The Problem with Parallax…and Freebase

Earlier this week, David François Huynh unveiled Parallax, a new search front-end for Freebase.

I tried out the site, and was initially impressed by the ability to quickly generate charts and maps. “Ooh, neat.”

And then I promptly left.

I wasn’t the only one to have this reaction, it seems. A couple days later, Read Write Web’s Marshall Kirkpatrick wrote up a piece that used the shortcomings of Parallax as evidence of Freebase’s failings:

"The example above is nice, but let's say I want to find out something
 about black women scientists. No luck. History of the internet? 
Not much information there. Venture Capitalists? Blank profile pages....
We love the idea of the semantic web, but give it's grand daddy website
 a usable UI like Parallax and we're left questioning just how much there 
really is inside Freebase anyway."

Freebase has some challenges, and the largest one is data induction. As more topics and relationships are introduced to Freebase, it becomes more useful to end-users and developers.

Induction is a major problem, and will likely take years for a significant amount of the world’s semantic information to be induced into Freebase.

But there’s another major problem that is keeping the bounce rate of both Freebase and Parallax too high.

People don’t know what to do once they reach the site.

Granted, Freebase has improved drastically in this regard, and their homepage now offers much more direction than it did just a few months ago.

But Freebase is obviously resisting the temptation to create a Popular/Recent Ranking on its front page. I’m sure that more than a few of the company’s 70 employees have considered a digg-like homepage. But the Freebase DNA resists the lowest common denominator. It resists the reality that 80% of site interactions are likely to be limited to 20% of content.

And the site usability suffers for this.

Even without the drastic move of a new homepage, perhaps there are some ways that Freebase (and Parallax) could improve search results with incremental UI improvements.

Linear ranking is one way to accomplish this, using the aforementioned indexes of popularity or recent activity. But there are also simple visual metaphors that could help searches yield more fruitful results, with just a glance.

UI-Patterns describes tag clouds as “a list of tags, where the font size of each tag is larger or bigger depending on its weight.”

Perhaps visual metaphors could help both Parallax and Freebase show search users where the action is, even if a simple “tag cloud” wouldn’t be appropriate. For instance, topics with more filled out properties could be highlighted, while empty topics could be smaller and grayed out.

I’ve tried to create about a dozen charts on Parallax, and most of them failed for lack of data. If i had a way to improve my “glanceability” factor of what I could easily chart, I’d be much less likely to leave the site out of frustration.