From the NYT: “Supporters of non-text-based representations of Web search results got a boost this week as Groxis, the makers of Grokker, released a version of the software that runs as a Java plug-in for browsers. Previously, the software, which returns search results in a circular “map,” was only available as a separate, $49 application. The company will now depend on revenue from advertisements placed next to search results by search engine Yahoo.”
For quite some time, Greg Newby (Arctic Region Supercomputing Center and Project Gutenberg) and I have separately envisioned a search engine that employs a virtual info-cell that contains all possible information, lets new information in based on relevance, and provides weighted connectivity to other related topics. On first glance, Grokker appears to be such an animal, but is really a better results visualization tool, i.e. it displays results in a limited space in the form of a zoomable tree structure (circles within circles – zoom to a particular circle). What I thought were information-sensitive areas of overlap / Venn diagrams turned out to be visually-appealing pods of Google-esque results.
Grokker also borrows from the Wiki mindset in that the user can add links within information circles, move information between circles and organize the layout to a certain extent, should said user choose to spend a few hours wrapping his/her head around what goes where. It’s a value-add, but not much of one if all you’re doing is cleaning house, in my opinion. Do you think the “cleaning house” capability is good or bad? What if someone came along and created/rearranged/deleted links based on their own logic and screwed up the search engine’s reliability? I’d like to hear your thoughts on this larger data organization issue.
I tried Grokker on “maitri” and was quite satisfied with the way the results were organized under Space Awareness, Healthcare, Foundation, Universal Love, Bay Area and General. Of course, when I explored the Space Awareness section, I discovered that this had naught to do with space stations, but is a method of identifying one’s chakras. What about the Maitri Antarctic science outpost? Harumph!
How does this software fare with a topic of scientific study? Grokker on “structural geology” (a topic with which I am quite familiar) is a mess because thousands of results came up in the wrong pods. For instance, a lot of material was derived from classroom instruction or class webpages. Wonderful – that’s actually a great source of science information. Unfortunately, all of this material was organized willy-nilly under useless subheadings – a listing of jobs under Faults, a lot of real structural geology pages under Other Resources.
These experimental anecdotes show that results weighting still relies on meta tags, the care with which owners market the contents of their site, and page hits. The Grokker engineers still haven’t figured out WHY that page is visited, in other words, for WHAT information. Additionally, there is no way to push the search further in a particular direction – e.g. faults – (minus) software + (plus) kinematics + (plus) field. While what I expected was more refinement in the form of less results, what I got was a large jumble of useful and irrelevant information. Some users like and may benefit from such reaches, but I don’t, especially not when the results are grabbed out of the ether without relevance weighting. Fear not – there is room yet for the brainchild of Newby’s and my desire.
Since D is search engine king with queries in the realms of computer hardware, food, libations and vacation, I’ll have him send us a report on how Grokker handles his requests (update: see comment below).
Please try Grokker in place of Google and let me know of your experience.
OK, here’s my opinion. Used both for two full workdays. Note that I downloaded “mygrokker” trial to play with.
For “work”, e.g. seriously have to find something fast, or relevant, or specific: GOOGLE. Why? Because of dominance of “relevance” and the
“organization” layer of grokker slows you down.
Well, I grokked Anne McCaffrey, and (after I installed the latest version of java so that the darned thing would actually work) the results were pretty much in a completely uselessly disorganized bunch of rings. I’m not impressed with how the links are reported by only the domain-level url (pretty useless when the page/site its referencing is part of geocities, for example), and I couldn’t figure out how to actually follow a link to go anywhere (expect for typing it in a separate browser window). On the upside, I was happy to see my websites mentioned all over the place in the results. That book I read on search engine optimization did me good!
I found grokker lacking in results and presentation. I like lists; vivisimo works for me because it takes a lot of results and organizes them into easily-digestable lists. I use search engines frequently and I use them well. My parents, for instance, do not. I found
grokker’s navigation too ponderous and I think that my parents would be completely lost. Who is this aimed at helping? Yes, people are visually oriented, but unless those colored circles can be instantly recognized as meaning something specific, this type of search has little benefit. It might work better if they had easily recognizable icons, but how many icons are you going to have and how long will it take for people to recognize them? Maybe something in between, like you see on the slashdot main page?
D