.

No answers, just questions.

Be Sociable, Share!
  • http://www.lukew.com LukeW

    Dave, I posted an outline of some of the sites doing clustering a while back: Visual Information for Search Results. So there’s lots of examples out there. But the big guys seem to feel the technology is not there yet- though coming soon.

  • http://www.whynotsneeze.com/blog charles h

    What’s annoying to me about the Clusty/Vivisimo clusters is simply the fact that they’re extracted from the text, so given the variability of language, there is conceptual overlap, differing degrees of granularity, trivial concepts, and so on. The Google model seems to be better in that it focuses on “high probability” clusters.

    I think within specific domains — as on a corporate intranet — the Vivisimo clustering would be more consistent.

  • Kelly G

    I did not find google’s new results interface as helpful as Clusty’s interface.
    I prefer to see the clustering as a list of options that I can then select from. Google still requires the user to scroll down through a list to locate the particular “cluster” you wish to see. Users do NOT like to scroll; many studies have shown that users do not go beyond the first page and often not beyond the first couple of results! Of course, any of these approaches are less effective when you are searching millions of records.

    Another helpful interface is faceted searching. However, faceted browsing is difficult to do outside of a rather specific area.


The archives run deep. Feel free to search older content using topic keywords.