Tom Chi  

Grokker 2 in the News

January 6th, 2004 by Tom Chi :: see related comic

CNN ran a story yesterday on Grokker 2. It offers an interesting way to visualize and interact with search results. You can download it from the Groxis site list.gif

I’m a big fan of data visualization so I was excited to try out Grokker. As it stands, however, it has a long way to go before it can compete with Google. First off, the “circles within circles” interface makes it difficult to scan results quickly, and the positioning of circles seems arbitrary. The circles *do* provide visual clues (bigger circle = more results), but the circle scaling appears to be logorithmic, or some other non-obvious scale. The application also offers a “squares within squares view”, which addresses some of the text scannability/readability problems, but doesn’t offer as many visual cues.

Moving on… The colors are random. Actually to be accurate, results turn gray when they are excluded by subsequent filtering, but other than that the coloring is random. There is also very little communicated in the positioning of the circles. Position could have easily been used to express relative rank, or any number of informative properties. Overall, the map is fun to play with, but it doesn’t offer much in the way of data visualization (it’s hard to be good when you don’t make sensible use of color and space).

As for the quality of the actual results, it did do a good job at offering breadth. A search for “tom chi” turned up some DJ listings for me in the Seattle Weekly that I hadn’t seen before. On the downside, about 40% of the results were to Vietnamese sites - many of which returned 404s. These circles appeared quite large, but offered little information. Taskwise, it also took me longer to find things, though I did find it quite enjoyable to zoom in and out of circles. The most sucessful part of the interface is probably the filtering box on the bottom. It was very speedy at paring down the results against additional search terms.

Overall though, Google has nothing to worry about. Grokker certainly isn’t faster to use, and the visualization provides little to improve my understanding of the search space. With a new visualization approach (better use of color and space) and interaction model (simpler to scan, elevate filtering), they might eventually become a compelling alternative, but there is a long way to go.

I’d be interested in hearing comments from others who have installed this software. What do you like about it? What can be improved? Maybe Groxis can get some free interaction design advice.

7 Responses to “Grokker 2 in the News”
Ryan Shuya wrote:

I installed Grokker, and I too was impressed with the filtering and “circles in circles” concept. I did come across the same pitfalls, ie, the circles are hard to scan, and my results were not Google-esque. I find that when I use a search engine, I don’t care about categories; I want to click the first link and get the information I require. Google is very good at this, and I rarely have to scroll the results to useful information.

gary wrote:

There will soon be a Google plug-in. The 404s are from the search engines, not Grokker. Grokker is for (re)searching. If you are looking for something specifically, then you want to Google. But when you want to did, you use Grokker. Remember, Google only goes down several hundred and then stops. Grokker can go down several thousand. If you don’t like circles, try the tiles (square).

Juan Casares wrote:

The vivisimo clustering helps to divide and conquer as you search. It has no fancy visualization and a slow response time, but you can actually find things faster. (

Tom Chi wrote:

From Gary’s comment, it looks like the jist behind Grokker is not “getting an answer fast”, but more based on “researching a space thoroughly”. I’ll certainly keep an eye on it as it develops, though as per my review, there are still quite a few interface problems with it.

Bob wrote:

I have played with Grokker 2 myself and I can tell you that your original post contains several inaccuracies. First, the size of categories is meaningful: it tells you how many documents or web pages are actually contained within the category and its subcategories. So you can easily find out what concept is the most widely represented on the Web for your query. Second, the colorization can be turned on by clicking the color button in the filter panels at the bottom and it can tell you useful things as well. Third, the placement of categories and documents is alphabetical, not random. In fact, try zooming on a category then click its label and type a different one. You will see that the category’s position changes automatically as you type. I agree with Gary above: Grokker is for doing research, not quick search. I also use Google a lot. Both have their advantages and it is silly to try to put them against one another. I look forward to the Google plugin (supposedly in mid-February).

Tom Chi wrote:

To quote myself:

“circles *do* provide visual clues (bigger circle = more results), but the circle scaling appears to be logorithmic, or some other non-obvious scale.”

I never said that the size of the circles had no meaning. I just said that their scaling was non-obvious. Imagine if the Grokker design included a number next to each circle (in parenthesis) to indicating how many results were in that circle. That sort of design tweak could allow people to understand how the sizes of circles were assigned. As it stands, do circles scale logorithmically?…Some other scale? I can’t tell. This might sound trivial, but compare this to another possible visualization: a bar chart with the categories with the most hits at the top. That visualization would give a very clear indication of the relative size of result buckets… the circles give some clues, but are not as successful at communicating this data.

In terms of colors. What do the colors mean? It certainly wasn’t clear to me why a green circle and a blue circle were inside a red circle next to a very large mauve circle, etc. Usually when color is applied to visualization, there is some obvious system. For example changes in color value could map to search relevance or categories of colors (warm, cool), could map to some attribute (results with keywords in order, results with keywords scattered). As far as I can tell, the colors in Grokker are applied more as one would apply a “map-coloring” (in the mathematical sense). While a mathematical map-coloring allows you to see the border between regions, the specific color does not matter. This is all fine and good for math, but it’s a shame when it comes to data visualization, because color can be such a powerful axis with which to communicate.

As for the order being alphabetical?
Let’s look at the screen again:

A user scanning across the top will see: “Weblog / Others / Archive / Ca / League”. It does not immediately strike me as alphabetical. If you look in the bottom left you will also see “Trong / Tom / Viet”. Also not in order. That’s all you get. Many users will scan a couple spots quickly and conclude (incorrectly) that the results are not ordered. Even though the user makes the “mistake” it is the designer’s responsibility to address such confusion. BTW, the second ordering problem is due to the fact you are using unicode but not ignoring diacriticals for sort order. No need to get into a long discussion about that, but as a rule of thumb, the cross language results will come out better if you ignore diacriticals.

I want to stop for a second and say that I have nothing against Grokker. I think it’s pretty cool and that it may eventually be a great search or research tool. My evaluation mostly spoke toward the shortcomings in the approach to visualization and in the current user interface. This doesn’t mean the service is bad - (e.g. I like my camera even though its interface is not optimal). But given that Grokker is still evolving and has a chance to improve, I think we should tailor the comments toward helping it improve. With some luck, these comments will get over to the Grokker designers and they may rev the interface to a level of undeniable awesomeness. That would be a great outcome.

As for the comparison to Google, I only compared it in this way because the original CNN article labeled Grokker a “Google Killer.” My comments went along the lines of investigating that claim. If it is a “research” tool that does not fall into the same category as Google, then which tools do? I’d be happy to evaluate it against a tool in its category. Honestly though, I (and many others) use Google for in-depth research into topics, so I’m not sure why they couldn’t be compared.

rudy wrote:

I took a look at the tool and wrote an aricle about it
basically saying how I liked it and what I would like to see
Rudy D

Leave a Reply

OK/Cancel is a comic strip collaboration co-written and co-illustrated by Kevin Cheng and Tom Chi. Our subject matter focuses on interfaces, good and bad and the people behind the industry of building interfaces - usability specialists, interaction designers, human-computer interaction (HCI) experts, industrial designers, etc. (Who Links Here) ?