Paper Reading 23 – Automatic Generation of Research Trails in Web History

Comments
Patrick Frith
Paola Garza

Reference Information
Automatic Generation of Research Trails in Web History
 - Elin Rønby Pedersen, Karl Gyllstrom, Shengyin Gu, and Peter Jin Hong
 - IUI 2010, Hong Kong, China

Summary
Research trails are an attempt at tracking web activity more robustly than simply using bookmarks or view browsing history.  The designers noted how web research behavior is very different from more conventional means of research like in a library.  Users will start at a certain site with sometime a vague understanding of what they are exactly looking for.  Over the course of exploring pages, their path they took can be quite varied.  If they decide to take a break from this activity, there is no effective tool right now to make getting back into the research a simple task.  Users tend to forget where they were, and will waste time trying to find their place.

The research trails concept eliminates this problem by representing browsing history as a path of connected sites.  As the user is browsing, the algorithms the designers use pay attention to the content and decipher when one research path ends and when a new one begins.  Once finished exploring a research avenue, the user can reference their path and easily see where they started and specifically how and when they reached their final destination.  The authors noted in their user study the challenge of how to handle similar research trails.  A possible solution would be to merge similar trails together and open up the similarity threshold within the algorithm.

Google as a giant robot.

The designers were satisfied with the trailing methods robust ability to recognize gradual shifts and research direction and to handle those properly.  In order to make the product more ideal, they would like to add pertinent snapshots to various points along the user’s trail.  Another important feature addition would be allowing the user to tweak the trails by changing adjusting its figures and moving items around.  They noted how these features are important in the final product, but they were not necessary in order to demonstrate the research trails concept.

Discussion
This was a very fascinating and extremely useful concept they explained in this paper.  I have a feeling everyone has gone on some extreme Wikipedia tangent before.  You start by looking at one page and then you see three, four, or maybe more links you would like to investigate.  So you  open new tabs or windows for those.  And then each one of those can spawn multiple new topics to investigate, and so on.  Before you know it, you have filled up an entire afternoon of exploring.  This can be entertaining and educational, but what if you want to take a break or you have other important tasks to accomplish that day?  Bookmarking all of those windows or tabs for later use is an option, but what are the chances you will remember what in the world you were doing and where to start?

Honestly, you’ll probably give up or miss a major avenue of research.  Either that, or you will waste a great deal of time researching on where you left off in your research.  Yes, a random Wikipedia binge is often not extremely important, but it definitely could be.  This is exactly the kind of situation research trails are made for.  The program would be tracking your tendencies the entire time, so you could easily come back to exactly where you left off.  It would also make is easy to back track in case you wanted to at yet another tangent to the research.  On top of this, the algorithm is also designed to recognize subtle changes in research paths.  So, it should be able to recognize separate research tangents if they are distinct enough.  This would further the simplicity of picking up all of the various research trails if you need to set them aside for awhile.

In case it is not obvious, I am very enthusiastic about the idea of research trails.  If it is not already in use, I would imagine it is getting close to becoming available for people to try out.  Since two people from Google worked on this project, I would be willing to bet this will be yet another interesting innovation that will simply appear one day without warning just like the web snapshots within search results did.