Findability with tags: Facets, clusters, and pivot browsing

For a while I have been thinking of different ways of supporting finding information with tags that go beyond tag-clouds. There are three trends that are worth pointing out.

1) 7_06_facets.jpg
Faceted Browsing Metaphor
The first is faceted browse interfaces. A good example of such an interface is You can start browsing by picking any of the facets (price, region, type). For example, you might start by looking at wines from a particular region. Next, you decide to view by price. If you don’t like that option, you can change the facet, and start looking by type. You are in control every step of the way. In many ways, this could be the holy grail for tagging systems.

The metaphor that comes to mind for faceted browse is that of driving a car. You are in control, you decide when to turn right, or when to turn left, when to stop.

There are several different ways of going from tags to facets. RawSugar, Siderean (, and are different ways of going from tags to facets.

2) 7_06_recommenders.jpg
Recommenders & Clustering Metaphor
2) The second is algorithm driven approaches like clustering and recommendations (often driven by collaborative filtering). Although clustering and recommendations are very different from each other, I am putting the two together, because in terms of user experience, there are similarities. The control is taken away from the user and rests in the system. The system decides what items should be presented to the user (in the form of clusters or as personalized recommendations).

The metaphor that comes to mind is riding on a roller coaster. You are not completely in control of the experience. But its kind of fun, and you go along for the ride.

(Warning: take the navigation metaphors, like all metaphors, with a grain of salt.)

Given that clustered search results have never really caught on for web search, I had been skeptical whether clustering makes sense with tags. However, Flickr’s success with clustering makes me hopeful. has experimented with recommendations. I personally have not found them very useful, but I think thats due to certain interface issues.

3) 7_06_pivotbrowsing.jpg
Pivot Browsing Metaphor
3) The third possibility is one that is native to tag based systems – and can be termed “pivot browsing“. I first heard the term in a presentation by the folks at dogear and it captures very well the nature of browsing on bookmarking systems. They describe it as “The ability to reorient the view by clicking on tags or user names, called pivot browsing, provides a lightweight mechanism to navigate the aggregated bookmark collection.”

When everything (tag, username, number of people who have bookmarked an item) is a link, you can use any of those links to look around you. You can change direction at any moment. Its a lightweight browsing mechanism, less structured and more freeform than browsing traditional hierarchies or even browsing faceted systems.

The metaphor that comes to mind for pivot browsing is walking in the forests or some other open space, stopping to smell the pine, taking a break. You get the lay of the land as you walk around. The point is not just the destination, the point is the journey itself (anyone who has wasted time looking around on one of the tagging system will know what I mean).

You could also consider plain ol’ search as another way of finding information. While search is important, I am not sure that how different tag search is from traditional search (at least from the user’s perspective). If you disagree, please tell me why I am wrong!

Tag-Clouds and other tag visualizations: How could we discuss finding information in tagging systems and not mention tag-clouds. Its the classic visualization related to tags (if anything related to tags could be called classic!). And it serves a very important purpose. It lets important stuff (as defined by frequently used) bubble to the top. While there are a lot of criticisms of tag-clouds, overall I kind of like them. However, I do believe that both from a cloud-content and visualization perspective, much can be done to improve them.

To stretch the already stretched navigation metaphor, viewing a tag-cloud is like getting the 30-thousand foot view from the airplane. You can see some of the important landmarks, but you need to get much closer to really understand the lay of the land.

I have spent some time thinking about facets and tags and will focus on that in a series of posts. I recently spoke to the CTO of RawSugar, Frank Smajda, a startup that I have been following for some time, and the first post will be on Raw Sugar’s approach to garnering facets from tags. (Please let me know if you see any other promising approaches of getting at facets from tags).

7 thoughts on “Findability with tags: Facets, clusters, and pivot browsing

  1. is an excellent example of faceted browsing. By applying multiple facets, I can easily target my browsing to only those clothes that are A) my size, B) in my price range, C) for males, D) the type of clothes I’m interested in (brand, style or color).

    More importantly, it will remember these selections after I’ve added something to my shopping cart so I can Continue Shopping within the same facets.

    For this particular type of e-commerce, it works wonders.

  2. Yusef, I read your paper. Very interesting approach. I really like the idea, the tag-cloud could do with some visual changes. Thanks for pointing it out.

  3. A very interesting article related to search. I might not find the examples the best but I definetely understand the key message you’re bringing up and I’m definetely learning a bit more on methods of implementations to address different user scenarios. :)


  4. Thanks for this post, Rashmi.
    As for the Clustering/Recommendations: You are absolutely right that the user gives away the control to an algorithm. I therefore think the key lies in presenting the recommendations so it makes sense to the user. I think that’s why delicious’ recommendations don’t really work. While trying out delicious’ recommendations I was sitting there with the question “why is this recommended to me?”

Comments are closed.